My little DIY camera project

John Wyatt · March 5th, 2006, 11:49 AM

Noah -- I thought when an adapter includes an optical element as you described, it suggests that the adapter is involved in redirecting light because of the difference in original coverage, though I'm probably wrong about that. I've started the new lens thread called "Call for C-mount lens info", so maybe we can find out more about these great little lenses if others join in.

Regards,
John.

Noah Yuan-Vogel · April 5th, 2006, 12:18 PM

Sorry ive been bad about updating lately, I was traveling and hadnt made any significant progress. But to update whats going on with my camera, Sumix has sent me the spacer I need to fix the backfocus on my lenses with the behind the lens IR filter. For now I've decided that IR filter is enough (my computar lens actually came with another IR filter so that improves the situation a little). Neither nor both, however, are as good as the B+W IR/UV cut filter that i cant get to mount on my lenses.

I've been working on the program a little, I intended to test audio/video recording at the same time to test whether the camera's framerate will allow it to sync with audio, but i found that my computer just cant really handle doing both at once. This in turn has turned me toward focusing on ram recording since that may be required to reduce cpu usage and data writing lag and since I know i want to eventually end up with resolutions higher than 1024x576. Just today I finally picked up my camera handheld and recorded some shots to check out the rolling shutter artifact. Unfortunately it reminded me rolling shutter could be a big problem depending on what i am shooting. I think I might try to play around with some settings a bit and post some videos and see if people here can give me advice as to what is acceptable in terms of rolling shutter artifacting. I will keep looking further into using blanking instead of sumix's built in functions to control framerate, since it seems likely those functions are not at all written to minimize rolling shutter issues.

John Wyatt · June 14th, 2006, 10:54 AM

Noah - any news on your project? Have you shot any more tests with the M73 camera? Rolling shutter and RAM recording results would be of great interest...
Regards,
John.

Noah Yuan-Vogel · June 14th, 2006, 12:27 PM

Hey I'm glad there is still some interest. I have been working on the camera on and off, but really not much lately, although I did get adapters to allow me to mount my IR filter on all of my lenses, which has so far given me good results. I have been experimenting with rolling shutter a bit, but have come to the conclusion, its not great but it will do as long as i set framerates myself. If the siliconimaging HD camera runs its rolling shutter only twice as fast as its framerate, I figure that should be good enough for my camera. this means I should be able to handle a ~20MBps data stream at 24fps. (~40MPps performance at 48Mhz, i think this has to do with minimum possible horizontal blanking limits and control data overhead, the sensor's manual seems to confirm this somewhat, but I am still not sure why my actual pixel rate seems so hard to calculate, although it is consistent) Maybe some more if i allow a little more artifacting.
Actually the main thing i have worked on lately is the LUTs, which i find to be very cool, since I can get an 8-bit log image out of my 10-bit hardware. This means more highlight latitude (subjectively) and more apparent light sensitivity. Of course 10bits linear isnt a ton to start out with but its nice to have so much control over it.
Another thing that interests me is when I looked through the micron manual, its exciting the amount of control that an be had through register programming. I hope to get into this since sumix supplies functions that allow access to the cameras registers. for example I might be able to do horizontal line skipping to get an anamorphic image, so that i could get a 960x1080 anamorphic image that uses (nearly) the whole area of the chip and is very compatible with 1920x1080 HD and fits my datarate requirements. This wouldnt be too different from the imaging of the FX1/Z1 (except mines 1/2", 24p, uncompressed dtd, 8bit log, with better faster lenses, and cheaper).
Anyway, there isnt a lot of time to work on this lately and my computer has been acting up (it seems moroccan electricity may be damaging to computer equipment) but i want to at least try to get some nice handheld test footage off the balcony of my new apartment. I also now have a new website to post it on. I'll let you know when that happens.

Jason Rodriguez · June 15th, 2006, 02:18 PM

Hi Noah,

We're way over twice the frame-rate for the rolling shutter . . . while double is good, we're clocked much higher than that.

Wayne Morellini · June 16th, 2006, 04:10 AM

I think people are still subscribed to the thread, but we are just waiting to hear from you.

Noah Yuan-Vogel · June 16th, 2006, 07:16 AM

Jason, sorry, that was just what I gathered from the siliconimaging forums, Ari had told me that you were running at twice the pixelrate and capturing every other frame, which I imagine would lead to a rolling shutter difference of half a frame length, no? This made sense to me since I figured the camera runs at 75MHz or so which on an altasens translates to 150MP/s which would make sense for 12bit 1080/24p (~75MP/s). You're saying the silicon imaging camera actually runs at over 150MP/s? or am I missing something? I thought the altasens couldnt go past maybe 85MHz max.

