4:4:4 10bit single CMOS HD project

Ben Syverson · July 24th, 2004, 11:46 PM

Jason, the color is best pulled out by manipulating the RGB as YCbCr (YUV). That's what my new software chromaPop does -- gives you chroma gains.

You go from this (straight off the camera, no CC)

... to this (same image with levels applied, and then chromaPop)

I had been wondering how to get rich reds and bright colors... I guess it was all hiding in the chroma! :)

chromaPop also lets you apply the change in only the R&B channels -- sometimes you want to keep the G channel pristine, since it's doing the heavy lifting detail-wise.

- ben

Jason Rodriguez · July 25th, 2004, 12:01 AM

I have a vectorscope here though, and that swings image you sent basically had no color info except for cyan. I'm not sure if that was intentional or it was really dark or what, but playing around with the YUV channels isn't going to help that image. Your rotten tomates shot has much more color info, plenty to play with.

I think we need to shoot a MacBeth chart or something to actually get this stuff white-balanced.

Ben Syverson · July 25th, 2004, 02:48 AM

Here's the dollar bill shot. Although it's not exactly a completely fresh bill. There are a bunch of artifacts as a result of the moiré happening and my de-Bayering software not knowing what to do with it.

But you can see that "Secretary of the Treasury" is readily legible, as are some of the paper fibers. The camera is almost able to capture the insane grid behind Washington...

I'd love to see the same test shot on the HD10, the SI-1300 and maybe even the DVX-100...

- ben

Wayne Morellini · July 25th, 2004, 04:55 AM

Like the doller bill, maybe we should compare all the cameras with test charts, and window and 100W bulb streaking, lines and blowouts.

Good to see the images getting better. Could you try the head and shoulders shot again with the same lighting, to see how much the image has improved? How is low light sensitivity now? I am most worried about colour range and sensitivity, it seems too low to play with.

There has been some negatrive talk about the Fillfactory chip and USB in times past, so could I ask some questions to clear things up, your reply would be greatly appreciated.

Have you got the latest version of the chip?

How fast can you run the shutter and not drop below 24FPS in 8 bit and in 10 bit modes.

Can the software gaurantee a frame rate (i.e 24fps, 25fps)? Does sumix know of any front end pro video capture/control software, I have found the cameralink website over at the technical thread but haven't done a extensive search yet.

Obviously there is no pixel packing mode to better use the USB bandwith? Do they have frame buffer memory on the camera (to even out the USB2 bandwidth requirements)?

Real Range and light sensitvity estimates?

Sumix is supposed to be making a range of cameras to suit us, even with compression in the camera head. Have they mentioned anything to you about where ever there will be Gigabit E version, or Gigbit ethernet and USB2 versions with compression?

Thanks for giving me some hope for USB2.

Just done a cinema fov test on the doller of bill shot, and the bayer errors on the still looks acceptable for veiwing.

<<<-- Originally posted by Les Dit : Ben,
Resolution wise, none of the CMOS 1.3 cameras looks anywhere near that sharp. not even close.
It's obvious JVC has a kick ass demosaiker.
-Les -->>>

Thats a worry, the HD10 only has around 850K pixel actual resolution.

Thanks again

Wayne.

Wayne Morellini · July 25th, 2004, 05:32 AM

Obin

<<<-- Originally posted by Obin Olson : .if you have a hotspot that is close to 100% it will have the smear across the whole image -->>>

The smearing problem, could that be from using a faster lense supplying a greater range and more energy to the pixels.

Steve wrote some stuff about adjusting the camera settings to reduce problems a little while ago.

<<<-- Originally posted by Steve Nordhauser : Jason on smearing:
That smearing with oversaturated pixels is a Micron problem (all Micron 1.3Mpix cameras will have it). Definitely not the lenses. I bugged them about it and they said run slower, which negates what we want to do about rolling shutter. I was told (but haven't seen) that the 3.2Mpix Micron doesn't do that as much - different pixel architecture. The 3.2Mpix should be able to do both 720p and 1920x1080@24fps, 10 bit. If Obin and Rob want to upgrade in about a month I'll work it out with them. The SI-3300 will have less sensitivity though.
-->>>

So the next version is an improvement.

