View Full Version : 44.1 KHz vs 48 KHz Audio in Divx DVD Rips
iparout
13th August 2002, 17:12
Hi there everyone...
I am encoding to Divx 5 Pro using GKnot and the output quality is really excellent...
What I am concerned about is the audio. I encode to vbr mp3 128 kbps using the --alt-preset switch in Lame however, what I want to ask about is the sampling rate I should prefer for the audio...
Should I keep the sampling rate at 48 KHz or should I downsample to 44.1 KHz. Can someone briefly explain the pros and cons of each...
Finally, I added a poll just to see what most guys do when ripping a DVD to Divx.
Thanks in advance for any help.
DJ Bobo
13th August 2002, 19:05
for low bitrates such as 128k, you better downsample to get better audio quality.
You can use 48KHz safely starting from 160k (compression is there transparent enough, so it can compensate the higher sampling rate)
Downsampling (as such, independantly from compression) from 48KHz to 44,1KHz causes [b]no hearable[b] quality loss, since AC3s has a lowpass already by 20,3KHz, which is equivalent to 40,6KHz sampling and this is far enough from 44,1KHz, so you can downsample without any regrets ;)
iparout
13th August 2002, 19:56
Thanks for the info mate...
That was really helpful.
I never go above 128 kbps so I think downsampling is the way to do it...
Thanks again...
Koepi
13th August 2002, 20:09
I use 48kHz even for lowest bitate ogg vorbis sound.
By downsampling the sound you loose more quality than with encoding a higher frequency, I assure you (tested it several times).
Btw., since the most common soundcard should be SB Live!/Audigy, 48kHz is the frequency of choice anyways, since those chips upsample every other frequency before outputting it to 48kHz.
Regards,
Koepi
jggimi
13th August 2002, 20:10
From the GKnot Help Tab:known problem:
audio-bitrates below 128 (in lame) can cause sync-problems (and sound terrible), because lame is
automatically downsampling to 32 or even 22kHz. i highly recommend to use 128 or 160 and nothing else.
to avoid downsampling you have to add "--resample 48" to your lame commandline options if you encode at
low bitrates. also downsampling to 44.1 with BeSweet and ssrc should be no problem.
DJ Bobo
13th August 2002, 21:11
@ koepi
I don't use ogg, so I can't confirm. Talking about MP3, downsampling down to 44,1KHz helps getting better quality definitely.
BTW, I have a Yamaha sound card. And not all Sound Blaster models are locked on 48KHz output, the "newer" AC97 specification (which is about 2 years old) allow multiple sampling rate output, so many cards can switch between 48 & 44,1KHz, so this is no argument.
@ jggimi
I'll try to precise things:
1) downsampling down 44,1KHz or 32KHz doesn't affect synch, since Lame is still compressing MPEG-1 compliant. Only 22KHz switches to MPEG-2 or MPEG-2.5 compliant which causes asynch sound (because mux software doesn't handle MPEG-2 & 2.5 well, not because of Lame or because of downsampling).
2) From my experience, following MP3 bitrates are suited to following sampling rates:
80-96kbps -> 32KHz - it does sound muffled compared to the original 48KHz, but it sounds better than with 44 or 48KHz (artifactwise). Besides, one shouldn't expect too much when using such low bitrates.
112-128kbps -> 44,1KHz - nothing to complain about (no hearable quality loss because of downsampling, in the contrary, there is less artifacts compared to 48KHz)
160kbps and higher -> keep 48KHz and save your time
3) downsampling in Lame itself isn't bad. May be in theory, but in practice I can't detect any differences between it and WaveLab (which is equivalent to SSRC but faster) downsampling.
theReal
14th August 2002, 02:21
When you're using --alt-preset settings in LAME, the encoder uses a lowpass filter at 19kHz (any good mp3 encoder uses a lowpass filter at 20kHz maximum - if not lower).
