PDA

View Full Version : Little raw cam side project...



Soeren Mueller
11-27-2012, 06:12 AM
Hey guys.. just thought I'd share a little info about a project I'm working on while waiting for the BMC.. it certainly helps with that ;)

Last week I finally got the Point Grey USB3 camera head (normally for industrial use) which has a 1/2.5" Sony Exmor CMOS built in and takes lenses with C-Mount. I already started to work on my own software using their SDK before that, but of course couldn't be sure about the quality I'd get from the camera because of the lack of any demo footage/raw framegrabs.

http://soeren.de/misc/ptgrey.jpg

Over the weekend I experimented a little with capturing footage - directly to SSD the data rates are no problem, the sweet spot is seemingly 200 MB/sec: the electronics and/or sensor in the camera head can't handle more data troughput unfortunately. The 200 MB/sec is no problem at all for any decent single SSD.

So with that I'm currently able to get a 3200 x 1800 pixel raw 12-bit frame at 24fps or when using binning (pixels are connected to groups) I can use the whole sensor for 2048 x 1080 pixel raw 12-bit at max 60fps. At 3,2K the rolling shutter is about the same as with a 5D Mk2, at 2K using binning it is (thankfully) nearly non existent (roughly on par with Red One MX)!

The max resolution if you go with only 8 bit is 3728 x 2096 at 24fps (using almost the full sensor), but then you obviously got very little headroom for grading.
It's quite fun to have such a little camera head on a 5m cable - reminds me very much of the SI-2K.

The dynamic range is about 11.5 f-stops, so (of course) it's not in the BMC ballpark. But still it's such a great thing to get real raw data and without any sh*tty camera electronics or compression in the way even at 24db analog sensor gain (max) you get quite a decent picture - of course theres noise but it looks quite organic.

Another thing I have to take care of myself that way is bad/stuck pixels, every sensor has a few of them. That's the downside to just getting a sensor "dump" basically ;)

Right now I'm working on the code to safe the raw data for each frame in a Cinema DNG container so that I can use the usual "high quality" tools to process the footage. And taking a noise/bad pixel "footprint". We'll see how complicated that will get - if and what kind of profiles or further meta data is required for Adobe Camera Raw etc. to do a good de-bayer.

Keep you posted...

Frank Glencairn
11-27-2012, 06:19 AM
Quite Impressive - looks like, on a weekend you get as far, as the DBolex folks in a year :)

Soeren Mueller
11-27-2012, 07:01 AM
Hehe ;) to be fair that was the easy part.. I didn't design any electronics let a lone a housing for the whole thing.. (head + embedded pc + ssd caddy) .. much left to be done...

Abobakr
11-27-2012, 07:10 AM
This is impressive to hear about it.. interesting to hear about your progress.. Please continue!! Oh don't forget to document everything

adam777
11-27-2012, 07:12 AM
Sounds great ... looking forward to the footage :)

Brad Ferrell
11-27-2012, 08:35 AM
That's awesome. Keep us posted.

mhood
11-27-2012, 08:52 AM
I demand an update!!!

jesperperl
11-27-2012, 10:06 AM
I demand an update!!!

:D

adam777
11-28-2012, 06:32 AM
Is that the Flea3 FL3-U3-88S2C-C ?

$945 USD for the camera head

http://www.ptgreystore.com/products/213-flea3-88-mp-color-usb-30.aspx

http://www.ptgrey.com/products/flea3_usb3/Datasheet_Flea3_usb3.pdf

Soeren Mueller
11-28-2012, 11:11 AM
Exactly, that's the one!

Still working on the cDNG stuff.. for that I also had a closer look at the publicly available DNGs from the BMCC. There's a bunch of colorimetric info in there which I guess helps Adobe Camera Raw and the other DNG readers to correctly interpret and get the most out of the raw bayer data.
One interesting detail is that it even employs a LUT which is present in every BMCC DNG... this maps the 12 bit (1-4096) values in the raw data to 16 bit values (1-65536) ... although I don't know how strict the readers adhere to that. I guess I'll have to do some experimentation in using a similar LUT and see if it makes any differences in the ACR output or if it's just plain ignored ;)

I made a quick graph of the LUT that the BMCC writes to every of it's DNGs:

http://soeren.de/misc/bmcc_dng_lut.gif

Interestingly this maps more than half of the possible input values to the lower 8th of the output values. There's even a bunch of the 12 bit input that's mapped to exactly the same output values.. so if this LUT would strictly be adhered to the outcome would be more 11 bit like (even a little less). Which is kinda strange. The curve looks like it's used to get back more contrast into the very flat S-Log raw data, so perhaps if this LUT is used it's just the starting point for ACR & co, and by tweaking their parameters you can get to the original flat data again.... weird
And also there are two entries for Black Level and White Level that instruct the reader to clip below 256 and above 60074 (from the 16 bit output value) - again if adhered to but perhaps that's just the starting point for the default settings.

Bart
11-28-2012, 01:24 PM
When can I pre-order?

Soeren Mueller
11-29-2012, 08:43 AM
Hehe ;) .. it's not that kinda project unfortunately... at least nothing planned yet. Have to evaluate the performance of the Sensor. But so far it looks quite nice.

Update: I got the first Cinema DNG recording/converting working.. so far the DNGs only work in a few programs, Adobe Camera Raw won't read them and I have yet to find out why.. something is still missing it seems :/
And I tried to convert them to Cineform RAW using Cineform Studio... it works "a little".. unfortunately the Bayer pattern is different from the BMCC and so far it seems Cineform is always expecting that exact Bayer arrangement... so while import works the colors are pretty off ;) .. but the debayering and color correction in Cineform looks great - I already wrote a support ticket to the Cineform folks in regard to that, let's see what they say. Or perhaps I have to write a conversion for Bayer rearrangement (although I can't do that in realtime while recording as it'll burn too much CPU cycles).

So I have to figure out now what Adobe Camera Raw doesn't like about my DNGs.. really stupid trial & error stuff.

And we tried some really really sharp machine vision lenses from Kowa (8.5mm 1.8) and Tamron (8mm 1.4, 16mm 1.4, 25mm 1.6, 50mm 2.8) today.. all solid metal and really really sharp image! But these things are smaaaaaaallll ;)
Of course wide angle is hard with this chip size... tried a 4mm 1.4 which is about 24-26mm full frame/5D equivalent... gets already a little fisheye-e around the corners (but nothing GoPro like ;) .. but there are a few wide angle machine vision lenses that don't distort much... we're looking for them..

By the way.. although wide angle is hard to do.. the upside of a smaller chip is the amazing near focus range with these lenses. You can do macro shots with basically all of them.. getting stuff 3-4cm in front of the lens in perfect focus. ;)

Soeren Mueller
11-29-2012, 08:14 PM
Ok guys.. here it is.. some of the first cDNG that my beta cam software wrote that would even open in Adobe Camera Raw. The color science is still a little off as I haven't calibrated anything but lent the color matrix from the Decklink ;)

Only 3200 x 1800 at 8 bit RAW for this... 12 bit would have been no problem at all but it was a simple user error while recording some test footage while at Tamron. Still it's not too bad I think! (with 8 bit normally even 3728 x 2096 would have been possible!)

Give it a try in ACR or whatever and let me know what you think!

http://all-digital.de/projects/voocam/jeannie/dngs/test1/

I'll have more/better test footage ready in the next days!


EDIT.. I think this thread needs more pictures ;) little preview.. no "real" CC or anything (try the DNGs yourself)

http://all-digital.de/projects/voocam/jeannie/dngs/test1/preview.jpg

Abobakr
11-29-2012, 08:32 PM
Man.. Great job on what you are doing.. Really interesting to watch your progress.. Keep it going

Michael Bergstrom
11-30-2012, 03:18 AM
Very interesting following this. I saw this sensor block and thought it would be a great project base, but I lack the skills needed to program such a thing. Please keep us updated!

Michael Bergstrom
11-30-2012, 03:23 AM
Just played with the DNGs! Awesome! Lot's of fun and great feel to the image. I opened them up in all my editing programs no problem.

Soeren Mueller
11-30-2012, 08:00 AM
Thanks guys!

I uploaded a few wide angle test shots out of the window today in the highest possible res for motion:

http://all-digital.de/projects/voocam/jeannie/dngs/test1/

However it's totally overcast ugly flat cloudy here today (and I haven't cleaned my windows for a loooong time :rolleyes: )
Still it's interesting to see that despite the tiiiiiny pixels on this chip it resolves quite a fair amount of resolution. Lens was even stopped down to f8!

I'll try to get some "moving" shots once my software is more stable. Bad pixel management still missing...

White balance is still more of a gamble though - have to get a "real" color chart asap...


http://all-digital.de/projects/voocam/jeannie/dngs/test1/street6_3200x1800x12bit_preview_cc_720p.jpg

Soeren Mueller
11-30-2012, 06:06 PM
I also uploaded the first clip I recorded a week ago immediately after I got the camera head. This is really the worst this thing can possibly look like: I didn't have any C-Mount lens so it's a Canon EF lens handheld in front of the little cam head (also handheld).. it was on 2048x1080 (binning), overexposed _and_ had gain activated (thus grainy) plus light of course leaked in between the lens and sensor. So this is max jello, max grain, least resolution possible:

http://all-digital.de/projects/voocam/jeannie/dngs/test1/shaky_overexposed_2048x1080x12bit_720p.mp4

nickjbedford
11-30-2012, 06:15 PM
Very cool! I'm a software engineer in a recent past life and are pretty curious as to how the whole hardware/developing/testing works. Can you tell us some basics of how you're doing it?

Soeren Mueller
11-30-2012, 06:34 PM
Very cool! I'm a software engineer in a recent past life and are pretty curious as to how the whole hardware/developing/testing works. Can you tell us some basics of how you're doing it?

Dito @ software engineer.. but I wouldn't second the "past life" aspect, I still try to do it and enjoy it.. altough I wouldn't want to go back to "only" doing software.
Well I'm using MS Dev Studio 2008, writing everything in C++ and have just finished my own DNG writer/export classes that are optimized for speed. I'm using Point Greys SDK to enumerate (more than 1 possible, hello 3D) and control their camera head, set the sensor parameters and read the sensor data via USB3. The data is "preview debayered" for realtime monitoring and can be written to SSD at the same time. So far no real rocket science involved ;) .. building my own touch based interface a la Cinedeck to control everything.
Got a small embedded PC board that will run everything and now have to find a nice little highres touch TFT...

But there's obviously much left to be done. Next week I'll hit the local rental place to shoot a few color charts so I can finally get a "real" color matrix into the DNGs and hopefully rate the sensitivity under controlled lighting conditions (so I know which db gain is which ISO). And I still have to finish the bad pixel/noise print (for the different gain settings) generation code. In the end what's left is all the display stuff... over/underexposure Zebra, focus assist etc. But I'll leave that for last.
Parallel to that I have just had a look at the Cineform Encoder SDK and it should be pretty easy to integrate that so I can record in Cineform RAW directly, the biggest "hassle" being that I have to get at least a trial license for that from David Newman, so far I haven't heard anything back from them in regard to that...

The "greater idea" here being to built a flexible embedded platform that can be quickly adapted to other industrial camera heads as well - "bigger/better" sensors like the CMOSIS CMV12000 for example - without having to start at square one again.

Cheers :)

nickjbedford
11-30-2012, 06:53 PM
Thanks Soeren. I started in C++ and ended up in C# so they're my two strengths. I have about 8 years of programming experience. I ask because it would be really cool to actually do a similar thing and witness the acquisition of raw images from camera hardware through my own code. It would satisfy both my developer curiousity and my love for photography at the same time.

How would your describe the initial difficulties/learning curve in working with the hardware?

Cheers!

Soeren Mueller
11-30-2012, 06:53 PM
PS: To give you an idea about the size of the lenses... really well built metal stuff for machine vision applications.. it's amazing how sharp some of them are while being so tiny. It's really important to have high quality glass with this little sensor, the wide angle pics on my server were made with a cheap CCTV lens - heavy chromatic aberration, soft edges etc.

http://all-digital.de/projects/voocam/jeannie/dngs/test1/lenses.jpg

Soeren Mueller
11-30-2012, 07:03 PM
Thanks Soeren. I started in C++ and ended up in C# so they're my two strengths. I have about 8 years of programming experience. I ask because it would be really cool to actually do a similar thing and witness the acquisition of raw images from camera hardware through my own code. It would satisfy both my developer curiousity and my love for photography at the same time.

How would your describe the initial difficulties/learning curve in working with the hardware?

Cheers!

You're welcome Nick! I never touched C# .. I wrote my first line of C code 26 years ago (don't ask, I'm only in my thirties so you can imagine how I spent much of my childhood ;o) .. the learning curve depends I'd say.
If you have knowledge in the field of realtime/video processing or hardware access it helps I guess, but the SDK is well documented.

But really the main thing to get anything usable out of this is to get the settings right/fine tuned (hours of experimentation) and to get the raw data into a good post application. The sample code and super basic capture tool that come with the head only save/export in formats that no postpro software "understands".

Roman
11-30-2012, 07:40 PM
This looks like such a fun project. I used to write python for a while, mostly simple stuff, automating tasks for this online company I used to work for. Makes me want to test out Point Grey, never heard about this SDK before. Love your idea about creating an embedded platform for other heads / sensors that seem to popping up everywhere, every 5 minutes.

Would be cool to develop a community spec around a standard housing w/ internals that can accept a variety of off-the-shelf sensors. Camera modding and creation would become such a huge thing. I hope at some point in the future that assembling a camera with non-proprietary parts and firmware is as easy as assembling a computer with chosen parts.

adam777
11-30-2012, 08:28 PM
Hi Soeren.

