Vinyl to CD - again

[johnb]

I think you mean 44.1 ?

(sorry for the pedantry....................)

Whoops, slip of the fingers. Now corrected.

[David-G]

OldTechie, Johnb and McGG, thank you all very much for the helpful information. Very enlightening.

[Gordon]

Once again many thanks all for input. My friend has retired to consider - more complicated than he thought [as I had said] and he may have to temper his love of vinyl - he's afraid they are wearing out from playing which is to some extent true!

I used to have a Yamaha HD1000, a previous version of the later 1300 and 1500 models. Good machine on which I did all my vinyl transfers and the results sound fine to me. Sadly it died [no, not the hard drive] and was uneconomic to repair. Even though I have kept some of my vinyl back and, having backed them up, have used them to demonstrate to my friend what those results were - still thinks he can hear a difference!! I can't say I can. I'm afraid I cheated a bit and made one CD by going to MP3 at 160kBit/s and then back to wav - he could not tell the difference between the direct to CD and that via MP3 which makes me a bit dubious of what he thinks he hears but - mum's the word!!

If I had to do this again I would probably go for something like the TASCAM DR-100 costing a few hundred pounds. It's a battery/mains portable recorder with mics built in and records to an SD memory card and will do 96/24 coding among other standards. It has good meters too. But like some other machines it's spec is not sufficiently comprehensive eg about its clock stability. Ideally we'd be looking for something around 100 picoseconds RMS jitter in 44.1 kHz for 16 bits - and 256 times better than that for 24 bits - that's about 2 parts per million which a good crystal oscillator should be able to do. This level of performance is needed primarily for large amplitude, high frequency audio signals - if we were to only be concerned about average levels [say -10dB on max] and frequencies in the main musical spectrum [ie fundamentals and a few harmonics] then this tolrance could be reaxed a lot and probably is in run of the mill designs. The jitter spec is proportional to both amplitude and frequency and also is logarithmically dependent on the number of bits.

As for the post processing software I can find very little so far [have not had that much time to do the research] that tells how the filtering to 44.1 is done. From 96 to 48 it is simple but from 48 to 44.1 it is not trivial because of the odd ratios. One of the advantages of oversampling is noise shaping because it spreads the quantising distortion over a wider bandwidth so I'd like to know how they would do noise shaping from 24 to 16 bits which is not disclosed either.

[MrGongGong]

As for the post processing software I can find very little so far [have not had that much time to do the research] that tells how the filtering to 44.1 is done. From 96 to 48 it is simple but from 48 to 44.1 it is not trivial because of the odd ratios. One of the advantages of oversampling is noise shaping because it spreads the quantising distortion over a wider bandwidth so I'd like to know how they would do noise shaping from 24 to 16 bits which is not disclosed either.

I would ask the question on the CEC list (Canadian Electroacoustic Community) as it's full of folk who know this stuff.
I nearly bought at Tascam DR-100 but was more convinced by the sadly discontinued Sony PCM D50..........

[johnb]

Gordon,

If you want more control over resampling and bit-depth options it might be worth looking at SOX, the well regarded free command line utility:

Resampling options: http://sox.sourceforge.net/SoX/Resampling

It also has extensive dithering options.

SOX is a bit daunting but there is full documentation online and, of course, it can be used in a batch file (which makes it easier to use and easier to test out various options).

[Gordon]

I would ask the question on the CEC list (Canadian Electroacoustic Community) as it's full of folk who know this stuff.
I nearly bought at Tascam DR-100 but was more convinced by the sadly discontinued Sony PCM D50..........

Thanks. Neither of these machines have detailed specs. The Sony [spec -what there is of one - here: http://www.sony.co.uk/support/en/pro...CM-D50/manuals is a very similar machine to the TASCAM [spec here: http://tascam.com/content/downloads/...100_OM_web.pdf.

At LPCM-24 bits neither of the S/N specs are that impressive although by analogue standards of course they are, but the spec doesn't meet what one would expect from 16 bits well done [ca -107dB wrt full scale unweighted]. If the S/N is determined only by the digits then these figures, esp if they are weighted [93dB, IHF-A weighted as is claimed by Sony so unweighted the figure is even worse by several dB], should be better. But they don't declare their reference level which I assume to be full scale or something close to it.

