Digital that sounds like analog
Posted 10th December 2012 at 04:18 AM by abraxalito
Updated 15th February 2013 at 06:11 AM by abraxalito
Updated 15th February 2013 at 06:11 AM by abraxalito
For those who missed Frank (fas42)'s link on a thread I started then here's where I'm continuing my minimal oversampling DAC developments for the time being : Digital that sounds like analog
Total Comments 108
Comments
-
I doubt that their intention was to have that outcome, but they then discovered the behaviour occurred, and their solution was externalise the upsampling. Rather than, improve the engineering within the DAC to eliminate the cross-interference, a more involved exercise. Perhaps a more honest approach would have been to create a fully shielded module to do the upsampling, which fitted inside the DAC case, and offered that as an upgrade ...
So in that sense, yes, it was a marketing driven decision ...
FrankPosted 27th March 2013 at 06:32 AM by fas42 -
What does 'fully shielded' have to do with it, and why would fully shielding be necessary? You reckon that the primary cause of interference is radiated noise from the digital processing? If so I'm sure the base model DAC also has digital processing, after all its an S-D converter. Also why would the processing module be needed as the base model DAC alterady upsamples to 3MHz or so - the only difference could be the level of precision being used as I'm sure the DAC does not run faster when the upsampler's being employed, its just offloading processing out of the DAC and into the upsampler. The output symbol rate must stay the same.
Posted 27th March 2013 at 01:17 PM by abraxalito
Updated 27th March 2013 at 01:20 PM by abraxalito -
The 'fully shielded' is meaning to imply that the upsampling circuitry is completely isolated from the DAC in every sense -- it's as if the currently separate upsampling unit was squashed up to fit inside the DAC case. So, this means power supply, grounding, the works are distinct electrical entities, that just happen to be surrounded by the same casing.
Years ago, I remember reading about the first Stax CD player; it got some excellent reviews by people who were listening the "right way". And Stax cheated -- got a bargain basement Sanyo player, just plastic rubbish inside; slung a completely independent DAC underneath, with own power supplies, disabled the Sanyo's processing, and fed the appropriate signals, clock and digital stream, through adjacent holes punched in the casings. Largely ignored, it made the music lovers happy ...
The key phrase in your comment is "just offloading processing out of the DAC and into the upsampler" -- that's where the difference is being felt ...
FrankPosted 27th March 2013 at 11:17 PM by fas42 -
Ah so your argument is that the dCS is suffering from cross-talk between the digital and analog parts? Which for a DAC priced into 5 digits, is gross incompetence of design isn't it?
Also when the upsampler's being used is there a way to turn off the processing inside the DAC? Sounds to me like that would be too much of a consumer (or dealer) headache to involve them in that level of detail. I can't imagine how to implement it cleanly, can you? As far as I'm aware, the upsampler doesn't upsample all the way to 3MHz, that final part must still be in the DAC proper.Posted 28th March 2013 at 03:06 AM by abraxalito
Updated 28th March 2013 at 03:12 AM by abraxalito -
I don't know about incompetence, my feeling is that everyone is still learning. The designers dive into their bag of tricks to get as clean a sound as they feel they can achieve, partly going by what's currently fashionable to fiddle with, and they end up stopping at some point because they run out of time, money, patience of their employer, etc, etc - a product has to be got out on the street at some stage.
Turning processing on or off shouldn't be too hard, this would be a digital "decision" by the input circuitry locking on to the incoming data stream, from "measuring" the clock rate.
Until people get everything totally under control, have full knowledge of how all the interference effects occur, then the best policy is to separate as much high speed processing from the analogue side, or reduce it, as you can.
My experience with digital has been that it is a scurrilous thing - you think you have the quality of it under control one moment, but then you come back half an hour later, or do something theoretically completely unrelated to the playback mechanism, and find the sound has gone off completely - what the ... !!!!!
I keep my eyes open, but from where I sit I can't see anyone who has all the answers ...
FrankPosted 28th March 2013 at 10:40 AM by fas42 -
Sure, nobody has all the answers but that doesn't render everyone competent (or equally incompetent) at a stroke. Plenty of designers disbelieve in sound quality differences - Doug Self for one. I agree that SQ is improved by keeping HF out of the analog as far as humanly possible.
On the subject of the sound 'going off' I had exactly this experience over the past couple of days with my DAC. I had been changing the decoupling to the I/V amps but for some reason the sound felt 'off'. I fixed a couple of loose wires (when there are many DACs in the array one can go off without destroying the sound completely - its a hazard). But none of these fixes fixed the sound. I scratched my head and went into detective mode...