For people asking where Steve is:

<<<-- Originally posted by Steve Nordhauser : Hey gang, I'm going to be gone starting this evening through Sunday. Doing a camping folk festival - not even taking a computer, cell phone or video camera. Hopefully the 1920HD will be at the FG company by then, you guys will conclude that a system can be built the size of a cell phone and someone finds all the software under their pillow. That's my fantasy and I'm sticking to it until a better one comes along. -->>>
On the 21st, that was a lot of pages ago ;) So he should be back soon.

Steve, I have an idea for this.

Wayne Morellini · July 25th, 2004, 06:10 AM

<<<-- Originally posted by Juan M. M. Fiebelkorn : You don't understand me.I mean CONCENTRATE (put a lens between the SLR and the sensor), not WASTE.

Anyway you don't like what I'm saying :(

Please if someone here could explain him what I'm trying to say, it would be nice :) -->>>

My version doesn't have GG.

Please don't get offended Ben, if I repeat something you allready know, I don't know what you know, and it is usefull to repeat things for clarity and for the other readers.

Quote:

Originally posted by Wayne Morellini :somebody mentioned it on a XL1s 35mm adapter thread, and I also asked about it in a thread last year. By making an 35mm adpator with suitable codensor, instead of projection surface, you can pipe all the light down to the chip (no loss from the projection surface). You don't get SLR DOF, but you get the angle of veiw and extra bright image.

http://www.dvinfo.net/conf/showthrea...hreadid=22715&

Quote:

Originally posted by Jonathon Wilson :
I actually built one exactly as you described because I was frustrated at the amount of light loss with the ground glass. I replaced my ground glass with a 'relay lens' which changes the magnification at the focal point down to the size of my ccd. It worked beautifully -- incredibly bright image... but very long Video-looking depth of field. I basically had the focal length/field of view of my SLR Lens, but depth of field of video.

We face the same problems as detachable XL1s lense people face:
http://www.dvinfo.net/conf/showthrea...down+the+image

Have agood week. I eagerly wait for your next camera releases Steve.

Juan M. M. Fiebelkorn · July 25th, 2004, 10:56 AM

Ben,
If your lens is 30 years old or up, sorry to hear that.
Although it would be a really good lens, it would give some color distortions and artifacts under certain situation.This is just because it lacks a modern coating treatment.If it is 20 years old I´m sure it is fine.
But no problem it is what you got. :)

The tomatos shot is nice but has unbalanced color.Look at the bright zone left on the tomato.It is yellow/green that is why you don´t get a nice red.

Now that you work in YUV colorspace, there is a simple method to apply a lowpass filter on colors.
Take U and V planes, downsize to 640x360 and upsize again to 1280x720 using bicubic or Lanczos. (Bicubic sharpens or not depending on internal coefficients, Lanczos always sharpens)

Test it.I can´t send you an image now.

Jason Rodriguez · July 25th, 2004, 12:55 PM

I'm curious to know why these cameras, if they're operating at a certain frequency, can't keep a consistent frame-rate.

Ben, you were saying that your frame rates were all over the place. I was doing some math, and if you're varying by more than just a little bit, like .2 or .3 fps, you're going to be out of sync very fast-half a frame to a whole frame every second. I was planning on getting 2GB of RAM for 90 seconds of recording time, but at that rate, I'll be way out of sync by the end of 90 seconds, anywhere from 2-4 seconds from the frame rate flucuations you described!!

So if everything here is computer controlled, how come the fluctuating frame-rates? Is there a way to get a timing circuit on the camera that can lock the frame-sync? Anything else seems to spell big trouble for my projects which depend on sound-sync.

BTW, pitch-shifting audio when you're trying to edit mixed takes is an absolute nightmare, and any other editor except for myself on my own projects would throw the material back in my face! :-)

Eric Gorski · July 25th, 2004, 01:28 PM

hey ben,
how do quick pans look with the camera? could we maybe see a clip?

