Distortion
I'm somewhat hesitant to ask this, because it may have been addressed previously and, like a lot of topics, is potentially contentious. But, here goes.
A couple decades ago, a fellow from the BBC proposed yet another approach to distortion testing.
http://www.hopkins-research.com/information/electroacoustics/A new distortion measurement.pdf
In that article, he refers to an earlier classic paper by Brockbank and Wass that discussed intermodulation distortion products.
I'll quote from Dr. Belcher:
"One interesting result given by Brockbank and Wass is that if the programme signal is assumed to be represented by n tones, each of equal power, and if n is greater than 30, then the contribution to the total distortion power due to harmonic products is at least two orders of magnitude less than that due to intermodulation products. For complex signals such as those produced by speech and music, n is generally large enough for the distortion power level contributed by harmonic products to be negligible. They also derive an equation for the total distortion power, assuming that n is large enough for harmonic products to be ignored, which shows that the power of the distortion signal resides mainly in products generated by higher order terms in the power series."
So, assuming that is true (I'm sure somebody will set the record straight on that for me...), what level and character of harmonic distortion would be considered acceptable to produce suitably low levels of IMD for musical signals?
I'm somewhat hesitant to ask this, because it may have been addressed previously and, like a lot of topics, is potentially contentious. But, here goes.
A couple decades ago, a fellow from the BBC proposed yet another approach to distortion testing.
http://www.hopkins-research.com/information/electroacoustics/A new distortion measurement.pdf
In that article, he refers to an earlier classic paper by Brockbank and Wass that discussed intermodulation distortion products.
I'll quote from Dr. Belcher:
"One interesting result given by Brockbank and Wass is that if the programme signal is assumed to be represented by n tones, each of equal power, and if n is greater than 30, then the contribution to the total distortion power due to harmonic products is at least two orders of magnitude less than that due to intermodulation products. For complex signals such as those produced by speech and music, n is generally large enough for the distortion power level contributed by harmonic products to be negligible. They also derive an equation for the total distortion power, assuming that n is large enough for harmonic products to be ignored, which shows that the power of the distortion signal resides mainly in products generated by higher order terms in the power series."
So, assuming that is true (I'm sure somebody will set the record straight on that for me...), what level and character of harmonic distortion would be considered acceptable to produce suitably low levels of IMD for musical signals?
Same old same old.
Let me rephrase this in a more unassuming (realistic?) way:
"Every attempt to invoke the currently known and applicable laws of physics, the EE body of knowledge based on simplifying assumption of the previous, any STEM based rationale based on the previous times 3 (math are not approximations)..."
There is some truth to that, for example: https://www.diyaudio.com/forums/dig...8-modding-topping-dx3-pro-11.html#post5726922
Same old, same old...
It is known that audio reproduction requires some advances in the laws of physics, quantum electrodynamics is not fine enough to explain our hearing process.
The same EE body of knowledge (based on simplifying assumption of the previous) helped detecting gravitational waves, track spacecrafts 10 billions of miles away from Earth, via the Deep Space Network received signals (20 billion times weaker than the power of a wristwatch battery), etc... Obviously not sensitive enough to compete with our human hearing.
Etc...
Last edited:
That was sort of implied in my question. I guess poorly implied.
For example, if the threshold of audibility is -30 dB (first number that comes to mind - not based on anything but that), what level of harmonic distortion is required? From the original paper, it's clear that higher order distortion products contribute a lot toward the IMD power. But, just what is needed?
I should add that since the math shows that the IMD power is a couple orders of magnitude higher than the total harmonic distortion power, the IMD would be more audible.
Last edited:
It depends. The two examples mentioned in that link are ones that were designed to measure as well as possible, at the lowest possible cost. And, in fact they can't sound very good. By now I have examined enough dac circuit variations I can tell how the designs work to very large extent just by looking at pictures of the circuit boards, and of course, the prices makes sense in the context of what is on the boards. Kudos to the engineers though for close to acing the measurements at that price point.