Noah Yuan-Vogel · June 19th, 2006, 01:27 PM

Quick update:
I finally got around to playing with sensor registers directly through sumix's API. They do not provide any documentation on it, but it turned out to be surprisingly easy to do, and I successfully implemented anamorphic imaging in a matter of minutes (and after bugged sumix so much to implement it in their software). This means I can run at resolutions like 2048x1152 using the full frame skipping every other column to get a squished 1024x1152 image that is much easier to deal with in terms of data rates and using extra pixel clocks for vertical blanking (cutting down rolling shutter artifacts) . Anyway, I am excited about being able to use the whole size of the sensor and be able to run at resolutions compatible with 2k and 1080p. I've also been able to optimize the recording algorithm a bit, but still not enough to get more than 960x540 to disk at 24p with my slow slow processor. I am getting excited about my plans for upgrades, I got the idea to build a modular computer system that folds out to fit in a suitcase for a mobile editing bay and that folds up to about the right form factor to run on rails with my camera head for handheld work. I'll take some shots with the camera in the next week and post them on my new website, and I'll be sure to add updates when i return to the US and am able to start building my new and improved computer end (should end up being compatible with the siliconimaging camera head, just in case ever get enough money to afford one and get tired of working on my own camera)

Noah Yuan-Vogel · June 22nd, 2006, 10:13 AM

This isnt anything totally new or special, but I have a video now posted:

http://www.noahyv.com/videos/01_01_01_01_xvid.avi

Let me know what you think and if there are things I am not seeing that needs to be worked on. For one, its at 960x540 which is still a bit too much for my prototype 1ghz computer for the camera(you can tell since it drops one or two frames), in the future ill probably just post 720x540 videos until i upgrade my system. Also note this is a compressed representation of uncompressed video, so any compression artifacting you see are not issues with the raw video I am capturing, just the xvid compression. Note the gamma curves which i've implemented based on some film log response curves. This was shot with vertical blanking set to more than the vertical height of the frame so any rolling shutter artifacting in this video will probably be less than when I run the camera capturing a larger frame resolution.

Jason Rodriguez · June 22nd, 2006, 10:29 AM

Quote:

This means I can run at resolutions like 2048x1152 using the full frame skipping every other column to get a squished 1024x1152 image

Just curious, but will that create aliasing artifacts when you un-stretch the image? I'm assuming optically there will be "gaps" where those columns were suppose to be, so it's not exactly the same as using true optical anamorphics is it?

BTW, Your video looks great!

You should think about using a new Core Duo setup with the Yonah or Merom procs . . . there's a ton of processing performance on those machines (I mean we're doing a bunch of real-time signal processing AND CineForm compression at very high resolutions on these procs).

Noah Yuan-Vogel · June 22nd, 2006, 10:36 AM

Yeah that is definitely true. I have thought about that, and the sensor manual says that is supports binning (reading extra pixel quads and averaging them), but I have not gone that far into it and do not know what impact that would have on the pixelrate. I would imagine it is less aliasing than i have seen running at full 2x skipping (im pretty sure sumix's standard binning is just skipping).

Actually I have settled on a mobo and proc, i plan to go with the core duo and the ibase MB899F motherboard. Jason, any guess on whether that is compatible with the silicon imaging camera? The question is just the gigabit ethernet controller i suppose.

Jason Rodriguez · June 22nd, 2006, 01:11 PM

My suggestion if you want a mini-itx mobo that's ideal for the SI camera would be the Aopen i945GTt-VFA.

Now for our camera, if you can wait for Merom, I'd do that (overhead is always a good thing) . . . of course you could always just add a Merom later.

DDR2-667 is also must (compared to DDR2-533), and it's compatable with this board.

http://global.aopen.com.tw/products/mb/i945GTt-VFA.htm

Noah Yuan-Vogel · June 24th, 2006, 05:44 AM

Thanks for the suggestion, does that support HDTV component out, out of the box? I just liked that the other one i was looking at could support DVI, HDTV, VGA, TV and has full size ram modules (which may be more available and better value). Is the aopen board what SI is using for their setup?

Jason Rodriguez · June 29th, 2006, 09:27 AM

The Aopen does support HD component analog output out-of-the-box.

I would highly recommend it.