Matthew Miller · July 25th, 2004, 01:42 PM

Ben,
Have you tried to see what kind of framerate consistency you can get at lower resolutions? For example, have you tried 720x480 @24fps to see if the framerate still ends up being slightly different everytime you capture.
How many different computers have you tried capturing to? Does the camera get its power from the USB2.0 port?
Ultimately, I don't think the slightly erratic framerates are that big of a problem unless you plan on using really long takes (90 seconds) without any audio work. As an editor, I wouldn't shy away from mixed takes with such incrementally small differences in framerates. Even Adobe Audition has robust enough "time-stretching" filters to take care of the job these days.
I wouldn't suggest the camera rig for a long form documentary... but for cinema... I'm sure I don't need to say much more than ADR and Foley.

Jason Rodriguez · July 25th, 2004, 01:53 PM

I'm wondering if sync is coming from the USB port, or if it's coming from the Camera itself. If so, can you plug in a timing circuit to the external sync connector on the camera and run the camera in "snapshot" mode? That should get good sync with a crystal timing circuit.

Ben Syverson · July 25th, 2004, 03:08 PM

You guys are freaking out about the sync issue. It's a 90 second take, tops. Since it's double system, you'll have to line the clapper up with the audio anyway -- it's not too much more trouble to stretch it to fit a little. Like I mentioned before, it only varies by a small amount each time. Juan, your math is suspect. If it varies .2fps, how could you go out of sync by a whole frame in 1 second? The difference is .2frames.

The camera is driven over USB2 -- that's the only cable. It gets its power from the USB2 bus (although it needs very very little power -- it can actually run over USB1, even if the USB1 can't keep up with the data). The sync probably varies because the camera depends on the computer to tell it to take another frame. The computer doesn't always have the same taskload, so it will vary.

Wayne, interesting stuff on the 35mm adapter. Here's my question to Juan: how is using this setup better than C-mount lenses? This setup requires another piece of glass -- a "relay" lens -- of questionable optical merit. Any additional piece of glass between the film plane and the subject is "bad." So what makes you think that this would be better? Do you just loooooove SLR lenses? They're not carved out of god, man.

Like I said, I'm not sure how old the Angenieux is, but it's clean, smooth, and multicoated.

Eric, quick pans do indeed show some rolling shutter-ness. It doesn't bother me, but I'm not as picky as some people on this board...

@Juan: "The tomatos shot is nice but has unbalanced color.Look at the bright zone left on the tomato.It is yellow/green that is why you don´t get a nice red."

Case in point. This is not the Alatasens. It's an 8 bit camera. Keep that in mind.

- ben

Juan M. M. Fiebelkorn · July 25th, 2004, 03:45 PM

Ben,
Well it seems that no matter what I tell you, everything I say to you sounds wrong.
Anyway do as you like.This is not a competition.I'm telling you some things I've found through my years of experience working at feature films and at the post-production industry.
Within my 5 years of work to develop a Film Recorder system, I've researched many things and discover that some usual things are not as believed (at least not always).
My bussiness partner has around 25 years of experience as DP.He has made more than 5,000 commercials from here and the rest of the world, and agrees with me in what I'm saying.
If you say your Angenieux lens 30 years old is better than, for example, a 10 years old Pentax, its OK to me but I ought to tell you no.
I say SLR lenses cause they are cheap, nothing else, if you want to buy a $10,000 PL lens good for you.
Do you know what's the difference between a normal Karl Zeiss and the Ultra Prime lenses?

What do you think about the low pass method I posted before?

I don't understand what's the relation between a bad white balance and an Altasens chip.

Where did you get that a lens between a lens and the sensor is "Bad"?

Do you know how many lenses a Good angular lens have?
Do you know how much light do you loose because of that extra lens?

Wayne Morellini · July 25th, 2004, 03:53 PM

As I have started discussing in the technical thread, Steve was talking about USB requiring too much processing power and being unreliable

(thanks Intel, why didn't you just adopt a version of Firewire, ohhh what "Apple/TI product", rather than "Intel Microsoft product" how pathertic, thanks for making our lives a nightmare).

Have a look at cpu usage, it might be possible that USB itself has timing problems, or it is maxing out cpu time. You could work around it, but it will need specialist USB drivers (and system setup) like Steve does with GigbitE (ask Sumix). I have posted some links over at the technical thread.

Anhar Miah · July 25th, 2004, 03:54 PM

@ Ben

why not make they priority set to the highest and shutdown all other programs/ internet connections etc..

and have a "clean" defragemented harddrive?