Have you tested the camera with the free Flycapture SDK that you can download from PointGrey? :confused:

http://www.ptgrey.com/products/pgrflycapture/flycapture_camera_software.asp

rawCAM35
11-30-2012, 08:48 PM
Hi Soeren, have you tried Streampix ?, it records live un-compressed raw or compressed video directly to your PC in real time with machine vision cameras, works with CineFormRaw, it is powerful, it has been used for single camera recording and also dual camera ( synchronized ) for 3D, you have to check with Norpix if your camera is supported yet.

Norpix
http://www.norpix.com/products/streampix5/streampix5.php

Peter J. DeCrescenzo
11-30-2012, 09:00 PM
... I made a quick graph of the LUT that the BMCC writes to every of it's DNGs:

http://soeren.de/misc/bmcc_dng_lut.gif
...

No, no, no. That's the BMCC monthly manufacturing run rate.

We're currently at the lower left corner of the graph (fewer than 256 cameras per month).

:-)

Soeren Mueller
12-01-2012, 05:26 AM
Hi Soeren.

Have you tested the camera with the free Flycapture SDK that you can download from PointGrey? :confused:

http://www.ptgrey.com/products/pgrflycapture/flycapture_camera_software.asp

That's the SDK im using to control the camera ;) .. the capture application that comes with it is pretty basic and only useful for "experimentation" (finding the right register settings and such) and it writes only formats that you can't use in any of the common raw readers/postpro apps.

Soeren Mueller
12-01-2012, 05:28 AM
Hi Soeren, have you tried Streampix ?, it records live un-compressed raw or compressed video directly to your PC in real time with machine vision cameras, works with CineFormRaw, it is powerful, it has been used for single camera recording and also dual camera ( synchronized ) for 3D, you have to check with Norpix if your camera is supported yet.

Norpix
http://www.norpix.com/products/streampix5/streampix5.php

As far as I know Streampix doesn't yet support USB3 cams. And it offers me too little control.. I'm after a nice Cinedeck like touch interface which I've already worked on for some time. But if others want to experiment with this camera head it would be a viable alternative once Streampix supports it.

adam777
12-01-2012, 07:17 AM
Just found an industrial camera with:

5k x 5k @ 32 fps (and up to 100 fps), 25 megapixel using the CMV12000 chip

http://www.adimec.com/en/Service_Menu/Industrial_camera_products/High_performance_cameras_for_the_machine_vision_ap plications/SAPPHIRE_series_fast_high_resolution_CMOS_global_s hutter_cameras

.

Soeren Mueller
12-01-2012, 09:39 AM
Just found an industrial camera with:

5k x 5k @ 32 fps (and up to 100 fps), 25 megapixel using the CMV12000 chip

Yeah I already wrote to them.. haven't heard back yet... and I'll have to find a lightweight PCIe camera link card for that.. next milestone ;)

rawCAM35
12-01-2012, 01:30 PM
Yeah I already wrote to them.. haven't heard back yet... and I'll have to find a lightweight PCIe camera link card for that.. next milestone ;)

I emailed them over two weeks ago, so far no response, this is becoming the norm, so many companies care less to respond to customer emails.

Soeren Mueller
12-01-2012, 09:04 PM
So far I'm now able to directly record to cDNG from within my software while having realtime preview (albeit only black & white) with direct selection of the resolution/framerate presets etc. ... although a few dropped frames here and there, not really optimized yet.

I recorded a short clip a few min ago.. lit by a single tungsten bulb + litepanel micro to check mixed color temps. Converted the DNGs to Cineform RAW, I'm uploading the AVI right now - it's about 2.5 GB so it'll take a few hours and no fast download, but perhaps you want to have a look. It comes with a Cineform Look .fdb file, just double click that to have the clip looking fairly neutral.
It's 3728 x 2096 pixels, recorded at 8 bit. Still even at 8 bit imho much more fun to play with than anything compressed in regard to CC.. and the resolution is quite nice. And there's something in there even for bokeh lovers albeit the small sensor. (and yeah I already noticed I'm in serious need of some hand moisturizer after reviewing the frames)

http://all-digital.de/projects/voocam/jeannie/dngs/test1/video/

single_bulb_3728x2096x8bit_6db_protune_bl_0_045.av i + .fdb

Upload will be finished in about 4 hours.. good night ;)

PS: Lens was a Kowa 8.5mm 1:1.8 which is about the equivalent to 50mm on a 5D or ~35mm on S35

noirist
12-02-2012, 08:54 AM
Very cool! Thank you for sharing your work and the clip.

Soeren Mueller
12-03-2012, 01:48 PM
Thanks :) you're welcome!

New update today... I'll get the embedded PC board this week I plan to use so I can start to optimize my software for that. It's pretty small (10 x 7.2 cm) ...
And I started to implement the necessary functional calls to the Cineform Encoder SDK. So next week I can already see/say if realtime Cineform RAW encoding will be possible. Granted that I get at least a Cineform Encoder trial license until then...

Now gotta find the right touch TFT.. aiming for 7" to 9" at WXGA...

Soeren Mueller
12-04-2012, 07:19 PM
Not that many news today - just tested two really wide angle lenses. One 1.3mm and the other 1.7mm! But "corrected" ones so without the usual fisheye style - looks really interesting, especially for point of view stuff I guess.

Uploaded three little clips made with them.. again more of a low light test, nearly maximum gain (18db, 24 would be max) so it's a little grainy, still holds up imho because of uncompressed. The only lights were two 20 watt bulbs and the monitor - total darkness outside!

http://all-digital.de/projects/voocam/jeannie/dngs/test1/video/

(wide1, wide2 and wide3)

imdjay
12-05-2012, 12:05 AM
ya know, you could think about making the control/interface be run off an iphone/android.

gopro is doing it, the c300 is doing it. albeit wirelessly, it's a neat thought

nickjbedford
12-05-2012, 12:13 AM
Ahh, if I was sitting on a gold mine, I'd fully design my own perfect camera. That'd be an awesome project :D

Soeren Mueller
12-05-2012, 07:42 AM
the embedded board i get for evaluation this week has wireless lan too so... of course it'll get wireless control :D

Soeren Mueller
12-06-2012, 04:11 PM
Ok.. first drawback today... the embedded board I have for evaluation is a really great board and quite fast despite it's size, unfortunately it's USB3 ports can't really keep up with the needed data rate. :( (so only 23-24fps max @ 2k 12bit for example etc)

http://soeren.de/misc/via_p910.jpg

So sad.. because it would be an awesome board. Quadcore 64bit x86 compatible VIA with 2x SATA ports on 7.2 x 10cm... 8 GB ram, running Windows 7 x64 embedded... it's as small as my Samsung 830 SSD!

nickjbedford
12-06-2012, 04:29 PM
USB3 is 5gps theoretical maximum isn't it?

Soeren Mueller
12-06-2012, 04:34 PM
Yes, it is.. but the PT Grey camera head maxes out at ~200 MB/s anyways.. but the VIA board maxes out at ~130-140 MB/s :( .. perhaps it's more of a driver issue. I already wrote to the technical contact I have there... still hoping that perhaps more is possible. The board would otherwise be perfect. :-/

Soeren Mueller
12-06-2012, 05:36 PM
I won't give up.. I've contacted both PT Grey and VIA technical support... the USB3 has got to be faster somehow, I still hope for a driver fix. I benchmarked the SSDs connected to the board - via SATA 3gbps "only" (the board doesn't have SATA 6gbps) - sustained write speed 260 MB/sec, read 275 MB/sec.. great numbers! So recording could be easily handled. And the graphics hardware is powerful enough for preview debayer etc.

Keeping my fingers crossed... oh and with SSD and camera connected it's still within 2 ampere @ 12V! :)

Jorge De Silva
12-06-2012, 05:49 PM
This is going to be fun! :)

imdjay
12-06-2012, 05:54 PM
can you try 10bit to see if that works at higher fps?

Soeren Mueller
12-06-2012, 06:23 PM
you can only transmit 8, 12 or 16 bit (16 bit is the same information as 12 bit as the ADC only has 12 bit) ... will have to wait for feedback from point grey and via... would be awesome if it'll work.. perfect small package (once i found the "perfect" 1280x800 7in touch tft for it)

imdjay
12-06-2012, 06:40 PM
well, i just finished a roast chicken and nibbling on some good chocolate, i'll poke around for some screen options for ya

imdjay
12-06-2012, 06:43 PM
to start, there's the lilliputs, which are all limited to 800x480 native, but handle up to 1080:

http://lilliputweb.net/index.php?Controller=User_Product&action=Index&category_id=53

noirist
12-06-2012, 09:08 PM
I wonder if it's the driver/operating system. What operating system are you running? You need a PF_RING DNA for USB3.0 so the interrupts and memory copies don't kill performance. See http://www.ntop.org/products/pf_ring/dna/

Soeren Mueller
12-07-2012, 03:15 PM
thanks darren.. but i am looking for a bare Panel with an LVDS Connection so I can connect it directly to the board. there are some with 1280x800.. but most of them with glare surface.. Seeking mainly non-glare and with resistive touch.

noirist.. See above, Running win7 x64 emmbedded with the necessary drivers. The Point Grey SDK that provides the Camera Interface comes with a Special usb3 driver for exactly what you mentioned.. to provide faster, more direct hardware access but unfortunately it doesnt Support the via USB controller, Not yet at least. Thus i can only use the original drivers.. which are the bottleneck right now...

Soeren Mueller
12-09-2012, 04:23 PM
Took the board + cam with me over the weekend, doing the first "real" tests of the software and system "in the field".. for early pre-beta state it performed quite well!
But I also wanted to shoot some full bandwith 2k 60fps and 3.2-3.8k highres stuff so I also connected it to my notebook but using my own software (will show some screenshots in a week or so) - I'll upload some more short clips in RAW DNG form shortly.. a few teasers found here, let me know what you think about the IQ (I mainly tested for skin tones this time, mostly natural daylight, couldn't take my sticks or anything else with me unfortunately):

http://all-digital.de/projects/voocam/jeannie/dngs/test1/2012_12_08/

nickjbedford
12-09-2012, 05:09 PM
Impressive detail in the image. Without having a high DR scene, it's hard to tell about that, but the bit depth in the file is very good. Colour is off what I would consider "normal" though, but that is probably calibration stuff, yeah? I had +80 magenta in the raw file, and the white balance was far higher (11,000K) than I would normally see it.

But it has the same feel in Adobe Camera Raw that the BMCC raw files have. Extremely malleable, able to be made "filmic" in quality. It makes you wonder what the hell every other manufacturer is thinking shooting to baked compressed 8-bit codecs. Raw offers so much more for the colourist and it makes the image that much more photographic in nature when processed right.

These have moderately high contrast, a boost of perhaps half a stop or more of exposure, lowered highlights, much lower blacks and negative vibrance.

http://img703.imageshack.us/img703/5513/image1uk.jpg

http://img94.imageshack.us/img94/3972/image2jaj.jpg

Soeren Mueller
12-09-2012, 06:02 PM
Yupp, color calibration data is still more or less missing (there is some written to the DNGs but it's totally off sorry ;o) ... when I get around that in the next 2 weeks I could have the necessary matrices in the DNG so the color temperature settings in Adobe etc. will work. Currently you can basically only start with the white balance manual picker tool...

And yeah - my sentiments exactly... when you experiment with this stuff and suddenly get quite "cinematic" images without all the stupid 8-bit onboard color processing "magic" going on it made me wonder where the problem is - and after seeing what data transfer rates you get with cheap off the shelf hardware it's now pretty clear to me that they just don't _want_ to make such a cam for us (besides BM which are currently trying, too bad they had bad luck with the sensor/glass issue)!
It's already kinda "cheap" when you put all the work in yourself, but it would be even more cheap to mass produce something nearly on par with the BMCC, most work would go into the "color science"/calibration part and the user interface/software... (nearly on par = without the sCMOS it wouldn't have the DR of course) ... the head I'm using has a max DR of 11.5 stops "only"

Seems they consider the market for something like that too little or whatever... I have no idea what their problem is...


PS: Also I'm still missing most of the monitoring options so I have a hard time judging exposure correctly and most of the time overexpose.. so that's also a bunch of work left to do.. but I'm on it ;)

nickjbedford
12-09-2012, 06:58 PM
Exactly. I can bet you bottom dollar that the sensors in most of these cameras are quite alright for a cinematic image, but they choose to use the most useless codecs or require thousands of dollars in "upgrades" to do so (cough Sony cough).

Or they shove 12 stops through a log transform into... 8 bits, when you need to deliver in 8-bits. Surely the manufacturers of 14-bit DSLRs understand why you need >delivery bit depth in grading?!

nickjbedford
12-09-2012, 07:40 PM
More people need to experience raw :)

This is from a light 4WD trip I did with my brother and dad in November. It's his 4WD btw.

Original settings on top, massaged still high quality image from Lightroom. And they wonder why I don't like shooting any video on my 5D Mark III.

http://img33.imageshack.us/img33/3864/screenshot20121210at103.jpg

adam777
12-09-2012, 07:50 PM
Yupp, color calibration data is still more or less missing (there is some written to the DNGs but it's totally off sorry ;o) ... when I get around that in the next 2 weeks I could have the necessary matrices in the DNG so the color temperature settings in Adobe etc. will work. Currently you can basically only start with the white balance manual picker tool...

