Saving as MP3 creates "fuzziness

[Soundoff]

Today I recorded my voice using a good microphone and GoldWave 5.5, saving the file in WAV. After a few minor edits, I used File--Save As and saved it as an MP3 at 128 kbps. No problem.

But when I played back the MP3, there was a slight, but noticeable, fuzziness to the voice. I played a short segment of the original WAV and the same segment in the MP3, and the difference was pretty noticeable. The fuzziness isn't bad, just a slight fuzz that you might not notice unless you compared the two. Plus, it might not be noticeable on music or other mass sound, but with a single voice it was pretty obvious.

So..... I downloaded the latest lame_enc.dll file, just in case, but that made no difference, saving as MP3 still added a fuzz to the voice.

Then I used Windows's own conversion tool to convert the WAV to MP3 at 128kbps and there didn't seem to be any added fuzz.

Back in GoldWave, I saved the original WAV file to a WMA format, and there was no fuzz in that.

I tried saving the WAV to MP3 on two different Windows XP machines using GW 5.5, with the same results. I'd attach two short samples, but.... there doesn't seem to be a way to attach files in the forum. Drop me a note if you're interested and want to hear what I'm talking about. I'm interested in hearing what could be producing this. Again, the fuzziness might not be noticeable with music or crowds, but it was with a single voice. Thanks,

(By the way, I'm not a frequent user of GoldWave, but I bought it almost 5 years ago and have always loved using it, I'm a big fan.)

[DougDbug]

Strange... The LAME encoder has a good reputation. But, if you're getting better results with the Windows supplied encoder, by all means, use it!!!!

128kbps should be high enough to get "transparent" encoding of voice, but you might need a higher bitrate (or higher quality VBR setting, etc.) Of course, MP3 is lossy compression and it can degrade sound quality. You can find some recommended LAME settings at HydrogenAudio.org. The guys at Hydrogen Audio have done lots of blind listening tests, and with high-quality settings the MP3 should sound identical to the original WAV file (with most program material).

What kinds of "minor editing" are you doing?

After a few minor edits, I used File--Save As and saved it as an MP3 at 128 kbps....

...Then I used Windows's own conversion tool to convert the WAV to MP3 at 128kbps and there didn't seem to be any added fuzz....

....Back in GoldWave, I saved the original WAV file to a WMA format, and there was no fuzz in that.

It sounds like you are comparing the original WAV file to the edited MP3 file? Of course, you need to compare the edited WAV with the edited MP3, etc.

[Soundoff]

Using GoldWave 5.55 on Windows XP Professional, dual-core processor, 3.5 GB RAM.
----
Thanks Doug. Yes, the few minor edits I made were on the original file that I saved as a WAV (added some silence at the beginning, fade-in, that's about it). After those minor edits, I then saved it as an MP3, so that file was exactly the same as the WAV. That let me go back and listen to the same segment of each to compare them.

I agree about the 128 kbps being of more than sufficient quality, and I've had great success on music files, etc, in the past. But with just the voice, that fuzziness was hard to miss. Typically, we should be able to drop the bit rate when it's just voice.

Your comments about the lame encoder remind me that I've never seen any detailed, lame-specific settings in GW that affect the way encoder works. It seems completely transparent; I don't even see a setting in GW that specifies which encoder to use. I just have that lame DLL file in the proper folder and GW finds it.

When I choose File--Save As to create an "MPEG Audio (MP3)" file, I can choose the typical bitrates, stereo, etc, from the drop-down list, or I can click the Attributes link to the left of that list. That brings up the MPEG Audio box which shows the attributes as I would expect, such as:

Sampling rate: 44100
Channels: Stereo
Bitrate range: 128000 [no range, just that one value]
VBR quality: Off
Presets [none]
MPEG Settings: [Only the Original option is checked]

Nothing weird about that, right? So where might I tweak any other lame settings, if there are any? Also, when GW downloaded the newer version of the lame DLL, the resulting MP3 file sounded the same as the first time I saved it using the older DLL; the voice still had that fuzziness to it.

I'd much rather save MP3 files in GW and not have to worry about this weirdness, so I'm very interested in finding out what's going on in this case. If you or anyone else watching this thread want to try it on your own systems, just record your voice for 5 or 10 seconds, save it as a WAV, then Save As an MP3 at 128 kbps. Then have both files open in GW and play one, then the other, then the first again, and see if they sound exactly the same. If they are the same, with no fuzziness, then this will be very curious, indeed. I need to dust my RAM or polish my hard disk.

Gene

[DougDbug]

I don't know if this will help.... Maybe the Windows encoder is taking advantage of some other MP3 options to get the most out of the given bitrate????

Is this a mono file? You might try mono or joint stereo instead of stereo. For a given bitrate Joint Stereo or Mono are the best, Dual Mono is the worst.

You can also try a variable bitrate of V5 or V6. Acording to Hydrogen audio, V5 or V6 should give you an average bitrate somewhere near 128kbps. (Set the minimum at 32,000 and the maximum at 320,000.)

For a given (average) bitrate, variable bitrate (VBR) gives you better quality than constant bitrate (CBR). For example, CBR uses the same bitrate for silence as for complicated musical parts.

With LAME's VBR, you select a "quality level" (V0 is best, V9 is worst), and LAME chooses bitrate moment-to-moment depending on the complexity of the sound. At a given setting, spoken voice will get a lower bitrate (smaller files) than music.