(I still think the solid state hard drives are the way to go)

Also @ Ben once you have your captured file (AVI?), is it ready to play? i'm a bit confused with all these bayer stuff that i've read earlier,

i'm thinking of getting this cmos camera:

http://www.compumodules.com/image-pr...S-camera.shtml

FINAL QUESTION:

If i was too buy the above camera (CMOS ULTIMA II, COLOUR 1.3 MEGAPIXEL USB2)

And lens + USB cable, what software will i need to start capturing video
P.S i have a 2.6GHz Toshiba laptop (i hope its fast enough)

July 25th, 2004, 01:42 PM	#1060
Matthew Miller New Boot Join Date: Jun 2004 Location: Pittsburgh, PA Posts: 20	Ben, Have you tried to see what kind of framerate consistency you can get at lower resolutions? For example, have you tried 720x480 @24fps to see if the framerate still ends up being slightly different everytime you capture. How many different computers have you tried capturing to? Does the camera get its power from the USB2.0 port? Ultimately, I don't think the slightly erratic framerates are that big of a problem unless you plan on using really long takes (90 seconds) without any audio work. As an editor, I wouldn't shy away from mixed takes with such incrementally small differences in framerates. Even Adobe Audition has robust enough "time-stretching" filters to take care of the job these days. I wouldn't suggest the camera rig for a long form documentary... but for cinema... I'm sure I don't need to say much more than ADR and Foley. __________________ This is where my optional signature could have been.

July 25th, 2004, 03:54 PM	#1065
Anhar Miah Major Player Join Date: Dec 2003 Location: United Kindom, England Posts: 290	silly, thought @ Ben why not make they priority set to the highest and shutdown all other programs/ internet connections etc.. and have a "clean" defragemented harddrive? (I still think the solid state hard drives are the way to go) Also @ Ben once you have your captured file (AVI?), is it ready to play? i'm a bit confused with all these bayer stuff that i've read earlier, i'm thinking of getting this cmos camera: http://www.compumodules.com/image-pr...S-camera.shtml FINAL QUESTION: If i was too buy the above camera (CMOS ULTIMA II, COLOUR 1.3 MEGAPIXEL USB2) And lens + USB cable, what software will i need to start capturing video P.S i have a 2.6GHz Toshiba laptop (i hope its fast enough)

July 24th, 2004, 11:46 PM	#1051
Ben Syverson Major Player Join Date: Mar 2004 Location: chicago Posts: 434	Jason, the color is best pulled out by manipulating the RGB as YCbCr (YUV). That's what my new software chromaPop does -- gives you chroma gains. You go from this (straight off the camera, no CC) ... to this (same image with levels applied, and then chromaPop) I had been wondering how to get rich reds and bright colors... I guess it was all hiding in the chroma! :) chromaPop also lets you apply the change in only the R&B channels -- sometimes you want to keep the G channel pristine, since it's doing the heavy lifting detail-wise. - ben

July 25th, 2004, 12:01 AM	#1052
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	I have a vectorscope here though, and that swings image you sent basically had no color info except for cyan. I'm not sure if that was intentional or it was really dark or what, but playing around with the YUV channels isn't going to help that image. Your rotten tomates shot has much more color info, plenty to play with. I think we need to shoot a MacBeth chart or something to actually get this stuff white-balanced.

July 25th, 2004, 02:48 AM	#1053
Ben Syverson Major Player Join Date: Mar 2004 Location: chicago Posts: 434	Here's the dollar bill shot. Although it's not exactly a completely fresh bill. There are a bunch of artifacts as a result of the moiré happening and my de-Bayering software not knowing what to do with it. But you can see that "Secretary of the Treasury" is readily legible, as are some of the paper fibers. The camera is almost able to capture the insane grid behind Washington... I'd love to see the same test shot on the HD10, the SI-1300 and maybe even the DVX-100... - ben

