Digitising LPs - a question for Graham

Gary,

not to put the dampener on your technical enthusiasm at all, but my reference for improved quality of vinyl digitisation is the original recording.

If I play both on my system I can tell how good my rip (and my DAC) are.

I am not sure how one might use spectral analysis for relative SQ analysis.

If I cannot hear an improvement when I play back and compare to the original, then, to all practical purposes there is none.

If it sounds better (than the original) then there is a problem 

Have fun anyway.

tg,

 I agree, based on my experience digitizing audio over the past 10 years or so, obtaining an acceptable input signal for recording is the greatest challenge. I have worked with tapes, cassette and reel to reel; and vinyl using various ADCs with a constant issue related to input signal level and noise.

 My experience has convinced me, it is an impedance mismatch issue between our phono preamp outputs and whatever type of ADC we connect with. Your point about the buffer is a good approach to try. It could provide the impedance match or correction allowing signal from your phono to fit, so to speak, nicely into the window provided by your ADC, therefore cleanly digitizing the signal maintaining minimal loss.

 I have been trying to find the matching capability with the output of the phono preamp. What we do, digitizing vinyl, does not generate enough interest in the marketplace for a vendor to spend their resources developing products specifically for our application, especially within your “enthusiast level” focus. Therefore, it is up to us to find the solution and apply it.  Just look at the compromises we make to do this, yes I appreciate the challenge, but I also have years’ worth of vinyl to deal with.  I have modified my phono preamps over the years as my ADCs have changed, but I’ve never felt confident I ‘m getting the best sound.

 I have been very interested in Graham’s discussions concerning a phono preamps design and the theory behind it. His insight has helped answer numerous questions I have had over the past years concerning problems with level matching between my phono output and the ADC.

 I look forward to discussing this down the road.

Tom

tg wrote: In perusing and experimenting with various methods of doing this, I have occasionally come across the idea (which I have not tried) of recording direct without RIAA equalisation and then applying that equalisation in the digital domain. I would be interested, as I imagine, would be others, in Grahams thoughts on this approach and its merits or failings.

The output of a magnetic cartridge as established already is a rising one where the highest frequency in the accepted audio range - 20Hz to 20kHz has 1,000 times the amplitude of the lowest frequency. Because of the "tuck and nip" of the RIAA recording curve the actual output playing a record will give a range of 100:1, or in decibels: 40dB.

The analogue input to an ADC (analogue to digital converter) chip is 2 volts peak to peak in the case of a Texas Instruments PCM2902. In r.m.s. that is approximately 700mV. It should be similar for most chips as all now seem to work on a 3.3V supply.

700mV r.m.s. is therefore the maximum the ADC input will take before clipping. The input would have to be adjusted down to cater for the dynamic peaks of the record. If reduced by -14dB which seems to be the "pre loudness wars" input level, the input at -14dB is 140mV.

140mV is therefore the input level to the ADC input at 20kHz and therefore the input at 20Hz is 1/100th of 140mV which is 1.4mV. In decibels that is -54dB.

If the dynamic range of the ADC is 16 bits there is 96dB dynamic range. At -54dB there is 42dB left.

In this 42dB has to fit the low frequency (20Hz) dynamic range less the peaks. Beyond that is the noise floor which is different from an analogue noise floor - in digital nothing can exist at or lower than the noise floor. If 42dB is considered sufficient then noise wise the idea seems OK.

However, please do not forget that the distortion increases in proportion with falling digital signal level and may be in the region of 10% on the quietest lowest bass notes, whilst being much more respectable as frequency increases.