Ha, ha, so much ignorance.Mirlo said:
It's like someone wanting to cure cancer without having any medical background.
Hans
Just out of curiosity, what is the intended application for such a low distortion amplifier?
I mean, once you reach the point of inaudible distortion, is the goal the same as for most typical engineers - to keep pushing the boundaries just because you can?
Maybe the intended application is more toward the realm of instrumentation (no, not guitar! 😀)?
Perhaps you want to sell the technology to the good folks at LIGO for their endeavours in finding black hole collisions etc? But they seem to have been happily finding them using plain vanilla op-amps..
So (and this is an honest question) why go so low?
I mean, once you reach the point of inaudible distortion, is the goal the same as for most typical engineers - to keep pushing the boundaries just because you can?
Maybe the intended application is more toward the realm of instrumentation (no, not guitar! 😀)?
Perhaps you want to sell the technology to the good folks at LIGO for their endeavours in finding black hole collisions etc? But they seem to have been happily finding them using plain vanilla op-amps..
So (and this is an honest question) why go so low?
The clearly stated goal was to get some kind of "certification" from respected Forum members so that could be shown to investors and entice them to part with some of their money, "presumably" the bait at the end of the hook being some kind of exclusive/patentable idea which could generate Audiophile interest and a nice cash flow.
Yeah, sure. Once I get my DM up and running and find some spare time again I'll show you and we can either debunk or confirm.
Yes, this. You are much better at words than I. Totally agree.
Well I don't know what the real world number is but in spice the gain of the EC is about 100 quadrillion when building the full scale version. I use a combination of techniques and unusual configurations to get to this point. But it is surprisingly stable in general for what it is. As you've seen from my progression of screenshots I can increase the gain and performance incrementally, so the last screenshot I posted was only about 40% of the gain I can set it at and it is already under the noise floor of the analyzer at -100db+ performance improvement. It becomes increasingly complex and annoying to build as I go up though. Even without using the DM I can still guestimate where the dice lie on the other side of the noise line, however with the DM I have been able to see harmonic peaks down to -150db. Scott seems to question whether my DM is actually seeing down this far but I suppose I'll find out after I buy a new wall wart.
Not quite. I just want a way to measure it. I never said anything about imposing recommendations from anyone here.
Last edited:
Good idea, the most productive thing to is repeat what you did and show us your interpretations of what you see. I already pointed out that your "0dB" was really -20dB THD (it was right on the QA display) the fundamental was simply 20dB off scale. The simplistic view that some EC scheme simply reduces everything by -150dB without device or BW limitations is wrong.
Please make note of all actual levels in each instrument/device at each stage of the experiment. The QA will overload dramatically all of a sudden.
SPICE is another fantasy, you can make a twin tee with -200dB notch and -300dB distortion. BTW -300dB is the numerical noise floor of 64 bit FP math. -300dB distortion on say a 1kHz sine wave at 100mA is essentially individual electrons moving in lock step.
Last edited:
I want to address the audio gear marketing issue, and apologize in advance for the rather lengthy comment which follows.
In general, high-end audio is very much about overkill. It's a seperate, but interesting discussion as to why, exactly, unimpeachable performance metrics don't seem to reliably predict a commensurate subjective listening experience. Which, it seems, the metrics should reliably predict. If they did, then we could all simply buy any value priced mass-market AV receiver or disc player, and be done with worrying about the hardware to focus only on enjoying the music. Of course, that is exactly the process many music lovers take, and are left to wonder about all of the high-end market nuttiness.
At any rate, back to the marketing issue. In any industry, a business not planning to compete on having the lowest commodity type pricing must then plan to compete on having some differentiating factor. For vendors in the high-end audio industry, competing is all about differentiation. So, if a vendor can rightly claim to have some best specification or another, that then gives them a competitive advantage in the market. It gives potential customers a reason to buy from them versus buying from some other vendor, and potential investors a reason to invest.
