VHS Capture Size Debate, The Huge Experiment

[sammie]

Ok folks here are the results of my size of capture experiments this weekend. The experiment came about from my noticing that the CVD resolution files I had been capturing weren’t near as good as a full frame DVD resolution capture I had done. The debate was over could that be since there supposedly is no additional information obtainable from a VHS source beyond CVD resolution. Reducing the results to couple of simple sentences. The higher the resolution you capture at, the better the final results. The higher the bitrate you encode at the better the results. I didn’t try an exhaustive array of sizes and bitrates (this takes a long time guys!) just typical “standard” combinations.

I used my ATI 8500 all in wonder, and ATA100 RAID array capture drive, fast P4, etc. Clips were captured with VirtualDub and encoded with TMPEGEnc, MPEG2, 2 pass highest quality (except for the VCD). So, my results may be somewhat skewed in that I don’t have to worry about dropped frames. The resolutions I tried were VCD, CVD, SVCD and DVD. Bitrates I tried were 1150 CBR, 1800 CBR, 2000 VBR 3000 VBR, 4000 VBR and 5000VBR. I didn’t try all bit rates with all sizes just those that were more or less appropriate.

The results were pretty much each increase in size or bitrate resulted in a noticeable increase in visible quality. I don’t have a good way of quantitizing this but they were all noticeable. One item of note was that capturing at sVCD or DVD resolution then converting back to CVD gave results that were ultimately a lot closer to a DVD final resolution than capturing and encoding at a CVD resolution. The CVD / SVCD formats did seem to level out at around 4000 kbit/sec, but there was still an increase in quality at higher bitrates.

Obviously there is an advantage to capturing at higher resolutions. Based on some extensive reading (had to do something while I was doing all this encoding) there absolutely is additional information above 352 in VHS tapes. Maybe not a lot depending on the quality of your capture system, but always some and with a good system likely significant amounts. My suggestions are to capture at the highest resolution you can without dropping frames. If you want to get more on a disc then resize down to CVD or SVCD resolution (they were better at 2 – 3000 than DVD resolution). I’ve never noticed the artifacts of resizing noted in other posts. I have seen some over sharpening if you use the wrong settings on your resize filter, but never ballooning or skewing. The best compromise seems to be kind or an XSVCD with around 3000 VBR encoding if your standalone will play it. I’m worried about the overall compatibility of SVCD (never mind XSVCD so I would probably use CVD if I really needed to get more on a disk). If not CVD at the same bitrate is a close second. One thing I noted was that for CVD and DVD at around 5000 the DVD resolution was still better so as long as your aren’t bit starved and are using a highquality multipass VBR encode, bitrate per pixel isn’t everything. I didn’t try an XSVCD at around 5000 because my standalone player freaks out at SVCD’s above about 3500. Since I have a DVD burner and bulk DVD-r’s are under a buck now, I’ll just be doing straight DVD’s from now on.

That was my 2 cents. Flame away if you must but all this was this was the unanimous conclusion of 6 viewers who are now heartily sick of the same 10 minutes of A Muppet Christmas Carol. I didn't try high motion vs. low motion at border line bit rates. I'm done for a while. Maybe someone else wants to try high motion at 4000, 5000, 6000 DVD vs. CVD if you want to debate that.

[bde]

Ah yes... when trying out different bitrates, I would use different movies to test the settings on... otherwise, you do get sick of the same 10 minutes of a movie. Oh well... I have to say that your post was a very good one to read, and it helps to read this the day after buying a new $80 tv capture card... of which is the best for me..

[sammie]

One other point I forgot to make. VCD, SVCD and low bitrate CVD are definitely bit starved on a VHS capture. What i found to be almost as important as anything above is using a good quality cleaner prior to encoding ESPECIALLY at low bitrates. FlaxenVHS for VirtualDub or Convolution3D for AVISYNTH work really well for me.

[Anonymous65]

You might find some extra info here http://steve.kittelsen.com/vcd/

[sammie]

Arrgghh!!! Nothing like a wasted weekend. Good reference though and the test patterns really show it up well. Actually, I did a few still captures, but the differences are even more apparent in the video than for stills.

Now if someone had pointed that out before, when everyone was telling me I couldn't be right.....

[fmctm1sw]

This I like. I like when folks back up their idea with experimentation and facts. I wish I had more time to invest in this kind of thing. I did a test I titled "capturing over 352x480 making vcd's?" and backed it up with an experiment but it didn't seem to gain any support. My thought was a frame is a frame is a frame.... etc....