Capturing VHS tapes with an ATI Radeon 9200 VIVO

[kassandro]

Usually I am just recording DVB streams, which is a fairly easy task. But I have also several hundred old VHS tapes (at least 6 years old) and I would like to convert at least a few of them to DivX/XviD. Now I tried VirtualVCR, VirtualDub 1.6.4, IUVCR and Flycap. With VirtualDub 1.6.4 I got excellent A/V sync but I lost up to 1% of all frames, independent of resolution and with cpu usage of only 20-30%. With VirtualVCR I couldn't get A/V sync no matter how I tweaked it and I was also regularily loosing frames. IUVCR reported no frame losses at all, but it obviously did cheat. When I choose the alternative method for frame counting it reported also lots of frame losses. A/V sync was poor as well. Finally I tried Flycap. It lost only frames at the very beginning of the tape and at the video end (not the end of the tape), but I don't know whether flycap is also cheating. A/V sync seemed to be acceptable, when Audio was chosen as master. After all these tests I started with real work. I took PicVideo Version 3 to capture the luma with quality 8 and the chroma with the lowest quality 100 (it was a b&w video) with a resolution of 720x576. Audio was mono with a 32KHZ sampling rate. As expected I lost only frames at the beginning and the end. I ended up with 5 avis. 4 were 2 GB and the last was 1.4 GB. For the first 4 avis a/v sync was acceptable. But the 5th avi was way out of sync at the end. After about 5 audio reencoding attempts of the 5th avi with Vdub I accomplished a reasonable a/v sync. Then I thought that the rest is childs play, but when I tried the Avisynth script

input = avisource("capture.avi") + avisource("capture1.avi") + ...

VdubMod rejected the script. Because all the avis had slightly different frame rates. Finally the rediculous

input= avisource("capture.avi").AssumeFPS(25.0,true).ResampleAudio(32000)
+ avisource("capture1.avi").AssumeFPS(25.0,true).ResampleAudio(32000) + ..

was accepted by VdubMod and the rest was easy. I did spend an entire weekend for one single movie. Is this a problem of the ATI theater chip in the VIVO card? Is it crap, or what should I do? Most analog capturers have tv cards instead. For the time being I have given up analog capturing. While the video quality of b&w movies is fairly good, I am getting a quite regular noise pattern with color VHS tapes. The following pic is typical:
http://home.pages.at/kassandro/pics/albers.png
What is going on here?

[reboot]

First, set Video as master stream. If frames are dropped, the audio should be dynamically adjusted to fit. This works in VirtualVCR, not sure about others.
Second, you have two things, interlacing, and the noise at the bottom is from the VHS tape's audio track.
This will NOT show up on a TV, but is ugly in DivX encodes.
Use the cropping feature of virtualdub to remove it.

[Boulder]

That screenshot doesn't look like it's from an interlaced source - there's no visible combing. The lines in the picture are more like some interference.

Fizick has made an Avisynth plugin for removing such interference noise.

[kassandro]

Originally posted by Boulder
That screenshot doesn't look like it's from an interlaced source - there's no visible combing. The lines in the picture are more like some interference.

The strange thing is, that this unpleasant pattern - one can't really call it noise - only occures with color films and it is not a sole chroma problem. If I take away the chroma, this pattern is still there. However, it may be that chroma signal is causing this pattern. When the film is b&w the chroma signal may be too low to cause such disturbances.
Actually I don't have the original cable adapter for this card. I bought the graphics card in late 2003 for about 50€ on Ebay without the cable adapter. I primarily needed a graphics card with a DVI connector for a TFT display, which I did buy in early 2004. When I upgraded my system I bought a similar cable adapter from MSI, which had the same 9 pin mini-DIN connector. Unfortunately it worked only with S-Video in/out (I have tested this with my DVD player). It didn't word with composite in/out. Thus I bought an adapter to connect a VCR to a TV with S-Video input and instead of the TV I connected it with that cable adapter. Now I have good quality b&w but poor quality color capturing. When I capture from my DVD player (this is of course only for testing) then I do not have this unpleasant pattern, but I didn't use the composite->S-Video Adapter. I do have also a lot of frame drops, though. To be certain that the cable connection is the problem, I have to connect the DVD player with the composite connector to the graphics card. I will make a test next weekend.



Fizick has made an Avisynth plugin for removing such interference noise.
I now remember something like that. Anyway, I will make a look on his web site.

Originally posted by reboot

First, set Video as master stream. If frames are dropped, the audio should be dynamically adjusted to fit. This works in VirtualVCR, not sure about others.

I will try this with VirtualVCR next weekend.

