Dosboxxer wrote on 2020-10-11, 23:15:

Not sure if this helps, but I want to capture pixel perfect captures and I've been reading up on Integer-ratio scaling.

This article seems to help shed a lot of light on the subject.

https://tanalin.com/en/articles/integer-scaling/

Integer ratio scaling and line doubling/tripling, etc are essentially the same thing . The OSSC is a great tool for generating integer-ratio scaled video from a vintage/retro PC or console's output and minimizing (sometimes eliminating) the amount of actual non-integer scaling that a typical monitor (usually do not fo integer-ratio scaling) has to do, which results in a sharper image .

As for capture, high quality pixel-perfect capture can be done with a card like the E1S directly . Alternatively, you can combine an OSSC and E1S, if you like (which is what I am doing, for now) .

swaaye wrote on 2020-10-10, 20:54:

VCS and Vision are fluid.

By the way, if you enjoy VCS, consider starring the repo on GitHub.

vvbee wrote on 2020-10-12, 20:12:

By the way, if you enjoy VCS, consider starring the repo on GitHub.

Done. 😀

darry wrote on 2020-09-27, 21:57:

Got my E1S . It works great, but I need to set refresh and resolution manually for the capture pin in Virtualdub2, even though t […]
Show full quote

Got my E1S . It works great, but I need to set refresh and resolution manually for the capture pin in Virtualdub2, even though the capture filter detects and displays the proper values .

Upon capturing 1600x1200 from Voodoo 3 in passthrough through OSSC, I see VGA ghosting in the capture . This ghosting is practically invisible on my display monitor, but easy to see in the capture I am using an HDMI spliiter) . Capturing 1280x960 in the same setup give a perfect ghosting free image . Capturing 1600x1200 from my FX 5900's DVI out (running through same HDMI spliiter), is also perfect .

I have ordered another VGA cable, hopefully that will fix it. Then again, capturing over 1024x768 from a Voodoo 3 will probably no happen very often .

EDIT : My 1280x960 is actually line-doubled 640x480 .
640x480 line doubled to 1280x960 :

1280x960.png
1280x960_detail_e1s.png

1600x1200:
1600x1200.png
1600x1200_detail_e1s.png

I have received the VGA cables and a first-party OSSC from videogameperfection.com (3-day shipping from Ireland to Canada in covidic times is a beautiful thing) .

a) VGA cables change nothing, so I reverted to my original one, to keep a level playing field
b) running the Voodoo 3's output through a 350MHz amplified VGA splitter before feeding the OSSC makes the ghosting slight worse .
c) The new OSSC changes nothing in terms of ghosting (maybe less noise, but TBH I was not really checking for that)
d) connecting the VGA cable directly to the monitor, I see no ghosting, not that it is very noticeable through the OSSC to begin with on 25" monitor (equivalent to approx. a 21" 1600x1200 one)
e) running the VGA cable directly to the E1S for a capture test (had bring the machines closer together), the ghosting is even worse than through an OSSC to the E1S' DVI input . Maybe it's my VGA to DVI-I adapter's fault (only one I could find was bundled with brand new Crestron/Extron/CYP (forget which) brand new scaler unit, so it should be high quality) .
f) Yes, I have tried adjusting phase (best picture is at 45 degrees) and yes Powerstrip timings exactly match the OSSC's .

Unanswered questions :
1) Is this some kind of impedance mismatch ?
2) Does my monitor apply some kind of noise reduction or filtering on its VGA input that masks the ghosting ? Or could it actually be adjusting its input impedance ?
3) Am I asking too much from a Voodoo 3 or does mine simply have a marginal VGA output ? Is it time to start looking at faults design issues on the Voodoo's VGA output ?

Almost all LCDs with an analog input low pass filter the signal before digitizing it. The OSSC does not

maxtherabbit wrote on 2020-10-19, 00:53:

Almost all LCDs with an analog input low pass filter the signal before digitizing it. The OSSC does not

The OSSC does have a switchable 95 MHz LPF on its VGA input, but that obviously makes the picture blurry at 1600x1200 and does nothing for the ghosting anyway .