And yeah - my sentiments exactly... when you experiment with this stuff and suddenly get quite "cinematic" images without all the stupid 8-bit onboard color processing "magic" going on it made me wonder where the problem is - and after seeing what data transfer rates you get with cheap off the shelf hardware it's now pretty clear to me that they just don't _want_ to make such a cam for us (besides BM which are currently trying, too bad they had bad luck with the sensor/glass issue)!
It's already kinda "cheap" when you put all the work in yourself, but it would be even more cheap to mass produce something nearly on par with the BMCC, most work would go into the "color science"/calibration part and the user interface/software... (nearly on par = without the sCMOS it wouldn't have the DR of course) ... the head I'm using has a max DR of 11.5 stops "only"

Seems they consider the market for something like that too little or whatever... I have no idea what their problem is...


PS: Also I'm still missing most of the monitoring options so I have a hard time judging exposure correctly and most of the time overexpose.. so that's also a bunch of work left to do.. but I'm on it ;)

Red one MX has 11.5 stops of DR so congratulations on achieving that!

The increased resolution makes a huge difference.

Just use an external monitor with false colour or histogram to judge exposure. Why try to reinvent the wheel.


Nice work! :)

adam777
12-10-2012, 03:11 AM
vacuum cooled, which further explains the price difference (5-6 times!)

Hi Soeren

Would a low profile CPU cooler work for your camera if you ever look at the CMV12000 (4k sensor) or CMV20000 (5k sensor) ?

http://www.cmosis.com/products/standard_products



Update: The Peltier heat pump that BM uses looks cheaper and more efficient than a cpu cooler.

David
12-10-2012, 04:02 AM
More people need to experience raw :)

This is from a light 4WD trip I did with my brother and dad in November. It's his 4WD btw.

Original settings on top, massaged still high quality image from Lightroom. And they wonder why I don't like shooting any video on my 5D Mark III.

http://img33.imageshack.us/img33/3864/screenshot20121210at103.jpg


Made this from the top half of the jpeg in seconds. It took longer to crop, save and upload this tiny file than it did to do this quick grade .This is from an 8 bit jpg compressed file for the web. There was actually quite a bit of information there and could have pushed it much further if it wasn't for the high jpg compression which was causing posterization. 8 bit is not 12 or 14 bits by any stretch, but it's not the chump it's made out to be. I wasn't trying to exactly match your image just illustrate that there is a fair bit of DR that is unused by most people who have 8 bit files and think the information is lost when it's right there in front of them. No power windows, no dodging and burning, no feathered masks, no nonsense, just a simple grade that could be easily applied to a sequence.

http://img543.imageshack.us/img543/3864/screenshot20121210at103.jpg

nickjbedford
12-10-2012, 06:23 AM
Impressive, David. I just did it myself in Photoshop and you're right, there is more information than I suspected. The biggest thing about raw is that you can do it and not experience posterisation after grading.

I've had success pulling back compact camera JPEGs, though not to the extent that raw files can (my dad sometimes gets me to 'tweak' his photos), I do run into mucky shadows if I'm not careful.

David
12-10-2012, 06:33 AM
By the way Nick you are an impressive young man. Your work on your sight is really good but what is most impressive about you is your attitude towards learning. You are a sponge soaking up knowledge. Don't ever loose that.

nickjbedford
12-10-2012, 07:04 AM
By the way Nick you are an impressive young man. Your work on your sight is really good but what is most impressive about you is your attitude towards learning. You are a sponge soaking up knowledge. Don't ever loose that.

Thanks David. Means a lot! I try to always be a sponge when it comes to learning something.

CaptainHook
12-10-2012, 03:41 PM
Soeren > What you're doing is awesome, keep it up and good luck! :)

Soeren Mueller
12-10-2012, 07:09 PM
Thanks guys! :) ... I won't stop until I have a working portable little raw cam!

Uploaded a few mp4s.. downscaled to 1080p, but two of them are 2K 60fps playing as 24fps (but I still have some skipped/dropped frames at 60fps due to my not-yet-optimized-programming) ... you clearly see the resolution drop but imho it's still usable.. has some 16mm grain/quality to it - and shows how much less rolling shutter the binning mode has!

http://all-digital.de/projects/voocam/jeannie/dngs/test1/2012_12_08/

Sorry for shaky cam/wobble.. didn't have any sort of tripod with me and this little ice cube is much more fiddly/fragile than any DSLR out there! And exposure/focus are still quite tricky with the limited monitoring options currently...

I've got a commercial shoot coming up tomorrow with an EPIC and perhaps I find a few minutes to finally mount this thing on the dolly besides it as well... ;)

Soeren Mueller
12-14-2012, 07:08 PM
Funny... same time with the BMC update.. good news from my little project as well - yesterday and today I had a remote session (screen sharing on the embedded board) with a service/support technician from Point Grey and we were now able to patch their USB driver so that it works with the VIA USB3 controller!
This is really awesome as now "all of a sudden" I can get the full frame rate/resolution same as on the Intel based notebook! So the small 10 x 7.2 cm VIA board is again in play as the main platform for this whole thing.

And I've got a bunch of the usual test charts (resolution and color charts) laying around from our local camera rental place which I will shoot over the next few days to finally get the color calibration going!

Yay.. really motivated again to get the software to the next level...

Ryan Paige
12-14-2012, 07:24 PM
Awesome.

imdjay
12-14-2012, 10:59 PM
think this little box has the processing power for what you're doing?:

http://www.tomshardware.com/reviews/vision-ht-321b-jbc700-q180,3231-5.html

runs off DC, usb 3, etc... but not sure of its power draw

which board are you running it off of now? while im sure you've got some lovely plans to build it all into a case. i might just like to try a modular approach.

Soeren Mueller
12-15-2012, 02:51 AM
Should have enough power.. but I'm not sure if its USB3 controller would be supported. The New Intel NUC should work too (but you'd Need to add a Mini PCIe SATA Controller + Mini PCIe USB3 controller) and it has considerably higher power draw.

I'm using the VIA P910 which is a pico ITX Board so even smaller than the Zotac and NUC.. its only 3.93 x 2.84 inch! Its 1GHz quad core 64bit (x86 compatible) with USB3, SATA, Audio (and additional usb2 Ports) Takes max 8GB RAM.. Mini HDMI and VGA Out (plus LVDS Display connector on Board) - really powerful little bugger ;)

And its powered via little 12V DC connector. Capturing to SDD from the Cam Head works fine when connected to a 12V power supply rated at max 2A! So it should run quite some time off batteries :)
Works fine with my Tekkeon battery for example.

imdjay
12-15-2012, 09:39 AM
very neat little board

have you found a touchscreen panel yet?

it seems as if you may want to even try picking up a refurb/used nook color for under 100 bucks and hack that apart. very pretty screen. glossy yes, buy plenty of matte protection options

Soeren Mueller
12-17-2012, 09:41 AM
no touchscreen yet... we "stumbled" upon a few nice barebone panels for car use but i'm not sure about their quality yet... interesting idea about the nook!

i need the touchscreen soon.. the user interface comes along quite nicely.. and yes shame on me i borrowed a few ideas :(

http://all-digital.de/projects/voocam/jeannie/voocam_ui.jpg

rick.lang
12-17-2012, 04:10 PM
Soeren, this is all very interesting and all and we wish you every success with your amazing developments, but let's get to the important stuff! Who is your model? :o

CaptainHook
12-17-2012, 06:07 PM
Where do i pre-order? ;)

Soeren Mueller
12-18-2012, 07:05 AM
Soeren, this is all very interesting and all and we wish you every success with your amazing developments, but let's get to the important stuff! Who is your model? :o

Haha.. my lovely girlfriend... the only thing keeping me sane ;)

@CaptainHook - thx, let you know asap.. perhaps a small batch or something, need a good housing/body first..

getting a 7" 1280x800 touch tft this week!

Abobakr
12-18-2012, 07:39 AM
You are a genius, man!! I really wish you the greatest success with this development

adam777
12-18-2012, 07:48 AM
Where do i pre-order? ;)

Maybe you could build them one at a time and slowly sell them, gradually finding better sensors etc. Raw is the new revolution in cameras right now. Wish I could program as there's plenty of buyers throwing their money to the wind through crowd funding and preorders.

MIDDLEEASTMAMBO
12-18-2012, 08:38 AM
sounds great! would be also interested in a pre - order.

Phil
12-18-2012, 09:33 AM
no touchscreen yet... we "stumbled" upon a few nice barebone panels for car use but i'm not sure about their quality yet... interesting idea about the nook!

i need the touchscreen soon.. the user interface comes along quite nicely.. and yes shame on me i borrowed a few ideas :(

http://all-digital.de/projects/voocam/jeannie/voocam_ui.jpg

Ah the SI2K I have one for sale..S16mm 2k Cineform raw direct to a SSD

rick.lang
12-18-2012, 04:00 PM
Haha.. my lovely girlfriend... the only thing keeping me sane ;)

Well, listen my friend, I hope you get her under contract ASAP (and invite us all to the wedding reception)!

blahey
12-19-2012, 12:15 PM
Keep up the good work, I feel like we are witnessing the first open source RAW Cam!

metaljesus
12-19-2012, 09:36 PM
Mind.

Blown.

nickjbedford
12-19-2012, 09:47 PM
So when can we buy the 3.5K Muellercam?

;)

teh
12-20-2012, 01:57 PM
With the lack of news about the camera I'm really starting to lose my trust in your company. This is no way to treat your customers...


;)

Soeren Mueller
12-24-2012, 06:27 AM
Haha :) .. merry Xmas everyone!

Trying to post some pics with the TFT connected.. really crisp image on the little 7" with 1280x800! No touch yet :( working on it...

imdjay
12-24-2012, 10:54 AM
which lcd did you go for?

skibovision
12-29-2012, 09:51 PM
Let me suggest SP-20000MCL from jai. That was the camera I would use

skibovision
12-30-2012, 10:49 PM
Oh and for those of you wondering the camera I mentioned above is 5k 30fps at almost full frame (same size as d800 video recording area)...and its 8,000 USD.

imdjay
01-05-2013, 02:55 PM
how goes it?

another thought:

http://www.amazon.com/dp/B004YCWBIY/?tag=dlevi-20

you should be able to plug the camera directly into that and be all set. a tad too big yes, but also much more turnkey

skibovision
01-07-2013, 02:06 PM
Ok so here is a little update on some things I've found.

1. There is a camera coming out that will be USB3 capable of (roughly, still waiting on benchmarks) 5120x2880 @ 24fps in the spring based on the camera link version I talked about in my last post. I just got off the phone with the rep I have been dealing with at 1stVision (JAI distributor). The camera will be roughly 8,000 similar to the camera link version but will not require a frame grabber (saves a ton on power and cost).

2. As I said the camera is roughly capable of putting out 5120x2880 @ 24fps. That works out to (5120x2880x12bitsx24fps) / 8bits = 530,841,600 bytes or 530MB/s of Data. This is in bayer raw format and I'm not sure what this would mean exactly if it would be in DNG but the point is we have a problem with Data throughput.

The pico-itx board from via we have been discussing on this thread has 2 usb3 ports and 2 sata ports along with a proprietary connector which equals "about 2 lanes of pcie" according to the guy I spoke with at via...this is being verified.


@Soeren are dng's smaller or larger than bayer raw files? would the board you have been discussing be capable of handling that type of data?

I think that this camera would be perfect for this type of project because:

1. It will allow for greater than 4k resolution so we will have the benefits of downscaling and creating content for the emerging 4k market (items like this http://abcnews.go.com/blogs/technology/2013/01/lenovo-ideacentre-horizon-27-inch-tablet-turns-into-27-inch-table/)

2. We will be able to have a turnkey 5k digital cinema capture system for likely less than 10k. The closest is resolution would be epic which is brain only at 20k.

3. It will be the first non-consumer digital Full Frame cinema camera

4. We can window it down to super 35 and super 16 for faster frame rates/ increased lens choices.

5. I know its a lot of data but we could easily treat offloaded data like a film mag, dump dng's to another computer, and convert to cineform RAW (10:1 compression) and then delete original raw dng's if we like.

imdjay
01-07-2013, 07:24 PM
nice. or could have a look at cineform integration

Soeren Mueller
01-07-2013, 08:02 PM
@skibovision the via board wouldn't be able to handle this data rate - in fact the data rate you are talking about would be too much even for usb3!

the point grey usb3 cams only reach the full through put with their own drivers and even on a high end i7 notebook the tops rate is ~200-250mb/s.
i have to remind you, this is more complicated than just copying data at max speed because we are talking about realtime stuff.. the camera captures at a fixed clock speed/scan rate and there is a very small timing window in which the scanlines have to be read... or else you encounter a partly retransfer and stuff like that which quickly results in dropped frames. the via board/processor is quite powerful for it's size but still is hard on the edge handling the 190-200mb/s. (you shouldn't think of it as mb/s, more as scanlines/frames per time)

oh and the dngs are exactly the raw bayer data size plus a few header/data bytes.. perhaps 200-300 bytes "overhead" compared to the raw image data. (or a few 1000 bytes more of you were to add a lut or something)

the display i'm currently using is the innolux n070icg - 1280 x 800 on 7".. quite bright and really good quality, but i haven't been able yet to properly connect the touch part, also i'm waiting on feedback from a swiss company that's specialized in stuff like that (for industrial use as well)... it's not easy to find the right connectors to be able to use the lvds connection on the via board (there's no real standard regarding the pins etc). i'm still working on the software and it's coming along quite nicely. cold bootup time is roughly 30 secs then my capture software runs and is usable.
but finding the right touch interface/adapter etc. is really more complicated than i imagined... sigh

@imdjay these atom based systems aren't powerful enough unfortunately for high data rate usb3 capture.

and yeah of course i'm interested in bigger sensor sizes as well, but first i have to "finish" this thing for it's main application currently where i need the ice cube sized sensor with the very lightweight and small c-mount lenses. i think i can show you what this is about in 2-3 weeks, stay tuned ;)

still testing different lenses.. we just had a 1.3 and 1.7mm wide angle on loan.. recorded a few test shots that i'll be uploading here:

http://all-digital.de/projects/voocam/jeannie/dngs/test1/2013_01_02/

skibovision
01-07-2013, 11:02 PM
thank you for getting back to me. I had actually planned on doing something like this a few months ago but abandoned it due to power issues. I was working with 1st vision and norpix to find a solution. I am not a programmer so where I left off was planning to put together the following system and operating it simply by running streampix in windows:

Mini-ITX board running i7 processor
16GB ram
camera link frame grabber
removeable sataIII SSD's
400x800 usb touchscreen monitor controlling camera and streampix
all mounted inside of an enclosure I was designing with a machinist I have used for other photo projects.