July 25th, 2004, 04:55 AM	#1054
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Like the doller bill, maybe we should compare all the cameras with test charts, and window and 100W bulb streaking, lines and blowouts. Good to see the images getting better. Could you try the head and shoulders shot again with the same lighting, to see how much the image has improved? How is low light sensitivity now? I am most worried about colour range and sensitivity, it seems too low to play with. There has been some negatrive talk about the Fillfactory chip and USB in times past, so could I ask some questions to clear things up, your reply would be greatly appreciated. Have you got the latest version of the chip? How fast can you run the shutter and not drop below 24FPS in 8 bit and in 10 bit modes. Can the software gaurantee a frame rate (i.e 24fps, 25fps)? Does sumix know of any front end pro video capture/control software, I have found the cameralink website over at the technical thread but haven't done a extensive search yet. Obviously there is no pixel packing mode to better use the USB bandwith? Do they have frame buffer memory on the camera (to even out the USB2 bandwidth requirements)? Real Range and light sensitvity estimates? Sumix is supposed to be making a range of cameras to suit us, even with compression in the camera head. Have they mentioned anything to you about where ever there will be Gigabit E version, or Gigbit ethernet and USB2 versions with compression? Thanks for giving me some hope for USB2. Just done a cinema fov test on the doller of bill shot, and the bayer errors on the still looks acceptable for veiwing. <<<-- Originally posted by Les Dit : Ben, Resolution wise, none of the CMOS 1.3 cameras looks anywhere near that sharp. not even close. It's obvious JVC has a kick ass demosaiker. -Les -->>> Thats a worry, the HD10 only has around 850K pixel actual resolution. Thanks again Wayne.

July 25th, 2004, 05:32 AM	#1055
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Obin <<<-- Originally posted by Obin Olson : .if you have a hotspot that is close to 100% it will have the smear across the whole image -->>> The smearing problem, could that be from using a faster lense supplying a greater range and more energy to the pixels. Steve wrote some stuff about adjusting the camera settings to reduce problems a little while ago. <<<-- Originally posted by Steve Nordhauser : Jason on smearing: That smearing with oversaturated pixels is a Micron problem (all Micron 1.3Mpix cameras will have it). Definitely not the lenses. I bugged them about it and they said run slower, which negates what we want to do about rolling shutter. I was told (but haven't seen) that the 3.2Mpix Micron doesn't do that as much - different pixel architecture. The 3.2Mpix should be able to do both 720p and 1920x1080@24fps, 10 bit. If Obin and Rob want to upgrade in about a month I'll work it out with them. The SI-3300 will have less sensitivity though. -->>> So the next version is an improvement. For people asking where Steve is: <<<-- Originally posted by Steve Nordhauser : Hey gang, I'm going to be gone starting this evening through Sunday. Doing a camping folk festival - not even taking a computer, cell phone or video camera. Hopefully the 1920HD will be at the FG company by then, you guys will conclude that a system can be built the size of a cell phone and someone finds all the software under their pillow. That's my fantasy and I'm sticking to it until a better one comes along. -->>> On the 21st, that was a lot of pages ago ;) So he should be back soon. Steve, I have an idea for this.

July 25th, 2004, 10:56 AM	#1057
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	Ben, If your lens is 30 years old or up, sorry to hear that. Although it would be a really good lens, it would give some color distortions and artifacts under certain situation.This is just because it lacks a modern coating treatment.If it is 20 years old I´m sure it is fine. But no problem it is what you got. :) The tomatos shot is nice but has unbalanced color.Look at the bright zone left on the tomato.It is yellow/green that is why you don´t get a nice red. Now that you work in YUV colorspace, there is a simple method to apply a lowpass filter on colors. Take U and V planes, downsize to 640x360 and upsize again to 1280x720 using bicubic or Lanczos. (Bicubic sharpens or not depending on internal coefficients, Lanczos always sharpens) Test it.I can´t send you an image now.