John Wyatt · July 1st, 2006, 03:02 AM

Noah -- I've ordered an M73. Hoping to familiarise myself with the software having the camera connected to my dektop, later will buy a laptop. I'm looking around for an Angenieux 15mm lens, but have a Pentax K/C-mount adapter to use a 28mm SLR lens in the meantime to start some tests. I'll keep you posted on how I get on. I believe the software is capable of recording a still image sequence, and I would like to experiment with this at 25 pictures a second (at whatever frame size works within the various bandwidth limitations).
John.

June 14th, 2006, 10:54 AM	#93
John Wyatt Regular Crew Join Date: Jan 2006 Location: West Country, UK Posts: 141	News? Noah - any news on your project? Have you shot any more tests with the M73 camera? Rolling shutter and RAM recording results would be of great interest... Regards, John.

March 5th, 2006, 11:49 AM	#91
John Wyatt Regular Crew Join Date: Jan 2006 Location: West Country, UK Posts: 141	Noah -- I thought when an adapter includes an optical element as you described, it suggests that the adapter is involved in redirecting light because of the difference in original coverage, though I'm probably wrong about that. I've started the new lens thread called "Call for C-mount lens info", so maybe we can find out more about these great little lenses if others join in. Regards, John.

April 5th, 2006, 12:18 PM	#92
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Sorry ive been bad about updating lately, I was traveling and hadnt made any significant progress. But to update whats going on with my camera, Sumix has sent me the spacer I need to fix the backfocus on my lenses with the behind the lens IR filter. For now I've decided that IR filter is enough (my computar lens actually came with another IR filter so that improves the situation a little). Neither nor both, however, are as good as the B+W IR/UV cut filter that i cant get to mount on my lenses. I've been working on the program a little, I intended to test audio/video recording at the same time to test whether the camera's framerate will allow it to sync with audio, but i found that my computer just cant really handle doing both at once. This in turn has turned me toward focusing on ram recording since that may be required to reduce cpu usage and data writing lag and since I know i want to eventually end up with resolutions higher than 1024x576. Just today I finally picked up my camera handheld and recorded some shots to check out the rolling shutter artifact. Unfortunately it reminded me rolling shutter could be a big problem depending on what i am shooting. I think I might try to play around with some settings a bit and post some videos and see if people here can give me advice as to what is acceptable in terms of rolling shutter artifacting. I will keep looking further into using blanking instead of sumix's built in functions to control framerate, since it seems likely those functions are not at all written to minimize rolling shutter issues.

June 14th, 2006, 12:27 PM	#94
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Hey I'm glad there is still some interest. I have been working on the camera on and off, but really not much lately, although I did get adapters to allow me to mount my IR filter on all of my lenses, which has so far given me good results. I have been experimenting with rolling shutter a bit, but have come to the conclusion, its not great but it will do as long as i set framerates myself. If the siliconimaging HD camera runs its rolling shutter only twice as fast as its framerate, I figure that should be good enough for my camera. this means I should be able to handle a ~20MBps data stream at 24fps. (~40MPps performance at 48Mhz, i think this has to do with minimum possible horizontal blanking limits and control data overhead, the sensor's manual seems to confirm this somewhat, but I am still not sure why my actual pixel rate seems so hard to calculate, although it is consistent) Maybe some more if i allow a little more artifacting. Actually the main thing i have worked on lately is the LUTs, which i find to be very cool, since I can get an 8-bit log image out of my 10-bit hardware. This means more highlight latitude (subjectively) and more apparent light sensitivity. Of course 10bits linear isnt a ton to start out with but its nice to have so much control over it. Another thing that interests me is when I looked through the micron manual, its exciting the amount of control that an be had through register programming. I hope to get into this since sumix supplies functions that allow access to the cameras registers. for example I might be able to do horizontal line skipping to get an anamorphic image, so that i could get a 960x1080 anamorphic image that uses (nearly) the whole area of the chip and is very compatible with 1920x1080 HD and fits my datarate requirements. This wouldnt be too different from the imaging of the FX1/Z1 (except mines 1/2", 24p, uncompressed dtd, 8bit log, with better faster lenses, and cheaper). Anyway, there isnt a lot of time to work on this lately and my computer has been acting up (it seems moroccan electricity may be damaging to computer equipment) but i want to at least try to get some nice handheld test footage off the balcony of my new apartment. I also now have a new website to post it on. I'll let you know when that happens.

June 15th, 2006, 02:18 PM	#95
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	Hi Noah, We're way over twice the frame-rate for the rolling shutter . . . while double is good, we're clocked much higher than that.

June 16th, 2006, 04:10 AM	#96
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I think people are still subscribed to the thread, but we are just waiting to hear from you.