So, what is going between measurements and listening then? I will qualify my response right off as IMHO. For one thing we still don't know how to correlate measurements very well with aural perceptions as processed in the brains of different people. It could also be that we aren't measuring all the things we could be. S-D dacs are inherently very non-linear and I don't mean that in the sense of something like a power amp that may have a fairly stationary, but somewhat curved voltage transfer function. S-D dacs can be non-linear like fairly bizzare DSP algorithms. Measuring with steady tones doesn't reveal anything about how an S-D modulator reacts to dynamics.
Something like even more radically bizarre DSP goes on in human brain processing of sound. Modern science still has a lot to learn about that. That is probably the same reason that as a practicality Jakob2 advises us that broad sweeping conclusions cannot be drawn from the results of any human perceptual experiment. Unfortunately or not, human brains also are always automatically trying to extrapolate generalities from specifics. Referring to the two system model of cognition, the results of such automatic processing running in System 1 often appear in the form of leaps of realization in System 2, conscious awareness. We just 'see' that the generality is true, and we generally accept such 'obvious realizations' without question, and do so despite the fact that they are often wrong. Such 'realizations' are only much more likely to be right not particularly in smart or intelligent people, but in trained experts in their particular field of expertise. Unfortunately, expertise in one field does not automatically carry over to any other parts of life in the forming of judgements nor in decision making.
Last edited:
I´d say if a comparable amount on high quality audio sensory tests would be spent like in the food industry for their tests, chances would be quite high to settle a lot of these things.
Afair expenses for gravitational waives detection projects were combined over the years somewhere between 1.1 - 1.5 billion $ .
Afair expenses for gravitational waives detection projects were combined over the years somewhere between 1.1 - 1.5 billion $ .
Last edited:
That would be sad indeed, for both sides. It's already done, known as THX and MQA.
Last edited:
So we are back to this again 'I can hear it but you cant measure it'
This is the mystique stuff I was referring to earlier.
Anyway, subjectivists are from Venus and objectivists are from Mars and pragmatists are from the moon so we will never agree.
Peace.
Clowns & Jokers....
Clowns to the left of me, jokers to the right . . . | Stereophile.com
"The dichotomy between what is measured and what is heard has resurfaced in recent months. Jon Iverson discussed"....
//
Clowns to the left of me, jokers to the right . . . | Stereophile.com
"The dichotomy between what is measured and what is heard has resurfaced in recent months. Jon Iverson discussed"....
//
Staying with what I said... for low level signals (not PA) and for IC Opamps.... How many good IC Opamps have access to comp pin? Still need output buffer for HPA and other low Z apps.
Without that access, i am forced to go discrete circuitry.
I always do all those other things as well, -- layout, grounding, component parts selection etc for best results.
THx-RNMarsh
Last edited:
And still no definition of 'not very good'. We've been around this circle already this week.
Same old stuff works very well. Same for if you are designing for lowest noise. same old noise formulas there too. Or any other long known subject. I keep it simple and the distortion is very low. Sounds great and it works for me like that.
THx-RNMarsh
This is very good. I'd like to hear it compared to any particular high VFB IC Opamp. However, this is discrete and not an IC Opamp. Which is the way I go also. But certainly not a simple circuit in parts count. But thats OK.
What is the buffer PN? Cant see it on your schematic. And, the model of the buffer in your SIM? It Measured very poorly.
THx-RNMarsh
Last edited:
You can try this.
(there are other techniques as well where you run the actual opamp itself at high gain)
But, if you want low loop gain, why are you not happy using a NE5534? It has much lower loop gains (10^5 IIRC) than modern opamps, access to comp pins and you can tailor the loop gain to exactly what you want.
I am not anti-discrete, but I don't buy that 'discrete is better' for any reason.
(R2 in the atrached diagram allows you to bias the opamp output stage into class A i.e. Ibias ~0.6/R2)
Attachments
- Status
- Not open for further replies.