It doesn't much matter whether some 'best in the industry' specification is vast overkill, as most customers lack the knowledge to properly weigh such things, an so, look to convenient figures-of-merit, such as THD to assess which products are better. This is akin to the PC processor clock speed war that Intel used to wage and largely compete on. Knowledgable PC customers realized that CPU speed was not the only factor limiting subjective PC responsiveness. RAM access times, memory transfer bandwidth, cache size and levels, etc. are all factors. However, not only did most PC customers not understand that, they didn't want to understand that. They didn't have the time nor inclination to understand that. Nor should they have to. They've got other things to think about. So, in the minds of most such customers, simple concise performance metrics, such as CPU clock speed, are looked to for assessing which products are 'best'. If a vendor can establish a brand identity of excelling on certain product metrics, their brand then becomes a proxy in the customer's mind for which products are 'best' on the market, whether that's accurate or not. In pursuit of that brand identity goal, engineering departments are, indeed, pushed to keep extending the boundaries of performance metrics that have become established in the customer's mind as defining which products are 'better' than others.
In truth, the life of all customers would be so much easier if a small set of available metrics did accurately predicted a user's EXPERIENCE with that product. Many will argue that such an set of metrics has long existed for audio products. Others will argue that such objective metrics don't reliably enough PREDICT an user's after purchase experience. This seeming disconnect between objective hardware metrics and subjective human experience drives a good bit of the high-end audio market, I suspect. Other factors drive it as well but, I suspect, to a lesser degree. There are other market based reasons, aside from differentiation, as to why the specialty audio business is so easy to enter yet so difficult to stay in long term, but won't address them here.
Last edited:
That's an easy one. The paper from Samuel linked in the 1st post discusses that, he had exactly the same issue as you have. He ended up measuring to -180dB.
I used his method in my measurements later in the thread. I got about -160dB.
It requires all of one (1) resistor. If you are uncomfortable using that method with your amp, ask here, but study it first. And no need to disclose details of your amp or your EC.
Jan
My references are to the original thread, Samuel Groner's super opamp, where this thread split off from. Sorry.
Jan
Yeah, for audio I could care less, but not by much. 🙂 Other than the pursuit of more zeros for its own fun, as you write.
On a personal/professional basis, if I could *trivially* get read electronics to have thd+n in the <-160dB, ESPECIALLY with sub-fA input bias currents, I could build some very cool electrochemical sensor arrays with fast settling times.
So, yes, I'd love essentially perfect read electronics for scientific/medical instrumentation. Then I could be confident any spuriae were from other sources, and not the electronics, which would be enabling.
I should say that in a lot of ways I'm already at a point where the read electronics are not a limiter, but it's not trivial, and the latter might mean I could scale up to higher parallelism and faster read times over an array.
Last edited:
to get below 150 db you probably need to have the amp placed in isolation system deep underground and prevent any radio-waves, A and B and Gamma radiations and vibrations from perturbing the device...
Markw4 said something interesting, I can correlate where I am at now in my findings of how to judge the quality of a system...
Sure, the THD is the number 1 indicator of a good system and generally applicable. Then this must be seen relative to some power levels, both very low under 0.1 Watt and at some substantial power like 40... and relative to bandwidth at the extremes.
Then I find that the response to capacitive and reactive loads is a must have data, this would otherwise be audible...
Then the square waves at low frequency tells a lot of the bass response and power supply quality, applicable to any device again from opamps to power amps.
IMD, is very relevant IF the design has little feedback as a quality factor, IMD, with proper feedback amp wont reveal any 'audible' problems, and it will be very low.
The best test should be square waves combined with their up to 11th harmonics.
The test consist of sending perturbation signals, such as noise, other frequency square waves.
You compare the multi-signal square wave with /and without the interference, in a dynamic analysis with only one oscilloscope signal take (on/off)
then in continuous, with and without perturbation.
the difference between the 'burst' and the continuous, is what should happen in a dynamic signal.
