I ran some more tests at 1600x1200 on the Voodoo 3 --> OSSC--> E1S using a Startech 1 foot VGA cable with coaxial cable inside and a ferrite bead : https://www.amazon.ca/gp/product/B003MQIEW0/ .

There is definitely less ringing, but I subjectively see a bit more noise (maybe the OSSC is picking more noise from the PC due to its proximity to it). This noise will not really show up on screen shots, though .

Short of hanging the OSSC directly off the Voodoo 3 using a VGA male to male adapter, I don't see how I could get a shorter VGA cable to work. The question still remains as to whether all my VGA cables are somehow crappy and causing an impedance mismatch, or if the Voodoo 3 itself is at fault and using a shorter cable just reduces the scope of the issue (signal reflection).

EDIT: I just realized that this thread has been going on for almost 10 years ! It looks to be among the older still active threads here .

I use Epiphan cards and made some years ago a tool for such optimizations. It is much easier to tune the stuff when you have the numbers at hand and don't have to judge some noise in screenshots.
It was posted in this thread here: Re: VGA Capture Thread

Beside tuning the phase it also gives a standard deviation for pixel noise, so one could exchange e.g. cables and see if the numbers drop.

darry wrote on 2020-10-21, 07:01:

the Voodoo 3 itself is at fault and using a shorter cable just reduces the scope of the issue (signal reflection).

This is my operating theory

elianda wrote on 2020-10-21, 09:34:

I use Epiphan cards and made some years ago a tool for such optimizations. It is much easier to tune the stuff when you have the numbers at hand and don't have to judge some noise in screenshots.
It was posted in this thread here: Re: VGA Capture Thread

Beside tuning the phase it also gives a standard deviation for pixel noise, so one could exchange e.g. cables and see if the numbers drop.

That tool looks great for evaluating noise and making critical adjustments quantifiably, rather than eyeballing it . Unfortunately, I do not (yet) have Epiphan hardware that I could use (a bit ironic since Epiphan Video is at least partly based in Canada).

Adjusting the OSSC's phase using a real-time 4x magnified view through a DVI capture card (E1S) does seem to work rather well for me .

If am not mistaken, the OSSC Pro's ADC has the ability to automatically tune the phase, which will make life easier for its users (assuming OSSC Pro implements that functionality) .

My main issue, for now, is the ringing/ghosting I get with the Voodoo 3, probably due to an impedance mismatch, which I am starting to suspect is caused by the card itself and not a cabling issue . When my final cable orders arrive, if the issue is still present with those said cables, I do not know what I will do .

maxtherabbit wrote on 2020-10-21, 12:41:

darry wrote on 2020-10-21, 07:01:

the Voodoo 3 itself is at fault and using a shorter cable just reduces the scope of the issue (signal reflection).

This is my operating theory

Assuming that cables themselves are not the issue (I have some more on order, including the C2G one you suggested), there are a 2 scenarios :

a) the Voodoo 3 itself, though using another card was only marginally better. Admittedly, both cards could be "bad" (a sampling of 2 is not conclusive, 🤣)
b) the OSSC's VGA connector and/or path to ADC could be at issue . I have tried 2 OSSC units (official and third party) with the same results and connecting directly to the E1S via VGA looked worse (admittedly could be due to a marginal DVI-I to HD (VGA) adapter, so I should probably re-test with another one). I doubt the ADC circuits/connectors on both OSSCs and the E1S are "bad" in exactly the same way, but it's not impossible

Unfortinately, I have neither the knowledge, the skill nor the equipment to test impedance of a capture card's input or a VGA card's output .

My limited understanding of how this works (please correct me if wrong and educate me) leads me to believe that, for a signal reflection to occur and to be affected by cable length, the impedance of both the OSSC's input (destination) and the Voodoo 3's VGA output(source) would need to be higher than that of the VGA cable used (in order for the signal to partly bounce first on the destination, then partly back at the source and to finally arrive once again at the destination in an attenuated and delayed form).

In other words, for the issue to occur, either both the OSSC and Voodoo 3 would need to have improper (too high) impedance OR the cables (or their connectors) would need to have too low an impedance .

An option I have been considering is finding a way to feed an SOG signal into the OSSC over its RCA connectors using an adapter like this one https://www.raphnet.net/electronique/sync-on- … on-green_en.php to see if it works better .

