Strange noises when recording to mp3

[lee321987]

   I'm getting noise whenever I encode to mp3, regardless of the bitrate.  The noise sounds like what I hear when I encode (on my computer) at very low bitrates (e.g. 8kbps).
   I have two Sansas - c240 and c250, and I'm only getting this effect on the c250.  This noise is not present if I encode to the other available formats (AIFF, WavPack, PCM).

   Some strange things I have noticed:
--  While recording on the c240, with Gain set to 0.0 dB, during silence the peak meter rests very close to the second dot (between the left and right meters), but if I turn the Gain UP to 6.0 dB the meter rests off the screen.
--  While recording on the c250, with Gain set to 0.0 dB, the peak meter rests between the 4th and 5th dots (this is regardless of the format I'm encoding to, but as I said, playback is noiseless if I use another format).

   Any ideas?

[Davide-NYC]

Can you attach an example MP3? I'm very curious to have a look/listen.

[lee321987]

   I tried, mp3 is not an allowed file extension.  I wonder if I'd be allowed to post an mp3 that I have changed the extention to .txt...?

[karashata]

The forum moderators don't like people doing that, the restrictions are there to save space and bandwidth.

Upload the file to a file-sharing site and post the link here.

[lee321987]

   Thanks.
   Ok, here's the link, you'll have to turn the volume up pretty loud to hear the noise, BUT BE CAREFUL --- THE FIRST SECOND AND LAST TWO SECONDS CONTAIN MY BUTTON PRESSES WHICH ARE MUCH LOUDER THAN THE NOISE!!!

http://www.mediafire.com/?ldhcrf9ntzw

[Davide-NYC]

Check out the weird DC offset. I've posted screenshots from Audacity showing the start and end of the file. Maybe this have something to do with the power section? (I don't know, just throwing ideas around) JHMikeS to the rescue?

[lee321987]

   I get the offset with both players, and in all formats (even with the OF).
   Good eye.
   Would you say my description of the noise ("like a VERY low bit rate encoding") is accurate?
   BTW this distortion is occurring even when the LCD is off.

[dreamlayers]

It effectively is a low bitrate MP3.  Sure, the frame size is 160 kbps, but if you look at the file in a hex editor you'll see most bytes are FF.  When I used mp3packer on it to losslessly compress it, the resulting file is 38,845 bytes (packed from 571,024) and EncSpot says the average bitrate is 10 kbps!

Note that the noise is very faint.  It's practically a straight line on those graphs and I need to turn up the volume to hear it.  My Gmini 220 running the original firmware has the same issue.  I wonder if MP3 encoders are meant to accurately encode low volumes.  Maybe the psychoacoustic model decides it's mostly inaudible, and it devotes very few bits to it, which results in what you see and hear here.

No idea about that DC offset.  I've never seen an MP3 with a DC offset.

[lee321987]

   Ok, I just did two more recordings - a 16 kbps, and a 160 kbps, and after using mp3Packer EncSpot reports 14 and 15 kbps (respectively).  Also I compressed a 160 kbps form my c240 (the "good" one), and it turned out 134 kbps.
   So the c250 is accepting my choice of bit rate, but only encoding 16 kbps worth of data...?
   Both of my players are using the same OF and the same version of Rockbox.

   What could cause this?

[Llorean]

Did you say you experience similar problems in the OF earlier?

[lee321987]

   Not the noise, just the DC offset, but the WAVs made by the OF and Rockbox do not contain this noise.

[dreamlayers]

So apparently the DC offset is actually present in data coming from the ADC and not from MP3 encoding.  Perhaps the DC offset is a problem for the MP3 encoder?

Edit: Just out of curiosity I generated a file with faint white noise and a DC offset and then encoded it with LAME 3.97.  The resulting file had irregular bass pulses (which were not present in the original) but the white noise sounded otherwise okay.   I think I made my noise more intense than your noise.  Go ahead, try to encode those WAV files with LAME and see how it messes them up.

[lee321987]

   Ok, I think you're right.  Remember I said the c240 doesn't produce this noise?  I just compared (with Audacity) the size of the offset on both players.  On the c240's file the top of the (noise only) part of the wave form was just about at 0.  The top of the (noise only) part of the c250 file's wave form was at about 0.05. 

   What exactly is this number (the scale just to the left of the wave form in Audacity)?  Some variation of dB?

   Would it even be possible to program a correction for this offset into Rockbox?

   Thanks for the help guys,

Will

[Davide-NYC]

   Not the noise, just the DC offset, but the WAVs made by the OF and Rockbox do not contain this noise.

Since the noise is very low DB and this does not happen with WAV files I'm going to presume it is caused by the encoder. The DC offset, on the other hand, I would presume is attributable to the hardware, since it appears with both the OF and RB.

Some questions just to clarify exactly what is going on:

• Do WAV files have this DC offset as well?
• Does this offset exist on both of your C2x0 units?
• Does this offset happen when recording with Sansa E2x0 units?

[Llorean]

The Rockbox MP3 encoder is, if I recall, fairly low quality. It was chosen because it's fast, rather than being tuned for quality, and should only be used if you absolutely do not have the space for higher quality file types.