Markw4 said something interesting, I can correlate where I am at now in my findings of how to judge the quality of a system...
Sure, the THD is the number 1 indicator of a good system and generally applicable. Then this must be seen relative to some power levels, both very low under 0.1 Watt and at some substantial power like 40... and relative to bandwidth at the extremes.
Then I find that the response to capacitive and reactive loads is a must have data, this would otherwise be audible...
Then the square waves at low frequency tells a lot of the bass response and power supply quality, applicable to any device again from opamps to power amps.
IMD, is very relevant IF the design has little feedback as a quality factor, IMD, with proper feedback amp wont reveal any 'audible' problems, and it will be very low.
The best test should be square waves combined with their up to 11th harmonics.
The test consist of sending perturbation signals, such as noise, other frequency square waves.
You compare the multi-signal square wave with /and without the interference, in a dynamic analysis with only one oscilloscope signal take (on/off)
then in continuous, with and without perturbation.
the difference between the 'burst' and the continuous, is what should happen in a dynamic signal.
...and a lot of liquid nitrogen!
Am i wrong or somebody screwed the title of this topic ? suggestion: +290 db noise!
It's affecting the ecosystem , i already see the birds going south!
Last edited:
A gain of 100 quadrillion?
A quadrillion is 1 with 15 zeros behind it.
100 quadrillion would be 100,000,000,000,000,000.
If you input one uV (One Microvolt, I.E., 0.000001V ) you would get 100,000,000,000 (one hundred billion) volts out.
That is absolutely not possible. (On this planet anyway...)
The only reason I bring this up is to please check your math a little better. This is the root of the issue ; -290db is also an impossible number, and you keep saying that -150db is trivial, when it's a significantly difficult under the best circumstances.
Last edited:
Well I said in plain English that I don't know what the real number is and that is just the spice number. That is the open loop gain of the circuit when built and measured in spice @ 20khz.
Anyway I read the paper here Low-Distortion, Low-Noise Composite Operational Amplifier - Semantic Scholar and I couldn't find anything about testing methodology. I'm assuming this is not the full paper?
Oh nevermind, I missed Jans last post. I found it. I don't understand what it means by testing the amplifier under high noise gain.
Anyway I read the paper here Low-Distortion, Low-Noise Composite Operational Amplifier - Semantic Scholar and I couldn't find anything about testing methodology. I'm assuming this is not the full paper?
Oh nevermind, I missed Jans last post. I found it. I don't understand what it means by testing the amplifier under high noise gain.
You seem to be Google challenged. Let me help you.
The noise gain is the gain from the +input with the -input grounded. If you put an R between the +in and -in pins you increase the noise gain while the signal gain remains the same. The noise gain determines the distortion reduction by feedback.
So if you have an amp with gain = 20, and you can't measure the distortion, increase the noise gain by 40dB and the distortion becomes 40dB larger and easier to measure. You then subtract 40dB from the result.
Jan
The noise gain is the gain from the +input with the -input grounded. If you put an R between the +in and -in pins you increase the noise gain while the signal gain remains the same. The noise gain determines the distortion reduction by feedback.
So if you have an amp with gain = 20, and you can't measure the distortion, increase the noise gain by 40dB and the distortion becomes 40dB larger and easier to measure. You then subtract 40dB from the result.
Jan
There's something to consider here, the amp has to be stable at the noise gain used. The S/P-composite has a phase margin of -10deg(!) at 60dB according to the paper and therefore is really on the edge of oscillation at this gain, square wave response looking really bad. According to the graph, with 42dB or so I think it would oscillate right away (PM=0).
Works in inverting mode as well as in non-inverting mode (always connect the resistor from (+)in to (-)in).
See OPA211 datasheet for example.
So what happens in a scenario where one of the pins isn't grounded? For example what if you place a resistor between the + and - pins on the second opamp in the composite shown in the paper? Neither pin is truly grounded, does it still increase the noise gain while maintaining the signal gain of the second opamp?
- Status
- Not open for further replies.