High Definition with Elphel model 333 camera

Andrey Filippov · April 17th, 2006, 11:47 AM

Quote:

Originally Posted by Frank Hool

Answers to my own questions:
image plane = 6.55mm*4.92mm
registration distance = 17.52mm
FO taper = no
am i right?

yes, you are right

Forrest Schultz · April 17th, 2006, 07:36 PM

I do not even know how this is possible, But i was able to stream

1600 x 896 frame size @ 24fps. (16:9 aspect ratio, note: 900 had to 896)

and 1920 x 816 @ 24fps (2.35:1 aspect ratio)

This is unreal ! but i checked everything in the avi file on a timescale and made sure there are 24 frames for every second of footage and there is! it is true 24 at 1600 x 896 and 1920 x 816

the only thing i see that is the only downside of using more of the sensor is that the electronic rolling shutter artifact is more noticable.(at 1280 by 720, your only using a small portion of the sensor, so the electronic rolling shutter can sweep through unoticed. but even at 1600 x 896 the rolling shutter isnt too bad. but for 1920 x 816 you see the impact alot because of the widescreen. but for shots without quick whips , or shaky handheld footage. the rolling shutter can't be too bad.

I am just amazed that the camera can actually stream these sizes at these speeds. i find it unbeleivable. is there a reason why this is possible Andrey? thanks

Andrey Filippov · April 17th, 2006, 09:02 PM

Quote:

Originally Posted by Forrest Schultz

I do not even know how this is possible, But i was able to stream

1600 x 896 frame size @ 24fps. (16:9 aspect ratio, note: 900 had to 896)

can be up to 26.93fps

Quote:

Originally Posted by Forrest Schultz

and 1920 x 816 @ 24fps (2.35:1 aspect ratio)

up to 24.57, that's correct

BTW - I've just made some FPGA mods to provide exact (to 1 usec) timestamp for each frame (measures at the start of the first line in each frame) - just need to add software to deliver this info through the stream.

Quote:

Originally Posted by Forrest Schultz

is there a reason why this is possible Andrey? thanks

333 can fully support the 3MPix Micron MT9T001 - there are timing calculation formulae in the datasheet (http://download.micron.com/pdf/datas...01_3100_DS.pdf)

The new 5MPix 96MHz will need an upgrade in the camera to keep up with the maximal frame rate (pixel rate can be 96MHz even with 333), it can also benefit from 12 (vs.10) bits - model 323 camera that uses older 313 boards receives 14 bit data.

Frank Hool · April 18th, 2006, 12:54 AM

one more question, Andrey. Do You have any experiencies using non-C-mount lens with such camera? In other words is it possible to use there FO taper front of ccd to get larger image plane. Wich would be very helpful to use there 35mm photolens in full functionality.

Wayne Morellini · April 18th, 2006, 01:17 AM

Did I say something wrong?:

Speed:
So, are you saying that the Ogg codec can do 70Mbit per second (and 9MB/s) that definitely helps?

Kodak sensor:
..will it do a 720p or 1080p frame at 25fps?

There has been talk of upcoming 5Ghz programmable gate array technology, is that any close to a commercial product?

The bayer differential compressing algorithm concept?

Thanks

Wayne.

Konstantin Serafimov · April 18th, 2006, 03:25 AM

i apologize for nit picking here, but, is the MT9T001 the only micron's sensor that does global shutter? So why do you have rolling shutter ussues at widescreen rez then?

And, honestly, all the footage placed in this thread is either overburn or noisy due to low light. Is it camera, or shooter or setup difficulties? In this form, it slightly misses main reason of sample footage - to show the device in its best.

to go a bit conceptual... i think from point of this forum, there are 2 features absolutely nesessary to make a good industrial camera head usable for digital filmmaking. First is direct recording of quality stream to harddrive and second is a way to monitor the picture on a device attached to the camera.