The issue I had was that even if I could set the computer up to run on DC power it would drain a 240wh v mount in an hour. Combine this with the added cost of streampix, frame grabber, and camera and I was well over 15k. That by itself made it not worth it. My contacts at the manufacturer assure me that if we could find a pico-itx board with sata III ports and find an SSD that would allow us 530MB/s sustained write times (intel 500 series maxes out at about 520MB/s) then we will be that much closer.

On another note I would really like to try something like you have only I would combine the camera with a 35mm dof adapter and relay lens to give myself the proper FOV using my existing lenses. I know people aren't fond of these machines now that we have dslr's but it will allow me to get that look without shelling out 8,000 just for the camera head.

skibovision
01-08-2013, 08:18 PM
nice. or could have a look at cineform integration

I think this would be a waste of resources. The guys at cineform have been bending over backwards and operating on these forums directly making sure we can easily convert dng to cineform if we want to. We can use less powerful internal computing if we capture directly to DNG.


@Soeren

I did not mean to derail things with the last 2 posts about things that I have found to not work so I will ask you something about your current camera. If the camera is spec'ed to give you 4096x2160 @21fps what bit depth would that be? I am curious as to what the top level of throughput is supposed to be.

4096x2160 @21fps 8 bit = 185MB/s
4096x2160 @21fps 10 bit= 233MB/s
4096x2160 @21fps 12 bit = 278MB/s

The theoretical limit of usb3 is 5Gb/s = 5,000,000,000 bits/s = 625,000,000 Bytes/s = 625MB/s. If the cameras limit ends up being 4096x2160 12 bit @21fps then that would roughly work to 185,794,560 pixels per second at 12 bits. If the camera can pull that off then hopefully by reducing the vertical lines scanned to a 2.4:1 aspect ratio would get us 4096x1706 12bit @ 24fps for a total of 167,772,160 pixels for second. I know that frame rate increases by shrinking the ROI doesn't scale at 100% efficiency but we would only need it to work at 91% to pull this off if my understanding is correct.

Is my understanding correct and is 4k 2.4 24fps 12 bit video something we can expect from this camera? If so then that would mean the data rate would be something like 250MB/s ((((4096*1706*12bit*24fps)/8)/1000000)) which would be pretty crazy considering it would only allow us 16 minutes of footage on a 250GB SSD. The data rate can be dealt with though by converting to cineform raw at roughly a 10:1 compression.

Soeren Mueller
01-09-2013, 08:59 AM
@skibovision .. sorry, short on time currently... but most of the info is already in this thread, albeit a bit scattered i have to confess. you should be able to find the max res @ 24/25 fps... hm wait i'll give you a short list:

2048 x 1080 x 12bit, 60fps (199MB/s, binning, full sensor)

3120 x 1754 x 12bit, 25fps (205MB/s, windowed)
3200 x 1800 x 12bit, 24fps (207MB/s, windowed)

3584 x 2016 x 8bit, 25fps (181MB/s, windowed)
3728 x 2096 x 8bit, 24fps (188MB/s, windowed but only very slightly of course)

and the theoretical max data rate doesn't apply.. as i already wrote it's all about the timings when reading the data so the 200-250mb/s is the limit when dealing with a usb3 camera, above that you start to get timing problems which means either corrupted or dropped frames which obiviously isn't an option for us. for their normal use case industrial automation this isn't nearly as critical most of the time ;)

if you'll have a look at the link i posted a few times you'll find a few rough first sample recordings of video at 2k and also at 3.2k-3.7k...

skibovision
01-09-2013, 10:02 PM
I enjoyed playing with the dngs, there is so much you can do with the image. here is a quick grade I did to try to get just basic natural looking colors

1975

I used just camera raw in adobe and color balance in after effects

jtbullet
01-21-2013, 04:09 PM
I am interested in getting one of these devices. Are there any external devices that can be used to capture this raw footage? Something like an atomos ninja or samurai?

jtbullet
01-24-2013, 10:16 AM
An idea...what about using an ipad mini to control the computer via remote connect? The display quality is good, the touch is already built in, and it (most importantly) allows you to sever your camera from the display...people can buy their own ipad. Many people already have them. This makes you able to offer the package at a much lower price. It also let's people who have their own ideas about displays, do their own thing.

In the end, you can do a real kickstarter, letting people know these cameras are ready to ship, in quantity, using off the shelf components...tell them they can feel free to buy their own hardware if they want. For those people they would simply back your fundraiser at the proper level and get the software for free. You could get a nice cage designed that would house the camera, and have a bolt on spot for a specific capture pc...you could work with Emm at cheesycam.com to get that part done.

Please don't let this project be dead already! I had been going through the motions searching for a capture solution for a prosilica gige camera. The samples you have shown are fantastic. I have the money for the camera - I just need the software! Set up a donation system...allow us access to your software, we will help you fine tune it, create marketing materials for it...basically be beta testers.

Charlie Doom
01-24-2013, 12:37 PM
Let me just say this is one of the cooler projects I've come across. I didn't think it was even possible for someone to make a digital raw camera without tons of resources, let alone in their apartment! I have been looking into building some custom lenses which doesn't seem that difficult to get started in, but very hard to perfect! Anyways, just wanted to congratulate you on this project and celebrate your hard work.

jtbullet
01-24-2013, 01:02 PM
Yeah, I'm in agreement with you on this one. Sadly, except for you and me, all I hear are crickets for a couple of weeks. I'm ready to buy the camera and computer, I just need a little heads up on the software end of things. Streampix does in fact support usb 3 cameras. I verified it with the manufacturers. Unfortunately, the software is $1000, so I'm stuck for the moment. Hopefully Soeren will get back in here soon and get us excited again. I'm hoping the ipad mini as remote controller for the pc is possible, as I have one, and it's always nice to be able to justify a gadget with a real world use ;)

rubiocesar83
01-28-2013, 04:49 PM
Hi;

My name is Cesar Rubio, I was directed to this interesting thread by a member here.

I will write my POV here publicly to see if it helps, please don’t feel offended if those are not particularly aligned to yours.

I’ve been working with Machine Vision Cameras (MVCs) since 2006, so I know a thing or two about them already (mostly for Stereoscopic 3-D applications).

I’ve seen projects like this come and go since then…why? Its not as easy as it looks.

And it requires a LOT of expense by these brave experimenters, that have to hand out a lot of cash in their projects….most just give up over time.

First, the camera itself has to meet certain minimum requirements in the whole pipeline workflow process.

What do I mean by this?

Is just not about “resolution/frame rates”…its about lenses and bandwidth required.

There is no use to me if I get a 4K camera if my lenses wont resolve that much resolution right?

MVCs lenses are tinny, and they don’t resolve as much as bigger glass would do…its just physics laws.

Most lenses resolve about 80-100 pl/mm (line pair per millimeter) in the corners…so that means that we need to see also the pixels size in the sensor.

I wrote an article here about that:

http://dna-rubio-s3d.blogspot.mx/2012/04/mvcs-lenses-quality-ratings.html

A rule of thumb would be to use no less than 5 microns pixel size sensors with most good quality lenses, or your lens wont resolve properly and you will also end up with lens diffraction problems.

The pixels size also dictates another very important subject in cameras, sensitivity.

I don’t care if a camera can shoot up to ISO/ASA 12,800 or more.

What I do care though is what is the native sensitivity at 0dB gain (for good looking clean images), usually with current sensor technology, with a pixel size like that it would be around ISO 640/800 which is fine with me.

So taking this in consideration as well, if we want more resolution in a camera, we need a bigger sensor size in order to keep the same pixel size, it makes sense doesn’t it? 

You can expect 1080p/2K resolution tops in a 2/3 inch sensor size, if you want 3K then you have to jump to 4/3 inch, 4K to S35mm and so on…

But that does not stop there, the more resolution the more bandwidth is needed and disk space.

You can not possible compress to Cineform Raw for example beyond 2K/24p resolution with a small form factor computer. (it needs a lot of CPU workload and the processor gets HOT Pretty quickly!, which means you have to use FANS at all speed, that are loud no matter how “silent” they claim they are…)

And Uncompressed Raw takes lots of space, I would not go beyond 3K/24p with it, its already almost 7 GB per minute of footage.

Then in post you have to de-Bayer it, if you want the best possible IQ, then it must be to Uncompressed RGB (AVI or MOV), but that will increase the file size by x3!

Sometimes I go the compression route de-Bayering to Cineform Filmscan2 4:4:4 (the maximum quality with this codec), and it keeps the file size around the original Uncompressed Raw ones…and they are still HUGE!

And we are talking of 8 bit files here, 12 bit will require 1.5 more times the bandwidth and file sizes.

I have made comparison tests between 8,12 and 14 bit, and found not visual difference between the 3 (in Uncompressed RGB). So I settled with 8 bit.

And no matter what you do in post: de-Bayer to Uncompressed RGB, compress to Cineform or whatever codec you like, or even trans-code to Cinema DNG, it requires TIME! A lot of it if you want the best quality, which you should do, there is no point in having the best gear if you are going to de-Bayer to the fastest (low quality) setting to save time….

Anyways, I don’t want to discourage anyone with all of this, is just an honest “re-cap” if you will of what is needed in order to make beautiful images with these cams.

Don’t just go for the cheapest, since in the long run it will cost you more…get the best you can get and that it will last you many years of work…even if its expansive at the beginning, its cheaper in the long run since you would not need to upgrade as often.

Most people don’t want to be messing around with a computer (even if its small) for recordings, I mostly use this cams for 3-D because of their compact size, but for 2-D I would rather get a proper camera like the BMD Camera (that is in my list to get as well…).

Here are some pics of a “test” shoot that I made with a DIY camera set up, it could be the first and only Uncompressed Raw 720/24p camera ever (back in 2009):

https://plus.google.com/u/0/photos/117102957642252969487/albums/5669704850652758929?authkey=CMb0jpCJr6qZZg


It’s a Prosilica GC-1380c (2/3 inch CCD) recording at 720/24p Uncompressed Raw 8 bit with Streampix 4.

It produces beautiful images…unfortunately I am traveling right now and I don’t have any images to show you at this point (the self taken shot on the mirror was using compression to Cineform Raw on the fly, and its not very good quality since this codec prefers 12 bit footage for better results).

Cheers,
CR.

rawCAM35
01-28-2013, 05:48 PM
http://www.bmcuser.com/showthread.php?2526-Just-wondering-has-anyone-tried-homebrewing-a-camera-like-this-using-streampix&highlight=Cesar+Rubio

jtbullet
01-28-2013, 08:58 PM
rawcam35, you're the one who pointed me to cesar! Many thanks. I am going to try this using a pt grey flea and streampix. I'm going to be looking for the proper lens while I await the components. Thanks for sending me to cesar, he has a LOT of knowledge.

Soeren Mueller
01-31-2013, 01:22 PM
Sorry for the silence currently, too much work *sigh* .. need 48 hr days ;)

@cesar.. thanks for the input, although this is really off topic for this thread imho. I guess most of us here know about sensor size vs. resolution/sensitivity, the storage requirements for raw footage etc etc - not much difference between the BMC and my project.

My software captures directly to Cinema DNG, no transcoding of any form required - today this is all pretty different from a few years ago because of two things basically:

- ultra small form factor computers/mainboards/processors available, even for "little money"
- high storage / fast SSDs available, also for "little money" (compared to a few years ago)

So I can easily capture 3.8k RAW cinema DNGs to a very light weight and small 256GB SSD and can capture up to 22 minutes! Post production etc. is the same as with raw BMC footage, no different requirements there.

Another goal of my project is to add a small 7" high res touch screen to the board + SSDs and put all of it in a compact housing - of course this only makes sense when you have a really nice touch interface as well - and the screenshot I posted is from a working prototype, so the interface is already working.

This whole thing is pretty adaptable to different USB3 camera heads as well. The one I'm currently using has a Sony CMOS sensor which is pretty much the same sensor used by Sony in many of their newer small form factor consumer cameras.
I chose the small sensor because this project was not conceived to be a BMC alternative or something, but because it is needed for a "specialty" use case where a particular small and light weight raw camera is needed!
(the board I'm using runs on 12V, has low power consumption and is about 4.3" x 2.8" (!) in size with a weight of ~210g - ~300g with 2 SSDs (!))

And because of this it is not some kind of "come and go" project - it IS already working! The touch screen and it's connection is not yet optimal and a new/different one is required plus all the housing and user interface finalization etc etc (and color science part for "correct" DNG handling)

Hope to be able to post a more detailed update soon!

rubiocesar83
01-31-2013, 01:36 PM
Congratulations on your progress Soeren.

Yes, probably I sounded too negative right?

I apologize for that and keep up the good work my friend.

I am all thumbs up for DIY systems...