EDIT: Or maybe getting one of these : https://www.kramerav.com/ca/api/ProductPdfApi/VP-100 A bit expensive just to try, though .

Using the RCA input of the OSSC for 162MHz pixel clock is a bad move. The VGA input goes directly to the TVP7002 which can digitize up to 165MHz. However the other two inputs are buffered by a THS series buffer first which only has enough bandwidth for ~150MHz pixel clocks.

I seriously doubt the issue is the OSSC anyway. It is a well designed device and sink termination isn't exactly rocket surgery

maxtherabbit wrote on 2020-10-21, 16:16:

Using the RCA input of the OSSC for 162MHz pixel clock is a bad move. The VGA input goes directly to the TVP7002 which can digitize up to 165MHz. However the other two inputs are buffered by a THS series buffer first which only has enough bandwidth for ~150MHz pixel clocks.

Thanks for the info, I will scratch that idea .

Any suggestions about other things to try, even expensive/exotic ones ?

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced.
Then you could check for the ghosting/reflection and eliminate or deduce that it is the capture end that's at fault.

A Radeon 9500/9700/9800/X800/X850 has some of best RAMDACs in the business so they could also help determine if the problem lies with the Voodoo3. Failing that a Matrox G400?
I use IXOS VGA cables that seem to be going cheap atm but can recommend Gefen/Extron/Kramer ones too and anything that uses Canare coaxial cable.

Hoping you have more success anyway!

Hanamichi wrote on 2020-10-21, 16:36:

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced. Then you could […]
Show full quote

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced.
Then you could check for the ghosting/reflection and eliminate or deduce that it is the capture end that's at fault.

A Radeon 9500/9700/9800/X800/X850 has some of best RAMDACs in the business so they could also help determine if the problem lies with the Voodoo3. Failing that a Matrox G400?
I use IXOS VGA cables that seem to be going cheap atm but can recommend Gefen/Extron/Kramer ones too and anything that uses Canare coaxial cable.

Hoping you have more success anyway!

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully still chugging along at one of my mother's neighbours . 1600x1200 was not all that sharp on that, AFAICR (cable I was using at the time was a quality HD15 to RGBHV BNC one) but had no visible ghosting.

Thank you for the suggestions . I do have a Radeon 9700, an X850 (somewhere) and a Matrox G550. I will probably gives those a shot if the cable avenue is not conclusive .

As for the cables themselves, I am looking at finding (relatively) affordable quality ones from Gefen/Extron/Kramer/BlueJeans and I have heard good things about Canare products, quality wise .

EDIT: IXOS cables seem to be hard to come by in my neck of the woods .

I've found that the Extron RGB-HDMI 300 A coupled with a HDMI scaler (I went with a TESmart HDMI Splitter) will do the job no problem at all and allow for seamless capture of both DOS and Win9X content.

Gentlepoke wrote on 2020-10-21, 19:49:

I've found that the Extron RGB-HDMI 300 A coupled with a HDMI scaler (I went with a TESmart HDMI Splitter) will do the job no problem at all and allow for seamless capture of both DOS and Win9X content.

I have an Extron RGB DVI 300 (basically the same as the HDMI one except for the connector). The reason that I am not using it is because it does not pass through 70Hz DOS modes, it converts them to 60Hz . Additionally, it's scaling is not as sharp as an OSSC's line doubling .

darry wrote on 2020-10-21, 17:50:

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully s […]
Show full quote

Hanamichi wrote on 2020-10-21, 16:36:

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced. Then you could […]
Show full quote

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced.
Then you could check for the ghosting/reflection and eliminate or deduce that it is the capture end that's at fault.

A Radeon 9500/9700/9800/X800/X850 has some of best RAMDACs in the business so they could also help determine if the problem lies with the Voodoo3. Failing that a Matrox G400?
I use IXOS VGA cables that seem to be going cheap atm but can recommend Gefen/Extron/Kramer ones too and anything that uses Canare coaxial cable.

Hoping you have more success anyway!

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully still chugging along at one of my mother's neighbours . 1600x1200 was not all that sharp on that, AFAICR (cable I was using at the time was a quality HD15 to RGBHV BNC one) but had no visible ghosting.

