Skip to main content
Topic: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12? (Read 5795 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #25
Undigging the topic, just want to add .02
I guess what was confusing me was how the WMP mp3s actually retained that >20.5kHz data, when that typically is not be expected, when both the WMP mp3 and Foobar2000 mp3 have about the same file size.
I think that's not retaining the data, but... clipping. Those frequency "peaks" go up to the Nyquist frequently, which is really unusual in correctly mastered audio. They just look like clipping.
Every lossy encoding, including MP3, more or less slightly changes peak level. If in the original audio peaks were close to the full scale, there's high risk that MP3 encoding will clip them! Probably that happened with WMP's encoding.
sox -e float -b 32 -V4 -D gain -3 rate -v 48000 norm -1
opusenc --bitrate 128

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #26
mp3 allows encoding sample values above full scale and they don't have to be clipped. it's up to the decoder and the player to deal with them in some way.
Keep calm and opusenc --bitrate 128

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #27
mp3 allows encoding sample values above full scale and they don't have to be clipped. it's up to the decoder and the player to deal with them in some way.
But spectrum analyser acts like a visual "player", so unfortunately it has to handle clipping just like a regular audio player - chop off peaks that are over.
sox -e float -b 32 -V4 -D gain -3 rate -v 48000 norm -1
opusenc --bitrate 128

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #28
No it hasn't. There are certainly some that don't.
Keep calm and opusenc --bitrate 128

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #29
just an slgiht off topic advise

If you are going to use  CBR 320kbps MP3  encoding please do yourself a favor and stop wasting space on padding
Use m3repacker on it to remove the useless padding.
Sven Bent - Denmark

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #30
But spectrum analyser acts like a visual "player", so unfortunately it has to handle clipping just like a regular audio player - chop off peaks that are over.
When you're retrieving samples from some sort of acquisition device, non-sensical data (such as floats at infinity), are simply labeled as such. You can then decide to accept them assuming the maximum value after some conversion, or disregard them altogether.
In places like TV studios, clipping in audio is dealt with in a number of ways. Some filter to zero (for instance when someone drops a microphone and it breaks), others are allowed to be a bit more forgiving (such as someone touching their lapel mic). This is often done automatically, such that no annoying signal is encoded or sent through to the viewers.
The kind of spectrum analysers that come with audio players, are for little more than entertainment. Reminiscent of consumer equalizers with a bunch of cool blinking lights, or little apps making a voiceprint.

Having said that, spectrum analysers and waterfall diagrams are used all the time with Ham radio.

You can't generalize how it has to be dealt with or displayed, etc. Depends on the application.

 

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #31
No it hasn't. There are certainly some that don't.
You can't generalize how it has to be dealt with or displayed, etc. Depends on the application.
Please show me some spectrogram software that doesn't clip the samples. Yes, I know that spectrogram software can deal with clipped audio in other ways than chopping off the top (e.g. reducing the gain, limiting etc. and then performing analysis), but I think they usually chop off clipped peaks.
sox -e float -b 32 -V4 -D gain -3 rate -v 48000 norm -1
opusenc --bitrate 128

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #32
Please show me some spectrogram software that doesn't clip the samples.
Izotope RX2 and RX5 do not clip when calculating the spectrogram or the spectrum analysis graphs. If you boost a sine wave, it shows a positive peak and no harmonics. In the amplitude range of the spectrogram one can zoom out to +10. Spectrum Analyzer in Sound Forge doesn't clip. In Cool Edit Pro 2.1 neither the spectrum analyzer, nor the spectrogram clip.

However, the programs do not load compressed files to float. It has to be PCM input.

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #33
Please show me some spectrogram software that doesn't clip the samples. Yes, I know that spectrogram software can deal with clipped audio in other ways than chopping off the top (e.g. reducing the gain, limiting etc. and then performing analysis), but I think they usually chop off clipped peaks.
Baudline will display them as "errors" (anything that is not a regular double float will me marked as such).
You can mask them out, depending on the format.

GnuRadio will set 'NaN' when the signal of your SDR is overpowering the input stage, but that depends on the chipset.

Re: Which has a better MP3 encoder, Foobar2000 or Windows Media Player 12?

Reply #34
OK, I was half right and half wrong.
I was wrong saying that most spectrogram software can't handle clipped audio - they can without showing harmonics (tested with clipped 32-bit float track on Audacity and Spek).
However, I was right saying that clipped audio and lossy codecs aren't the best combination - I fed the exact 32-bit float track I used earlier to LAME, then I decoded it and resulting WAV was 16-bit PCM (so it had to be clipped), so I tried loading MP3 directly to Audacity and Spek (spoiler: not very effective).
However, the programs do not load compressed files to float. It has to be PCM input.
It seems that j7n was right from the beginning :)
sox -e float -b 32 -V4 -D gain -3 rate -v 48000 norm -1
opusenc --bitrate 128

 
SimplePortal 1.0.0 RC1 © 2008-2019