I will try with another VGA card and see if it's still as bad .

darry wrote on 2020-10-19, 01:14:

maxtherabbit wrote on 2020-10-19, 00:53:

Almost all LCDs with an analog input low pass filter the signal before digitizing it. The OSSC does not

The OSSC does have a switchable 95 MHz LPF on its VGA input, but that obviously makes the picture blurry at 1600x1200 and does nothing for the ghosting anyway .

I will try with another VGA card and see if it's still as bad .

I compared the Voodoo 3 with a Geforce FX 5900XT . Please ignore the slightly shifted image on the FX 5900XT pictures. I had omitted to manually tweak the back porch/sync length settings .
The FX 5900XT still has ghosting, but is noticeably better .

FX 5900XT :

Voodoo 3 :

Comparison 1 :

Comparison 2 :

EDIT : I tested with this cable https://media.startech.com/cms/pdfs/mxt101mmh … 3_datasheet.pdf , but reverted to my usual 6' 6" one for practical purposes when I saw no difference in picture quality .

darry wrote on 2020-10-19, 01:14:

maxtherabbit wrote on 2020-10-19, 00:53:

Almost all LCDs with an analog input low pass filter the signal before digitizing it. The OSSC does not

The OSSC does have a switchable 95 MHz LPF on its VGA input, but that obviously makes the picture blurry at 1600x1200 and does nothing for the ghosting anyway .

I will try with another VGA card and see if it's still as bad .

It's not exactly 95MHz, and it's also a POS. If you read the actual TVP7002 datasheet it's very vague about what the actual corner frequency is, and seems to indicate that it attenuates by some amount all the way down to 50MHz. It seems to just be a garbage filter overall and I avoid using it at all costs. It's also off by default on the OSSC unless you specifically turn it on.

maxtherabbit wrote on 2020-10-19, 13:25:

darry wrote on 2020-10-19, 01:14:

maxtherabbit wrote on 2020-10-19, 00:53:

Almost all LCDs with an analog input low pass filter the signal before digitizing it. The OSSC does not

The OSSC does have a switchable 95 MHz LPF on its VGA input, but that obviously makes the picture blurry at 1600x1200 and does nothing for the ghosting anyway .

I will try with another VGA card and see if it's still as bad .

It's not exactly 95MHz, and it's also a POS. If you read the actual TVP7002 datasheet it's very vague about what the actual corner frequency is, and seems to indicate that it attenuates by some amount all the way down to 50MHz. It seems to just be a garbage filter overall and I avoid using it at all costs. It's also off by default on the OSSC unless you specifically turn it on.

It did not go as far as the datasheet, but it is labelled as 95MHz in the OSSC and I when I did try it, it made the image softer and did not affect the ghosting, or should I say ringing .

After reading some more, I am really starting to think that this is an impedance issue (signal reflection) and not something that can be low pass filtered.

Unfortunately, either way, I will likely need to learn to live with it , unless someone has some suggestions regarding things to try .

Yes I'd be inclined to agree that it is a reflection issue. You said you tried different cables and nothing changed, but are you certain that the new cables were actually impedance controlled?

maxtherabbit wrote on 2020-10-19, 14:25:

Yes I'd be inclined to agree that it is a reflection issue. You said you tried different cables and nothing changed, but are you certain that the new cables were actually impedance controlled?

According to Startech's spec sheet, at least this one is supposed to be :
https://media.startech.com/cms/pdfs/mxt101mmh … 3_datasheet.pdf

There is always the option of ordering something from https://www.bluejeanscable.com/store/rgbhv/index.htm .

At 60+ US$ + shipping, taxes, duties (though it is made in USA, so NAFTA or whatever it's called should apply) for a 3' cable, it's a bit steep .

I wouldn't mind paying this much for a high quality custom made cable, if I could be certain the issue will be resolved or reduced significantly.

https://www.ebay.com/itm/Cables-2-Go-VGA-Cabl … 12/122635458833

I can personally vouch for the C2G VGA cables

maxtherabbit wrote on 2020-10-19, 17:34:

https://www.ebay.com/itm/Cables-2-Go-VGA-Cabl … 12/122635458833

I can personally vouch for the C2G VGA cables

Thanks for the suggestion.

I ordered a 50211 and a 50212 by C2G from Amazon .

I will post back on whether they help or not .

