During the Q&A of the last webinar dCS (Andy McHarg) mentioned that they were thinking about the issue of ‘inter-sample overs’, which is where the peak of the musical waveform exceeds 0 dB between two successive samples (each of which are themselves OK). I first read about this in an ‘Application Note’ of the chief engineer of Benchmark Media, which can be found here: https://benchmarkmedia.com/blogs/application_notes/intersample-overs-in-cd-recordings. This is what they say:
“Here at Benchmark, we believe that most oversampling D/A converters produce audible artifacts when playing 44.1 kHz recordings. These artifacts are caused by inadequate headroom in the oversampling interpolation filters. Fortunately, this artifact is preventable.”
I don’t know to what extent the DAC technology used by dCS might be affected by such an issue (the DACs of Benchmark all use ESS chips I believe), but it must be at least a potential problem otherwise Andy McHarg and others wouldn’t be thinking about it.
What do Forum members know about this? I’d like to learn more. Benchmark think the problem is widespread. They happen to use a particular Steely Dan track for listening tests, and it suffers from about 3.7 of these ‘overs’ per second, they say. Are these already handled gracefully by our Varese(1)/Vivaldi/Rossini/Bartok/Lina DACs, or are they something to consider in future?
[quote=“Urbanluthier, post:2, topic:6902, full:true”]
Don’t think inter sample overs are an issue with any dcs dac
[/quote]e
I would like to believe that but I do not recall the subject being raised specifically in relation to dCS aside from during the Varese webinar. As @Simon_C says, if all was OK why would Andy and his colleagues be thinking about it ?
If I am correct, isn’t it a piece of collateral damage from the loudness wars?
I am not sure that the old thread that @Ermos has kindly posted helps much with the issue from a reproduction standpoint rather than how inter-sample overs are caused. Also much of the thread ( to which I am guilty of contributing) is on other subjects.
I think you are right: the average level of the signal is set so high that some peaks are not quite accommodated within the permitted range. Still, it seems to be fixable according to Benchmark - unlike some mastering horrors.
Every D/A chip and SRC chip that we have tested here at Benchmark has an intersample clipping problem! To the best of our knowledge, no chip manufacturer has adequately addressed this problem. For this reason, virtually every audio device on the market has an intersample overload problem. This problem is most noticeable when playing 44.1 kHz sample rates.
Still, we do not know whether they have tested dCS DACs.
This was also mentioned by David on the first call of the day. IIRC correctly in the context of something in the fw backlog for current products. Maybe something we can look forward to in the next fw update?
Thought this was covered in the Bartok review exchange. David Steven’s noted dCS dacs deal with inter-sample overs correctly. Can’t find the reference however. So it would be great to hear from dcs on how inter-sample overs are dealt with throughout the dCS product line
You are quite right, at least that the issue has been raised here before. Erno spotted that same Benchmark article in Dec 2020, and wrote about it in a thread called “Stupid question”. (BTW: I think that title was referring to something else. I haven’t had time to read the topic yet.)
Inter-sample overs are a bit of an interesting one. The assumption is that the DAC shouldn’t clip on an inter-sample over, but that isn’t necessarily correct. From the playback side, you have no idea whether the mastering engineer was hearing clipping on inter-sample overs or not (if they were present). The question we are considering is if they were hearing clipping, is there an argument that the playback device should clip as well to better match what the engineer heard?
Our DACs handle them very well currently and do not clip, but the argument could be made that for recordings where inter-sample overs are present, or at least some of these recordings, it formed part of the overall sound the mastering engineer heard so you may want them to clip.
That’s the question that is forming our thinking around this, so it will be interesting to see what Andy and the team come up with…
Interesting- wondering if filtering options can engineered to allow the end user to choose between clipped and unclipped. Curious to see what dCS does next!
But is the sonic signature of clipping sufficiently consistent across systems that one can be certain that provoking clipping in one device will indeed “better match what the engineer heard” in another completely different one? Maybe it is…
Fascinating! As @Urbanluthier says, it will be really interesting to see where this all lands.
Quite. Of course what the engineer heard or intended is impossible to ascertain. In fact which engineer? The original recording , mastering for production, glass master ( for CD) ? To confuse matters copies from any stage can be circulated to senior figures at the record label who revert with their own comments on what the sound should be like and request changes.
Good luck on this one as there is no clear answer ( or even a clear question).
There is nothing new about 44k1 overshoots accumulating themselves and effectively going over peak bits, then causing trouble further down the line. It was highlighted by Philips when Bitstream first came on the scene a long time ago, because in their case filter ripple caused overshoots into potential clipping further down the line.
As a precaution ever since those distant days all masters leaving our premises never peak at a nominal 0dB - we always set an absolute maximum of -0.4dB. Losing half a decibel or even a whole dB in the cause of avoiding clipping in some designs of dacs is no big deal.
Some non-classical material runs levels hard into the bloodline of peak bits intentionally.
As to whether dCS dac designs have problems with 44k1 inter-sample overshoots I have never experienced hearing anything. If other brands’ d/a products have troubles, it is their problem (!) - I certainly would not be surprised. Some of them sound filthy and like breaking glass, and lossy packing of overdriven material is likely to make it sound even worse.
