Rob Watts

Yes all my WTA filters are extremely fast roll-off; the 49,152 tap WTA filter is at -1dB within 75Hz of FS/2; the 1M tap WTA filter in the M scaler gets to 4Hz of FS/2. The M scaler WTA coefficients are identical to an ideal sinc function coefficients to a better than 16 bit accuracy. This implies that under all circumstances the M scaler 16FS WTA filter will recover the bandwidth limited un-sampled analogue signal in the ADC to a better than 16 bit accuracy.

I am a fan of recovering the original bandwidth limited analogue signal before it was sampled with as small a change as is possible, as this approach gives the best sound quality, and the best measured performance. Fortunately, theory tells us exactly how to do this (and there is absolutely only one way to do this); simply use an infinite ringing sinc function for the interpolation filter. This is a symmetrical linear phase FIR filter but categorically is not a slow roll off filter, but an ideal brick wall filter, which has an infinite amount of pre-ringing; but it will perfectly reconstruct the original un-sampled bandwidth limited analogue signal within the ADC, without any changes whatsoever. Clearly we have a paradox here; a filter which is guaranteed to return the original analogue signal rings for an infinite period of time, clearly not returning the original Dirac impulse. The answer to this paradox is that an ideal Dirac impulse is an illegal signal from sampling theory point of view - because it is not bandwidth limited. A Dirac impulse has the same level at FS/2 as at 1 kHz; but sampling theory categorically requires that the signal is bandwidth limited so that at exactly FS/2 the levels are zero. So with a bandwidth limited impulse fed into an ideal sinc function filter, then the interpolation filter, would, even with it's infinite ringing behaviour, reconstruct the bandwidth limited impulse perfectly - without any extra pre or post ringing at all. The audio industry collectively is completely incorrect in using illegal (from a sampling theory POV) non-bandwidth limited impulses to infer the sound quality of digital systems; unfortunately, simple dumb diagrams showing "unnatural" pre-ringing is easy fodder for marketing.

It is isochronous asynchronous USB operation. I use the full term isochronous asynchronous as asynchronous jars with me, as within the FPGA it is synchronous - as the timing comes from the FPGA. But I have indeed been guilty of dropping the asynchronous part. In future I will use the full term isochronous asynchronous, as just saying isochronous is not correct. Or maybe I will try to get used to the term asynchronous USB equals synchronous FPGA! As the designer of the FPGA (not the USB) my perspective is always from the synchronous FPGA side. Rob

Oddly, I was thinking about our chance encounter a few hours ago - my far east trip was very full on, and your comment about getting tickets to Beethoven's 9th was playing on my mind - I ought to add a few extra days on trips so I can enjoy live music! Hugo is indeed gaining traction with mastering engineers - one told me that Hugo had the ability to tell you immediately what was wrong with a recording, but at the same time communicate what is musically right about it. Indeed, I often listen to BBC Radio 3, and they often have archive recordings from the 30's - what struck me with Hugo was that one could perceive what was wrong easily - distortion, noise, poor eq. etc. - but also hear what was musically right about the performance. I have never designed an audio product before where I achieved such a big improvement in transparency and a step change in musicality at the same time. Often a perceived step change in transparency results in less recordings one can enjoy, rather than more! Rob

I agree very much with what Ted says. Simple statements like "native DSD is best", when it involves audio, are invariably way off the mark. Audio reproduction is a highly complex multidimensional problem, with a considerable number of unknowns, and since we have no good idea how the brain processes ear data to create the wonderful illusion of an audio landscape, there are a huge number of unknown unknowns... The vast complexity of audio distortions degrading the listening experience is of course what makes this subject so fascinating. I would also take with a pinch of salt any statement that isn't backed up by robust and careful AB listening tests. I too have been guilty of this - making assumptions like that level of performance is OK because the distortion (or deviation from ideal) is way to small to hear. My biggest advances as an audio designer have been from using AB listening tests to test assumptions. And many times what I thought was the wrong answer is the one that sounds better. But these are the times when your understanding expands which leads you into uncharted waters, and hence closer to the goal of musicality from audio.