Berg/Beethoven Violin Concertos (Faust/Abbado)

[thrand1]

Not being familiar with the Berg piece, I purchased this primarily for Faust's highly regarded performance on the Beethoven. Excellent sound quality that really enables new observations on what may be an already familiar piece. Downloaded from eClassical, but also available at Qobuz.

 

foobar2000 1.3.2 / Dynamic Range Meter 1.1.1

log date: 2014-12-19 22:37:05

 

--------------------------------------------------------------------------------

Analyzed: Faust, Isabelle, Claudio Abbado, Orchestra Mozart / Berg & Beethoven: Violin Concertos

--------------------------------------------------------------------------------

 

DR Peak RMS Duration Track

--------------------------------------------------------------------------------

DR14 -0.01 dB -19.37 dB 11:51 1/5-(Violin Concerto - 'To the Memory of an Angel') - I. Andante - Allegretto

DR13 -0.01 dB -18.60 dB 16:14 2/5-(Violin Concerto - 'To the Memory of an Angel') - II. Allegro - Adagio

DR13 -0.01 dB -19.15 dB 22:56 3/5-(Violin Concerto in D Major Op. 61) - I. Allegro ma non troppo - Adagio

DR17 -1.27 dB -25.73 dB 9:21 4/5-(Violin Concerto in D Major Op. 61) - II. Larghetto

DR12 -0.01 dB -17.63 dB 8:34 5/5-(Violin Concerto in D Major Op. 61) - III. Rondo allegro

--------------------------------------------------------------------------------

 

Number of tracks: 5

Official DR value: DR14

 

Samplerate: 44100 Hz

Channels: 2

Bits per sample: 24

Bitrate: 1400 kbps

Codec: FLAC

================================================================================

[Jay-dub]

Here's a screenshot of a spectrogram (2:57 to 3:05 in the slow movement of the Beethoven).

Lines just below the top of the diagram (the Nyquist frequency that move in mirror image to the other harmonics of the violin note are a sure sign of aliasing. In this screenshot you clearly see aliasing down to about 21.4 kHz, the position of the marker, or a little lower. This is typical of recordings made with an ADC chip that uses a half-band antialiasing filter. If you play it back on a standard DAC with its half-band interpolating filter, you will keep that aliasing, plus some above the Nyquist frequency. I like my music to be completely free of aliasing, so on the basis of this spectrogram, I put the recording through a 20.6 kHz lowpass with the top of the transition band at 22.1 (sox sinc -a 60 -t 3000 -20600). I find that this kind of processing makes the sound much more analogue-like, but I have to find a different cutoff for each recording.

[astrotoy]

I'm interested in what you are doing, but I have no idea how to do it. What software do you use to do this or is there a setting on my DAC - I have both a BADA2 and an Exasound E28? I also have Pyramix and Izotope RX3Advanced, but I only use them for specific purposes. Also, I looked up aliasing in wiki, but it was fairly difficult for me to understand exactly what it does. Is there a simple explanation somewhere?

 

Thanks, Larry

[Jay-dub]

This spectrogram was made by Sonic Visualiser. Here's what's going on in here. A solo violin is the easiest instrument to use for detecting aliasing on a spectrogram; that's why I zoomed in on one. If the fundamental frequency of the violin note is x, then the overtones are 2 x, 3 x, 4 x, in general n x. The 44.1 kHz sampling makes for aliased overtones at 44100 - n x. When the violin notes go up, all the overtones go up, but the aliased overtones go down. So that's what I look for in the spectrogram. I noticed aliased overtones down to 21400 Hz. This is a specific artifact that is unlike anything close to the audio band in analogue or high sample rate recordings. Now, some ARSC's may roll off around here, but most traditional DAC's will preserve all the content up to the Nyquist frequency: that's what "half-band" interpolating filters do. Even though some DAC's have user-selectable custom filters, I strongly doubt that you want to change your DAC's settings from one album to the next. I find it easier to just store a back up of the files that I ripped (or in this case downloaded) and make a new version that I keep in iTunes. I just use the free command-line tool SoX, though I'm pretty sure there are better tools out there. I ran the following Ruby script to do the processing on all files in the album simultaneously:

 

thr = []Dir.mkdir "resampled" unless File.directory? "resampled"
Dir.foreach(".") do |x|
 next unless x.match(/aiff?\Z/)
 thr << Thread.new do
   system("sox",
           x,
       "resampled/"+ x,
       'gain','-0.1',
       'rate','-v','-s','48000',
       'sinc','-a','60','-t','3000','-20600')
   puts x + ": done"
 end


end
thr.each { |i| i.join }

 

The spectrogram of the same point in the processed files looks like this:

 

It will play back with the same gentle rolloff and freedom from aliasing regardless of the exact filter on your DAC.

 

BTW, if you went to college with Robert Levin, that means you were also at school with my dad. Small world.

[astrotoy]

BTW, if you went to college with Robert Levin, that means you were also at school with my dad. Small world.

 

Thanks for the explanation and the help. I will try to figure this out. I'm glad you didn't say that I went to college with your granddad!

 

Larry

[Jud]

In this screenshot you clearly see aliasing down to about 21.4 kHz, the position of the marker, or a little lower. This is typical of recordings made with an ADC chip that uses a half-band antialiasing filter. If you play it back on a standard DAC with its half-band interpolating filter, you will keep that aliasing, plus some above the Nyquist frequency. I like my music to be completely free of aliasing, so on the basis of this spectrogram, I put the recording through a 20.6 kHz lowpass with the top of the transition band at 22.1 (sox sinc -a 60 -t 3000 -20600). I find that this kind of processing makes the sound much more analogue-like, but I have to find a different cutoff for each recording.

 

I just get a general feeling from some of what Miska has said that one of the things most or all of his filters do is help with this same problem.

[Jay-dub]

I just get a general feeling from some of what Miska has said that one of the things most or all of his filters do is help with this same problem.

 

No doubt. And of course the filter that I chose (sinc -a 60 -t 3000 -20600) is just a very generic filter that gets the job done. I strongly doubt that anyone will find it horrible sounding, but it isn't necessarily optimized to human hearing quite as much as Miska's filters. In particular, many will prefer apodizing filters, which you can try in SoX by using

sinc -a 60 -I -t 3000 -20600

or

sinc -a 60 -M -t 3000 -20600

instead.