Firstly how 'off' did it sound? The answer was 'very DSD-like'. The key feature of DSD being lack of dynamics - caused by so much HF energy. So I wondered what could be oscillating at high freq to give the same subjective effect. My suspicions landed on the current source feeding the reference (TL431) voltage setting the DAC compliance point at 2.5V. That's an LM317 wired as CCS and I've had stability probs with those before as CCSs. So I fitted a series resistor on its output and bingo the old valve-like timbre and depth was back. Interesting that it hadn't given that problem before though, still haven't figured out what changed to provoke the CCS to instability.Posted 28th March 2013 at 02:30 PM by abraxalito -
Yes, that type of 'offness' is exactly what I've spent many years chasing down. It's the first thing that I discern when listening to most systems, and what has driven me into bouts of tweaking frenzy, especially over the last couple of years.
Now, what's interesting is that you believe that the problem is that high frequency oscillation, whether as part of the source signal or instability in the circuit somewhere, is at the core of the issue. Is that an intuitive insight, or have you made measurements at any stage which clearly points to that?
I'm certainly very cognizant of the need to eliminate spurious high frequency noise on power supply rails at all points to prevent "greyness" intruding ... the trouble is that the linkage of cause and effect at times can be mind-numbingly difficult to understand ...
A good example: a friend I've been lending advice to for a number of years I visited a week ago, his SQ has gone up and down like the tides, as he keeps fiddling! His recent digital, sourced from a portable media player, and buffered through a simple valve circuit prior to the amp, has sounded excellent at times. Yet, this time it was definitely off ... what had changed? Well, he thought he was doing the right thing by putting the valve buffer into a grounded metal box, shielding it ... but, my intuition and experience said this was a wrong move and so it turned out to be: we biffed the box, and the good sound returned. I certainly can't say what precisely what was going on ... only by having the best test instrumentation on hand, and really knowing how to use it without the presence of the probe itself in turn changing the circuit behaviour can one really nail the situation, I feel ...
FrankPosted 28th March 2013 at 11:28 PM by fas42 -
I've made measurements with my scope that demonstrates HF oscillations in CCS-connected LM317/337s. When I first noticed it I also played around and found that adding some series resistance to the output fixed it. So here we had a case off 'offness' which was fixed by the same solution. What's the chance that it wasn't in fact HF oscillation in the regulator (true enough I didn't connect a scope, just used my intuitive understanding to reason that the resistor might work in fixing up the sound)?
Posted 29th March 2013 at 01:38 AM by abraxalito -
I meant HF in a general sense, in "all" situations where digital is below par. I certainly have had plenty of examples of such, and can imagine that there are multiple "weak" points in the circuitry ready to break into oscillation under various stimulii.
Years ago, this was a common occurrence for me -- the sound would degrade after a while. And the simple solution was to do a power reset, worked every time ... until the cycle was repeated.
Of interest, is your system susceptible to mobile network frequencies? Start with clean sound, switch on your mobile and bring it close to the circuits, does the treble develop an edgy, unsavoury quality - switch it off, and SQ is restored?
FrankPosted 29th March 2013 at 03:48 AM by fas42 -
Definitely my DAC is susceptible to GSM, that's the AD605's front end, very low noise and hence very prone to RF upsets. I'm going to try another chip, AD8129 to see if this one's better.
As regards HF, yes in general its the case that more HF means poorer sound. I started out playing with grounding, not knowing why star grounding gave better SQ. Now I realize its probably because star grounding means cleaner grounds at HF. In dealing with the HF problems I adopt a double-pronged approach. Firstly, cleaner grounds and power supplies. Secondly, more HF-resistant electronics - by which I mean get rid of the LTPs.
Forgot to add that passive filtering of signal lines whenever they could be subject to RF ingress - like inputs and outputs is essential too. For this, I've adopted the inductor (and ferrite bead) as my chief ally. Next up - transformers...Posted 29th March 2013 at 04:24 AM by abraxalito
Updated 29th March 2013 at 04:43 AM by abraxalito -
Thanks for alerting me to the AFA devices, not something I was aware of before ... still more things to think about!
And, just "discovered" the "Using the AD844 as an I/V" thread, some interesting commentary there -- I especially note George's last post, [URL]https://www.diyaudio.com/forums/digital-source/227677-using-ad844-i-v-34.html#post3425948[/URL] ... this is sounding familiar, :) ...
What this demonstrates for me, yet again, is that there is no "one way" - a variety of different techniques and combinations will get one over the line; meaning that the SQ is at least at the level where all the normal considerations of whether it sounds "analogue" or "digital" disappear -- it just sounds like the real deal ...
If all of the very best techniques and ideas are all combined, the sound will be capable of becoming unbelievably stunning -- I've had glimpses of this happening, and it has helped me maintain my enthusiasm when I've hit the inevitable "dog days" ... ;)
FrankPosted 29th March 2013 at 11:38 AM by fas42 -
Yeah AFAs are the flavour of the month/year for me at the moment, can't get enough of these. I have an idea for testing out the dynamics hypothesis with transformers - the idea being to lighten the load on the amp by a step-down trafo. I figure I might even try a null test - record the amp output waveform with differing load impedances and see how much the output noise changes by subtracting the unloaded output from the loaded. But on reflection I suppose this might just get me the output impedance of the amp factored in...