Thank you for the suggestions . I do have a Radeon 9700, an X850 (somewhere) and a Matrox G550. I will probably gives those a shot if the cable avenue is not conclusive .

As for the cables themselves, I am looking at finding (relatively) affordable quality ones from Gefen/Extron/Kramer/BlueJeans and I have heard good things about Canare products, quality wise .

EDIT: IXOS cables seem to be hard to come by in my neck of the woods .

Ah that's right IXOS is probably aimed at the UK and Europe. I've seen some bad and some good KVM monitor cables that might be cheap.

I am hoping to utilise an AJA IO HD as a capture device with a hackintosh but there is little mention of supported 4:3 resolution upper and lower limits in the manual.

I received my C2G 3' cable and tested it, after re-ajusting phase, as usual .
There is still ringing/ghosting, but it is significantly weaker (probably closer to 75 ohm) than my other cables, but its offset is bigger than that of the 1' cable (longer cable implies longer delay before reflection arrives at destination) . I feel that I am on the right path here, but not quite there yet . That BlueJeans cable made from Belden cable stock is starting to look tempting, even at its high price .

Hanamichi wrote on 2020-10-23, 15:16:

darry wrote on 2020-10-21, 17:50:

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully s […]
Show full quote

Hanamichi wrote on 2020-10-21, 16:36:

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced. Then you could […]
Show full quote

Do you have a high end Trinitron/Diamondtron CRT? Or maybe a lower spec'd one you could run 1600x1200 interlaced.
Then you could check for the ghosting/reflection and eliminate or deduce that it is the capture end that's at fault.

A Radeon 9500/9700/9800/X800/X850 has some of best RAMDACs in the business so they could also help determine if the problem lies with the Voodoo3. Failing that a Matrox G400?
I use IXOS VGA cables that seem to be going cheap atm but can recommend Gefen/Extron/Kramer ones too and anything that uses Canare coaxial cable.

Hoping you have more success anyway!

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully still chugging along at one of my mother's neighbours . 1600x1200 was not all that sharp on that, AFAICR (cable I was using at the time was a quality HD15 to RGBHV BNC one) but had no visible ghosting.

Thank you for the suggestions . I do have a Radeon 9700, an X850 (somewhere) and a Matrox G550. I will probably gives those a shot if the cable avenue is not conclusive .

As for the cables themselves, I am looking at finding (relatively) affordable quality ones from Gefen/Extron/Kramer/BlueJeans and I have heard good things about Canare products, quality wise .

EDIT: IXOS cables seem to be hard to come by in my neck of the woods .

Ah that's right IXOS is probably aimed at the UK and Europe. I've seen some bad and some good KVM monitor cables that might be cheap.

I am hoping to utilise an AJA IO HD as a capture device with a hackintosh but there is little mention of supported 4:3 resolution upper and lower limits in the manual.

If you already have AJA IO HD , why not, but I hope you realize that it does not natively support RGB video capture or even non sub-sampled video. Then manual says RGB output is possible, but YPbPr is the unit's native format and there are only references to 4:2:2 , never 4:4:4 , so the RGB must be generated from 4:2:2 YPbPr . It might not support 70Hz input either (nothing about this in the manual) . Finally, the unit runs over Firewire 800 , which is limited to 800Mb/second, which is 100MB/second . Uncompressed YUV 4:2:2 video at 640x400@70Hz is about 35MB/second and uncompressed YUV 4:2:2 video at 1024x768@60Hz is about 90MB/second, which means anything above 1024x768 will need to be compressed using a lossy codec ("visually lossless", if you believe the marketing).

darry wrote on 2020-10-24, 00:08:

Hanamichi wrote on 2020-10-23, 15:16:

darry wrote on 2020-10-21, 17:50:

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully s […]
Show full quote

I divested myself of all CRTs a few years ago . My last one was a 19" Viewsonic P95F with a Diamondtron tube that is hopefully still chugging along at one of my mother's neighbours . 1600x1200 was not all that sharp on that, AFAICR (cable I was using at the time was a quality HD15 to RGBHV BNC one) but had no visible ghosting.

Thank you for the suggestions . I do have a Radeon 9700, an X850 (somewhere) and a Matrox G550. I will probably gives those a shot if the cable avenue is not conclusive .

