Hi all,

I'm hoping someone out there has this CD / lossless rip, and can spare a little time:

If you do, I'd be very grateful if they could either:

1) Upload a small (< 30 secs) segment (in a lossless format) of "Don't Go Home With Your Hard-on", or
2) Show a spectral analysis of the track.

This is in relation to this thread.

Many thanks in advance.

C.

This is from a FLAC file from a borrowed CD. From memory this was the original CD release, i.e. not a reissue.

EAC spectral analysis; no obvious signs of lowpassing:

Excellent, thanks Nerkenheimer, and welcome to HA.

I wonder if the resident LAME experts would know if LAME 3.98 at ABR 256 would likely change this spectrum to this.

C.

I'm hardly an expert, but - yes, pretty much.

Here's the next best thing to an audio snippet, which I'm a bit reluctant to send.

Firstly, the LAME output screen to ensure we're roughly on the same page...

...and the EAC spectral view from the MP3 decoded to WAV:


Compare and contrast with the ex-lossless one I sent earlier. As you can see it has the characteristic MP3 16kHz lowpass - though in this case there isn't much above 16k to chop anyway.

I can't purchase MP3s from amazon.com (I tried once and they h8 furriners, it seems), so I'm wondering: what MP3 encoder do they use, and what settings? You could probably answer this with EncSpot and/or Mr QuestionMan.

Amazon seems to use one of three settings:

1. Some version of the FhG encoder at 256kbps stereo
2. Lame 3.97 at -V 0 using joint stereo
3. Lame 3.98.2 using abr 256

Amazon doesn't stick to just one setting with their mp3 files. This is really my only complaint about their service as the OCD part of me wants all my downloads to be encoded using the same encoder and setting.

Yeah, that's ideal - very similar to my spectrum. You've basically answered my question. That just leaves Kornchild's question (which I'm interested in), namely why does LAME do this (i.e. low pass at 16kHz) when -V 4 has a higher low pass, yet will (normally) yields lower bitrates?


With this particular file they used LAME 3.98b ABR 256 (Encspot reports 250 kbps - but I have a feeling Amazon target 256 on ABR - this one just required less bits), but as Kornchild stated it varies.

Big thanks for solving the issue.

It's odd though, as on the DOS screen shot it states (as Encspot did too) that the low pass filter is c. 19700 Hz.

C.

Could anyone verify for me that the little bit of signal above 16 kHz is probably dither applied during decoding? I notice that it is strongest at 22 kHz and gets weaker as you approach 16 kHz. The only part that I can't explain is the increase in high frequencies at the end as the music is fading out.

Notice too, that this isn't on the Amazon download.

C.

Admittedly, I chose a song that'd provide SB21 bloat for this demonstration.
Lossless:

LAME 3.98.2 --preset 256:


If this were actually a lowpass at 16 kHz, there'd be virtually nothing above 16 kHz.
I still think that the 16 kHz dropoff is LAME not wasting bits on SB21 unless actually necessary.
EDIT: And another screenshot of the mp3. Note that LAME reports "Using polyphase lowpass filter, transition band: 19383 Hz - 19916 Hz"

Another edit:
The song is Red Hot Chili Peppers - Snow ((Hey Oh))

This could be consistent with abr behaving much as though -Y was enabled, as noted in another thread.