I have in my listening room RIGHT NOW both a NOS DAC (Border Patrol DAC SEi) and a delta sigma (Stello DA100). I have a vested interest in the Stello as it doesn't have the odd output impedance of the Border Patrol and has optical input for my TV. So my bias wouldn't be towards NOS ...
I can assure you from MULTIPLE ABABAB trials on the same tracks over multiple days and even times of the day (just gotta swap the SPDIF between the two and change input on amp remotely) that I hear (or perhaps FEEL) a difference between them. FEEL is the better term. My brain (objective, rational) mind says, they basically sound the same. But what I FEEL is that the NOS is more relaxed, more right. It's subconscious almost.
Yes, that's a good description. I feel that same relaxed quality when listening to NOS. For me, it's the most important difference between NOS and OS. The Echo experiment tests a particular hypothesis for what is the underlying circuit mechanism causing that feeling.
Does the experiment include tracks using the same ladder DAC chip with and without oversampling / upsampling? Or just NOS and delta sigma with upsampling? The phenomena in question causing this irritation are (1) digital filter and/or (2) some aspect of the delta sigma design. We have to first determine which is the root cause, although it could be both, honestly. You can't use delta sigma in NOS mode (as far as I know). So to isolate whether it's digital filtering, you have to compare the same ladder DAC chip with and without over/upsampling. I'm not sure what the difference between up- and oversampling is anymore.
I should have brought this up sooner, but at some moment in time I will make the results public (I don't know when, but certainly not in the next few days). Do you want me to mention the user names of the participants and their results or should it all be anonimized? Anonimized has the advantage that it is less likely to turn into a sort of contest (it isn't a contest, just an experiment) and the disadvantage that it's quite impersonal.
Last edited:
Our investigation has focused on non-SDM DACs. We quite intentionally were holding off analyzing SDM DACs in the context of this investigation to limit the number of variables, until near the end. I will note, however, that contributor dddac (Doede's ) has been adamant that SDM DACs without (bypassed) an interpolation-filter bypassed sounds like fully-multi-bit DACs do when NOS. While this very point has been discussed in the thread, it appears that we've identified pedestrian and ubiquitous FIR OS filter design as the reason for the sound of NOS.
Not really. We've found (admittedly, with a low level of statistical confidence due to participation numbers) that a very high performance, software based resampling engine, the PGGB, delivers 88.2 upsampled (from 44.1) playback that essentially sounds the same as 44.1 NOS playback. Pointing a finger directly at the OS interpolation-filters of most DACs as the reason. Of course, our lacking total control of the experiments (they aren't scientific), means there very well be some other DAC sub-system responsible, but the OS filter is most implicated at this point. What we haven't yet identified is the underlying mechanism responsible. The objective of the Echo experiment is to test a hypothesis regarding one possible root cause.
Last edited:
That would suggest that the starting premise for this project is that 44k1 NOS is right.
Surely there ought to be a reference setup for such testing ?
That was not the starting premise. The starting premise was that many of us hear a characteristic difference between the playback of OS and NOS, when it seems there should be none. Not that one sounds better than the other.
We've identified a means for removing that characteristic difference by utilizing a very high performance oversampling engine. Which logically suggests that the reason why the difference existed was somehow due to inadequate performance among common FIR OS interpolation-filters. It's particularly interesting that very high performance OS interpolation-filtering sounds more like NOS playback, rather than more like common OS playback.
The desire to answer the subjective difference question was not merely intellectual curiosity. It was with the further hope of then discovering a means for producing digital playback that was subjectively superior to either OS or NOS. Which I hear, as each having it's own particular subjective advantages over the other.
Certainly, there is. However, that would run straight in to the practical problem of ensuring that everyone participating in the experiment (all hobbyists, mind you) have that exact same reference set-up. Since that's obviously impractical, we've had to trudge forward as best we can, with the hope that the gods of music are guiding our efforts a bit.
Even should we never verifiably identify the root mechanism behind the original thread question, it would be a greater success if we only found a means of creating playback which is subjectively superior to both 'typical' OS, and NOS.
Last edited:
Ultimately the results of testing will be revealed, whereupon any individual can draw conclusions whatever they please, being in consensus or otherwise. Currently I have gone through two different DAC's, three different plug in line stages and two sets of RCA cables. Everyone has their own setups, reference or otherwise, making consensus difficult. What is the makeup of a reference system? IMO to engage in a reference setup seems would stall all engagement, as seemingly would also have little judgmental value to those whose systems are ultimately variant from it.
Although this testing seems a shot in the dark to establish some global unequivocal truth, the files themselves are valuable tools to draw individual conclusions in contrast to global conclusions. This is always up for discussions as well, as you appear commenting. It is questionable to support or denounce the outcomes of testing as unequivocally true or false of merit, yet the idea seems to gather variant opinions as to establish some consensus to pursue further direction of inquiry. This is in contrast to completely random directions.
Although this testing seems a shot in the dark to establish some global unequivocal truth, the files themselves are valuable tools to draw individual conclusions in contrast to global conclusions. This is always up for discussions as well, as you appear commenting. It is questionable to support or denounce the outcomes of testing as unequivocally true or false of merit, yet the idea seems to gather variant opinions as to establish some consensus to pursue further direction of inquiry. This is in contrast to completely random directions.
My DAC was built to allow switching between NOS and OS during playback. It is based on PCM1794. This thread will get five stars for the reason I now can define what previously was just a feeling regarding the difference between the two modes. I'm about to submit my answers to Marcel. Just to add that when switched to OS, "A" fell somewhere between "B" and "C".
hi Marcel, In amazed that you know og my old work - feels like centuries ago. I think that the woosh woosh is due to tiny DC offsets. All these modulators exhibit FM modulated tones and the noise level seems to follow this modulation. My own 8th order DAC also had a bit of woosh woosh but it was mainly due to nonlinear fold down noise. As I improved the analog sections, the problem was reduced. It takes an awful lot of dither or chaos to kill the FM modulation. I have done later work on dynamic element matching that attacks this FM problem.
Cheers,
Lars
Hi Lars,
It's indeed offset related. In the latest version I just added a DC offset that shifts the repetition rate to 40 kHz or so when playing silence. It also appears to be related to imperfect settling.
By the way, I love the basin of attraction pictures in your PhD thesis, they are very beautiful.
Regards,
Marcel
It's indeed offset related. In the latest version I just added a DC offset that shifts the repetition rate to 40 kHz or so when playing silence. It also appears to be related to imperfect settling.
By the way, I love the basin of attraction pictures in your PhD thesis, they are very beautiful.
Regards,
Marcel
If there is no specific cut date, it is open to speculations regarding possible manipulation of results by delaying. It also brings a question why in the middle of collecting results stage there is a call for anynomous results?