As for the cables themselves, I am looking at finding (relatively) affordable quality ones from Gefen/Extron/Kramer/BlueJeans and I have heard good things about Canare products, quality wise .

EDIT: IXOS cables seem to be hard to come by in my neck of the woods .

Ah that's right IXOS is probably aimed at the UK and Europe. I've seen some bad and some good KVM monitor cables that might be cheap.

I am hoping to utilise an AJA IO HD as a capture device with a hackintosh but there is little mention of supported 4:3 resolution upper and lower limits in the manual.

If you already have AJA IO HD , why not, but I hope you realize that it does not natively support RGB video capture or even non sub-sampled video. Then manual says RGB output is possible, but YPbPr is the unit's native format and there are only references to 4:2:2 , never 4:4:4 , so the RGB must be generated from 4:2:2 YPbPr . It might not support 70Hz input either (nothing about this in the manual) . Finally, the unit runs over Firewire 800 , which is limited to 800Mb/second, which is 100MB/second . Uncompressed YUV 4:2:2 video at 640x400@70Hz is about 35MB/second and uncompressed YUV 4:2:2 video at 1024x768@60Hz is about 90MB/second, which means anything above 1024x768 will need to be compressed using a lossy codec ("visually lossless", if you believe the marketing).

Yeah it's just doing nothing in the garage. I bought it cheap for it's unique ability to de-embed 8ch HD Audio digitally to multiple SPDIF outputs hooked up to an older AV receiver.

It does seem to convert to YPbPr color space so not ideal for screenshots I agree. Only looking to capture at 800x600 max so thought it was worth trying out and nice to free up a pcie slot. I was thinking 10bit 4:2:2 would be good enough for Youtube as once uploaded it would end up 10/8bit 4:2:0.

I guess capturing RGB 24bit or higher and integer scaling 4x then uploading is the best way?

Hopefully I have fixed my Epiphan DVI2PCIe board I have so I can assess the IO HD performance against the best.

Hanamichi wrote on 2020-10-24, 01:27:

Yeah it's just doing nothing in the garage. I bought it cheap for it's unique ability to de-embed 8ch HD Audio digitally to mult […]
Show full quote

darry wrote on 2020-10-24, 00:08:

Hanamichi wrote on 2020-10-23, 15:16:

Ah that's right IXOS is probably aimed at the UK and Europe. I've seen some bad and some good KVM monitor cables that might be cheap.

I am hoping to utilise an AJA IO HD as a capture device with a hackintosh but there is little mention of supported 4:3 resolution upper and lower limits in the manual.

If you already have AJA IO HD , why not, but I hope you realize that it does not natively support RGB video capture or even non sub-sampled video. Then manual says RGB output is possible, but YPbPr is the unit's native format and there are only references to 4:2:2 , never 4:4:4 , so the RGB must be generated from 4:2:2 YPbPr . It might not support 70Hz input either (nothing about this in the manual) . Finally, the unit runs over Firewire 800 , which is limited to 800Mb/second, which is 100MB/second . Uncompressed YUV 4:2:2 video at 640x400@70Hz is about 35MB/second and uncompressed YUV 4:2:2 video at 1024x768@60Hz is about 90MB/second, which means anything above 1024x768 will need to be compressed using a lossy codec ("visually lossless", if you believe the marketing).

Yeah it's just doing nothing in the garage. I bought it cheap for it's unique ability to de-embed 8ch HD Audio digitally to multiple SPDIF outputs hooked up to an older AV receiver.

It does seem to convert to YPbPr color space so not ideal for screenshots I agree. Only looking to capture at 800x600 max so thought it was worth trying out and nice to free up a pcie slot. I was thinking 10bit 4:2:2 would be good enough for Youtube as once uploaded it would end up 10/8bit 4:2:0.

I guess capturing RGB 24bit or higher and integer scaling 4x then uploading is the best way?

Hopefully I have fixed my Epiphan DVI2PCIe board I have so I can assess the IO HD performance against the best.

DISCLAIMER : I am not an expert in video or image processing so my understanding may well be incorrect, please correct me if wrong .

