CD and Pre Emphasis

[OldTechie]

Reading the link in Gordon's original post I see it was suggested that iTunes would de-emphasise a rip, but mess up the stereo image when doing so. Maybe iTunes had copied the SOX parameters that had the left/right problems noted on the link I found.

I have just tried ripping my new disc with iTunes (to ALAC on a PC). I compared it with my dbPowerAmp FLAC rip with its flag set using foobar2000's ABX comparison feature (which allows me to play both in sync and switch between them.) I can hear no difference at all. If iTunes was once wrong I reckon it is probably OK now (or the error is at frequencies I can no longer hear.)

[Gordon]

It seems the pre-emphasis is not a single pole affair - it is 50uS (3,183Hz) boost but with a 15uS (10,610Hz) levelling off again.

Quite right OT!! In my defence I will say though that my #1 does give the two time constants used by the PE network. Being pedantic , there is only 1 pole, the other time constant is a zero!! At the PE stage there is a pole at 15us and a zero at 50; at the DE there is a zero at 15us and a pole at 50!!

I'm left with a question about this concept. Presumably it was designed with the idea that the pre-emphasis was applied in the analogue domain, and that the de-emphasis should be in the analogue domain. So it is digitised after pre-emphasis, and dither is added to compensate for the 16 bit quantisation problems.

Yes, I believe that was so originally. The idea was to deal with early DACs that only had 14 bits [the first Philips machine did]. The DE would suppress some of the HF quantisation distortion [QD]. PE and DE can be applied in the digital domain. You can see PE as a very simple form of Noise Shaping.

The dither is often shaped and mainly in the HF region. If we de-emphasize it in the digital domain, we attenuate the dither before applying it to the DAC.

Dither [D] and Noise Shaping [NS] are not quite the same thing. D is a small signal normally added at an ADC input to prevent quantisation defects at low signal levels. There seems little use in applying it after because the damage will have been done. However, under some conditions, instead one could do some number crunching after the ADC to make sure that the low level LSBs are randomised to the same extent which leads us back to adding something to the audio thus “corrupting” those LSBs.

IF NS is to be used then the D and NS can be combined. However if the audio is sampled at 44.1 and coded to 16 bits then the PE may be best done in the analogue domain because the ADC gets the benefit of a larger HF signal. My guess though is that it is all done digitally taking advantage of a source format using over-sampling and 24 bits where the dither amount compared to 16 is much smaller – like 48dB smaller!! – so much so you might do without it because there is enough noise from the microphones.

Once D is applied it becomes inextricably part of the audio UNLESS the dither signal is defined and so could be removed at the DAC. This is not done with CD although it could have been if the designers at Sony had been more aware of future signal processing possibilities. It is low level but in a 16 bit system may be audible to some in which case attenuating it with the DE would be good thing – this assumes that the D spectrum has significant energy above the DE pole frequency. It isn’t supposed to be audible in the first place.

So do we get a different effect to doing the de-emphasis in the analogue domain? Of course older CDs probably don't have clever shaped dithering (or maybe no dithering at all) so I'm sure it is not of any practical significance.

I think you are right that older CDs will not have any complex NS and so will not sound so good – really old ones may even have been made with iffy ADCs. NS is done at the start of the CD production process and typically uses a source using 24 bits say to produce a 16 bit version that is better than just quantising to 16 at the ADC. PE, lifting the HF, could be applied at the same time using the digital 24 bit source so that the NS also affects the PE/DE by shifting QD into the HF where the signal is now bigger.

Similarly over sampling at source is also used to enable the enhancement of the resultant 44.1/16 signal – see here at posts #27 and 28 for some more detail. The value of PE/DE is reduced when NS shifts most of the QD into the super-sonic region ie beyond 22 kHz. If the NS is effective then the need for PE/DE may go away.

Given adequate number ranges in the computations there is no reason why the digital version should be any better or worse than the analogue. There could be a cost advantage to the analogue method if the handful of components is cheaper than a few extra bits of silicon to do the sums in the DAC chips. The CD player designer does need to know if the chips do it or not because he needs a flag if they don’t!! Of more concern in a DAC is the design of the over-sampling/interpolation filters and the control of clock jitter.

Downloads that use 24 bits and high sampling rates can be burned to a CD-R but need some software that will insert a PE flag correctly [default to none] and also reduce the sample rates and the 24 bits to 16 in a suitable manner. I doubt that many such programs will do sophisticated noise shaping or sample rate changing but then perhaps I am wrong.