If I mentioned the "shortcomings" of a project like this is, to try to inform interested people into getting the BEST equipment they can the first time, since usually cheaper products means that you have to buy twice, or more...

CR.

Soeren Mueller
01-31-2013, 01:44 PM
No worries Cesar! :)

I understood it that way too and I guess it's always good to stay "level headed". I never wanted to get folks expectations up too high or something.. perhaps I should have made it much more clear that it's (currently) not meant as a BMCC / Digital Bolex alternative or anything...

First have to get it working for this special use case and then I'll see what to do with the software/capture base etc.pp.

rubiocesar83
01-31-2013, 02:26 PM
Another important consideration is the file sizes, although higher resolutions are usually better, we have to find something manageable in terms of workflow with current technology.

I have settled with 3K/24p Uncompressed Raw 8 bit for now...the files sizes area already HUGE, at almost 7GB per minute of footage!

CR.

skibovision
02-01-2013, 05:14 PM
Another important consideration is the file sizes, although higher resolutions are usually better, we have to find something manageable in terms of workflow with current technology.

I have settled with 3K/24p Uncompressed Raw 8 bit for now...the files sizes area already HUGE, at almost 7GB per minute of footage!

CR.

If you ask me file size is not important at all. For a while I had been planning a streampix based camera system working with a 20MP @30fps camera. The design included 4 256GB drives in a raid 0. My initial calculations told me that would get me about 30 min of bayer raw. My plan was to offload that to a drive and then transcode to cineformraw later (through streampix). The issues I came up against was power, throughput, and cost. The cost was the straw that broke the camels back. For the 5k setup I was looking at a 15,000USD investment. For the same amount of money I decided I would be better off buying the zeiss zf 15, 25, 35, 50, 85, and 135

I think soerens software is the key to beating this because it allows you to save 1,500 on software + all the money saved on modules. I look forward to seeing where his project takes him.

jtbullet
02-01-2013, 10:10 PM
That's assuming Soeren gives the software away ;) Streampix has also combined the lut/gamma/color remapping plugin...so it's 1000 for that... and now it transcodes to adobe dng offline....plus the free upgrades for life...hundreds of already supported cameras...iridas look file import...I'd say both methods are great and viable.

When you look at this as a camera system rather than a camera...1000$ is not such a bad amount.

But...free is better, of course;)

rubiocesar83
02-02-2013, 12:24 PM
If you ask me file size is not important at all. For a while I had been planning a streampix based camera system working with a 20MP @30fps camera. The design included 4 256GB drives in a raid 0. My initial calculations told me that would get me about 30 min of bayer raw. My plan was to offload that to a drive and then transcode to cineformraw later (through streampix). The issues I came up against was power, throughput, and cost. The cost was the straw that broke the camels back. For the 5k setup I was looking at a 15,000USD investment. For the same amount of money I decided I would be better off buying the zeiss zf 15, 25, 35, 50, 85, and 135

I think soerens software is the key to beating this because it allows you to save 1,500 on software + all the money saved on modules. I look forward to seeing where his project takes him.

Hi skibovision:

I think that a 5K Uncompressed Raw camera at $15,000 USD is actually CHEAP!

Look at these cameras comparison pricing here (http://wolfcrow.com/blog/the-battle-for-4k-raw-a-red-epic-vs-canon-c500-sony-f55-price-comparison/).

Regarding transcoding to Cineform Raw, there is no point of that, then WHY to record Uncompressed Raw in the first place?

I did that before but did not like the quality very much, then tried de-Bayer to Cineform Filmscan2 444 (the highest settings with that codec) to keep about the same sizes of my Uncompressed Raw files (Uncompressed RGB is triple in size).

But that was before CinemaDNG, from now on I will transcode to CinemaDNG too keep the best IQ in an Uncompressed format for extensive grading and color correction.

Its looking promising so far, have a look at some tests (http://dna-rubio-s3d.blogspot.mx/2013/02/dynamic-rangeand-12-bit.html) I am doing right now with DNG files.

CR.

rubiocesar83
02-02-2013, 12:28 PM
That's assuming Soeren gives the software away ;) Streampix has also combined the lut/gamma/color remapping plugin...so it's 1000 for that... and now it transcodes to adobe dng offline....plus the free upgrades for life...hundreds of already supported cameras...iridas look file import...I'd say both methods are great and viable.

When you look at this as a camera system rather than a camera...1000$ is not such a bad amount.

But...free is better, of course;)

The Norpix guys (developers of Streampix) work DAILY to improve the software...its a small group (like 3 people) but they work full time on that...

And they have been doing that for YEARS in and out, its impossible to make this kind of high end software for free...unless it was an open source project of course, but it is not.

These guys deserve to be paid for their hard work, as much as any one of us do too for ours.

Thank you,
CR.

jtbullet
02-02-2013, 05:30 PM
I think Soeren should be paid for his work as well...but streampix is definitely a part of a long term strategy for me. That said, I am looking forward to both solutions...there's room in the marketplace for all! It encourages better applications!

rubiocesar83
02-03-2013, 02:18 AM
I think Soeren should be paid for his work as well...but streampix is definitely a part of a long term strategy for me. That said, I am looking forward to both solutions...there's room in the marketplace for all! It encourages better applications!

Agreed.

CR.

rubiocesar83
02-20-2013, 09:27 AM
Hey guys check this out, in case you want to go GIGANTIC at a IMAX screen with your work!

5K Uncompressed Raw camera project:

http://dna-rubio-s3d.blogspot.mx/2013/02/want-4k-you-got-it.html

CR.

noirist
02-20-2013, 06:50 PM
At that price ($18k), why not buy a Red with F mount? You can buy a brand-new ready-to-shoot Scarlet-X for $15k and Epic-X for $26k.


Hey guys check this out, in case you want to go GIGANTIC at a IMAX screen with your work!

5K Uncompressed Raw camera project:

http://dna-rubio-s3d.blogspot.mx/2013/02/want-4k-you-got-it.html

CR.

rubiocesar83
02-20-2013, 07:26 PM
At that price ($18k), why not buy a Red with F mount? You can buy a brand-new ready-to-shoot Scarlet-X for $15k and Epic-X for $26k.

Scarlet is only 4K with 6:1 compression ratio.

Epic is 5K with 3:1 compression ratio (the best).

This camera is 5K Uncompressed Raw, with GLOBAL SHUTTER, just like the new Sony F55.

Those are HUGE differences in my book...

CR.

rubiocesar83
02-20-2013, 08:35 PM
BTW if you want to know the REAL costs of Scarlet and Epic, see these ones please.

Scarlet (http://wolfcrow.com/blog/the-battle-for-cheap-4k-a-red-scarlet-vs-canon-c300-vs-1dc-vs-sony-f5-price-comparison/)

Epic (http://wolfcrow.com/blog/the-battle-for-4k-raw-a-red-epic-vs-canon-c500-sony-f55-price-comparison/)

Not cheap...

CR.

noirist
02-20-2013, 08:58 PM
The Red configurations on the wolfcrow blog are definitely not budget. For example, you don't need a redmote to operate the Scarlet or Epic; the side module is sufficient. Nor do you really need a day's worth of proprietary media to shoot for a day when you're on a budget. I think it's possible to configure a budget Scarlet-X for $15k and a budget Epic-X for $26k.


BTW if you want to know the REAL costs of Scarlet and Epic, see these ones please.

Scarlet (http://wolfcrow.com/blog/the-battle-for-cheap-4k-a-red-scarlet-vs-canon-c300-vs-1dc-vs-sony-f5-price-comparison/)

Epic (http://wolfcrow.com/blog/the-battle-for-4k-raw-a-red-epic-vs-canon-c500-sony-f55-price-comparison/)

Not cheap...

CR.

noirist
02-20-2013, 09:00 PM
P.S. When is the JAI SP-20000 going to be available? And what would your proposed configuration look like in a rig?

rubiocesar83
02-21-2013, 02:57 AM
P.S. When is the JAI SP-20000 going to be available? And what would your proposed configuration look like in a rig?

Let me check with those guys, and as soon as I hear from them I will inform you all here.

And another very important feature of a systems like this is that is totally silent , no fans in the camera head nor in the recorder...

I have heard of fan noise issues with the Red cameras.

CR.

mrholman
02-22-2013, 03:47 AM
This is impressive to hear about it.. interesting to hear about your progress.. Please continue!! Oh don't forget to document everything

This. Good stuff.

rubiocesar83
02-25-2013, 07:34 PM
Ok I've got word from the JAI guys regarding the SP-20000 camera.

The retail price will be only $8,000 USD (instead of $10K as stated before...)

And the USB-3 camera interface, should be able to do 5120 x 2708 pxs (1.89:1 aspect ratio) at 24 fps in Raw 8 bit mode.

CXP interface versions should do more frame rates/bit depth, and the price will be at $8600 USD.

They will have pre-production units available within the next 30 days, with production beginning in the second quarter of the year.

CR.

Soeren Mueller
03-10-2013, 09:16 AM
Hey Rubio, regarding 8 bit vs 10/12 bit... I think there definately is a difference between the bit depths even a visible one if you're going to do more than only a little post processing!

Is the SP-20000 by any chance using the CMOSIS CMV20000 or is this a totally different sensor? Sounds pretty similar from the data sheets.

Regarding my project: crazy but the most exhausting part is finding the right touch screen that's usable outdoors and provides a good user experience.

Oh.. and I'm about to get the CMV4000 and CMV12000 for evaluation =) .. let's see what kind of pictures they can produce albeit being able to "only" capture 10 f-stops dynamic range according to their tech specs.

rubiocesar83
03-10-2013, 12:01 PM
Hey Rubio, regarding 8 bit vs 10/12 bit... I think there definately is a difference between the bit depths even a visible one if you're going to do more than only a little post processing!

Well in my previous tests using CCD cameras from AVT (Pike F-210c), I did NOT see any difference between 8,12 and even 14 bit. But I have not made tests with a CMOS camera about that yet (I should have done what when I had a Sumix camera with a CMOS sensor for testing one time, but I didn't)

I hope I can get my hands in a Dalsa Genie TS-C2500 (http://www.dna-rubio-s3d.com/oc_store/product_info.php?cPath=25&products_id=54) to perform such tests with it.



Is the SP-20000 by any chance using the CMOSIS CMV20000 or is this a totally different sensor? Sounds pretty similar from the data sheets.

Yes it is, in the bottom of the cameras explanation in this post (http://dna-rubio-s3d.blogspot.mx/search?updated-max=2013-02-20T11:33:00-06:00&max-results=7) you can download their PDF data sheet.


Regarding my project: crazy but the most exhausting part is finding the right touch screen that's usable outdoors and provides a good user experience.

And the REAL problem is to get one that has enough native resolution to handle at least 1/4 of your cameras resolution. With StreamPix 5 if you want to see a clear B&W (not squared or pixelated image), you must re-size -if not in full resolution- to a quarter of the image, so for 5K we are talking of a 2560 pxs wide monitor...and AFAIK there IN NONE of those touch screen monitors yet. Even the non touchscreen are HUGE!

Or you can use the de-Bayer on the fly preview, which shows a color image in any screen resolution, BUT it requires a TON of CPU and GPU memory!


Oh.. and I'm about to get the CMV4000 and CMV12000 for evaluation =) .. let's see what kind of pictures they can produce albeit being able to "only" capture 10 f-stops dynamic range according to their tech specs.

Looking forward to see those, what cameras are you thinking to get with such sensors Soeren?

CR.

Soeren Mueller
03-12-2013, 04:43 PM
Well in my previous tests using CCD cameras from AVT (Pike F-210c), I did NOT see any difference between 8,12 and even 14 bit. But I have not made tests with a CMOS camera about that yet (I should have done what when I had a Sumix camera with a CMOS sensor for testing one time, but I didn't)

Ah ok.. well I never tested with CCD cameras yet - but of course uncompressed raw is even nice with 8 bit already, especially compared to the usual compressed formats. But yeah with the CMOS heads I tested you can definately see that there's a "finer" latitude that is not just noise. Especially on very fine gradients!

Ok, regarding the SP-20000 - well I'm trying to get my hands on one as well but it's in early sampling stage afaik so I don't care that much right now and just try to ignore it, but it definitely seems interesting. ;)


And the REAL problem is to get one that has enough native resolution to handle at least 1/4 of your cameras resolution. With StreamPix 5 if you want to see a clear B&W (not squared or pixelated image), you must re-size -if not in full resolution- to a quarter of the image, so for 5K we are talking of a 2560 pxs wide monitor...and AFAIK there IN NONE of those touch screen monitors yet. Even the non touchscreen are HUGE!
Or you can use the de-Bayer on the fly preview, which shows a color image in any screen resolution, BUT it requires a TON of CPU and GPU memory!

Well... as I'm sort of in "total" control I didn't encounter such problems until now. Of course I don't plan for "real" debayering anyways, the VIA P910 is just not powerfull enough. Currently at least I have it running (was Prio 1) as a realtime task to save the RAW data stream as Cinema DNG. And it's possible for me now to make an "on the fly" B&W preview out of that for display. The P910 has a quite powerful (at least more than I thought) GPU on board that handles the transforming/scaling quite nicely! I'm not yet in the final optimization stages so it's just sort of a "blindfolded" recording platform, with non-realtime monitoring but I'm heavily working on that!

