This is difficult, complex territory to attempt to address at all. However, I think we do know some things and we have some directions that might produce fruitful research.
First, I would like to mention a groundbreaking music theory book by David Huron, called Sweet Anticipation - Music and the Psychology of Expectation. The book is not flawless, and it is somewhat speculative, but I think it's the best we have for now. As usual for certain books I recommend, the serious reader should read and study the whole book including all the footnotes at the end to gain the most.
Having said that, please let me make a try at addressing, or start some conversation about, some of the issues you raise.
When it comes to music and emotion, we don't have deterministic models. Scientifically, what we do have is some statistical understanding at the social psychology level of abstraction (or modeling). Probably we would like to understand more at the neurological level of abstraction, but that's not where most of the action has been for the past 10 - 20 years. Its been in social cognitive psychology. The book I mentioned attempts to take on a new understanding of music theory at that level.
So, to get on with it, if there were the objective tests for accuracy you asked about, I think you would probably already know about them if they currently existed, so no, I don't think we have that at this point (or I would probably have studied it too).
Regarding a universal better sound, I think there is may be something like a sweet spot that is somewhat genre dependent, and that may be arrived at at different stages of music production and reproduction. And to the extent there may be one or more sweet spots, they are probably statistical preferences of the music listening population, and might be mathematically viewed or modeled as local optima of some sound quality cost function. In addition, I think we want to hear some harmonic content and even some IMD. To consider IMD a little more, some of it occurs in nonlinear mixing in acoustic musical instruments, otherwise an out of tune piano or guitar would not produce beat notes, or rich chord textures. To put it another way, if you make music with a synth that produces pure sine waves, in order to get texture to chords, you would need to add some type of nonlinearity, some IMD. But in the end, by the time the sound gets to the loudspeakers, too much IMD and/or THD may be well past any sweet spot.
On the topic of adjusting open loop sound quality by ear, there was the word "minefield." But, I'm not so sure it is a minefield any more than music production is. Perhaps the most important thing for a good mix engineer is to have some musical taste that correlates with what most of the music listening public likes to hear. In that case, you basically just turn the knobs until it sounds good, and hidden processes in your brain, that are often much more powerful than anything conscious awareness can do, do almost all the work. If an amplifier designer does more or less the same thing, but at a very subtle level, adjusting distortion products maybe around or below .01%, then some people who can hear that may be very pleased, at least they may be likely to be pleased along some probability distribution.
I understand its complicated and not highly satisfying, but I have skimmed over a little about what we do know.
Regarding the issue of ending up arguing about the merits of probably inconsequential differences in speaker grills, I don't think that is a necessary end point of not having highly deterministic models for guidance and well principled understanding. It is what you end up hearing in certain circles, and where the usual tendencies for cognitive bias effects and tribal behavior tend to bloom. Often that happens in forums, where people don't know each other, can't see each other's faces as they talk, and can't hear tone of voice and timing as words are delivered. I would try not to get too discouraged by it. Maybe things will improve somewhat in online forums in the future as we move past old text based communications.
Last edited:
Agreed, it would be a logical fallacy if somebody did say that. I was commenting on the existence of something else, the overconfidence bias.
Last edited:
What you mean tried to work silver into the fab process or maybe sent samples out to The Absolute Sound/Stereophile and ran new $100,000 mask sets and fab lots until the sound was "approved".
I'm sure some marketing/apps folks will spin you some spin the reality is for any major consumer manufacturer it's numbers. Zero feedback class A outputs with 70dB THD don't cut it. Let take JC's most recent comment (he's wrong BTW the 797 never goes into pure class B operation, I'll get back to that in a moment) about class A outputs. If anyone wants to claim they're designing for audio let them show how they significantly compromised the quiescent current to stay more class A, that is run several extra mA's in the output just to make the "audiophile community" happy.
A diamond output tries to maintain a geometric mean relationship in the two output transistors. Modern high speed devices with low non-linear base charge or quasi-saturation can drive a 600 Ohm load at 20V p-p and never FULLY cut off at 500uA of standing current. The transfer function remains smooth and continuous through crossover. You then have evil feedback or some form of error correction to reduce the effect.
Experience with PA's driving speakers that can get down to a couple of Ohms with low ft power transistors with huge stored base charge does not apply here. Sticking with op-amps here, IC PA's only have audio as an application so by definition they were designed for audio. As for real op-amps not much more than the 5534/5532 come to mind. It has several compromises in its DC performance that might benefit audio performance.
Last edited:
Scott,
National over the years has done specific for audio opamps. The 4562 was actually subjected to listening tests. Of course for marketing reasons it did get relabeled.
Now as to the AD797 being so popular, perhaps you can give us optimum component values to support it as a buffer or 30 dB gain amplifier.
National over the years has done specific for audio opamps. The 4562 was actually subjected to listening tests. Of course for marketing reasons it did get relabeled.
Now as to the AD797 being so popular, perhaps you can give us optimum component values to support it as a buffer or 30 dB gain amplifier.
No I wasn't asking about "stereophile validated", more like "the engineers designed it for audio as the primary application". I.e. not video buffer, instrumentation, electrometers etc.
The 5534 was on my mind too. There are others, like the 4562 that Simon mentioned. NJR probably has a few too, and I suspect that TI's "SoundPlus" ones may qualify.
There are a few debatable ones like OPA604. We can only guess, but if anyone has any insights please let us know.
The 5534 was on my mind too. There are others, like the 4562 that Simon mentioned. NJR probably has a few too, and I suspect that TI's "SoundPlus" ones may qualify.
There are a few debatable ones like OPA604. We can only guess, but if anyone has any insights please let us know.
Unless listening tests were used as a design tool, it's just stories. As mentioned "I think the sound could be improved let's delay release and do another mask set". Post facto listening to validate some product pitch is not design.
