What is 'incompetent digital' ?

[amirm]

Adding on to Cosmik's post, just like Audio Engineering Society, in this forum we accept both engineering excellence through better measurements and listening test results. No measurement paper is thrown out of AES because it was not accompanied with listening tests. Members can decide for themselves is engineering excellence is what they want to pay for or not.

 

[Ethan Winer]

We don't know if any shortcomings of one - say, a less than optimum filter implementation - are being masked / compensated by the implementation of the other.

I guess that's possible in theory, but I find it difficult to envision a scenario where A/D would harm the fidelity in an audible way, and then a D/A would hide that and make the combination transparent again. (Other than frequency response where a loss can be countered by a boost.) Can you even describe the type of artifacts that could cause such masking?

--Ethan

 

[RayDunzl]

I suppose the only truly Incompetent Digital I can think of would be some otherwise possibly interesting "Internet Radio" stations that stream at 32kb/s.

 

[Blumlein 88]

There was a 7.1 Denon AVR a few years back with incompetent digital design. The DACs and digital inputs were quite fine. The ADC for analog connections was incompetent. It sounded quite poor and you could hear hiss like cassette tape. In fact it measured just shy of 60 db SNR on the analog inputs. Unbelievably some fellow who owned one opened it up and determined why that would be. Well off in a corner of the device on the corner of a board they had placed the ADC chip in an area of noise from other circuits, and used an incredibly skimpy trace for power. The fix was relatively simple. A grounded copper foil shield between the noisy circuitry on the daughter-board and the ADC chip. A couple of short 1 1/4 inch jumpers to a thicker power trace, and increasing two very small capacitors on the PS trace by 300% to a still small, but adequate value. Voila', the analog input now had about 100 db SNR, and sounded fine.

Certainly there have been incompetent designs in PC soundcards. Sample converting everything very poorly to 48 khz and other noise issues. One I had reason to check out once had okay jitter on one channel, but the other channel was polluted with huge amounts of audible and measurable jitter. So much so it raised the noise floor and made the card sound wrong used in stereo. I don't know if it was a faulty chip or what. Obviously the clock used for both channels was being polluted on the one channel and not the other. I have seen stand alone DACs where the channels didn't clock right together either though nothing this bad.

 

[Ken Newton]

I guess that's possible in theory, but I find it difficult to envision a scenario where A/D would harm the fidelity in an audible way, and then a D/A would hide that and make the combination transparent again. (Other than frequency response where a loss can be countered by a boost.) Can you even describe the type of artifacts that could cause such masking?

--Ethan

Personally, I find subjectively complementary colorations, whether intentionally or unintentionally designed, to be not that rare a happenstance among various pieces of gear. Some term this effect, system synergy.

I don't take us to be attempting to reveal objective frequency response variations, which are easily measured. I take us to be attempting to reveal subjective sound characteristics not so easily identified via measurement. Bypass test such as this have long been conducted to 'prove' the subjective perfection of this affordable amplifier or that mass market receiver - Hafler's Transnova amp comes readily to mind.

If we are assuming that DAC units will have differing subjective transparency, which we are, then we have to also assume this equally true for ADC units. Therefore, it would be faulty logic to conclude that because some given ADC to DAC pair happened to sound transparent that the ADC unit alone could be taken to always be blamelessly transparent if utilized to test other DAC units.

 

[Blumlein 88]

Personally, I find subjectively complementary colorations, whether intentional or unintentional, to be not that rare a happenstance among various pieces of gear.

I don't take us to be attempting to reveal objective frequency response variations, which are easily measured. I take us to be attempting to reveal subjective sound characteristics not so easily identified via measurement. Bypass test such as this have long been conducted to 'prove' the subjective perfection of this affordable amplifier or that mass market receiver - Hafler's Transnova amp comes readily to mind.

We are assuming here that DAC units will have differing subjective transparency, which we are, then we have to also assume this equally true for ADC units. Therefore, it would be faulty logic to conclude that because some given ADC to DAC pair happened to sound transparent that the ADC unit alone could be taken to always be blamelessly transparent if utilized to test other DAC units.

http://audiosciencereview.com/forum/index.php?threads/choose-the-preferred-dac-from-these-5.386/

Try to vote in the poll where I use one ADC to record 5 DACs. See if you find them differing in transparency. It should be an interesting experience for you I think.

 

[Blumlein 88]

Personally, I find subjectively complementary colorations, whether intentionally or unintentionally designed, to be not that rare a happenstance among various pieces of gear. Some term this effect, system synergy.

I don't take us to be attempting to reveal objective frequency response variations, which are easily measured. I take us to be attempting to reveal subjective sound characteristics not so easily identified via measurement. Bypass test such as this have long been conducted to 'prove' the subjective perfection of this affordable amplifier or that mass market receiver - Hafler's Transnova amp comes readily to mind.