It looks really good until now that's why I'm sort of "desperate" to get a good touchscreen for it. The 1280x800 7" is really nice... it's no problem to have a sharp downscaled preview or switch to a 1:1 "true-pixel" window. But it's "only" a display until now and I still have to use a mouse. The display is available in a "tablet PC" version too with integrated capacitive touchscreen - but none of the asian supliers has any real tech specs/info on it. :( So without the right touch controller it's pretty much worthless!
And most of the industrial PC solution providers I spoke too until now.. they have nice small screens but most of them only max 1024 x 600 or something in that size. Jeez...
But at least I got one on route to me currently that is totally usable outside under bright sunlight - quite amazing! But not really highres :/

So if anyone has a tipp where to look for a good touchscreen.. let me know!



Looking forward to see those, what cameras are you thinking to get with such sensors Soeren?

Unfortunately I'm not allowed to tell yet (I'm not the one paying ;o) ... but they are made by a company here in Germany. I'll receive the CMV4000 one shortly and will post the first recorded samples here for you to play with ASAP! I'm really really curious about the real life dynamic range of the CMOSIS sensors!

The CMV12000 isn't really available yet on the new USB3 systems... but they are looking into it, as soon as there's a new development I'll post it here!

Cheers
-Sören

rubiocesar83
03-12-2013, 08:10 PM
Have you seen some of these touch screens (http://smallpc.com/lcdmon-enclosed.php) yet?

CR.

adam777
03-12-2013, 09:02 PM
Ah ok.. well I never tested with CCD cameras yet - but of course uncompressed raw is even nice with 8 bit already, especially compared to the usual compressed formats. But yeah with the CMOS heads I tested you can definately see that there's a "finer" latitude that is not just noise. Especially on very fine gradients!

Ok, regarding the SP-20000 - well I'm trying to get my hands on one as well but it's in early sampling stage afaik so I don't care that much right now and just try to ignore it, but it definitely seems interesting. ;)



Well... as I'm sort of in "total" control I didn't encounter such problems until now. Of course I don't plan for "real" debayering anyways, the VIA P910 is just not powerfull enough. Currently at least I have it running (was Prio 1) as a realtime task to save the RAW data stream as Cinema DNG. And it's possible for me now to make an "on the fly" B&W preview out of that for display. The P910 has a quite powerful (at least more than I thought) GPU on board that handles the transforming/scaling quite nicely! I'm not yet in the final optimization stages so it's just sort of a "blindfolded" recording platform, with non-realtime monitoring but I'm heavily working on that!

It looks really good until now that's why I'm sort of "desperate" to get a good touchscreen for it. The 1280x800 7" is really nice... it's no problem to have a sharp downscaled preview or switch to a 1:1 "true-pixel" window. But it's "only" a display until now and I still have to use a mouse. The display is available in a "tablet PC" version too with integrated capacitive touchscreen - but none of the asian supliers has any real tech specs/info on it. :( So without the right touch controller it's pretty much worthless!
And most of the industrial PC solution providers I spoke too until now.. they have nice small screens but most of them only max 1024 x 600 or something in that size. Jeez...
But at least I got one on route to me currently that is totally usable outside under bright sunlight - quite amazing! But not really highres :/

So if anyone has a tipp where to look for a good touchscreen.. let me know!




Unfortunately I'm not allowed to tell yet (I'm not the one paying ;o) ... but they are made by a company here in Germany. I'll receive the CMV4000 one shortly and will post the first recorded samples here for you to play with ASAP! I'm really really curious about the real life dynamic range of the CMOSIS sensors!

The CMV12000 isn't really available yet on the new USB3 systems... but they are looking into it, as soon as there's a new development I'll post it here!

Cheers
-Sören

Congrats on the progress of your project! You may yet beat Apertus in developing a CMV12000, or even better a CMV20000 based camera :)

rubiocesar83
03-13-2013, 01:02 PM
The CMV12000 isn't really available yet on the new USB3 systems... but they are looking into it, as soon as there's a new development I'll post it here!

Cheers
-Sören

Soeren:

You may want to give it a look at this Dalsa 4K camera (http://www.dna-rubio-s3d.com/oc_store/product_info.php?cPath=25&products_id=36), is not USB-3 interface, but it has mini Camara Link one.

And it produces beautiful images, you can see an image recorded at 3K resolution then down-sampled it to 2K with that camera here (http://dna-rubio-s3d.blogspot.mx/2013/02/dynamic-rangeand-12-bit.html).

Thanks,
Cesar Rubio.

rawCAM35
03-13-2013, 02:31 PM
Unfortunately I'm not allowed to tell yet (I'm not the one paying ;o) ... but they are made by a company here in Germany. I'll receive the CMV4000 one shortly and will post the first recorded samples here for you to play with ASAP! I'm really really curious about the real life dynamic range of the CMOSIS sensors!
The CMV12000 isn't really available yet on the new USB3 systems... but they are looking into it, as soon as there's a new development I'll post it here!
Cheers
-Sören

Actually I asked Ximea last year if they looked at the CMV 120000, they said that they are considering it

Soeren Mueller
03-13-2013, 02:50 PM
Actually I asked Ximea last year if they looked at the CMV 120000, they said that they are considering it

That's not the problem.. but availability is. I know that some of these cam head manufacturers are waiting half a year or more for sensors from them for small batches! So I guess BM are not the only ones stuck with non-reliable sensor manufacturers ;)

Rubio thanks.. but this platform is only about USB3 for now. So I'm only interested in _available_ USB3 sensors that are useable for video in "real life". I want to keep everything as simple, lean and straight forward as possible.
Any other way I'd end up fiddling for years with something in my garage.. not interested in that, too many people been there done that ;)

rawCAM35
03-13-2013, 03:00 PM
So I guess BM are not the only ones stuck with non-reliable sensor manufacturers ;)

The BM sensor manufacturer is very reliable, the problem was the glass supplier, that has nothing to do with the sensor manufacturer, companies have to wait for long time when a sensor is in demand.

Soeren Mueller
03-13-2013, 03:09 PM
The BM sensor manufacturer is very reliable, the problem was the glass supplier, that has nothing to do with the sensor manufacturer, companies have to wait for long time when a sensor is in demand.

You know that it was more complicated than that.

mhood
03-13-2013, 03:10 PM
The BM sensor manufacturer is very reliable, the problem was the glass supplier, that has nothing to do with the sensor manufacturer, companies have to wait for long time when a sensor is in demand.

IIRC, the sensor manufacturer changed glass suppliers without any notification to BMD. Hardly a "very reliable" thing to do...

rawCAM35
03-13-2013, 03:26 PM
IIRC, the sensor manufacturer changed glass suppliers without any notification to BMD. Hardly a "very reliable" thing to do...

Agree with you 100%, but my imprison was that another company is gluing the glass on the sensor not the sensor manufacturer, correct me if I am wrong.

Peter J. DeCrescenzo
03-13-2013, 04:47 PM
Grant's 11/8/12 BMCC status update:
http://herefortheweather.wordpress.com/2012/11/08/bmcc-grant-pettys-6th-shipping-status-update-message/

"...

The reason is the contaminated glass issue in many ways distracted [the sensor supplier] from the problems their manufacturer is having bonding the glass to the sensor itself. The sensor supplier now has two sources of glass, and both of them are showing the same problems. The parts without glass are ok, and the problems appear when the glass is bonded to the sensor. If the glass is clean then it’s really the company bonding on the glass that are introducing contamination.

Now the amazing part is that the first batch of sensors we got that we used for developing the camera and that were fine when we started production were manufactured by a completely different company to the second and subsequent batches of sensors. I could not believe this news when I heard it today as it explains a lot.

Our current understanding is that the company that has been bonding on the glass is crap and they have been contaminating the glass when bonding it. Because the sensor suppliers test process was also bad, it meant that no one really knew what was going on and it’s been weeks and weeks of confusion.

The sensor supplier is getting some new sensors made at the original supplier, which we should get test data back on late this week. Once we see this we will know if the original supplier can make the parts without contamination and so we can start building cameras again. I don’t know why they changed glass bonding companies.

..."

shijan
03-14-2013, 02:51 PM
I'll receive the CMV4000 one shortly and will post the first recorded samples here for you to play with ASAP! I'm really really curious about the real life dynamic range of the CMOSIS sensors!

It will be very interesting if you test it! past year i tried to start almost same DIY project with XIMEA's camera based on CMOSIS sensor but after long search and talk with sensor developers i drop it because they sure me that CMV2000/4000 are not very suitable for broadcast usage and those nonlinear HDR mode in sensor is targeted to monochrome sensors because it terrible shift colors.

Soeren Mueller
03-18-2013, 06:19 AM
Actually I asked Ximea last year if they looked at the CMV 120000, they said that they are considering it

And they still are, but it seems not to be Prio 1 for any of those industrial camera manufacturers. Add to that the fact that Ximea are literally a tiny company and that the CMV 12k is in a whole different ballpark in comparison to the "older" CMV 2k/4k. And there are other camera head (USB3) manufacturers that are working on models with the CMV 12k. I have spoken to (I'd say) nearly all of them and all are telling me the same: there's a big difference between CMOSIS announcing a sensor or announcing it's availablility and the sensor really shipping. Right now the CMV2000/4000 are for the first time really available in steady/bigger numbers and in newer revisions. This seems to be another recurring theme with the CMOSIS story, every cam head manufacturer was hinting at quality problems/several bugs (heavy horizontal lines, "black sun" issue, too much noise with first rev designs etc. etc.).
One was clearly stating to me for example that for our use case the current/new CMV4000 revision would be the first one where we'd be happy with anyways, the ones before that.. horizontal lines etc and other "bugs" were quite common.

So as I said, I've got a CMV4k USB3 head en route to me that is capable of 2048 x 2048 pixel at 90fps - the sensor itself is 1". (if you go lower resolution by using a crop/window you can achieve even higher fps)

The cam head I'm getting is set to global shutter fixed, can't make use of the rolling shutter mode...

Soeren Mueller
03-18-2013, 06:24 AM
It will be very interesting if you test it! past year i tried to start almost same DIY project with XIMEA's camera based on CMOSIS sensor but after long search and talk with sensor developers i drop it because they sure me that CMV2000/4000 are not very suitable for broadcast usage and those nonlinear HDR mode in sensor is targeted to monochrome sensors because it terrible shift colors.

Yeah I'm really curios too! I had quite lengthly talks with several manufacturers the last few weeks. A common theme is quality problems or buggy designs with first revision CMOSIS sensors so perhaps that is what they were telling you?
The manufacturer that makes the camera head that I am about to receive for example told me that this new revision is the first which they would have sold me without doubts anyways, before that I wouldn't have been happy with the images - they told me.
But this is all speculation until I can shoot tests with it.

Oh and by the way.. Anthony Dod Mantle made heavy use of the CMOSIS CMV2000 on the new Ron Howard movie "Rush" (the Indiecam indieGS2K uses the CMV2000). And he was reportedly quite happy with the image quality he got. So it can't be too shabby ;) .. let's hope!

The CMV2000 is basicly the same as the CMV4000 but with a ~16:9 active sensor surface, the CMV4k is sort of the "square" version of it.
So with the CMV4000 you can capture 2048x2048 1:1 image (1" sensor size) at max 90fps or a 2048x1088 window/crop portion at even higher fps (theoretical 340 fps but the USB3 interface even at 8bit bit depth will max out at absolute max ~200fps at that res).

rubiocesar83
03-18-2013, 10:58 AM
It will be very interesting if you test it! past year i tried to start almost same DIY project with XIMEA's camera based on CMOSIS sensor but after long search and talk with sensor developers i drop it because they sure me that CMV2000/4000 are not very suitable for broadcast usage and those nonlinear HDR mode in sensor is targeted to monochrome sensors because it terrible shift colors.

This is true, that's why CMOSIS is working in a 3rd version (or revision) of such sensors to correct those problems, they will be ready by summer of this year (2013).

CR.

Soeren Mueller
03-22-2013, 03:12 AM
This is true, that's why CMOSIS is working in a 3rd version (or revision) of such sensors to correct those problems, they will be ready by summer of this year (2013).

CR.

The current selling/shipping revision is the revision number 3 ;) .. but it's said to certainly not be the last one. Still has fixed pattern noise problems which the camera head manufacturers try to "compensate" through their FPGAs with calibration etc. I'm currently implementing this in my own software for black calibration and stuck/bad pixel compensation. The latter I can at least hand off to Cinema DNG which supports a table of bad pixel coordinates. ;)

I'll receive my CMV4000 based camera tomorrow - I'll report asap!

Still no word on any camera with the CMV12K, there was one prototype to be released first quarter of this year but the company sort of went M.I.A. .. at least nobody is really able to get information on the current state of things from them. :(

rubiocesar83
03-22-2013, 02:29 PM
The current selling/shipping revision is the revision number 3 ;) .. but it's said to certainly not be the last one. Still has fixed pattern noise problems which the camera head manufacturers try to "compensate" through their FPGAs with calibration etc. I'm currently implementing this in my own software for black calibration and stuck/bad pixel compensation. The latter I can at least hand off to Cinema DNG which supports a table of bad pixel coordinates. ;) (

So waht then Soeren?

A camera head with a Kodak 2/3 inch CCD instead? like the Prosilica GT/GX 1910 (http://www.alliedvisiontec.com/apac/products/cameras/gigabit-ethernet/prosilica-gt/gt1910.html), or maybe one with a Sony 2/3 inch CCD sensor like the 1920 (http://www.alliedvisiontec.com/apac/products/cameras/gigabit-ethernet/prosilica-gt/gt1920.html)?

CR.

Soeren Mueller
03-22-2013, 03:57 PM
So waht then Soeren?

What do you mean? As I stated previously this project is strictly USB3. And the Sony 1/2.5" sensor in the Point Grey is quite good for it's size, perfect for the original use case.
And if I'm lucky the CMV4000 will be a good alternative for cases where we need >60fps and/or a bigger sensor.

