Wearable Computer System for HD Capture

[Adriano Apefos]

Originally Posted by Kevin Kondra

Hi Jose,

1. The USB interface has a maximum throughput of 60MB/s,

I dont think usb can handle 60MBps. The maximun I can get with external HDD or DVD burner is just 6 (six) MBps. The 480Mbps that gives the 60MBps information is just theoretician in my opinion.

[David Newman]

Over USB2 or FireWire 400 you can sustain between 34-38MBytes/s if the drive is is fast enough. We regularly capture 20-25MB/s data directly to FireWire or USB2 drives from HDMI/HDSI sources. The fastest 2.5" drive will sustain up to 30MB/s, most will do 20MB+/s across the entire surface -- the Silicon Imaging SI-2K camera uses a 2.5" drive for all its video captured in CineForm RAW, that can be up to 15MB/s.

[Adriano Apefos]

I just would like to see test results from this equipment about capturing MJPG, NEO and Sheer using both usb and sata conections in the 2.5 drive to be sure about drop frames. A real test is the best answer. Thanks.

[Richard Leadbetter]

Er, David has done real tests. He's the CTO of CineForm.

His bandwidth measurements are accurate which means that basically only BM and CineForm will work. Sheervideo's stated compression is only 2:1 - way beyond the means of either a USB/Firewire or even 2.5" SATA drive.

If you read the thread earlier you will see that Kevin's reported frame loss using the MJPEG Blackmagic codec was due to inefficient use of the dual core CPU. MJPEG comes in at around 18MB/s, by the way.

Personally I do not see much point pursuing pristine HDMI quality capture if you're going to degrade that quality with the MJPEG codec. In my tests, even CineForm's lowest quality setting significantly outperformed MJPEG.

[Adriano Apefos]

Ok, so the system needs a more powerfull cpu!

[Adriano Apefos]

I saw the png files and the neo is very better than the other

[Kevin Kondra]

Wow, that's a lot of activity here. TO answer some questions:

1. I did all of my tests with the HD (Seagate Momentus 7200.2) connected with the USB cable, not the eSATA (which I could have). The HD will sustain a write speed (for the whole disk surface) of just over 30 MB/s with USB, so it is not the bottleneck for capturing video.

2. The BM codec skipped frames because it was cpu bound, not disk bound, as has been stated. My estimate is 50% more cpu power is need to avoid skipping since 32% of frames were dropped. At this time Intel doesn't make something that fast. The data rate for the MJPEG codec was around 15MB/s. Most people forget that JPEG has variable compression ratios, but the BM codec has a fixed ratio. Codecs that can set the ratio to less compression may provide better quality and require less cpu power. Any suggestions?

3. The BM intensity card is attached with a flexible PCIE extension cable. They are available for 1x,4x,8x, and 16x slots.

It seems everyone much prefers the quality of Cineform compression over MJPEG, so I'll stick to using it. In fact, I'll do some more comparisons and try using a higher quality setting if the cpu can handle it.

[Kevin Kondra]

I could also set up a striped disk set (might have to use 3.5" disks) to see if the system can handle using HuffYUV to record a lossless video stream in the backpack. Is this worthwhile or interesting to anyone?

[Kevin Kondra]

Originally Posted by Adriano Apefos

Hi, Kevin,

Congratulations. Your system is perfect in my opinion. I would like to know how did you fit the intensity card in horizontal position. Is there an extension PCIe cable? And about the 20MBps BM codec, does it work in the windows xp? I think this datarate can give artifacts results as low as in minidv? Did you try it?

Thanks,

Adriano.

I'm note sure what you mean with artifacts as low as in minidv.

[Adriano Apefos]

The minidv standard definition has a compression 5:1 this compression generates some mosquito noise but it is intraframe, so it does not have the blocking, the temporal and the posterization artifacts found in MPEG2 HDV. I was thinking about if it was possible to capture using the BM MJPG at 20MBps , so it would be about 6:1, near the minidv ratio and maybe it could give a good quality.

Another point: did you try to turn off the video and sound preview when capturing? it helps avoid drop frames.

And about the onboard graphics card? maybe it is another reason to low performance.

A good question to answer is: wich CPU can handle the BM codec without drop frames in a full size ATX motherboard????? If a full size motherboard can handle the capture with a 2.0ghz core 2 duo, so the reason for drop frames is not the cpu... off board motherboards with audio and video cards are allways better for video capture and editing.

thanks.

[Adriano Apefos]

I would apreciate you to do a test using the sheervideo codec for capture. I read about it in the website and it has a data rate about 40 MBPS, about 3:1 compression and it would be a good test to a two disk raid in sata. they promiss a fast performance, and a lossless quality, so your system would be nice to prove if it is usefull.

thanks

[David Newman]

While Sheer can produce a 3:1 compression, the source image would need to be very low noise otherwise we will get a much higher data rates, more typical is 2:1 but sometimes you get even less compression. The is an issue with lossless compression, some source images are just no very compressable.

Regarding MJPEG compression, all DCT compression can experience blocking, the is no only a inter frame compression artifact.

[Anmol Mishra]

1. Has anyone found a mini-ITX board with SATA II AND RAID0 ?
2. Can I add GigE and Firewire 800 :-) to requirements ?
3. PCI-X riser can be used..Or PCI-X extension cable.
4. A comparison of mini-ITX boards
http://www.mp3car.com/vbulletin/gene...tx-boards.html
5.

[Richard Leadbetter]

When you say PCI-X, do you actually mean PCI Express? PCI-X is something completely different.

[Anmol Mishra]

PCI-Express not PCI X