At 10nA Ib the 4562 would have Ib comp, the lack of which gives the 5534 distinct advantages in audio. Doug Self has made this abundantly clear to IC designers for a long time. 1.6pA/rt-Hz vs .4 pA/rt-Hz is possibly audible in a MM phono stage. I see no advantages offered in audio with Ib comp especially with low gain AC coupled circuits.
Last edited:
My source on the listening test was the late Bob Pease in one of his columns.
The problem the 5534 has in high end audio is that it is too cheap to be any good. Doug Self has done a lot of opamp rolling with 'dedicated' audio units costing 20 or 50 times as much as the 5534 on a per-opamp basis. Yet, again and again he got back to the 5534 (or its dual, the 5532), although he conceded in I think SSAD 2nd Ed. that the 4562 gets very close, albeit at a multiple price.
I spend half a day at an AP 2722 at Grimm audio selecting opamps for my autoranger and the toss up was 5534 and OPA2134; I did select the 2134 solely on the grounds that its FET input caused much less offset with higher source impedances. But the 2134 cost I think about 10x as much as the 5534.
Edit: want to comment on Mark4w's post above, which I find insightful. I want to mention another factor Mark, and that is that putting an opamp in a circuit seems 'too easy'. Of course, doing a good job really isn't easy, which you quickly realize once you try it, but for a lot of folks it seems that way. Something that is much harder to do like a discrete stage is therefor 'better'. Similar to the fact that you have more satisfaction from something that cost significantly in money and/or effort than something you get for free.
Jan
I spend half a day at an AP 2722 at Grimm audio selecting opamps for my autoranger and the toss up was 5534 and OPA2134; I did select the 2134 solely on the grounds that its FET input caused much less offset with higher source impedances. But the 2134 cost I think about 10x as much as the 5534.
Edit: want to comment on Mark4w's post above, which I find insightful. I want to mention another factor Mark, and that is that putting an opamp in a circuit seems 'too easy'. Of course, doing a good job really isn't easy, which you quickly realize once you try it, but for a lot of folks it seems that way. Something that is much harder to do like a discrete stage is therefor 'better'. Similar to the fact that you have more satisfaction from something that cost significantly in money and/or effort than something you get for free.
Jan
Last edited:
Dac's are different they can be prototyped via FPGA in large part. I would like to see a story of change affected by listening without, for instance, saying something is not quite right and looking for a shortcoming in measured performance. Or simply measuring performance vs simulation.
http://www.diyaudio.com/forums/digi...-claim-dac-pushes-boundary-audio-science.html
Funny that within minutes 24-30 he mentions listening tests where people were able to hear unmeasurable differences.
I wonder if there are any available examples of the opamp folks similarly engaging with the audiophile public.
P.S. Another link to the video in case the link in the mentioned post doesn't work for you:
https://www.youtube.com/watch?v=8Mn5PrnZV-k
Funny that within minutes 24-30 he mentions listening tests where people were able to hear unmeasurable differences.
I wonder if there are any available examples of the opamp folks similarly engaging with the audiophile public.
P.S. Another link to the video in case the link in the mentioned post doesn't work for you:
https://www.youtube.com/watch?v=8Mn5PrnZV-k
Last edited:
If you look at the specs of the ESS parts, they clearly are designed to be the best measuring.
If I were selling $75 audio DAC ICs I would like to have a good narrative to go along with that too...
Isn't it a part of business to have a good narrative when trying to sell anything? Such things are studied and taught at business schools.
Similarly, maybe it helps to have a good narrative when arguing against believing somebody saying things one has already decided in advance must be untrue? In other words, don't both sides in a dispute have an interest in constructing plausible narratives?
Irony again, Martin is a good friend for more than 30yr. and he thinks the AD797 is the greatest op-amp ever
. In fact the ESS apps guy for Asia asked ADI if I could come on one of their sales trips so we could have dinner and talk design. The Chinese take hi-rez cell phone audio even through ear buds very seriously mainly as a vehicle to sell data bandwidth.On the listening tests did Martin say HE heard the difference, I suspect the protocol might not pass muster either. IME the possibility anyone in a senior position in Asia being wrong in a rigorous DBT would be considered simply impossible and the height of rudeness. That is all listening we did was sighted with predetermined result.
Being time to retire I actually forgot that I worked on the AD1862's internal I to V op-amp in 1987. Designed to meet the numbers and be right first time all the way.
Last edited:
If the numbers were just under AK, for instance, no amount of certified GEB endorsements would earn a sale unless of course you slash priced it.
Now, that rings true. He did say he could not hear it, and cited tinnitus as a reason. But, very curiously, he said all the executives at ESS could hear it, which seems like a very odd coincidence for a group of old guys.
What I call distortions in audio, is any (inter-)modulation of the signal. The paradox is, some distortions are more audible but harder visible on standard measurements, and vice verse. That's why different amplification means introduce more of measurable, less of audible distortions, and vice verse. This steers up heated debates, if more distortions sound better or not.
What I am always trying to achieve, less of audible distortions. As less as possible. And it results in the sound that everyone prefer, expect some Cargo-Cult Audiophiles that judge quality by audibility of "signatures" of "famous equipment".
Of course marketing is to be expected. My point is, what do you expect them to say? "Our new DAC measures the best but probably sounds indistinguishable to properly implemented top chips from TI, AD, and AKM?"
I think the marketing strategy has to touch on "the sound" because the excellent measured numbers and integrated ASRC are already making the alt-audiophile crowd nervous.
How else are you going to get proponents of open-loop circuits and multi-bit R2R to gladly shell out cash for a super expensive sigma delta DAC with tons of digital witchcraft built in?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.