So far I'm very happy (unless the CMV4000 suddenly provides really bad images, but we'll see). My software will be future proof so once a CMV12K USB3 head comes around I'll happily make use of it...

rubiocesar83
03-22-2013, 09:08 PM
What do you mean? As I stated previously this project is strictly USB3. And the Sony 1/2.5" sensor in the Point Grey is quite good for it's size, perfect for the original use case.
And if I'm lucky the CMV4000 will be a good alternative for cases where we need >60fps and/or a bigger sensor.

So far I'm very happy (unless the CMV4000 suddenly provides really bad images, but we'll see). My software will be future proof so once a CMV12K USB3 head comes around I'll happily make use of it...

I was NOT referring to the camera interface, but the CCD SENSORS those cameras have...

Easily implemented in a USB-3 camera interface as well...

CR.

Soeren Mueller
03-23-2013, 08:31 AM
I was NOT referring to the camera interface, but the CCD SENSORS those cameras have...

Easily implemented in a USB-3 camera interface as well...

But not my job or cup of tea.. if somebody does it and provides it with USB3 interface, fine. For the time being I concentrate on whats readily available, that's kinda the key to the project (plus adding the right custom software).

Btw just sourced/found a really good industrial touch TFT that I'm going to use!

Aprox. 8" with 1280x768 pixels, LVDS interface like I need it for the VIA board and really good contrast/viewing angle - and high brightness :)

Soeren Mueller
03-23-2013, 08:32 AM
Look what the cat just dragged in:

http://soeren.de/misc/ximea_cmv4000.jpg

Ximea CMV4000 camera head ready for action! (pictured with Nikkor 50mm and the Point Grey head in the back)

shijan
03-24-2013, 10:26 PM
so BIG lens :D! is it native c-mount or larger format lens with adapter?

Soeren Mueller
03-26-2013, 05:45 AM
@shijan ... it's a native Nikon mount... Nikkor 50mm 1.4 lens on a (not really good) Nikon to C-Mount adaptor

So.. time for an update:

First impressions of the image quality of the CMV4000: very nice! Quite sharp but natural picture with good DOF (1" sensor) and the global shutter is definitely a big plus!
The software/API from Ximea is not really as "rich" though as I had hoped so without quite a bunch of implementing/coding it's not easy to capture RAW video for testing purposes :( .. but I try to upload a few test frames soon.
Another problem with this sensor/head: Even being the newest Revision (3) of the CMV4000 fixed pattern noise is quite prominent! They are working on a software fix for this, but it's not a good solution for my use case because the VIA P-910 is already at it's limit regarding CPU/bandwidth so I would need the processing for this to be happing in the FPGA on the sensor head. There is another industrial camera manufacturer who has such a model already available! I'll get my hands on it in 1-2 weeks! (presumably they also have a better API/SDK)

Somebody asked me why I went with the CMV4000 instead of the CMV2000 which has a much more commen aspect ratio. Well to make it short: heads with the CMV2000 aren't as available currently and the CMV4000 sort of has everything the 2000 has as well! You can always read out a region of interest (ROI) (sort of a cropped down area) from the 4000 that is the same as the 2000 sensor area and get the same high readout rate/FPS. And you have the option to go 4:3 or even 1:1 for anamorphics which is a big plus imho.

rubiocesar83
03-26-2013, 11:25 AM
@shijan ... it's a native Nikon mount... Nikkor 50mm 1.4 lens on a (not really good) Nikon to C-Mount adaptor

So.. time for an update:

First impressions of the image quality of the CMV4000: very nice! Quite sharp but natural picture with good DOF (1" sensor) and the global shutter is definitely a big plus!
The software/API from Ximea is not really as "rich" though as I had hoped so without quite a bunch of implementing/coding it's not easy to capture RAW video for testing purposes :( .. but I try to upload a few test frames soon.
Another problem with this sensor/head: Even being the newest Revision (3) of the CMV4000 fixed pattern noise is quite prominent! They are working on a software fix for this, but it's not a good solution for my use case because the VIA P-910 is already at it's limit regarding CPU/bandwidth so I would need the processing for this to be happing in the FPGA on the sensor head. There is another industrial camera manufacturer who has such a model already available! I'll get my hands on it in 1-2 weeks! (presumably they also have a better API/SDK)

Somebody asked me why I went with the CMV4000 instead of the CMV2000 which has a much more commen aspect ratio. Well to make it short: heads with the CMV2000 aren't as available currently and the CMV4000 sort of has everything the 2000 has as well! You can always read out a region of interest (ROI) (sort of a cropped down area) from the 4000 that is the same as the 2000 sensor area and get the same high readout rate/FPS. And you have the option to go 4:3 or even 1:1 for anamorphics which is a big plus imho.

Good Soeren, why just for the sake of testing purposes, you don't ask the Norpix guys a 15 days StreaPix 5 test trial to get some video frames you can post so we can see what this sensor is capable of?

You can either after record the Uncompressed Raw sequences, de-Bayer them to Uncompressed RGB (MOV or AVI), or trans-code them to DNG files...

Thanks,
Cesar Rubio.

shijan
03-27-2013, 01:08 AM
yep i believe that Norpix trial is the way to go. Soeren, nice to hear your first impressions. Can you tell us something about HDR mode? how those 120 db of dynamic range looks like? are they real? do they made image more filmic or too unnatural HDRish? it's the most disturbing thing as for me at this moment...

Soeren Mueller
03-27-2013, 03:44 AM
You can either after record the Uncompressed Raw sequences, de-Bayer them to Uncompressed RGB (MOV or AVI), or trans-code them to DNG files...

Cesar, you get that my project is mainly about implementing my own direct-to-DNG recording software (which is working) and direct-to-CineformRAW (longer term), right?

Don't worry, I'll share grabs but currently my main goal is to see if I can go with the Ximea head or should asap switch to another one. That's why I'm currently doing more coding (evaluating their API) and not so much "real" sensor testing.

shijan, I'll look into it, although I don't expect any of these sensors to have a real "usable" HDR mode for videography purposes.

larsbentley
04-12-2013, 12:17 PM
Soeren,

Any updates on your project? How's that Ximea camera working out? We're all watching and waiting.

Lars

IanFanning
04-16-2013, 10:37 AM
Soeren,

Any updates on your project? How's that Ximea camera working out? We're all watching and waiting.

Lars

Lars,

Few months ago I needed small camera head capable to push 150fps overcranking and it was luring to try Cmosis CMV2000 or CMV4000 sensors machine vision camera heads having pretty large pixels, high-res and fast sensors with nice specs on the paper. Tested two heads, albeit none of these were actually used in production primarily due to problems with capture and image quality.
First head I got was the Ximea MQ042CG-CM head with CMV4000 sensor and USB 3.0 interface, luckily overnighted from machine vision distributor, however with hefty price tag. Hefty price tag I mean the one compared to the BMC Cinema camera where you’ll get real stuff in the camera and post. However the experience with Ximea head and their ultra-minimalistic camera viewer app constantly crashing upon frame grab and without option to grab anything other than 8bit bmp not talking about stream capture, was just plain frustration. On top of it, banding and irregular strange noise was clearly apparent in the shadows with the zero gain and fast 1/200s exposure in the 8bit files with curve slightly lifted to open shadows. Leaving head running at full fps for half an hour, white defective pixels and paper like background texture started to appear with the exposure of 1/30s and below. After week of unsuccessful wresting with support guys and camera software it went back to the distributor to be exchanged for cheaper Basler aca-2000-340 with CMV2000, which arrived two weeks later together with Camera link cables and PCI interface demo setup.
There was extra cost for interface board and cables, but still within budget having the option to return it within 30 days. To my great positive surprise with Basler head there were not even slightest traces of the banding noise, very fine film like grain with 10x gain and no white pixels at all, even if the head got warm pretty quickly. Supplied viewer and utilities were like day and night compared to these of Ximea camera head, no crashes and stable high frame rates. Judging the grabbed frames I would say there was higher quality sensor grade used in the Basler head, but talking later to Basler support, they revealed the Basler head has fully fledged hardware on several printed boards inside of the head, doing banding noise and white pixels correction and complete image processing thingies like de-bayering and gamma correction while the Ximea head transfer data straight from the sensor to the PC interface. This is why the Ximea head is smaller – actually it is only one printed board where most of it is occupied with the sensor alone and no processing, image processing or sensors corrections takes place there.
On the other hand, camera link cables and interfaces used with the Basler head are real pain making the camera harder to use on the set, but overall Basler head proved the CMV2000 sensor is definitely capable to capture excellent picture.
I’ll find and post some similar subject test frame grabs from both heads so you can see the difference between Basler and Ximea heads. It also appears there are many more vendors coming with the USB 3.0 CMV2000 and CMV4000 based cameras for less than half of the Ximea or Basler heads price like these of IDS (Germany) and Point Grey. Point Grey CMV2000 based head with USB 3.0 interface is my favorite for the next production where such small head will be needed, similar to Basler, they have full hardware in camera and the price checked this week is even lower than Basler or Ximea.

-IF

larsbentley
04-17-2013, 11:43 PM
Ian,

I'd love to see the test frames when you get a chance. I sent you a message too with some other questions.

Lars

Soeren Mueller
04-18-2013, 12:42 PM
Hi Lars/Ian!

Well what Ian describes I experienced as well - however this is not due to different (better) Sensors being used - all the CMV Sensors currently suffer from quite heavy "fixed pattern noise" (this is Revision 3 already, from what I heard it was even worse before!)
The big difference being: the Ximea head does currently no fixed pattern noise correction! Only bad/hot/stuck pixel correction, this is what Ian was seeing (I saw it as well). And of course their "capture" software is pretty limited. Although that's not a big issue for me as I have my own capture platform anyways, but for quick testing it would be better to have at least something comparable to the really nice Point Grey FlyCapture program.

So Ximea is still working to implement the fixed pattern noise removal/"fixing" but it will be done _in software_ on the host system, which is why the Ximea head isn't an option for me as the VIA P910 is already pretty much maxed out even with clever programming as is... there's not much margin left by any means to do something like that in software.
However IDS (and I guess Basler as well) are already doing fixed pattern noise "fixing", and not only that but they are doing it (at least for IDS that is the case) _in_ the camera head/on its FPGA, so the host system doesn't have additional load from that.
I'm about to receive the IDS CMV4000 head in the next few days, then I will finally provide test frames/footage.

Until now it looks really nice I have to say, the sensor has a nice/pleasing image, although I am now already sort of used to the dynamic range of my BMC and compared to that the CMV sensors are obviously much more limited. But I have to do some real testing before I can come to any conclusions. Until now I unfortunately only had time to play around a little with it.
But the high FPS and global shutter are really nice. And you get a 2048 x 1152 RAW DNG in 16:9 aspect ratio while the sensor "only" has a crop factor of about 1.2-1.3 compared to the BMC, so they are pretty much on par when it comes to DOF etc.

Of course Ian is right regarding the price tag. However my project is about a solution where the new BMC Pocket would come in handy. Until a few days ago nothing like that was somewhere "visible" on the "horizon". But I continue with my project albeit it currently feels a little "redundant" after the announcement of the Pocket BMC :(
But then who knows when BM will really be able to deliver ;)
My own "Pocket BMC" a least already records some real RAW frames to a SSD without having to wait several months ;)

Cheers
-Sören

rubiocesar83
04-18-2013, 07:10 PM
Hi Lars/Ian!

Well what Ian describes I experienced as well - however this is not due to different (better) Sensors being used - all the CMV Sensors currently suffer from quite heavy "fixed pattern noise" (this is Revision 3 already, from what I heard it was even worse before!)
The big difference being: the Ximea head does currently no fixed pattern noise correction! Only bad/hot/stuck pixel correction, this is what Ian was seeing (I saw it as well). And of course their "capture" software is pretty limited. Although that's not a big issue for me as I have my own capture platform anyways, but for quick testing it would be better to have at least something comparable to the really nice Point Grey FlyCapture program.

So Ximea is still working to implement the fixed pattern noise removal/"fixing" but it will be done _in software_ on the host system, which is why the Ximea head isn't an option for me as the VIA P910 is already pretty much maxed out even with clever programming as is... there's not much margin left by any means to do something like that in software.
However IDS (and I guess Basler as well) are already doing fixed pattern noise "fixing", and not only that but they are doing it (at least for IDS that is the case) _in_ the camera head/on its FPGA, so the host system doesn't have additional load from that.
I'm about to receive the IDS CMV4000 head in the next few days, then I will finally provide test frames/footage.

Until now it looks really nice I have to say, the sensor has a nice/pleasing image, although I am now already sort of used to the dynamic range of my BMC and compared to that the CMV sensors are obviously much more limited. But I have to do some real testing before I can come to any conclusions. Until now I unfortunately only had time to play around a little with it.
But the high FPS and global shutter are really nice. And you get a 2048 x 1152 RAW DNG in 16:9 aspect ratio while the sensor "only" has a crop factor of about 1.2-1.3 compared to the BMC, so they are pretty much on par when it comes to DOF etc.

Of course Ian is right regarding the price tag. However my project is about a solution where the new BMC Pocket would come in handy. Until a few days ago nothing like that was somewhere "visible" on the "horizon". But I continue with my project albeit it currently feels a little "redundant" after the announcement of the Pocket BMC :(
But then who knows when BM will really be able to deliver ;)
My own "Pocket BMC" a least already records some real RAW frames to a SSD without having to wait several months ;)

Cheers
-Sören

Dont give up Soeren, what you are doing is GREAT for 3-D Stereoscopic recordings.

The BMD cameras cannot be sync yet, and even if they can later, they will need still a beamsplittetr rig...