As far as i understand, the present camera is a kind of sbc with fpga, so ideally it should be sbc with vga-out and ide/sata, where fpga doing debayer and resize function for preview. But it is different concept (from network camera) and camera board redisign. Not good. And even that ideal system, would require 2 person team to operate. 1 to control the camera settings via laptop and 1 to actually point and shoot. To eliminate this problem there must be all in one place: lense, head, recorder and screen. Why not attach a compact capable laptop to the head then, like a Tablet PC? And a battery bank.

But in general, how could you summarize the reasons to use such a camera in real life shooting. Except very special areas, like permanent multicam video system in small music club, where it is probably excelent option to take already, and a toy for filmmaker wannabe who loves cameras more than above mentioned fimmaking? Iam not offending anybody here, of course, except myself :).

Wayne Morellini · April 18th, 2006, 04:48 AM

Konstantin,

I have covered most of these points a few times. even with Andrey by email in times past. And yes I have noted what you said, but I was being kind not to point it out, because it does not prove how much it can do (and sensor and compression has a lot to do with it).

What I have been discussing here is ways to maximise performance to get it upto quality output, and defining what the limits are. Including a practical bayer scheme.

The problem is that it is a security camera (not industrial)so it is made for good security quality, and limited. A little extra quality and we can get to at least pro video quality. The Kodak sensor might deliver a lot better performance, but I note that Micron has extended latitude features now (there MT9V032 wide vga security chip pdf shows 32K lux and 2 lux test charts in the same shot. So there is hope for this to turn up on the HD chips.

Computer:
If you want a small platform, then car-PCs (about the size of a DVD drive) Origami and UMD, small cheap Tablet PC's, are worth looking at. There might be one out there with higher res screen and microphone input/s.

Have fun with it, we need your sort of thinking around here.

Thanks

Wayne.

Forrest Schultz · April 18th, 2006, 11:57 AM

Yes, you are correct that none of the test footage was performed under ideal condintions. and also, i recently found the zoom on the lens i am using has to be turned just a nudge away from 13.5mm to actually be at true infinity focus, where everything is in focus. i think the reason being is that, i can go past 13.5mm to macro function, and i was too close to macro. and note my lens is at f1.8 with it open.

Can i ask if you have seen all the footage posted on the link? there is one shot called guitar chair. and i wouldnt call that blown out. i had some focus issues perhaps on it, but nothing much else.

and of course we can say there are improvements to be made, but that can always be said. What Andrey has here is a high definition network camera that he probally didnt plan to be used for this purpose. And he has made a great camera capable of beautful picture. and i applaud him for that. He has been very generous and helpful in making this project go foward. I will have to try a little harder to get shots out that can please you a little better. And its not a bad thing, it keeps me on my toes, which is good. I am going to shoot some stuff right now infact. ill see what i can do. thanks

Oscar Spierenburg · April 18th, 2006, 01:30 PM

Yes... but please don't make this forum into a commercial discussion thing. We are discussing the first attempts of using a security camera for film making.
Forrest is sharing his experiments here, he's not selling a camera or something. Please share every footage, good and bad.

I haven't received my camera yet, but they say it should arrive in a few days.

Forrest Schultz · April 18th, 2006, 01:39 PM

Yes, thank you Oscar. We should discuss and show what the camera is capable of, not "what the camera should be capable of in a perfect world"

Thats what this thread is about, to show progress of using this as a film camera. And here is some more footage i shot this morning.

The setup was this: Bright Sunny morning. The subject is under the shade of the tree in my backyard, the sky in the background was basically white to my eyes, with a little blue tint, but it was very bright. I tried my best to show that the camera can get a subject in the shade, while still preserving the colors of a bright background such as the sky.
here is the file: http://www.savefile.com/files.php?fid=2614781

It as at full 1600 x 896 resolution but i convereted it to wmv for space. Slower computers (like mine) might have a bit of trouble playing back the high bit rate of the video. if anyone has this problem, and wants to see a lower rez version. let me know.