If we are assuming that DAC units will have differing subjective transparency, which we are, then we have to also assume this equally true for ADC units. Therefore, it would be faulty logic to conclude that because some given ADC to DAC pair happened to sound transparent that the ADC unit alone could be taken to always be blamelessly transparent if utilized to test other DAC units.

I believe in the case of the Hafler it wasn't bypass testing rather null testing. The signal thru the amp compared to a piece of wire. They obtained pretty good results, but the thing with amps is you need to do it with the loudspeaker attached. I have done bypass testing of amps years back, and not very many would be transparent in that situation. I might need to try it again with some modern amps and see what happens. The best performing in a bypass test in the few I tested that way were amps from Spectral.

The Swedish Audio Society has done such tests. I believe they only found two amps to be transparent in bypass testing when loaded with a speaker load. One of the Brystons and an Audio Research SS model.

 

[Ethan Winer]

Personally, I find subjectively complementary colorations, whether intentionally or unintentionally designed, to be not that rare a happenstance among various pieces of gear. Some term this effect, system synergy.

Synergy means the sum adds up to more than the individual parts. The best an audio device can do is pass sound with no audible change. So synergy is a misguided notion. If two devices somehow counter each other as I mentioned above about frequency response, then that's complementary, not synergistic.

I take us to be attempting to reveal subjective sound characteristics not so easily identified via measurement.

I can't think of anything that can be heard but not measured. If you have a specific example I'd love to hear it. But don't forget that the null test reveals all differences between devices, including those you might not be looking for or even know about:

AES Audio Myths - Null Test

--Ethan

 

[NorthSky]

 

[Ken Newton]

Synergy means the sum adds up to more than the individual parts. The best an audio device can do is pass sound with no audible change. So synergy is a misguided notion. If two devices somehow counter each other as I mentioned above about frequency response, then that's complementary, not synergistic.

If two components of a system chain are individually judged as being subjectively non-transparent, but are judged as being more transparent if paired together, that seems a case of the total being more than the sum of the parts. What seems a bit misguided is to focus on whether this effect should be termed complementary or synergistic, rather than the salient point that the sound is subjectively more transparent with two particular components paired together than if they are not. A long known measurable example of this effect, not involving frequency response, is how even order distortion can be cancelled via complementary transfer curves.

I can't think of anything that can be heard but not measured. If you have a specific example I'd love to hear it. But don't forget that the null test reveals all differences between devices, including those you might not be looking for or even know about:

AES Audio Myths - Null Test

--Ethan

I don't believe in audio magic. I believe that if the sound is subjectively different then the signal is also different. Signal differences can be measured, IF we know what to measure, under what conditions it should be measured, and how to interpret that measurement. The most ready example I can think of is THD measurement. There was a time when it was assumed that a single-dimensional total percentage figure told what was needed to known about a component's harmonic distortion. That faulty assumption gave way to the realization that a two-dimensional spectrum plot of the distortion matters. I suspect even that will give way to the realization that a three-dimensional plot of spectrum versus level matters. DAC jitter is an good example of a parameter that once wasn't recognized at all as potentially having subjective consequence now recognized (not by all, I know) as such.

If simply passing a signal unchanged is the best a component can do, that doesn't necessarily follow that this is the optimum behavior for it within in a given system context. That would assume all of the other components in that system chain are also transparent. For systems, a holistic perspective is proper. Complementary component colorations exist in my experience, you may have a differing experience.

 

[Ethan Winer]

If two components of a system chain are individually judged as being subjectively non-transparent, but are judged as being more transparent if paired together, that seems a case of the total being more than the sum of the parts.

I imagine the idea that this could ever happen would go away quickly with a proper blind test. What sort of coloration could be audible such that when combined in series both colorations go away? Please be specific.

A long known measurable example of this effect, not involving frequency response, is how even order distortion can be cancelled via complementary transfer curves.

Ah yes, pre-distortion. I designed such a circuit 35 years ago:

Pre-Distortion Techniques

But what are the chances that one audio device will have soft clipping onset at a very specific level, and another will have the precisely opposite effect at exactly the same level? How would you even get "anti-clipping" in a circuit unless you were aiming for that specific curve? I truly doubt any hi-fi products that claim "synergy" have such a curve.

That faulty assumption gave way to the realization that a two-dimensional spectrum plot of the distortion matters. I suspect even that will give way to the realization that a three-dimensional plot of spectrum versus level matters.

I'm quite certain that even back in the 1950s audio engineers knew that THD at 1 KHz didn't tell the whole story. It's well known that some devices, for example transformers, have more distortion at low frequencies. Perhaps hi-fi magazine writers were ignorant, though that's still the case. We also knew about IM distortion, which is even more damaging than HD, long ago. From THIS article at the fabulous Rane library:

"Intermodulation distortion testing was first adopted in the U.S. as a practical procedure in the motion picture industry in 1939 by the Society of Motion Picture Engineers (SMPE -- no "T" [television] yet) and made into a standard in 1941."