[Boulder]

The vertical stripes which appear throughout the image look like bad grounding or something like that, and Fizick's DeFreq (that was it if I remember correctly) should really help. I have those stripes on my second TV which is attached to the same "antenna out" as my TV card. Luckily the stripes don't appear in the captures:)

Regarding frame drops in VDub, have you tried enabling "Correct video timing if it is off from real-time"? (Capture->Timing)

And yes, VIVO's are mostly crap;) I'd look for a second-hand capture card, any BT878 one should be good. I recently bought mine for 30e and got 90e for the PVR-250 I sold:p

[kassandro]

Originally posted by Boulder

Regarding frame drops in VDub, have you tried enabling "Correct video timing if it is off from real-time"? (Capture->Timing)

No. The following option were enabled in "Capture Sync Options":
1.Drop frames when captured frames are too close together
2.Insert null frames when captured frames are too far apart
3.Sync Audio to Video by resampling the audio to a faster or slower rate.
Because there were no a/v sync problems, I probably didn't change anything there. I don't know what 1. means. Does Vdub really measure the distance of two frames? I also did capture without audio in order to see, whether the frame drops were caused by audio, but the frame drops did persist. The captures with Flycap had a frame rate around 25.08. I thought that Vdub drops frames to lower the rate from 25.08 to 25, but 0.08 is only 0.3% and I loose almost 1% of all frames with Vdub.



And yes, VIVO's are mostly crap;) I'd look for a second-hand capture card, any BT878 one should be good. I recently bought mine for 30e and got 90e for the PVR-250 I sold:p
I thought that with a separate TV card you have not only PCI2memory traffic but also PCI2AGP traffic. With a capture chip on the graphic card you can completely avoid the PCI2AGP traffic. Thus VIVO seemed the more natural approach to me. On the other hand the BT878 has established itsself as the standard video capturing chip and there is even a generic VfW driver for this chip and probably most capturing software is tuned for this most common chip.
As analog TV is phased out here, the BT878 cards should be very cheap on Ebay, probably cheaper than a correct VIVO cable adapter.
I'll probably make a few tests of the new ideas later in the evening, when the current encoding process is finished.

[Boulder]

Originally posted by kassandro
The captures with Flycap had a frame rate around 25.08. I thought that Vdub drops frames to lower the rate from 25.08 to 25, but 0.08 is only 0.3% and I loose almost 1% of all frames with Vdub.

It drops them if the option I asked about is enabled. I actually keep it disabled since it drops a frame per 3-4 minutes on my system and I don't need that. The captures work fine and appear to be 25fps for both video and audio even though the video clock apparently runs over 25fps.

I thought that with a separate TV card you have not only PCI2memory traffic but also PCI2AGP traffic. With a capture chip on the graphic card you can completely avoid the PCI2AGP traffic. Thus VIVO seemed the more natural approach to me. On the other hand the BT878 has established itsself as the standard video capturing chip and there is even a generic VfW driver for this chip and probably most capturing software is tuned for this most common chip.
As analog TV is phased out here, the BT878 cards should be very cheap on Ebay, probably cheaper than a correct VIVO cable adapter.
I'll probably make a few tests of the new ideas later in the evening, when the current encoding process is finished.
The amount of traffic in the buses is not worth worrying with any modern machine. Besides, you can always disable preview. On my system I have zero dropped frames with preview on or off, and I can also do other stuff while capturing, although it's never recommended.

By the way, what chipset does your motherboard have?

The thing about a dedicated capture card is that the tuner is better than one on any VIVO card.

Some people say that Terratec's Cinergy 400 is one to consider, but it's on a Philips chip so I don't have any experiences regarding the quality. I've had three different BT878 cards so I guess that tells what I like;)

[kassandro]

Originally posted by Boulder
It drops them if the option I asked about is enabled. I actually keep it disabled since it drops a frame per 3-4 minutes on my system and I don't need that. The captures work fine and appear to be 25fps for both video and audio even though the video clock apparently runs over 25fps.

Surprisingly for me the option "Correct video timing if it is off from real-time" did lower the drop rate quite a bit. I now loose 1-2 frames at the beginning and then 1 frame per every 5000-10000 captured frames, which is acceptable. A/V sync stays excellent. Reducing the resolution or disabling preview doesn't improve the situation. Using ordinary preview cpu usage is about 40% and with overlay about 25%. With lower resolutions cpu usage drops. After disabling your option I am getting again an unacceptable drop rate of 0.7-0.8% like two weeks ago. Thus your option is quite valuable for me. Even if Flycap really would loose no frames, it is not fun to use because of the a/v sync unsecurity.



The amount of traffic in the buses is not worth worrying with any modern machine. Besides, you can always disable preview. On my system I have zero dropped frames with preview on or off, and I can also do other stuff while capturing, although it's never recommended.

No, while I am writing this stuff, I am lossing about 1 frame per every 500 frames, still well below what I had two weeks ago.


