He is saying that level setting is achieved in digital domain. (so, no possibility of digital overload), the level is not achieved by analogue means. Doing it this way surely produces a more transparent output. (it allows headroom)
Digital overload is nasty! and since Chord dacs oversample heavily, there is a chance of digital overshoot, if there is no overhead.
A couple of problems with the above:
1. Obviously, the analogue output level must be achieved in the analogue domain. The output of the D/A conversion process has to be created at a nominal voltage level. So in the case of say a 3V output level, the output is amplified in the analogue domain to 3V. If we also want say a 2V output, then we must attenuate that 3V output, either by using some sort of analogue “pad” or in Chord’s case, by attenuating the digital input level.
2. Yes, digital overload is nasty and is very likely to occur when oversampling, due to “inter-sample peaks” (ISPs). But, this is a typical example of a common audiophile marketing strategy: Mention a *potential* problem with some aspect of the reproduction process, then explain how your clever design solves that problem and is therefore superior to other, cheaper, non-audiophile designs. However, this is BS marketing because it fails to mention that this potential problem has not only already been addressed but has in fact been standard practice for many years, even with cheap DACs. In this case, oversampling DACs already have built-in headroom to accommodate ISPs. I’m sure there are some exceptions, probably amongst the earliest, cheaper CD players but for around 30 years or so, pretty much all DACs allow ~3dB or more of headroom before oversampling. In other words, this statement is correct: “
there is a chance of digital overshoot, if there is no overhead.” - but it’s irrelevant because there pretty much always IS “overhead”!
Therefore, using more digital attenuation than would already be applied as standard by even cheap DAC chips, “surely” does NOT “produce a more transparent output”!
It is not just clipping, at 0dB digital, the noise shaper goes sideways a bit - I am not an expert, but the graphs say so.
But the graphs do not say so. The graphs indicate some clipping distortion but not that it’s been caused by the “noise-shaper going sideways”, it could be a flaw in the upsampling process, the filter or some other part of the process. However, the level of this distortion isn’t even reproducible by HPs/Speakers, let alone be audible.
A simple demonstration that noise-shaping doesn’t go sideways at 0dB is the SACD format. SACD only has 1bit, to achieve a dynamic range greater than 6dB (approx 120dB) it relies on aggressive noise-shaping. But obviously, being 1bit, with only two quantisation levels (effectively 0dB and -6dB) then the signal is at 0dB about half the time and if your assertion were true, the noise-shaping would therefore be almost constantly “sideways” and causing such high levels of distortion that the SACD format would be useless, but in reality of course there’s no audible sign of this “sideways” distortion.
That’s another fairly common audiophile marketing strategy: “It isn’t really understood” and/or “hasn’t really be researched” or “has only recently been researched but needs more digging into”, etc. This is a blatant marketing lie which is often applied to the (non) issue of jitter but is also quite common in other areas as well. In fact, this has been well “dug into” since at least the early 1990’s.
Using the digital level settings ensures cleaner output in audio band and beyond, OK you may lose a dB or two in SINAD, but as you correctly said before, what does that matter? Once past a certain line, SINAD becomes irrelevant.
Using digital level settings does NOT ensure “
cleaner output in the audio band and beyond”, it ensures noisier or far noisier output, depending on the amount of digital attenuation. As explained in my previous post.
BTW, I’m not accusing you of blatant lying or marketing falsehoods, just of repeating them (presumably without realising?).
G