July 25th, 2004, 12:55 PM	#1058
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	I'm curious to know why these cameras, if they're operating at a certain frequency, can't keep a consistent frame-rate. Ben, you were saying that your frame rates were all over the place. I was doing some math, and if you're varying by more than just a little bit, like .2 or .3 fps, you're going to be out of sync very fast-half a frame to a whole frame every second. I was planning on getting 2GB of RAM for 90 seconds of recording time, but at that rate, I'll be way out of sync by the end of 90 seconds, anywhere from 2-4 seconds from the frame rate flucuations you described!! So if everything here is computer controlled, how come the fluctuating frame-rates? Is there a way to get a timing circuit on the camera that can lock the frame-sync? Anything else seems to spell big trouble for my projects which depend on sound-sync. BTW, pitch-shifting audio when you're trying to edit mixed takes is an absolute nightmare, and any other editor except for myself on my own projects would throw the material back in my face! :-)

July 25th, 2004, 01:28 PM	#1059
Eric Gorski Regular Crew Join Date: Jun 2004 Location: Western Oregon Posts: 138	hey ben, how do quick pans look with the camera? could we maybe see a clip?

July 25th, 2004, 01:53 PM	#1061
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	I'm wondering if sync is coming from the USB port, or if it's coming from the Camera itself. If so, can you plug in a timing circuit to the external sync connector on the camera and run the camera in "snapshot" mode? That should get good sync with a crystal timing circuit.

July 25th, 2004, 03:08 PM	#1062
Ben Syverson Major Player Join Date: Mar 2004 Location: chicago Posts: 434	You guys are freaking out about the sync issue. It's a 90 second take, tops. Since it's double system, you'll have to line the clapper up with the audio anyway -- it's not too much more trouble to stretch it to fit a little. Like I mentioned before, it only varies by a small amount each time. Juan, your math is suspect. If it varies .2fps, how could you go out of sync by a whole frame in 1 second? The difference is .2frames. The camera is driven over USB2 -- that's the only cable. It gets its power from the USB2 bus (although it needs very very little power -- it can actually run over USB1, even if the USB1 can't keep up with the data). The sync probably varies because the camera depends on the computer to tell it to take another frame. The computer doesn't always have the same taskload, so it will vary. Wayne, interesting stuff on the 35mm adapter. Here's my question to Juan: how is using this setup better than C-mount lenses? This setup requires another piece of glass -- a "relay" lens -- of questionable optical merit. Any additional piece of glass between the film plane and the subject is "bad." So what makes you think that this would be better? Do you just loooooove SLR lenses? They're not carved out of god, man. Like I said, I'm not sure how old the Angenieux is, but it's clean, smooth, and multicoated. Eric, quick pans do indeed show some rolling shutter-ness. It doesn't bother me, but I'm not as picky as some people on this board... @Juan: "The tomatos shot is nice but has unbalanced color.Look at the bright zone left on the tomato.It is yellow/green that is why you don´t get a nice red." Case in point. This is not the Alatasens. It's an 8 bit camera. Keep that in mind. - ben

July 25th, 2004, 03:45 PM	#1063
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	Ben, Well it seems that no matter what I tell you, everything I say to you sounds wrong. Anyway do as you like.This is not a competition.I'm telling you some things I've found through my years of experience working at feature films and at the post-production industry. Within my 5 years of work to develop a Film Recorder system, I've researched many things and discover that some usual things are not as believed (at least not always). My bussiness partner has around 25 years of experience as DP.He has made more than 5,000 commercials from here and the rest of the world, and agrees with me in what I'm saying. If you say your Angenieux lens 30 years old is better than, for example, a 10 years old Pentax, its OK to me but I ought to tell you no. I say SLR lenses cause they are cheap, nothing else, if you want to buy a $10,000 PL lens good for you. Do you know what's the difference between a normal Karl Zeiss and the Ultra Prime lenses? What do you think about the low pass method I posted before? I don't understand what's the relation between a bad white balance and an Altasens chip. Where did you get that a lens between a lens and the sensor is "Bad"? Do you know how many lenses a Good angular lens have? Do you know how much light do you loose because of that extra lens?

July 25th, 2004, 03:53 PM	#1064
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	As I have started discussing in the technical thread, Steve was talking about USB requiring too much processing power and being unreliable (thanks Intel, why didn't you just adopt a version of Firewire, ohhh what "Apple/TI product", rather than "Intel Microsoft product" how pathertic, thanks for making our lives a nightmare). Have a look at cpu usage, it might be possible that USB itself has timing problems, or it is maxing out cpu time. You could work around it, but it will need specialist USB drivers (and system setup) like Steve does with GigbitE (ask Sumix). I have posted some links over at the technical thread.