Posted 29th March 2013 at 12:37 PM by abraxalito -
Apart from the fact you keep calling it "noise" - for me if there is a variation between what goes in and what comes out then it's distortion, the fact that it is relatively low level doesn't change the nature of the beast - that sounds like a good idea.
My own experience is that a very "shabby" bit of amplifying circuitry will always respond positively if you feed it very clean fuel, I'm used to what is achieved doing this, and it really surprised me how relatively "gutless" most of the systems sounded at the recent hifi show, in spite of the massive, impressive looking amplifiers being used ...
The aim should be to get to clipping point of the amps, and even a bit beyond, with no tonal change whatsoever; if running at maximum volume for a while, then slowly reducing the volume there should be no subjective sense of relief, it should rather be as if you were starting to walk away from the performers ...
FrankPosted 30th March 2013 at 12:21 AM by fas42
Updated 30th March 2013 at 12:29 AM by fas42 -
I call it noise but I don't mind if you prefer 'distortion' - we are meaning the same thing. To me noise has a broad-band characteristic and distortion is discrete tones, but I agree that's fairly arbitrary. Sibilance for example is noise modulation to me, not distortion. So noise for you, if its added by the DAC and/or amp (rather than encoded in the source) becomes 'distortion' ?
As for the 'no tonal change whatsoever' I reckon that there's going to be tonal change based on the loading of the amp - that's based on my so far very limited experiments with trafos. I agree with the proponents of TVCs but to me it doesn't make much sense to put the TVC before the amp since the most difficult load is in fact the speaker, by a few orders of magnitude. Why strain out the gnat of the (relatively light) load on the preamp output presented by the amp's input but then swallow the camel of the speaker's load on the amp?
I've just this morning fired up two new trafos which have winding ratios approx 4:1 (before I was using 12:1). The 12:1 work great with my 32ohm Sony headphones but sound a bit quiet when used to drive speakers. The new 4:1 are sounding like a million dollars in terms of the transparency (Mozart string quartet playing). They're transforming the load on the amp from 8ohms to something like 120ohms in the mid-band. Even at the lower output level the soundstage depth is clearer. I suspect the primary reason active speakers sound better is that a drive unit alone is a more benign amp load than a crossover, especially when the amp input is bandlimited. Next up - active speakers with each unit driven by its own dedicated trafo....Posted 30th March 2013 at 01:00 AM by abraxalito -
Yes, noise for me, is that which is random in nature, is not dependent on the signal content. Like tape hiss, vinyl crackles and pops. Acoustically, it separates into another "space" and the hearing mechanism can filter it out out relatively easily, as a subjectively distinct element ...
Yes, those output transformers are shifting the amps into voltage drive rather than current drive; the weakness in most amplifiers is when they have to deliver significant current, and the output devices are running into the class AB transition zone; this exposes any weaknesses in the power supply and topology generally.
Active speakers, all else being equal, will always do a better job for the reasons you stated. Again, amps and power supplies hate current spikes being called for ...
FrankPosted 30th March 2013 at 02:43 AM by fas42 -
Noting comments about the Viola gear -- at the Sydney hifi show one distributor stood out by having 3 rooms with clearly superior sound, obviously some knowledge by someone had been acquired along the way! And, one of those was using Viola amplication - simply, the sound was more "correct", less hifi ...
FrankPosted 1st April 2013 at 02:13 AM by fas42 -
I have a hunch that my transformers aren't helping for the reasons I thought - probably not related to the loading but rather they're improving the RFI immunity. In order to test this hypothesis I'm winding up some common-mode chokes to replace the trafos. What I'm hearing initially from the chokes is its reducing a certain 'tizziness' on piano which I've previously associated with transient instability in output drivers. I wonder if the 2m or so of speaker cable is acting as 'agent provocateur' for some parasitic oscillations in the output stage of the chipamp. Watch this space...
Posted 1st April 2013 at 05:37 AM by abraxalito -
To this day, I have not found that subjective problems in the sound end up being laid at the feet of the speaker drivers. Anomalies are always traced back to electronics misbehaviour, and always have been solved by cleaning up their act.
Are the speaker cables strongly twisted, or equivalent?
FrankPosted 1st April 2013 at 06:08 AM by fas42 -
Posted 1st April 2013 at 04:59 PM by abraxalito -
I verified it was not a simple loading issue by making a pair of 1:1 trafos and they gave the improved sound too. I'm now listening with some series L (about 2.5uH) and its about as good as with the trafos so to me this is sufficient evidence that the leakage inductance of the trafos was what was giving the sound I liked, nothing much to do with PSU loading. The chipamp really does need output inductors is the moral of the story, they're required even when not shown in the DS.
Posted 2nd April 2013 at 10:35 AM by abraxalito