This applies to computer generated graphics with hard edges, which is essentially the worse case scenario for sub-sampling, AFAIK :

Ideally, to negate the effect of the sub-sampling, you would need to integer scale by a a factor of 2 or more BEFORE video capture, otherwise you will just end up making the sub-sampling artifacts more noticeable . Since 320x200 is already line-doubled to 640x400, that should look OK . Everything else will be affected by sub-sampling unless line-doubled at the source or by an intermediate device, like an OSSC.

In then end, if it ends up as 4:2:0 on Youtube or elsewhere, capturing it as 4:2:2 or even 4:2:0 should not be a significant issue as long as you do not upscale (integer or otherwise) the video if it has already been sub-sampled . Upscaling sub-sampled video will just increase the visibility of the sub-sampling's image re-construction artifacts . Up-sampling video BEFORE it is sub-sampled will minimize the visual effects of the sub-sampling .

For example, if you take an RGB image and integer scale (line double) it by 2x on each axis (preserving aspect ratio), each original pixel will become 4 identical pixels . If you then 4:2:0 sub-sample this image, chroma will only be sampled once every 2 pixels on each axis, but since every other pixel is identical to the one preceding it (due to the integer scaling before the sub-sampling), there is no actual information lost (or nearly so, because the RGB to YPbPr conversion that occurs prior to sub-sampling is lossy and alignment between the integer scaling an sub-sampling can play a role as well ).

I am curious about the audio de-embedding . Does it not care about SCMS and is it low latency enough to allow real-time listening when synchronized to a video track ?

If you capture YUV the OSSC doubling is a good option.

If you capture RGB it is on your side where you do the color space conversion. As you upscale for youtube anyway, it makes sense to already provide the encoder with the RGB data and let it do the conversion.

Aspect ratio: In a general case there is no way to accomplish this with integer scaling. Just think of 720x400 text mode, which is shown as 4:3, so it does not have square pixels. What has proven for me suitable for YT is scaling up with integer factor and nearest neighbour scaling by e.g. 800%. This creates a 5760x3200 intermediate image. Then downscale to 1440x1080 with lanczos. This step changes the pixel aspect ratio to 1:1. The choice of algorithm defines how well the pixel edges are kept or not.
Overall the 720 was scaled now by 2 and the 400 by 2.7. If you convert now to YUV the effect is negligible.
Of course, if you source resolution increases then the factors decrease (targeting FHD) and degradation through YUV conversion gets more prominent again.

elianda wrote on 2020-10-24, 11:41:

Aspect ratio: In a general case there is no way to accomplish this with integer scaling. Just think of 720x400 text mode, which is shown as 4:3, so it does not have square pixels. What has proven for me suitable for YT is scaling up with integer factor and nearest neighbour scaling by e.g. 800%. This creates a 5760x3200 intermediate image. Then downscale to 1440x1080 with lanczos. This step changes the pixel aspect ratio to 1:1. The choice of algorithm defines how well the pixel edges are kept or not.

I don't know, 720 x 400 would show as 4:3 if you don't have a choice of display, but I personally never end up squishing it when viewing in a capture window.

By the way, have you done testing on the effects of interpolated scaling on temporal image quality on YouTube? I've noticed it tends to end up with a larger video file if you do interpolated rather than nearest neighbor, but also with sharp large pixels you get super bad blurring in motion. I wonder if with some pre-mixing of edges it would be more bearable.

Does anyone have experience with these no-name $30 VGA to USB 2.0 capture devices? https://aliexpress.com/item/4001339740771.html

here you can see PCB, but not much detail and the chip is under heatsink

adalbert wrote on 2020-11-13, 01:52:

Does anyone have experience with these no-name $30 VGA to USB 2.0 capture devices? https://aliexpress.com/item/4001339740771.html

This thing most certainly uses MJPEG or some other lossy compression algorithm at higher resolutions, otherwise even 1280x720@60Hz would be impossible over USB 2.0. Maybe it allows uncompressed sub-sampled 4:2:2 video (like YUY2) at lower resolutions, but that is far from certain.

It is also unlikely that it supports 70Hz input (though not impossible).

EDIT : the price looks low enough that risking it might seem like a good idea, but considering that 60$-100$ gets you a known working and good quality solution like the Datapath E1S (albeit PCI Express, not USB), I would personally not bother risking 30$, unless you want to review it .