Stupid question re: conversions and bitrates

[jbodin]

Okay, stupid question -- I've searched around a bit but I haven't found any posts that answer my question, so I figured I'd go ahead and show just how stupid I am.

I want to stick four (4) TV episodes (~ 42 minutes each) on one DVD. My video capture card and capture software (Emuzed Maui III capture card and associated MovieMill capture app) only does MPEG2 captures (hardware encoded) at either 720x480 full-D1 resolution, or 480x480 SVCD resolution. I've been doing my captures at 720x480 then converting them to 352x480 half-D1 resolution at a lower bitrate, and the results have been very satisfactory so far.

The question is this: Would I be better off doing my captures in full-D1 720x480 resolution at a lower bitrate which would allow me to get 2 hours and 48 minutes' worth of video on a DVD (the VideoHelp Bitrate Calculator tells me I'd have to do my capture at approximately 3400 kbit/s in order to do that, versus doing it at 6000 kbit/s that I'm currently using), or am I better off doing the video capture at full-D1 720x480 resolution at 6000 kbit/s and then re-sizing it to half-D1 352x480 resolution at a lower bitrate as I've been doing?

Does a capture and conversion to half-D1 352x480 resolution look worse than a straight full-D1 720x480 capture at a lower bitrate would look?

At this point, I've been pleased with the results of the capture/conversion efforts that I've done so far, but it occurred to me recently that I could just use the original 720x480 full-D1 captures as-is with no reduction if I simply lowered my capture bitrate . . . what I'm not sure of, though, is whether the resultant capture would be worse than the results I'm getting now, or if it would just work out to be a wash overall.

Using lower bitrate full-D1 captures would eliminate the time it takes to convert and re-encode the files, but if the results would be worse, I don't want to go that route. If it would be better, though, then all I'm doing is wasting my time with the half-D1 conversion effort.

Any input or advice would be greatly appreciated!

[mats.hogberg]

720x480 @ 3400 kbps is on the low side, but it all depends on what you capture. Talking heads would still look good, but a KungFu movie unwatchable.
Since you're resizing down anyway, why not cap @ 480x480 and resize down? (More bits/pixels=higher quality in the source.) Or try to cap at 480x480 and author with some authoring app that allows for out of specs DVD (like DVD-Lab can) or run it thru SVCD2DVDMPG before authoring. Most players will play 480x480 DVD fine.

/Mats

[lenti_75]

a half D1 at higher bitrate looks better than a full D1 at lower bitrate.

for a full D1 lower than 4000 looks really bad.

[jman98]

a half D1 at higher bitrate looks better than a full D1 at lower bitrate.

for a full D1 lower than 4000 looks really bad.

As a general rule, yes, you are right. However, if you have something like CCE and know what you are doing, it is possible to encode to lower bit rates and get very good results. I would not recommend capturing at 3400, but a capture at a higher bit rate, say 8000, and re-encoding down to 3400 with a good encoder should produce good results.

[mats.hogberg]

There is never anything gained (qualitywise) by encodeing multiple times (unless you're doing some clever filtering).
To try to be as clear as cornucopia:
You start with 100% quality which is then encoded to say 90% quality.
Then you encode the 90% q video at a lower bitrate, say 75% q: 0.9x0.75=67.5%

If you had encoded once with the 75% quality (bitrate) setting, you'd ended up with 1.0x0.75=75% quality.

/Mats

[jbodin]

Appreciate all the feedback and knowledge-sharing, guys -- MANY thanks!

I spent the evening last night doing quick 5-minute captures of the same recorded show (recorded on my PVR) at full-D1 and SVCD resolutions, then doing half-D1 conversions on both files; I also did a couple of full-D1 captures at different bitrates (one at 5000 and one at 3400), then I burnt two test discs -- one with the two re-encoded half-D1 samples and one with the two full-D1 samples.

Once I had my two test discs, I did back-to-back comparisons of the four output samples just to see which looked best to me, what looked acceptable, and what was beginning to show noticable degradation.

My eyes told me this:

Full-D1 captures at a bitrate of 5000 look every bit as good as the PVR source.

Full-D1 captures at 3400 show a small but noticeable amount of degradation compared to the PVR source.

Re-encoding a full-D1 capture to a half-D1 mpeg file produces good results as long as I make sure that the original interlacing and field order is preserved (I use GUI4ffmpeg to do my half-D1 conversions because it's fast and free and the output quality is very good . . . but I can't get it to retain the original interlacing and field order, which means I have to demux and use ReStream to restore the interlace and field order, then remux . . . not hard, not too time consuming, but more work than I enjoy doing).

Re-encoding an SVCD capture to half-D1 mpeg file produced the worst results -- not sure why, but the resultant sample showed some blockiness that I didn't see in any of the other samples. This would NOT be my preferred choice.

My final solution (at this point, anyway), is to either do full-D1 captures at a bitrate of 5000 and just put two 42-minute shows per disk, or drop the bitrate down to about 4500 and in order to fit three 42-minute shows per disk. As cheap as blank discs are these days, there's no compelling reason to spend the time and effort to re-size my captures down to half-D1 resolution just to fit more shows per disc. I still need to do a couple of tests using a full-D1 captures at a bitrate of 4500 or so just to see if that's acceptable, but my guess is that this will be a good compromise quality-wise which will still allow me to put three episodes (or a single episode and a double-length epsiode) on a single disc.

Again, though, thanks for the feedback, and for giving me some things to consider. As it turns out, I think I've been wasting a lot of effort with the half-D1 conversion process . . . certainly seems to be more trouble than it's worth, but at least it's been a fun learning experience.

[jbodin]

Re-encoding a full-D1 capture to a half-D1 mpeg file produces good results as long as I make sure that the original interlacing and field order is preserved (I use GUI4ffmpeg to do my half-D1 conversions because it's fast and free and the output quality is very good . . . but I can't get it to retain the original interlacing and field order, which means I have to demux and use ReStream to restore the interlace and field order, then remux . . . not hard, not too time consuming, but more work than I enjoy doing).

Found another solution -- I can use gui4ffmpeg to convert my full-D1 captures to half-D1 352x480 resolutions AND retain the interlace and "tff" field order -- this is still an extra step, but the resultant quality is good, and I don't have to use restream (which would add yet ANOTHER step). The solution is this command line switch:

-ilme force interlaced me support in encoder (MPEG2/MPEG4)

The "-ilme" switch forces interlaced motion estimation support -- using that switch, when I resize mpeg files using gui4ffmpeg, interlaced (top field first) files come out interlaced (top field first), rather than progressive. Now when I want to stick multiple episodes on a disk, I just do my capture, remove the commercials (using comskip and comclean), then run it through gui4ffmpeg to resize it down to half-D1 resolution, and I'm done.

[lordsmurf]

However, if you have something like CCE and know what you are doing, it is possible to encode to lower bit rates and get very good results.

"Know what you are doing" doesn't have anything to do with it. Neither does CCE, which is actually not all that great of an encoder when compared to Procoder, for highly compressed bitrates. Even PC would puke on bitrates that low.

There is no way around poor allocation. No knowledge, no hardware, no software. Satellite tv compresses down to about 480x480 with 2000-2500k and there is a good bit of noise still, and they are using machines that cost in the 10's of thousands of dollars (USD).

If you want low bitrates, using Half D1. It basically looks the same anyway, especially on captured traditional sources.

If you want more time on a disc, and insist on 720x480, use DL media.

Good encoders should properly detect field order, unless the MPEG source is wonky. That can happen.