Next best exponential DAC quality level?


I recently did a shoot out of three DACs using my Hint6 + routing each of the other DACs to analog input on the Hint6:

(1) Hint6: ESS Sabre32 -- Integrated 

(2) SMSL M500: ES9038PRO D/A   ~$400 

(3) Khadas ToneBoard(v1): ESS ES9038Q2M - ~$99

I played the same song passages on Amazon Music and was able to cycle through each Hint6 input corresponding to each DAC.

The result?  Very small difference in terms of rendering.  Maybe a more open sound stage with better overall balance using the Hint6 DAC.  The Khadas was more bass / midrange pronounced w/ a more narrow soundstage.  However, I wouldn't suggest that any were head-and-shoulders "better" over the others.  In fact, they were all pretty decent with only small nuances (certainly not worth the price differences.   

I decided to keep the Khadas for my small headphone listening area. 

But it got me thinking - how much would one have to spend to realize an exponential difference in quality?  Is the Khadas that good, or is DAC technology differences more nuanced than I originally thought (meaning, we're paying 10x for only 5% better).  

 

martinman

If you are are a retired electrical engineer, then you should understand the technology enough to know that it does not take that great a clock to achieve THD and IMD well below the tube output stage on your DAC and possibly other equipment.

Blanket statements about what distortion does cannot be made. It depends on the order of harmonics and the level. Tubes do tend to lend a distortion that generates pleasant higher frequencies and can result in both a slight perception of increased volume, more substantial highs, and greater "air", which helps with the perception of greater soundstage width. Not accuracy, but width. Frequency response anomalies can also give the impression of more width, and height (or less). These are not guesses. These are studied and known.

That you highlight ASR specifically, is a false premise. ASR takes measurements. The conclusions are make are all from the decades of research that has been done on how sound is perceived. ASR has no influence on interpretation in that regard.

Depth perception is almost exclusively a function of differential volume, and intentional reverb. If you are getting more that what is in the recording, then it could be pleasant, but not accurate. Nothing, absolutely nothing in your electronics takes as long to settle as your room, followed by your speakers. Again, frequency response anomalies including some reduced bass, if bass is slow to settle in your rooom or you have nodes can help.

w.r.t. the power supply, at least for the DAC, that is entirely a function of the DAC itself. If the DAC is well designed with good supply filtering, then it will achieve accurate performance well beyond the Audionote, purely by architecture. The one issue that does come into play with lower end DACs, and even higher end DACs, is ground conducted noise. That can be addressed easily for <$100.

W.r.t. the DAC and settling time, frequency response defines settling time 100%. Electrical engineers know this. Also w.r.t. the DAC, solid state DACs almost as a rule have much lower output impedance. If you were talking about speakers, settling time is measured, and if you are talking about amps, again, frequency response and phase defines settling time, and 4 ohm frequency response w.r.t. 8 ohm can tell you everything you need to know about output impedance. Invariably, for most speakers, most solid state amps are not remotely the cause of settling time, though with tube amps that can come into play. This is basic audio EE.

The article you linked only speakers to linear regulators, and though they spent much time, it is rather amateur in its process. The conclusion talks about RFI, EMI, etc. (all at frequencies well beyond what they tested), and then goes on to make a test with 1V on a 25V rail, and well beyond what would typically be experienced, and well beyond what you would see in a 5V USB wall-wort of even middling quality. 

What is above only provides crude details. My circuit, in a DAC, starts with a regulated power supply, then likely adds regulation, then adds RC or even RLC filters, with the net result that by the time you get to circuits that matter, power supply noise is near non-existent. Again basic EE stuff.

The proof is in the puddling so to speak though because the other thing that rejects power supply noise is feedback. Any test I have seen of Audionote DACs has indicated noise in the output at twice the line frequency plus harmonics, where many low cost DACs have this well under control even with low cost supplies. That tube output stage requires a whole lot more to keep it happy.

Best minds in the field? Do you truly believe that John Walton is one of the best minds in linear analog electronics? Really?

Who are you Cindyment? How do you know so much? 

Are you NW AV Guy? Amir? Archimago? 

I believe you meant bandwidth determines settling time not frequency response, but it is not 100%.

Phase angle, the load reactance and feedback also determine settling time.

I was referring to the engineers who designed the regulators and Jung, Swenson and the other companies designs represent the best minds out there.

The regulators absolutely need to have the bandwidth in digital circuits and to conflate human hearing with digital circuits is sort of disingenuous.

Possibly you have had seen base model AN units that only have RC or CLC power supplies and SRPP output stages but the upper tier models are CLC filtering into feedback regulated tube pass elements units with plate transformer opt’s.

Yes with the 44:1 opt the output z is only 30 ohms but that is more than adequate for 600 ohm balanced lines.

 

Yes I agree that full ground plane implementations are mandatory in high speed digital circuits but that is basic design implementations on nearly every digital circuit.

@seanheis1 

I have just been around a long time, and as opposed to talking the talk, I have walked the walk ... but not that long. Still got another decade of work in me or so.

@dht4me 

Frequency response and phase (which I said later on). Those two incorporate bandwidth which is an inadequate descriptor. Since we are talking about DACs, somewhat in isolation, but even if we were not, it does not matter, pre-amp loads are typically 10K-100K, non-reactive, and even $10 DACs have more than enough bandwidth, phase response and drive capability, again, technically more than the Audionote.

Jung, Swenson, etc. in the field of analog electronics, linear, are so far from the best minds. Competent perhaps, but if you are using what they say to determine what you are writing, then obviously at a system level ... well they should stick to analog.

What Audionote does in there power supplies or does not, really does not matter. All that matters is results. Measurements I have seen have shown more power supply harmonics then I would expect on any modern DAC, even <$100. Insisting on simple tube based output circuits makes it pretty hard to avoid that.

 

The regulators absolutely need to have the bandwidth in digital circuits and to conflate human hearing with digital circuits is sort of disingenuous.

 

Why? Tell me why, in full detail, and please explain what sort of result, say in picoseconds of jitter will result, and how much THD/IMD will result?

With any half way proper designed DAC, the DAC is being driven by a local clock either via buffer, USB, ASRC, etc. I can get exceptionally low phase noise with just some basic sense on the power supply side. It's amazing what a resistor, a ferrite bead, and a few ceramic capacitors can do. Since I have a stable clock, now I am down to logic edge speeds, or more specific, how fast I transition through the range of uncertainty, and that is going to be a few 100 picoseconds, now power supply noise will affect that, but if I am 0.1% noise on the power supply (and I can get better) then I am down in picoseconds worst case jitter, but because the edge speed is fast, and the important transitions actually very very few in audio, the odds of a noise peak being concurrent with a critical edge are low and hence RMS jitter contribution from a half way decent power architecture will have limited impact on performance. Of course, all of this assume the DAC itself has not implemented any techniques in the analog domain to reduce jitter. Most DACs chips do. What is in the Audionote does not, so it will be more sensitive again to design implementation.

We could talk about the DAC reference, but again, it is amazing what can be done with well chosen simple parts. There is a reason why companies buy test equipment. They know if they got it right or not. Not the right sound, which is much different, but they know if there are extraneous things happening they don't want to happen.