June 16th, 2006, 07:16 AM	#97
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Jason, sorry, that was just what I gathered from the siliconimaging forums, Ari had told me that you were running at twice the pixelrate and capturing every other frame, which I imagine would lead to a rolling shutter difference of half a frame length, no? This made sense to me since I figured the camera runs at 75MHz or so which on an altasens translates to 150MP/s which would make sense for 12bit 1080/24p (~75MP/s). You're saying the silicon imaging camera actually runs at over 150MP/s? or am I missing something? I thought the altasens couldnt go past maybe 85MHz max.

June 19th, 2006, 01:27 PM	#98
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Quick update: I finally got around to playing with sensor registers directly through sumix's API. They do not provide any documentation on it, but it turned out to be surprisingly easy to do, and I successfully implemented anamorphic imaging in a matter of minutes (and after bugged sumix so much to implement it in their software). This means I can run at resolutions like 2048x1152 using the full frame skipping every other column to get a squished 1024x1152 image that is much easier to deal with in terms of data rates and using extra pixel clocks for vertical blanking (cutting down rolling shutter artifacts) . Anyway, I am excited about being able to use the whole size of the sensor and be able to run at resolutions compatible with 2k and 1080p. I've also been able to optimize the recording algorithm a bit, but still not enough to get more than 960x540 to disk at 24p with my slow slow processor. I am getting excited about my plans for upgrades, I got the idea to build a modular computer system that folds out to fit in a suitcase for a mobile editing bay and that folds up to about the right form factor to run on rails with my camera head for handheld work. I'll take some shots with the camera in the next week and post them on my new website, and I'll be sure to add updates when i return to the US and am able to start building my new and improved computer end (should end up being compatible with the siliconimaging camera head, just in case ever get enough money to afford one and get tired of working on my own camera)

June 22nd, 2006, 10:13 AM	#99
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	This isnt anything totally new or special, but I have a video now posted: http://www.noahyv.com/videos/01_01_01_01_xvid.avi Let me know what you think and if there are things I am not seeing that needs to be worked on. For one, its at 960x540 which is still a bit too much for my prototype 1ghz computer for the camera(you can tell since it drops one or two frames), in the future ill probably just post 720x540 videos until i upgrade my system. Also note this is a compressed representation of uncompressed video, so any compression artifacting you see are not issues with the raw video I am capturing, just the xvid compression. Note the gamma curves which i've implemented based on some film log response curves. This was shot with vertical blanking set to more than the vertical height of the frame so any rolling shutter artifacting in this video will probably be less than when I run the camera capturing a larger frame resolution.

June 22nd, 2006, 10:36 AM	#101
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Yeah that is definitely true. I have thought about that, and the sensor manual says that is supports binning (reading extra pixel quads and averaging them), but I have not gone that far into it and do not know what impact that would have on the pixelrate. I would imagine it is less aliasing than i have seen running at full 2x skipping (im pretty sure sumix's standard binning is just skipping). Actually I have settled on a mobo and proc, i plan to go with the core duo and the ibase MB899F motherboard. Jason, any guess on whether that is compatible with the silicon imaging camera? The question is just the gigabit ethernet controller i suppose.

June 22nd, 2006, 01:11 PM	#102
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	My suggestion if you want a mini-itx mobo that's ideal for the SI camera would be the Aopen i945GTt-VFA. Now for our camera, if you can wait for Merom, I'd do that (overhead is always a good thing) . . . of course you could always just add a Merom later. DDR2-667 is also must (compared to DDR2-533), and it's compatable with this board. http://global.aopen.com.tw/products/mb/i945GTt-VFA.htm

June 24th, 2006, 05:44 AM	#103
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Thanks for the suggestion, does that support HDTV component out, out of the box? I just liked that the other one i was looking at could support DVI, HDTV, VGA, TV and has full size ram modules (which may be more available and better value). Is the aopen board what SI is using for their setup?

June 29th, 2006, 09:27 AM	#104
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	The Aopen does support HD component analog output out-of-the-box. I would highly recommend it.

July 1st, 2006, 03:02 AM	#105
John Wyatt Regular Crew Join Date: Jan 2006 Location: West Country, UK Posts: 141	Noah -- I've ordered an M73. Hoping to familiarise myself with the software having the camera connected to my dektop, later will buy a laptop. I'm looking around for an Angenieux 15mm lens, but have a Pentax K/C-mount adapter to use a 28mm SLR lens in the meantime to start some tests. I'll keep you posted on how I get on. I believe the software is capable of recording a still image sequence, and I would like to experiment with this at 25 pictures a second (at whatever frame size works within the various bandwidth limitations). John.