LAME also has another variable bitrate mode called ABR (Average Bit Rate), which allows you to choose the exact average bitrate, but GoldWave doesn't seem to access that option. (For most users the V-settings make more sense than forcing an average bitrate... Maybe most of your songs sound good at 128kbps, but some have artifacts and need a higher bitrate. With VBR, LAME can use a higher bitrate when necessary, and a lower bitrate when it can get away with it.)

[GoldWave Inc.]

There are a couple of things you can try. First, use Options | Plug-in | File | GoldWave* and try the "Very high quality" option and maybe even the "Joint stereo" option to see if those make any difference.

The other thing to try is to select MPEG encoding attributes with the letters "ACM" (if listed) when saving the file. That will use the Fraunhofer ACM MPEG encoder (if installed).

Chris

[Soundoff]

I tried the "Very high quality" option but I still hear some fuzziness when compared to the original WAV. I've now tried this on a second computer, my laptop, using the same version GW and Lame file. Same results, the MP3s are a bit fuzzy compared to the WAV. Both machines are Windows XP, one Pro the other Media Center (also like Pro, I believe. Oh, I played the MP3s on an MP3 player, and they sounded "fuzzy" there, too.

If anyone is interested in listening to the three files I've created on your own computer, they're available for download. Each is 7 seconds of voice, the beginning of a birthday greeting:

http://www.au2pb.com/files/GW-WAV.wav - (1.2 MB) The original recorded WAV file.

http://www.au2pb.com/files/GW-MP3-128.mp3 - (115 KB) #1 saved as an MP3, Layer-3, 44100 Hz, 128 kbps, stereo.

http://www.au2pb.com/files/GW-MP3-128-v ... uality.mp3 - (115 KB) Saved from #1 with the same MP3 settings as #2, except Goldwave's: Options | Plug-in | File | GoldWave was set to "Very high quality".

To me, #1 sounds fine, #2 has that "fuzz" to the voice, and #3 sounds more like #2 (MP3) than the WAV file. Any comments or suggestions would be appreciated. You can drop me an e-mail (I'm set up for "private" on the forum), or just post a reply here. Thanks for "listening".

Gene

[el_supremo]

I can hear the problem in your mp3 versions.
I loaded the wav file and saved it as Layer-3 ACM, 44100kHz, 128kbps, joint stereo and the result sounds much better than the other mp3 files.
The Fraunhofer ACM MPEG encoder was already installed on my system (Win XP Pro 32-bit).

Best Wishes
Pete

[Soundoff]

Okay, I did the same on my system, using Layerr-3 ACM, 44100 Hz, 128 kbps, joint stereo. The resulting MP3 sounds just like the WAV file, no "fuzziness." Then I saved the WAV as an ACM at 64 kbps, mono, and it still sounded like the WAV, no noticeable degradation, which is sort of expected for voice only.

Now I'm wondering what all this means. The lame encoder is the newest version. Is there any way to verify which MP3 encoder GoldWave is using, in case there are multiple ones on my computer? I don't see a menu option that displays which encoder will be used. Are there any downsides to using the ACM encoder vs Lame? I already know what one upside is.

Thanks for checking Pete, that's what was needed. If you feel like trying it again, if you have the Lame encoder on your system, you could save with it and see if it sounds fuzzy like mine did.

Gene

[el_supremo]

I played with the options described by Chris and setting "joint stereo" while leaving the quality at high produces a better sounding file.

Pete

[gravy6]

I have this same problem. If needed I do save with the Layer-3 ACM. But I've solved my problem by always saving in a higher kbps - 192 is my regular setting now. Nothing else seemed to work. I don't get the fuzziness with the higher quality settings.

[donrandall]

I listened to the first file. The only thing I noticed was a slight background noise which was easily and quickly and totally eliminated by the use of the Noise Reduction filter.

Now, just because there are some folks who occasionally use rather imprecise wording and somewhat mystical descriptions of their problems, I'll just have to ask a couple of questions: Is it possible that you are hearing that background noise? Is it possible that your use of the word "fuzziness" to describe what you are hearing is actually a reference to that background noise?

[Soundoff]

That's funny. Actually, though, "fuzziness" is pretty precise, in that speech gained a bit of fuzz around the sounds, something that might be harder to detect with music.

But remember that the first file was the original WAV file, right off the microphone, and it sounded fine. If there was any background noise, it was so slight as to be normally undetectable, at least on my system.

It was the second file that was simply the first file saved as an MP3 at 128 kbps, using the Lame encoder (most recent version). That's the file that took on the fuzziness.

Third file was also saved from first as an MP3, but using the Goldwave "Very high quality" option. To me, it sounds like the second file, still fuzzy.

Another possibility in all this is that it's actually the MP3 decoder that's adding the fuzziness when a file is played back. I've got a Linux machine in the closet that I ought to get running and play these files on it, would be interesting to see if they sound exactly the same as on my Windows machines.

[donrandall]

But remember that the first file was the original WAV file, right off the microphone, and it sounded fine.

Ahhh... Sorry about that. I didn't catch that the first time around. Anyway, the second recording did show some very slight degradation, but it didn't seem all that bad here on my stuff.

If there was any background noise, it was so slight as to be normally undetectable, at least on my system.

I measured it a about -52 dB. Not all that bad for a home recording, actually. The Goldwave Noise Filter cleaned it up right nice with no damage to the audio content. It sounded to me as though it may have been the fan in your computer that I was hearing.