It shouldn't make a difference for the size or quality of the mp3 what the theoretical sampling rate of the sound file is - all sounds in the range from 19kHz (respectively 38kHz sampling rate) upwards are discarded anyways, so why downsample?
However, some old soundcards don't play 48kHz at all, or make a crappy downsampling to 44.1kHz - for these soundcards, 44.1kHz is much better (but I highly doubt that anyone with a fairly new computer has such an old soundcard).
DSPguru
14th August 2002, 05:26
Originally posted by DJ Bobo
WaveLab (which is equivalent to SSRC but faster) downsampling. what's your defenition for 'equivalent downsampling' when comparing ssrc with wavelab ?
iparout
14th August 2002, 09:54
I myself have encountered no quality loss when downsampling from 48 k to 44.1 K, but on the other hand, 48 K didn't sound artifact-eous (is this a word ?!? lol) whatsoever...
Why I downsample, you may ask me... I don't know... :D
DJ Bobo
14th August 2002, 10:51
@ theReal
hehe, you missed the point. There is a lowpass by 19 or 20 or whatsoever, but the other frequencies consume bitrate, not much but enough to produce more artifacts by lower bitrates (which need every bit they can have!). Especially 128k and lower as stated above. I'm not talking in the air, I really tested extensively ;)
@ DSPguru
equivalent in quality
@ iparout
I can definitely detect more artifacts, and I have a trick to make artifacts sound louder. I have a 3D sound option, when I enable it, it's like artifacts come to life. Even with --alt-preset, I tested in on 128k one time with 48kHz (default) and one time with 44,1KHz. On the 48KHz version, the high frequencies are somehow choppy (at least more than the 44,1KHz).
theReal
14th August 2002, 15:02
I was just asking myself why anyone ever had the idea to use 48kHz for DVD audio when there is also a lowpass-filter at around 20kHz for AC3 sound.
I mean 44.1kHz was supposed to give enough headroom for music on audio CD's, do explosions and stuff really need that much more headroom to be free of clippings?
DJBobo, do you know how much bitrate is used by mp3 codecs per unused kHz? I always thought it was totally negligible.
DJ Bobo
14th August 2002, 20:49
@ theReal
Don't know, sorry
fu2k
14th August 2002, 23:37
Originally posted by theReal
I was just asking myself why anyone ever had the idea to use 48kHz for DVD audio when there is also a lowpass-filter at around 20kHz for AC3 sound.
Real filters are never "brick wall", they have a finite transition band. The sharper the cutoff, the more complex and expensive the filter is. DVDs may use a 48kHz sampling rate with the same 20kHz "cutoff" to allow the use of simpler/cheaper analog filters.
western shinma
15th August 2002, 21:39
theReal,
Audio recording for film and TV is usually done at 48 or 48.048 kHz, so there wasn't much of a reason for DVD to support 44.1 kHz audio.
ookzDVD
16th August 2002, 05:52
@all,
sorry, how about the _filesize_ issue ?
is there any ~8% size reduction from 48 -> 44.1 ?
DSPguru
16th August 2002, 07:26
Originally posted by DJ Bobo
@ DSPguru
equivalent in qualityi never read WaveLab's sources (thought they should be puclic, ain't they.. ?), but i doubt it's as good as shibatch's algorithm.
anyway,
i also doubt that your process of : Azid, WaveLab, RazorLame (+MP3GAIN?) takes faster then BeSweet's end-to-end process (azid.dll, shibatch.dll, lame_enc.dll), and i'm 101% sure that you'll get better quality with BeSweet.
lots of people that used to do a manual process are now using BeSweet, and all reoported speed imporvement and quality imporvement.
theReal
16th August 2002, 17:58
I've done a little comparison with 48kHz and 44.1kHz mp3 - if you're interested, read the results (http://www.computerfuzzies.de/temp/comparison48-44.pdf)
I guess I'll do this again some time with a different audio source, but I really don't expect the results to be that much different.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.