By the way, what chipset does your motherboard have?

It's a Via PT880. It supports Dual Channel access and after removing the Technotrend sat card the motherboard is very robust. In fact, it never crashed since then.


The thing about a dedicated capture card is that the tuner is better than one on any VIVO card.

My card has no tuner at all. There is only capture chip. It should not be confused with the ATI all in Wonder graphic card series, which have a tuner and even an mpeg2 hardware encoder, which I really wouldn't want. I think, all the other VIVO cards including those from NVIDIA do not have a tuner. Usually they only have a 9 pin mini-din connector and nothing. In particular, no antenna input. Clearly, if you capture analog TV and not VHS, then a tuner is preferable, because a composite connection with the TV will lower quality and PAL TVs rarely have S-Video output. On the hand, for capturing VHS that seems to be no disadvantage per se. Surprisingly Avery Lee, the Virtualdub author, has an ATI All in wonder graphics card with an ATI Theater chip. Hence the WDM support for this chip in Vdub 1.6.4 should be ok.


Some people say that Terratec's Cinergy 400 is one to consider, but it's on a Philips chip so I don't have any experiences regarding the quality. I've had three different BT878 cards so I guess that tells what I like;)
Over here Hauppauge seems to be the clear market leader with analog TV cards.

[Boulder]

Originally posted by kassandro
Surprisingly for me the option "Correct video timing if it is off from real-time" did lower the drop rate quite a bit. I now loose 1-2 frames at the beginning and then 1 frame per every 5000-10000 captured frames, which is acceptable. A/V sync stays excellent. Reducing the resolution or disabling preview doesn't improve the situation. Using ordinary preview cpu usage is about 40% and with overlay about 25%. With lower resolutions cpu usage drops. After disabling your option I am getting again an unacceptable drop rate of 0.7-0.8% like two weeks ago. Thus your option is quite valuable for me. Even if Flycap really would loose no frames, it is not fun to use because of the a/v sync unsecurity.

As the software is still very much experimental, you never know what's going to happen:D Good to hear it works for you.

It's a Via PT880. It supports Dual Channel access and after removing the Technotrend sat card the motherboard is very robust. In fact, it never crashed since then.

If that option above hadn't worked for you, I would have suspected the VIA chipset. Many people have had bad experiences with the Intel-VIA -combination in the past. I currently have a i865 chipset and my own experiences are only in the Duron/TB/Athlon XP + VIA environment, which was tricky at times.

[kassandro]

Originally posted by Boulder
As the software is still very much experimental, you never know what's going to happen:D Good to hear it works for you.

Yes, I will do some real work with it at eastern.

Originally posted by Boulder

If that option above hadn't worked for you, I would have suspected the VIA chipset. Many people have had bad experiences with the Intel-VIA -combination in the past. I currently have a i865 chipset and my own experiences are only in the Duron/TB/Athlon XP + VIA environment, which was tricky at times.
Around the year 2000 Via had a very embarassing problem with the PCI bus. It had a builtin break such that hard disk transfers were slowed down substantially. Obviously they must have solved it, otherwise they wouldn't be here anymore. I am just building from old parts a second PC (to be used mainly as a digital video recorder)around my old 1.3 GHZ Tualatin Celeron, a very well designed CPU, which hardly requires active cooling and which was the basis for the Pentium M, and my old Technotrend Sat card. I only need a reasonable graphics card. Probably the Radeon 9200 will move to this "old" system and then I will repeat the tests, this time with an Intel 815 chip set.
I would have loved to buy a mainboard with an i865 chip set. But there were obstacles. Firstly most of boards with i865 had also a builtin Intel graphics card with shared memory and none of them had a DVI port for my TFT display and connecting it to the VGA port makes 19 inch TFT displays look very poor. Although these mainbords had also an AGP slot, this AGP slot is difficult to use, because it is primarily for dual monitor use. There were no(!!!) Bios options to disable the onboard graphics card and for the AGP slot there was a very restrictive compatibility list. The Radeon 9200 was, however, on this list. Thus my only Intel choice were the very old i845 boards without dual channel or a non-Intel board and the only non-Intel boards with could compete with the i865 and i875 boards were those with Via PT 880 chip set, because the other Via chip sets and those from SIS didn't have dual channel memory. In tests the PT880 had about the same performance as the i865 and slightly less than the i875. The second problem was Prescott compatibility. While all these mainboards, even the old i845 boards could be upgraded with a new BIOS to host also Prescott cpus, my ASRock P4V88 was one of the fews, where this was guaranteed. After removing the Technotrend card, I must say that the new board is more robust than the old one. Probably the very rare crashes of the old i815 board were also caused by the Technotrend card, which seems to cause stability problems for any mainboard.