It implies that the analogue circuits are setting performance thus masking defects in the digits, in which case I would doubt the quality of the LSBs and/or the sampling clock stability as I have been bleating on about. Looking also at the distortion spec at 24 bits these are not what one might expect from a highly linear ADC/DAC pair either. Having said that, we would have taken both our arms off for machines like these in the old analogue tape days!! And, anyway, if we are dubbing from LP whose surface noise and dstortion is way above the noise floor of both these machines why should we worry?

[MrGongGong]

There are some very detailed tech reports on most of these machines if you dig a little .........

Solid State Sound (http://www.solidstatesound.co.uk/) are helpful
and there are some things here on the excellent London Sound Survey site

http://www.soundsurvey.org.uk/

There is a site I have used (but my brain has failed today ) that gives very detailed and comprehensive specs etc I think it might be found via this ?

http://www.musicofsound.co.nz/blog/

(also excellent IMV)

[Gordon]

....It also has extensive dithering options.

Thanks for links, interesting. But who needs dithering when you have a sloppy clock and/or a noisy input, way worse than a less than perfect digital conversion process??!!??

Doing a 48 kHz to 44.1 conversion properly of course needs the calculation/interpolation of 441 output samples in the same time period as the input 480 and that is a bit ofa bind. BUT...one horrible, horrible way to convert simply from 48000 to 44100 Hz is to lose 39 [ie 480-441] samples out of each 480 to leave 441 and then s-t-r-e-t-c-h those 441 to occupy the correct time!! Would anyone notice???

[OldTechie]

As for the post processing software I can find very little so far [have not had that much time to do the research] that tells how the filtering to 44.1 is done. From 96 to 48 it is simple but from 48 to 44.1 it is not trivial because of the odd ratios. One of the advantages of oversampling is noise shaping because it spreads the quantising distortion over a wider bandwidth so I'd like to know how they would do noise shaping from 24 to 16 bits which is not disclosed either.

A few references and thoughts:

• Comparative measurements: http://src.infinitewave.ca/ (Look at the Sadie results. Did you sell them your #38 system? Maybe Sadie adds the dither for ready for 16 bit conversion - it seems to be the right sort of level.)
• He's sure he's better than everyone else: http://www.weiss.ch/p2d/p2d.html ($805 from Vintage King)
• SSRC uses FFTs: http://shibatch.sourceforge.net/
• Secret Rabbit Code is a converter, and the site seems to have some glimpses of info: http://www.mega-nerd.com/SRC/
• I have seen posts where people claim that SoX sounds better than SSRC - but no ABX tests
• dBpoweramp uses SSRC and is easy for simple folk like me to use. It has a Windows menu extension so you can select a group of files in Windows explorer, right click and tell it to convert them and it will launch as many processes as you have processor cores on your machine, and convert all the files as fast as you machine can do them.
• foobar2000 (http://www.foobar2000.org) can convert files but by default uses the PPHS converter - this is chosen for minimal processor usage and reasonable conversion which is what you want if you need a playback conversion on a limited spec machine. However, it there is a SoX plugin for it (see the SoX site). I've not tried the SoX route.
• I think I've resolved how to do ABX tests to prove I really can hear the difference between the 24/96 and 16/44.1 versions of my Mahler 4. The problem with ABX tests is you have to repeat them at least 16 times, preferably 20 to get reliable statistics. I think the problem is rapid learning of what the correct sound is, so your brain fills in the missing bits for you on a repeat. I noticed that I always had a correct answer on the first test, and it rapidly deteriorated. By doing only one or two tests per day I get more reliable results. I got it down to it telling me there was less than 1% chance I was guessing before I accidentally closed the application - I'll have to start again!
• I've converted my 24/96 to 16/44.1 using dBpoweramp and compared it with the commercial conversion - and I think I prefer it, or maybe I can't really hear any difference. I definitely do not think it is worse. So if your friend is happy with commercial CD's, I'm sure he would be happy with SoX or SSRC conversions. (Of course if he does not like commercial CDs, it could be that he just likes the resonances in his cartridge, pre-amp mismatches etc., in which case he'll still love the conversions so long as you don't fix it.)

