Subject:Saving to MP3 causes echo effect

Posted by: Velcro Face

Date:11/21/2008 10:45:01 PM

In my work as a consultant, I record Skype conversations with my clients with a separate track for each side. I then combine them into a stereo file, which I process with the Channel Converter, blending 25% of the left side into the right, and vice versa. I find that this gives the conversation more "presence," as if it's happening in the same room, rather than at two distinct locations.

The problem is that, when I save it to an MP3 file, I end up with a slight reverberation effect, as if there are two copies of myself speaking the same words at the same time. This is especially pronounced at bitrates under 128 K. It also occurs if I save it as a WMA file. Simply downsampling the WAV file doesn't cause the reverb effect. Any ideas why this echo occurs and how I can avoid it? The obvious solution is to save it at a higher bitrate, but for an hour's consultation that results in a pretty large file.

--
Michael

Subject:RE: Saving to MP3 causes echo effect

Reply by: musicvid10

Date:11/21/2008 11:39:51 PM

If you're going to save anything in a compressed format under 128k, there are going to be some audible artifacts. You can have little files, or quality, but not both.

Do you really need stereo? Voice recordings at 16 bit 22khz Mono have given me hours of good archives (of seminars, etc.) without the need for large file sizes.

That being said, do a test render to AC3 stereo. I have absolutely no idea how it performs at low bitrates, but it's worth a try.

VBR and/or 'joint stereo' Lame Encoder rendering of your mp3's might give you a little more bang for your kb, too.

Subject:RE: Saving to MP3 causes echo effect

Reply by: Chienworks

Date:11/22/2008 7:21:01 AM

I save voice recordings as mono MP3 at 24kbps and the result is very clear and listenable. Yes, it's definitely distorted compared to the original, but not enough to make it unusable. There's no echo effect at all. At 48kbps voice sounds almost the same as the original. For a comparison, i have a Vonage voice line and highest quality uses 90kbps, medium uses 50kbps, and low uses 30kbps. All three settings sound very acceptable.

If you're getting the echo effect then something else is going on. Is there any chance that you don't have the two channels perfectly aligned after doing the 25% blend? Try the same MP3 saves without the blend, and also as a mono file. If these sound ok then it's your blending process that is causing the problem, which may not be apparent until after MP3 compression is applied.

Subject:RE: Saving to MP3 causes echo effect

Reply by: musicvid10

Date:11/22/2008 7:38:17 AM

I can always hear slight flanging from compression artifacts at 64kbs mp3, especially with stereo headphones. Possibly that is what the OP is hearing. As Kelly said, this is not terribly objectionable for speech-only recordings, but I generally don't go that low if it's something I plan on keeping or sharing.

Subject:RE: Saving to MP3 causes echo effect

Reply by: Velcro Face

Date:11/22/2008 10:12:38 AM

Apparently the artifacts introduced by compression are the source of the problem. The audio sounds fine as a low bitrate WAV file, even when both sides are blended. Why this should happen, though, is a mystery to me.

I'll try other formats, but I've got to distribute the file as a common enough type. Perhaps the only solution is to forget about blending the channels... or reduce it to mono. Thanks for your thoughts.

--
Michael

Subject:RE: Saving to MP3 causes echo effect

Reply by: musicvid10

Date:11/22/2008 3:31:31 PM

Changing to mono will save you a lot of space . . .

Subject:RE: Saving to MP3 causes echo effect

Reply by: rraud

Date:11/23/2008 2:43:34 PM

"Changing to mono will save you a lot of space"
Well... maybe.. but not like a .wav file, where the file size would be halved rendered to mono.
A 64kbps mono mp3 would have pretty much the same quality as a 128kbps stereo, all other factors being equal, but mono of course.

Subject:RE: Saving to MP3 causes echo effect

Reply by: Chienworks

Date:11/23/2008 3:20:17 PM

A 64Kbps mono file would be about half the size at about the same quality as a 128Kbps stereo file. You would save as much space this way with MP3 as you would with WAV.

Just keep in mind that a 128Kbps mono file will be the same size as a 128Kbps stereo file, but presumably much better quality.

Subject:RE: Saving to MP3 causes echo effect

Reply by: musicvid10

Date:11/23/2008 5:45:19 PM

Right -- 128kbs mono is still 128 kbs. So the trick is to halve the bitrate each time I halve the source size.

If I start with a 40MB .wav at 16/44 stereo, and render a 16/44 .mp3 at 128kbs stereo, I end up with about 3.64MB (11:1 compression).

If I render the same .wav to 16/44 at 64kbs mono, I've got half the file size with the same quality (as you both mentioned), or 1.82MB.

If I go one step further and halve the sample rate, which is perfectly OK for speech, I end up with an .mp3 at 16/22 32kbs mono with almost no discernible loss and a 910KB file. Now I've got 30 minutes of speech per1 MB of disc space with very good, but not perfect quality. Of course, if any of these archives needs to go onto a CD, I have to upsample it again to 44.1khz, which could introduce some noise.

Btw, I never go below 16/22 for archives, because the sound quality goes down the tubes at that point. To answer this post, I had to dig into my old archives to see what I'd actually done. I think I've got 40 gigs of this stuff stored by now, which will probably never be listened to again . . .

Subject:RE: Saving to MP3 causes echo effect

Reply by: Velcro Face

Date:11/24/2008 11:20:13 AM

I tried VBR compression, and the results sound much better. For a file size only slightly larger than that of a CBR 64K MP3 file, it sounds comparable to a CBR 128K MP3 file.

--
Michael

Subject:RE: Saving to MP3 causes echo effect

Reply by: musicvid10

Date:11/24/2008 1:28:42 PM

Glad to hear it.
If you need stereo, also try "joint stereo" when encoding if it is available.
It is supposed to work better on low-bitrate encodes, by not duplicating material that is common to both channels.