After playing with some of the "spectrum anyalizer" type visuals, I noticed that on a lot of my music, there was very little activity on the right side of the spectum where the higher frequencies are. I thought that I was hearing them but they wern't showing on the visual. Are the visuals ment to be very accurate? I decided to do a test. I used this fequency sweep to check the Spectral Plot visuals like Spectrometer 128. The results showed that the visual is not very acurate. Are others experiencing the same thing?

The Frequency sweep is located at:

http://www.ridges.dynip.com/images/response.mp3
Thanks,
Steve

using the 10 second linear sweep i found at: http://users.belgacom.net/dr.lex/macast/mp3test.html (rename .bin file to .mp3) it seems to sweep all the way accross the spectrum (stops about 1/2" from the right).

Yeah I noticed that, I think that maybe there needs to be a Log funtion for the upper frequencies. Also, if you pass a sinewave thru, it should only give one frequency spike at a time. There appears to be a bell curve of frequencies centered around the spike.
I'm not complaining at all, I didn't get the empeg to be a spectral analyser! The visuals are really only for show. and if we criticise too much Toby might take them away!

Murray 06000047
I don't think, therefore I am not.

Most mp3 encoders ie. xing throw out everything above 16khz, so you won't see the frequencies to the right side of the visual being lit up. I just looked at the file with encspot and I see it was created with xing. Try to find another test sweep or maybe make your own.

Sean

The spectrum analyser visuals are based on a linear fourier transform - The response of each 'spike' is therefore linear from about 40Hz at the lower end to 20kHz at the right hand side.

The loss of accuracy in the upper frequencies is certainly due to your mp3 compression. Later versions of the software will allow you to play wav files through your empeg, and therefore test the response of visuals without compression losses and artefacts.

Later versions of the software also include an improved fourier transform with a custom designed window that reduces the 'bell-curve' effect (although this is practically impossible to eliminate totally, and debatable whether it should be).

Looking at the 128 point fourier transform you will notice that treble response is actually very good, it is bass that is the problem >> Only the first 8 or so bins are used to represent frequencies from 40Hz to 320Hz, this is not nearly enough, and you'll notice that the lower end of the plot responds poorly with the average bin level being fairly high all the time.

I am currently working on an algorithm for calculating numerous linear fourier transforms and converting these into a single logarithmic transform. This will improve the bass response of the transform expanding it by around 4 times.

Noticed the same, this is how I found that 16KHz lowpass when using fraunhofer. Now I'm usin lame with the lopass switched off and hey, the visual goes all to the right! Positive side effekt, they sound much better (to me) since I turned off the lowpass while compressing. Negative side effet I had to redo all my CD's but this time RIGHT

I'm using LAME with audiograbber and didn't see a setting for that. Where do I turn off the lowpass?

Thanks,
Steve

Hi.

I don't exactly see an option to turn off the low pass filter, but you can either use the "-k" option to disable any filter (low or high pass), or use "--lowpass 22050" to set a low pass on the maximum frequency a CD could provide (44100 samples/second = 22050Hz Audio signal).

cu,
sven

proud MkII owner (12GB blue/green/smoked, was #080000113 is #090001010)