Further, the notion that there are still more facets of audio to be discovered is easy to disprove with the null test. I already explained that in post 348 above. Did you not see that section of my video? It's not a lot to wade through and it's highly informative.

DAC jitter is an good example of a parameter that once wasn't recognized at all as potentially having subjective consequence now recognized (not by all, I know) as such.

Aside from being broken, or as Amir pointed out certain HDMI audio devices, jitter is never audible. It's just too soft.

--Ethan

 

[Ken Newton]

I imagine the idea that this could ever happen would go away quickly with a proper blind test. What sort of coloration could be audible such that when combined in series both colorations go away? Please be specific.

You are certainly free to imagine.

For one thing, in my experience, subjective dynamics can be such a coloration. Although, we haven't yet agreed on the definition of subjective coloration. For another, subjective tonality. For yet one more, subjective stereo effects such as soundstage clarity. That's about as specific as I can give about what afe subjective affects. Those are some in my experience, but as I indicated earlier, your subjective milage may vary. Belief or disbelief about my experience is your prerogative.

Ah yes, pre-distortion. I designed such a circuit 35 years ago:

I'll accept that as your acknowledgement that even order distortion is a valid example proving the real potential for an complementary/synergisitc component effect.

But what are the chances that one audio device will have soft clipping onset at a very specific level, and another will have the precisely opposite effect at exactly the same level? How would you even get "anti-clipping" in a circuit unless you were aiming for that specific curve? I truly doubt any hi-fi products that claim "synergy" have such a curve.

The chances of perfect cancellation is very low. But perfection isn't the threshold for determining useful efficacy. The point is that some cancellation will likely result, giving at least some degree of improvement. For example, most any single ended circuit will produce lower even order distortion if given complementary symmetry, even without effort to match the halves. Some improvement is better than none, yes?

I'm quite certain that even back in the 1950s audio engineers knew that THD at 1 KHz didn't tell the whole story...

I feel quite certain a well. Objectively knew about, yes. Fully understood the subjective consequence, I suspect not so much. Evidence for that is the feedback driven THD wars of the 1970's and early 80's. Increased feedback improved single figure THD, while worsening the spectrum (increasing the order) of that distortion. A full understanding and concern over the subjective impact of higher order distortion products once seemed absent despite the objective knowledge of their added presence.

Further, the notion that there are still more facets of audio to be discovered is easy to disprove with the null test. I already explained that in post 348 above. Did you not see that section of my video? It's not a lot to wade through and it's highly informative.

Aside from being broken, or as Amir pointed out certain HDMI audio devices, jitter is never audible. It's just too soft.

--Ethan

I confess that I only briefly skimmed through your AES video. However, one slide I noticed showed jitter, listed under time-domain errors, as one of the four factors which can affect audio. Perhaps, you would reconcile that slide with your above declaration that jitter doesn't matter? I should clarify that I don't harbor a jitter fetish. While I think jitter matters, I also suspect it is a lower tier system issue, assuming it's not so severe as to provoke actual data errors. I'm uncertain as to what you mean by jitter being too soft to be audible, unless you meant as opposed to it provoking hard data errors.

I'll ponder your implied notion that null testing essentially proves that audio perfection is here and now.

 

[amirm]

THD became popular since it was very easy to build such a meter. And it indeed is sensitive to non-linearity. Its output is "psychoacoustically blind" but as a way of teasing out distortion, it served its purpose and hence its popularity.

 

[Don Hills]

I imagine the idea that this could ever happen would go away quickly with a proper blind test. What sort of coloration could be audible such that when combined in series both colorations go away? Please be specific ...

Apply 10 dB of boost at 1 KHz with a filter Q of 2. Audible.
Apply 10 dB of cut at 1 KHz with a filter Q of 2. Audible.
Now apply the two filters in series. Not audible.

 

[Blumlein 88]

Apply 10 dB of boost at 1 KHz with a filter Q of 2. Audible.
Apply 10 dB of cut at 1 KHz with a filter Q of 2. Audible.
Now apply the two filters in series. Not audible.

This reminds me at one time I had an ESL speaker. The low output impedance in the treble would normally have caused treble drop from the high output impedance of my tube amp. Somehow the xfmr and panel had a low Q resonance just over 20 khz. This almost fully flattened out the upper octave . Such things can sometimes happen.

 

[amirm]

We can think of scenarios where such cancellations occur. The challenge is to find them in real product and demonstrate them with blind listening tests. I don't think that has ever been found.

 

[Ethan Winer]

Apply 10 dB of boost at 1 KHz with a filter Q of 2. Audible.
Apply 10 dB of cut at 1 KHz with a filter Q of 2. Audible.
Now apply the two filters in series. Not audible.

Yes, I mentioned that in my Post 342 above. Still, that's complementary not synergy.