Konst Seraf · April 18th, 2006, 03:01 PM

well, i should apologize for being extra critical, the "chair" file is correctly exposed, and overall visually pleasent. But, there was one problem that become apparent in last the file, with girl. It is luck of crispness. May I consider that the lense you are using is not even an industrial purpose megapixel lense? You wrote in initial post that it was some combination of 35mm model with wideangle adaptor. It may be not adequate.

I had no intention to start a commercial discussion. But i think even being a thing in itself, the project like this still requires some perspective view. What is it good for and what is the goal? The only reason to use such a modular system in real shooting process is execptional video quality. Lets say noticebly better than present prosumer hdv. An interesting task.

Eric Gorski · April 18th, 2006, 03:27 PM

that last clip is very nice. the footage you're shooting reminds me alot of clips that a board member by the name of OBIN was shooting a year ago with a similar camera, possibly the same chip?

Forrest Schultz · April 18th, 2006, 03:37 PM

Konstantin, you do not need to apologize at all. Your comments were very helpful in the for the fact that i must continue to better the progress and quality.

The lens i am using now is an f1.8 zoom lens that is c-mount. It is a correct type of lens, but prehaps its glass isnt good enough for the the cameras resoultion. That might be the case. I just finished ordering a 25mm c-mount prime lens. with f1.4. With this lens, i am going to build the 35mm adapter onto the camera with this lens to serve as the relay lens.

I will post footage with that f1.4 to see how it handles the resolution also. thank you.

And thank you Eric. I can't remember what chip Obin used, but the process that were going is quite similar. The chip in this model camera is a 3MPix Micron MT9T001 1/2" CMOS.

Andrey Filippov · April 18th, 2006, 07:47 PM

Quote:

Originally Posted by Forrest Schultz

http://www.savefile.com/files.php?fid=2614781
It as at full 1600 x 896 resolution but i convereted it to wmv for space.

Seems some frames were still lost - Mplayer shows 21.75 fps average - 32 sec, 696 frames

I hope it will soon be easier to troubleshoot - each frame will have a precise timestamp

Forrest Schultz · April 18th, 2006, 08:41 PM

Oh wow, i didnt check that one. let me see if that has happened to some more recent footage i did. Thanks Andrey

EDIT: well it seems that things got a bit weird when i convereted to .wmv

The original footage coverted to avi from ogm via VirtualDub shows full 24 fps. 24 frames for every 1 second of footage. for some reason, the wmv version doesnt want to play right, also the wmv in windows media player states 29 sec duration. perhaps when it says 32 sec in mplayer it is not playing it right? i dont know.

April 18th, 2006, 01:17 AM	#65
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Did I say something wrong?: Speed: So, are you saying that the Ogg codec can do 70Mbit per second (and 9MB/s) that definitely helps? Kodak sensor: ..will it do a 720p or 1080p frame at 25fps? There has been talk of upcoming 5Ghz programmable gate array technology, is that any close to a commercial product? The bayer differential compressing algorithm concept? Thanks Wayne. Last edited by Wayne Morellini; April 18th, 2006 at 02:53 AM.

April 18th, 2006, 01:39 PM	#70
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	Yes, thank you Oscar. We should discuss and show what the camera is capable of, not "what the camera should be capable of in a perfect world" Thats what this thread is about, to show progress of using this as a film camera. And here is some more footage i shot this morning. The setup was this: Bright Sunny morning. The subject is under the shade of the tree in my backyard, the sky in the background was basically white to my eyes, with a little blue tint, but it was very bright. I tried my best to show that the camera can get a subject in the shade, while still preserving the colors of a bright background such as the sky. here is the file: http://www.savefile.com/files.php?fid=2614781 It as at full 1600 x 896 resolution but i convereted it to wmv for space. Slower computers (like mine) might have a bit of trouble playing back the high bit rate of the video. if anyone has this problem, and wants to see a lower rez version. let me know. Last edited by Forrest Schultz; April 18th, 2006 at 02:09 PM.