I am going to get some frames of these cameras (http://dream-genesis.blogspot.mx/2013/04/720p-mono-cameras.html) myself...

Cesar Rubio.

rubiocesar83
04-18-2013, 08:41 PM
Also BMD pocket camera can not do Uncompressed DNG, and you NEED that if you want to get more dynamic range, when pushing the highlights and shadow details in post...

CR.

noirist
04-18-2013, 08:59 PM
Also BMD pocket camera can not do Uncompressed DNG, and you NEED that if you want to get more dynamic range, when pushing the highlights and shadow details in post...

CR.
The BM Production Camera records lossy compressed DNG, but the BM Pocket Cinema Camera records mathematically lossless compressed DNG. The BMPCC footage is mathematically indistinguishable from uncompressed DNG.

rubiocesar83
04-19-2013, 01:30 AM
The BM Production Camera records lossy compressed DNG, but the BM Pocket Cinema Camera records mathematically lossless compressed DNG. The BMPCC footage is mathematically indistinguishable from uncompressed DNG.


Such thing DOES NOT exist..I have made LOTS of tests about that...

I will put it as simple as this, lets say you have your wallet with 10 bills, then you throw half of them to the street, then you look at what is left in your wallet again, would YOU call that mathematically-visually loss-less?

Not me.

CR.

MiguelFranco
04-19-2013, 10:05 AM
Also BMD pocket camera can not do Uncompressed DNG, and you NEED that if you want to get more dynamic range, when pushing the highlights and shadow details in post...

CR.

JB said more than once that you get the same dynamic range in ProRes that you get in RAW, so I think the same applies for compressed and uncompressed CinemaDNG...

rubiocesar83
04-19-2013, 01:18 PM
JB said more than once that you get the same dynamic range in ProRes that you get in RAW, so I think the same applies for compressed and uncompressed CinemaDNG...

I HIGHLY doubt that with Prores, you can push the files as in this extreme case (http://dna-rubio-s3d.blogspot.mx/2013/02/dynamic-rangeand-12-bit.html) here with the Uncompressed DNG footage.

Maybe they look close enough as naively shot, but NOT when pushed...

CR.

noirist
04-19-2013, 02:59 PM
Such thing DOES NOT exist..I have made LOTS of tests about that...

I will put it as simple as this, lets say you have your wallet with 10 bills, then you throw half of them to the street, then you look at what is left in your wallet again, would YOU call that mathematically-visually loss-less?

Not me.

CR.

"Mathematically lossless" means that when you uncompress the compressed file, you end up with the original file. Like when you zip a file, and then unzip it. The zipped file is different than the original file, and often smaller. But when you unzip the zipped file, it is identical to the original file.

rubiocesar83
04-19-2013, 03:22 PM
"Mathematically lossless" means that when you uncompress the compressed file, you end up with the original file. Like when you zip a file, and then unzip it. The zipped file is different than the original file, and often smaller. But when you unzip the zipped file, it is identical to the original file.

I have already discussed that with some so-called experts...

Its here if you want to read it

http://movies.groups.yahoo.com/group/3dtv/message/29055

If not, then I will make a simple explanation, while some limited size documents can be zipped and un-zipped later with no loss, that does NOT work for real time video recordings...

CR.

rubiocesar83
04-19-2013, 03:26 PM
But YOU don't have to believe me, nor the other guys either...DO YOUR OWN Tests on the matter...

That is the BEST Eye opener in things.

CR.

Soeren Mueller
04-19-2013, 04:29 PM
JB said more than once that you get the same dynamic range in ProRes that you get in RAW, so I think the same applies for compressed and uncompressed CinemaDNG...

Miguel, do you have a quote for that somewhere? If I remember correctly JB always said the exact opposite.. that with Prores and DNxHD you lose at least 1 stop compared to RAW!

Oh and btw sorry to rain in your guys parade but this discussion is quite off topic to my thread.. thanks ;)

I'll receive the IDS CMV4000 head on Monday or Tuesday and will post samples then!

Tried the Ximea CMV4000 with Streampix demo... works fine besides any of the RAW modes.. doesn't want to capture/show anything in any of these... sigh :( .. so perhaps I'll try to post some black and white or badly realtime debayered stuff... :(

rubiocesar83
04-19-2013, 06:34 PM
Miguel, do you have a quote for that somewhere? If I remember correctly JB always said the exact opposite.. that with Prores and DNxHD you lose at least 1 stop compared to RAW!

Oh and btw sorry to rain in your guys parade but this discussion is quite off topic to my thread.. thanks ;)

I'll receive the IDS CMV4000 head on Monday or Tuesday and will post samples then!

Tried the Ximea CMV4000 with Streampix demo... works fine besides any of the RAW modes.. doesn't want to capture/show anything in any of these... sigh :( .. so perhaps I'll try to post some black and white or badly realtime debayered stuff... :(


Why is off topic if WE are talking about Raw...and CinemaDNG uncompressed?...

And is OUR topic, of everyone participating in it.

I already pointed out to Norpix your statements here, to see their opinion on this...

Yes is their thread too, and Ximea, Point Grey and IDS since WE are discussing THEIR products too...

CR.

rubiocesar83
04-23-2013, 04:10 PM
This is the response I got from Norpix regarding the Xima camera recordings issues with StreamPix 5:

The Ximea Raw issue on the 4MP sensor is a bug we have reported to Ximea 3 month ago. The camera firmware returns a erroneous information upon a request for the maximum X size. It says 2048 while this is actually 2040. This crashes during DMA xfer because of mismatch image size and buffer size. Ximea recommended that we patch our software rather than them to fix the firmware. We are planning on releasing a patch this week.

CR.

Soeren Mueller
04-23-2013, 05:06 PM
Yes I already got their patch, I contacted them as well. Did some 3.8K test recordings today with the Sony sensor outside.. despite the rolling shutter (which isn't much of an issue when the sensor is on some sort of rig, in this case a crane) the images are really nice. Hope I can finally get around recording some CMV4000 footage the next few days.

Yesterday I got the IDS CMV4000 head so currently I have both - IDS and Ximea and can compare them.

Cesar, well I guess we have different viewpoints but to me this thread was intended to be primarily about my little embedded PC based project (and mainly about the Sony Point Grey head) and not as a "broad range" thread, there are better ones on DV-Info and other forums where every sort of sensor and software is covered. Especially I don't intend to discuss any Norpix based solution or products in this thread because this project is mainly thought out as an _alternative_ / different solution to that product. Thanks!

stip
04-23-2013, 05:20 PM
This thread is about Soeren's project.

rubiocesar83
04-23-2013, 05:22 PM
Yes I already got their patch, I contacted them as well. Did some 3.8K test recordings today with the Sony sensor outside.. despite the rolling shutter (which isn't much of an issue when the sensor is on some sort of rig, in this case a crane) the images are really nice. Hope I can finally get around recording some CMV4000 footage the next few days.

Yesterday I got the IDS CMV4000 head so currently I have both - IDS and Ximea and can compare them.

Cesar, well I guess we have different viewpoints but to me this thread was intended to be primarily about my little embedded PC based project (and mainly about the Sony Point Grey head) and not as a "broad range" thread, there are better ones on DV-Info and other forums where every sort of sensor and software is covered. Especially I don't intend to discuss any Norpix based solution or products in this thread because this project is mainly thought out as an _alternative_ / different solution to that product. Thanks!

Yes I understand Soeren.

I don't want to be disrespectful or an intruder on your project, but the last words you gave were like you already gave up on it since the BMD small camera announcement.

I told you NOT TO GIVE UP.

And I just want to mention the POSITIVE aspects of YOUR project for 3-D work....

For 2-D like I already TOLD you as well since I started participating in this thread, the first BMD 2.K camera is a BETTER option giving its price and capabilities.

And now with the 2 new camera announcement, well even MORE choices.

I would stick with the first camera due to the resolution and Uncompressed DNG capabilities, also the rolling shutter gives better DR than global shutters.

For 2-D work RS its fine...

Thanks,
CR.

rubiocesar83
04-23-2013, 05:28 PM
This thread is about Soeren's project.

Yes I understand.

But when he mentioned problems with StreamPix 5...well he himself took it to the other level.

Some people had requested some frame grabs with STP-5, and he said no because it was about his project.

But then he DID IT!

Now, is he going to let aside STP-5?

Why not compare both software so people can decide for themselves if its ok to purchase STP-5 or get the free? Soeren one...

CR.

larsbentley
04-23-2013, 05:28 PM
Yes, let's keep this about Soeren's project. Very interested in your work Soeren. I'd like to see some side by side comparisons of the two cmosis cameras and the sony sensor in low light. I'm wondering if dynamic range is much different, and if the larger pixel size of the cmosis sensors gives much better low light performance.

LARS

rubiocesar83
04-23-2013, 05:45 PM
Yes, let's keep this about Soeren's project. Very interested in your work Soeren. I'd like to see some side by side comparisons of the two cmosis cameras and the sony sensor in low light. I'm wondering if dynamic range is much different, and if the larger pixel size of the cmosis sensors gives much better low light performance.

LARS

Ok, but are you only interested in Soeren project because is FREE software?

What about if he decides to charge money for his HARD work he has put in it?...

CR.

noirist
04-23-2013, 07:43 PM
I'm interested in Soeren's project because he is providing very clear and interesting technical information about image sensors, lenses, and camera design. My interest is independent his business plans.

Cesar, Why don't you start your own thread where you can discuss your own projects?

John Brawley
04-23-2013, 08:42 PM
JB said more than once that you get the same dynamic range in ProRes that you get in RAW, so I think the same applies for compressed and uncompressed CinemaDNG...

No I don't think I've ever said that.

The ProRes is very good, and at the camera end it starts off with the same DR, but by the time it get's to being converted to ProRes there is a reduction.

jb

rubiocesar83
04-24-2013, 07:44 AM
No I don't think I've ever said that.

The ProRes is very good, and at the camera end it starts off with the same DR, but by the time it get's to being converted to ProRes there is a reduction.

jb

And you see the reduction right away off he bat, less sharpness-definition-, less color gamut, etc.

And when you start pushing the files in post you see GREATER difference.

I noticed that even with Cineform 444 files that are vast superior to ProRes.

And I did these test way back BEFORE the BMD camera was available.

Years ago...

But it does not matter what camera you use, it will be the SAME thing with ANY camera system.

I asked Marco Solorio to see what is his opinion on this regarding PropRes, and if he can achieve he same thing with it as here (http://www.onerivermedia.com/blog/?p=750#comment-10085) with Uncompressed CinemaDNG files.

And again, THIS is related too with Soeren project....

CR.

rubiocesar83
04-24-2013, 04:24 PM
In the meantime while we wait for the images of Soren am Ian, here (http://dream-genesis.blogspot.mx/2013/04/rolling-vs-global-shutter.html) are some grabs that IDS sent me.

They are good for the ones that want to compare Rolling vs Global Shutter modes.

CR.

jsfotografie
04-27-2013, 08:32 AM
Yes I already got their patch, I contacted them as well. Did some 3.8K test recordings today with the Sony sensor outside.. despite the rolling shutter (which isn't much of an issue when the sensor is on some sort of rig, in this case a crane) the images are really nice. Hope I can finally get around recording some CMV4000 footage the next few days.

Yesterday I got the IDS CMV4000 head so currently I have both - IDS and Ximea and can compare them.

Cesar, well I guess we have different viewpoints but to me this thread was intended to be primarily about my little embedded PC based project (and mainly about the Sony Point Grey head) and not as a "broad range" thread, there are better ones on DV-Info and other forums where every sort of sensor and software is covered. Especially I don't intend to discuss any Norpix based solution or products in this thread because this project is mainly thought out as an _alternative_ / different solution to that product. Thanks!


cant wait for the sample footage dude :-)

cheers janosch

nicolassiuda
05-28-2013, 02:44 PM
very Thread, congrat for your work you've already done ...

CRFilms
08-31-2013, 09:35 PM
Any updates? I'd read about the Flea3 camera on another forum, but nobody had any info on it. Found this thread a few months ago and was glad you gave it a try. I think the camera is a hundred bucks cheaper now than when you bought it and the model that maxes out at 2k 60fps capture is around $650. Also using a PC designed for automotive installation might be a good setup for mobile use: http://www.mini-box.com/Car-PC-Automotive-Computing-Solutions

Have you finished your capture software and are you going to give it away or charge for it? Once the Blackmagic pocket cam get's it's issues worked out, that would be the best bet for cheap raw, but your setup could be the cheapest 2k 60fps raw.

jsfotografie
09-01-2013, 01:43 AM
Any updates? I'd read about the Flea3 camera on another forum, but nobody had any info on it. Found this thread a few months ago and was glad you gave it a try. I think the camera is a hundred bucks cheaper now than when you bought it and the model that maxes out at 2k 60fps capture is around $650. Also using a PC designed for automotive installation might be a good setup for mobile use: http://www.mini-box.com/Car-PC-Automotive-Computing-Solutions

Have you finished your capture software and are you going to give it away or charge for it? Once the Blackmagic pocket cam get's it's issues worked out, that would be the best bet for cheap raw, but your setup could be the cheapest 2k 60fps raw.

dont want to hijack sörens thread but i build a similar camera http://www.eoshd.com/comments/topic/2874-my-miniraw-1-camera-needs-help/ and the car pc solution will be way to slow im doing 720p 120fps 12bit raw and for that i needed an raid 0 ssd raid and a beefy intel i7 3770T :-)
hope to here an update from sören :-)

cheers janosch

rubiocesar83
09-04-2013, 09:29 AM
So YOU abandoned the FIGHT already Soeren?...

Buenos Dias :-)

CR.