[David-G]

Thanks for links, interesting. But who needs dithering when you have a sloppy clock and/or a noisy input, way worse than a less than perfect digital conversion process??!!??

Doing a 48 kHz to 44.1 conversion properly of course needs the calculation/interpolation of 441 output samples in the same time period as the input 480 and that is a bit ofa bind. BUT...one horrible, horrible way to convert simply from 48000 to 44100 Hz is to lose 39 [ie 480-441] samples out of each 480 to leave 441 and then s-t-r-e-t-c-h those 441 to occupy the correct time!! Would anyone notice???

What happens if you just interpolate each sample? In other words, identify the 48K samples on either side of the new 44.1K sample, and interpolate? Is this not rather like converting a photograph from 4800 pixels to 4410?

[clive heath]

In the end it's what you are happy with that matters and whether diminishing returns set in when you try to improve. With a Thorens TD 165 deck and an Ortofon cartridge (MC15 super II) together with a pre-amp of my own design, I have achieved what seems to me to be a more-than-acceptable quality but I am happy to be disabused of this notion if listeners can pinpoint deficiencies. I tried the Behringer for recording but as suggested that was problematical so the built-in Realtek was used although the Behringer is fine for replay.The uploading has been MP3 320 kBit/s but could be WAV files. No rock, I'm afraid but Karajan's Mahler 4 is there and Munch's Saint-Saens Organ Symphony which is a sonic delight, airy acoustic, spacious recording, brilliantly played by the Boston Symphony (and several other major recordings).

http://www.cliveheathmusic.co.uk/vinyl.php

[Dave2002]

What happens if you just interpolate each sample? In other words, identify the 48K samples on either side of the new 44.1K sample, and interpolate? Is this not rather like converting a photograph from 4800 pixels to 4410?

Yes and no! In the film/video/TV world different approaches can be taken for dealing with the differences between 25 fps and the until recently standard for film of 24 fps, and the frame rate of 30fps for US TV. In the UK one approach is simply to play the film frames slightly faster, and there will be a slight rise in pitch of the soundtrack, which most people, unless they have really good absolute pitch, won't notice. This doesn't work for display on 30 fps TVs (US), so an approach of dropping frames is used. For analogue TVs (now largely defunct) some frames had the interleaving done with the odd lines from one frame, and even lines from an adjacent frame.

For audio there is always a slight degradation of sound quality when converting from 48kHz to 44.1kHz, but if done well this can be almost imperceptible. If done badly then significant artefacts are generated, though even then this might not be noticed by many. The artefacts might be shifted to higher frequencies, where they may be inaudible, but if further (often digital) processing is done they may inadvertently be brought back into the audible region in an undesirable way. There are also differences between real time interpolation and off-line interpolation, where data can be stored and really good algorithms used. Real time interpolation is needed for transmission systems with very low delay. CDs, DVDs etc. can all be processed off-line in the manufacturing process.

[MrGongGong]

The artefacts might be shifted to higher frequencies, where they may be inaudible, but if further (often digital) processing is done they may inadvertently be brought back into the audible region in an undesirable way. .

Pauline Oliveros writes about some of this in Software for People
describing how she taken to task for using inaudible frequencies but was insistent that by having them beat against each other their effects were perceptible


(a bit OT 0

[Dave2002]

In the end it's what you are happy with that matters ...

Up to a point, and it also depends on what you mean by happiness. If the processing introduces problems which are not apparent to the principal user then that can be fine, but if that user then plays the results to others who may be able to hear unwanted high or low frequencies, distortion, or other problems, then their discomfiture may reduce the happiness of that principal user.

[johnb]

As far as the performance of various resampling programmes it is interesting to compare the them using the link given on the SoX page I referred to earlier: http://src.infinitewave.ca/

There are a great many programmes, and options within the programmes, compared and it is evident that some are markedly 'cleaner' than others (as shown in the Sweep and 1kHz Tone graphs). To what extent the differences are audible is an entirely different kettle of fish! (Incidentally SoX seems to come out pretty well, especially with the "SoX 14.4 VHQ Linear Phase".)