April 17th, 2006, 07:36 PM	#62
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	I do not even know how this is possible, But i was able to stream 1600 x 896 frame size @ 24fps. (16:9 aspect ratio, note: 900 had to 896) and 1920 x 816 @ 24fps (2.35:1 aspect ratio) This is unreal ! but i checked everything in the avi file on a timescale and made sure there are 24 frames for every second of footage and there is! it is true 24 at 1600 x 896 and 1920 x 816 the only thing i see that is the only downside of using more of the sensor is that the electronic rolling shutter artifact is more noticable.(at 1280 by 720, your only using a small portion of the sensor, so the electronic rolling shutter can sweep through unoticed. but even at 1600 x 896 the rolling shutter isnt too bad. but for 1920 x 816 you see the impact alot because of the widescreen. but for shots without quick whips , or shaky handheld footage. the rolling shutter can't be too bad. I am just amazed that the camera can actually stream these sizes at these speeds. i find it unbeleivable. is there a reason why this is possible Andrey? thanks

April 18th, 2006, 12:54 AM	#64
Frank Hool Major Player Join Date: Apr 2006 Location: Barca Spain Posts: 384	one more question, Andrey. Do You have any experiencies using non-C-mount lens with such camera? In other words is it possible to use there FO taper front of ccd to get larger image plane. Wich would be very helpful to use there 35mm photolens in full functionality.

April 18th, 2006, 03:25 AM	#66
Konstantin Serafimov New Boot Join Date: Apr 2005 Location: Riga, Latvia Posts: 13	i apologize for nit picking here, but, is the MT9T001 the only micron's sensor that does global shutter? So why do you have rolling shutter ussues at widescreen rez then? And, honestly, all the footage placed in this thread is either overburn or noisy due to low light. Is it camera, or shooter or setup difficulties? In this form, it slightly misses main reason of sample footage - to show the device in its best. to go a bit conceptual... i think from point of this forum, there are 2 features absolutely nesessary to make a good industrial camera head usable for digital filmmaking. First is direct recording of quality stream to harddrive and second is a way to monitor the picture on a device attached to the camera. As far as i understand, the present camera is a kind of sbc with fpga, so ideally it should be sbc with vga-out and ide/sata, where fpga doing debayer and resize function for preview. But it is different concept (from network camera) and camera board redisign. Not good. And even that ideal system, would require 2 person team to operate. 1 to control the camera settings via laptop and 1 to actually point and shoot. To eliminate this problem there must be all in one place: lense, head, recorder and screen. Why not attach a compact capable laptop to the head then, like a Tablet PC? And a battery bank. But in general, how could you summarize the reasons to use such a camera in real life shooting. Except very special areas, like permanent multicam video system in small music club, where it is probably excelent option to take already, and a toy for filmmaker wannabe who loves cameras more than above mentioned fimmaking? Iam not offending anybody here, of course, except myself :).

April 18th, 2006, 04:48 AM	#67
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Konstantin, I have covered most of these points a few times. even with Andrey by email in times past. And yes I have noted what you said, but I was being kind not to point it out, because it does not prove how much it can do (and sensor and compression has a lot to do with it). What I have been discussing here is ways to maximise performance to get it upto quality output, and defining what the limits are. Including a practical bayer scheme. The problem is that it is a security camera (not industrial)so it is made for good security quality, and limited. A little extra quality and we can get to at least pro video quality. The Kodak sensor might deliver a lot better performance, but I note that Micron has extended latitude features now (there MT9V032 wide vga security chip pdf shows 32K lux and 2 lux test charts in the same shot. So there is hope for this to turn up on the HD chips. Computer: If you want a small platform, then car-PCs (about the size of a DVD drive) Origami and UMD, small cheap Tablet PC's, are worth looking at. There might be one out there with higher res screen and microphone input/s. Have fun with it, we need your sort of thinking around here. Thanks Wayne.

