Hugo and Dave don't use any kind of DAC chip, the analogue conversion is discrete using pulse array. The key benefit of pulse array - something I have not seen any other DAC technology achieve at all - is an analogue type distortion characteristic. By this I mean, as the signal gets smaller, the distortion gets smaller too. Indeed, I have posted before about Hugo's small signal performance - once you get to below -20 dBFS distortion disappears - no enharmonic, no harmonic distortion, and no noise floor modulation as the signal gets smaller. With Dave, it has even more remarkable performance - a noise floor that is measured at -180dB and is completely unchanged from 2.5v RMS output to no signal at all. And the benefit of an analogue character? Much smoother and more natural sound quality, with much better instrument separation and focus. Of course, some people like the sound of digital hardness - the aggression gets superficially confused with detail resolution - but it quickly tires with listening fatigue, and poor timbre variation, as all instruments sound hard, etched and up front. But if you like that sound, then fine, but its not for me.
On the digital filter front - original samples getting modified - actually the vast majority of FIR digital filters retain untouched the original samples, as they are known as half band filters. In this case, the coefficients are arranged so that one set is zero with one coefficient being 1, so the original sample is returned unchanged. The other set being used to create the new interpolated value. The key benefit of half band filters is that the computation is much easier, as nearly half the coefficients are zero, plus the filter can be folded so that the number of multiplications is a quarter of a non half band filter. When designing an audio DAC ASIC, the key part in terms of gate count is the multiplier, so reducing this gives a substantial improvement in die size, and hence cost. So traditional digital filters use a cascade of half band filters, each half band filter doubles up the oversampling - so a cascade of 3 half band filters will give you an 8 times over-sampled signal, with one sample being the unmodified original data. You can tell if the filter is like this as at FS/2 (22.05 kHz for CD) the attenuation is -6dB. The filters that are not like this are so called apodising filters, and my filter the WTA filter.
Going back eighteen years ago to the late 90's I was developing my own FIR filter using FPGA's. Initially, I was interested in increasing the FIR filter tap length as I knew from the mathematics of sampling theory that timing errors were reduced with increasing tap length. So the first test was to use half band Kaiser filters - going from 256 taps to 2048 taps gave an enormous sound quality improvement, so I had confirmed that tap length was indeed important subjectively. But at this point I was stuck; I knew that an infinite tap length filter with a sinc impulse response would return the original un-sampled signal perfectly - but the sinc function using only 16 bit accurate coefficients needs 1M tap FIR filter - and that would never happen, certainly not with 90's technology. So was it possible to improve the timing accuracy without using impossible tap lengths? After a lot of thinking and research, I thought there was a way - but it meant using a non half band filter, which would mean that the original sampled data would be modified. This was a big intellectual stumbling block - how can changing the original data be a good thing? But the trouble with audio is that neat simplistic ideas or preconceptions get in the way. Reality is always different, and reality can only be evaluated by a careful AB listening test. So I went ahead on this idea, and listened to the first WTA filter algorithm - and indeed it made a massive improvement in SQ - a 256 tap WTA sounded much better than 2048 tap half band Kaiser, even though the data is being modified. Why is this? The job of a DAC is NOT to reproduce the data it is given, but to reproduce the analogue signal before it is sampled. The WTA filter reconstructs the timing of the original transients much more accurately than using half band filters or filters that preserve the original data and it is timing of transients that is the most important SQ aspect.
So the moral of the tale? Don't let a simplistic technical story get in the way of enjoying music!
Rob