This is not ghosting, this is reflection issue (reflecting easily on liberally wires), I see this frequently with cheap VGA extension cables that does not use three coax cables for the RGB that is impedance of 75Ohms.

The true quality VGA cables is thick cable and has ferrite bead on each end.

https://www.google.com/search?q=quality+vga+c … =0dWvsdEDftvQvM

Cheers,

pentiumspeed wrote on 2020-10-19, 20:20:

This is not ghosting, this is reflection issue (reflecting easily on liberally wires), I see this frequently with cheap VGA exte […]
Show full quote

This is not ghosting, this is reflection issue (reflecting easily on liberally wires), I see this frequently with cheap VGA extension cables that does not use three coax cables for the RGB that is impedance of 75Ohms.

The true quality VGA cables is thick cable and has ferrite bead on each end.

https://www.google.com/search?q=quality+vga+c … =0dWvsdEDftvQvM

Cheers,

The word I most often heard in the 90s to describe the visual phenomenon was "ghosting", but the word has taken on a different meaning in the context of LCD . Another term, I have heard more recently is "ringing". Either way, whatever the currently in-fashion/official name for it, I agree that it is likely due to signal reflections caused by an impedance mismatch .

All the cables that I have tried have a thick cable and ferrite beads (at least one). The Startech one purports to be 75 ohm (EDIT : and coaxial). I do not have specs for my NOS thrift store one.

The C2G cables that I have on order do not have ferrite beads, but AFAIK, the ferrite beads help with noise, not with signal reflection related issues .

You can't go wrong with the C2G cables (made by legrand) but I hope I'm not sending you on a wild goose chase. I'm afraid the DAC on your voodoo may be to blame

maxtherabbit wrote on 2020-10-20, 01:14:

You can't go wrong with the C2G cables (made by legrand) but I hope I'm not sending you on a wild goose chase. I'm afraid the DAC on your voodoo may be to blame

You could well be right about the DAC of the Voodoo 3 . It may also be that the card's analogue output stage design is not that great at high frequencies . A Geforce FX 5900 XT (Gainward) is only marginally better at the same resolution . Maybe I should try more cards . Beside possibly a Matrox G550, which I have recently put back into storage, are there any recommendations of cards with a reputation for unimpeachably good quality analogue VGA output that goes at least to 1600x1200, preferably higher (I don't expect any consumer video card to look its best when running at its RAMDAC's limits) ?

This brings me to another question : who here (or elsewhere) has actually captured the analogue VGA output of consumer cards at 1600x1200 and actually magnified it enough to look for issues ? I'm sure I'm not the first one, but there can't be that many of us who have actually bothered . Practically anything worth capturing at that resolution is better done over DVI/HDMI .

Maybe this is all completely "normal" ? Maybe the OSSC and E1S should have come with BNC connectors (I would not have complained)?

I've digitized 1920x1080 from a geforce fx5200 via VGA on both my OSSC and data path vision-av sdi and it looks perfect. Didn't try 1600x1200 but I don't think the dot clock is much higher

maxtherabbit wrote on 2020-10-20, 02:00:

I've digitized 1920x1080 from a geforce fx5200 via VGA on both my OSSC and data path vision-av sdi and it looks perfect. Didn't try 1600x1200 but I don't think the dot clock is much higher

1600x1200 has a higher pixel clock than 1080p because the number of total pixels is higher . The number of active pixels at 1080p is actually lower . This actually me want to try to generate custom timings with severely reduced blanking (using Powerstrip), but unfortunately, I was not able to get below a 150ish MHz pixel clock (versus 162MHz) before I had trouble getting the OSSC and/or my monitor to sync . The resulting reduction in bandwidth did not result in better picture quality, however .

I will wait until I test some more cables before I run any more tests . I have a Chinese (designed and manufactured) FX 5500 I could try, but I doubt its VGA out is more than an afterthought, quality-wise .

By the way, do you have some sample lossless captures that you could share of that FX 5200 over VGA at 1920x1080 ?

I know 1600x1200 is higher bandwidth. Just stating the dot clock isn't *that* much higher. 148.5MHz vs 162MHz. Unfortunately I didn't save any of the captures from that PC it was just a test. I could do it again but it would be a while