April 18th, 2006, 11:57 AM	#68
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	Yes, you are correct that none of the test footage was performed under ideal condintions. and also, i recently found the zoom on the lens i am using has to be turned just a nudge away from 13.5mm to actually be at true infinity focus, where everything is in focus. i think the reason being is that, i can go past 13.5mm to macro function, and i was too close to macro. and note my lens is at f1.8 with it open. Can i ask if you have seen all the footage posted on the link? there is one shot called guitar chair. and i wouldnt call that blown out. i had some focus issues perhaps on it, but nothing much else. and of course we can say there are improvements to be made, but that can always be said. What Andrey has here is a high definition network camera that he probally didnt plan to be used for this purpose. And he has made a great camera capable of beautful picture. and i applaud him for that. He has been very generous and helpful in making this project go foward. I will have to try a little harder to get shots out that can please you a little better. And its not a bad thing, it keeps me on my toes, which is good. I am going to shoot some stuff right now infact. ill see what i can do. thanks

April 18th, 2006, 01:30 PM	#69
Oscar Spierenburg Major Player Join Date: Jan 2005 Location: (The Netherlands - Belgium) Posts: 735	Yes... but please don't make this forum into a commercial discussion thing. We are discussing the first attempts of using a security camera for film making. Forrest is sharing his experiments here, he's not selling a camera or something. Please share every footage, good and bad. I haven't received my camera yet, but they say it should arrive in a few days.

April 18th, 2006, 03:01 PM	#71
Konst Seraf New Boot Join Date: Apr 2005 Posts: 12	well, i should apologize for being extra critical, the "chair" file is correctly exposed, and overall visually pleasent. But, there was one problem that become apparent in last the file, with girl. It is luck of crispness. May I consider that the lense you are using is not even an industrial purpose megapixel lense? You wrote in initial post that it was some combination of 35mm model with wideangle adaptor. It may be not adequate. I had no intention to start a commercial discussion. But i think even being a thing in itself, the project like this still requires some perspective view. What is it good for and what is the goal? The only reason to use such a modular system in real shooting process is execptional video quality. Lets say noticebly better than present prosumer hdv. An interesting task.

April 18th, 2006, 03:27 PM	#72
Eric Gorski Regular Crew Join Date: Jun 2004 Location: Western Oregon Posts: 138	that last clip is very nice. the footage you're shooting reminds me alot of clips that a board member by the name of OBIN was shooting a year ago with a similar camera, possibly the same chip?

April 18th, 2006, 03:37 PM	#73
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	Konstantin, you do not need to apologize at all. Your comments were very helpful in the for the fact that i must continue to better the progress and quality. The lens i am using now is an f1.8 zoom lens that is c-mount. It is a correct type of lens, but prehaps its glass isnt good enough for the the cameras resoultion. That might be the case. I just finished ordering a 25mm c-mount prime lens. with f1.4. With this lens, i am going to build the 35mm adapter onto the camera with this lens to serve as the relay lens. I will post footage with that f1.4 to see how it handles the resolution also. thank you. And thank you Eric. I can't remember what chip Obin used, but the process that were going is quite similar. The chip in this model camera is a 3MPix Micron MT9T001 1/2" CMOS.

April 18th, 2006, 08:41 PM	#75
Forrest Schultz Major Player Join Date: Mar 2005 Location: Atwater, CA Posts: 246	Oh wow, i didnt check that one. let me see if that has happened to some more recent footage i did. Thanks Andrey EDIT: well it seems that things got a bit weird when i convereted to .wmv The original footage coverted to avi from ogm via VirtualDub shows full 24 fps. 24 frames for every 1 second of footage. for some reason, the wmv version doesnt want to play right, also the wmv in windows media player states 29 sec duration. perhaps when it says 32 sec in mplayer it is not playing it right? i dont know.