Musings on amp design... a thread split

Above was with gain of 28 dB, if I reduce to 20 dB, THD drops below -240 dB at low power, slightly modified circuit too.
-240dB at 10V output means 0.02nV at 20dB gain.
An extremely low noise amp may have 0,5nV/rtHz at its input giving 5nv/rthz at it's output when using 1Hz filter bins at 20dB gain.
That would theoretically limit THD to 20log(10v/5nV) = -174dB , 0.002ppm or 0.00,000,000,2% distortion.
In practice going beyond -120dB is already very hard to achieve as Halcro shows with their ultra low distortion amps.

So I'm puzzled how your -240dB should be interpreted.

Hans
 
Last edited:
Hans, I have already explained that it is impossible to measure the loss of information on pure sinusoidal signals.

How to measure exactly the losses I showed on numerous examples by the compensation method. I don't know of any other method. Everything was detailed in the article on distortion measurements.


As for plotting the dependence of THD on frequency, this is done by the MicroCap program in a step-by-step mode. The algorithm of work was written by Russian professors of the Smolensk University (the married couple Amelins).

In this case, the generator sequentially operates at frequencies from 250 kHz to 10 Hz, in steps of 0.9. You can set the number of generated sinusoids. In this case, using the example of the Rogov amplifier (Radio 2016-10), I showed the results of the calculation using two and ten sinusoids.

If you are familiar with the algorithm of the simulator, you should know that distortion measurements are carried out on the last period (in one case, on the 2nd period, in the other, on the 10th period).
I've read the article on distortion measurements that you gave me long ago, but I'm sorry, it didn't convince me with unqualified vector errors and speed errors.

I'm not at all familiar with the algorithm you refer to, so maybe you have some paper or you could explain yourself in more detail what it does.
Generating a number of sinusoids seems easy, but what happens then, what calculations are made on those signals, the most mysterious of them is High_Y(thd(harm((v(RL)).freq).1) vs freq.
I completely fail to understand that graph.

And since I don't have Radio 2016-10, could you provide the Rogov's circuit diagram ?

Hans
 
So I'm puzzled how your -240dB should be interpreted.
It is -240 dB THD, not THD+N. THD+N is average.

In steady state Microcap can simulate down to -250 dB, not steady state only -208 dB, with my settings.

For comparison I tried BB OPA627 as naked buffer with little gain and got this:

BB_OP627_thd+n_2Vin.jpg
 
It is -240 dB THD, not THD+N. THD+N is average.

In steady state Microcap can simulate down to -250 dB, not steady state only -208 dB, with my settings.

For comparison I tried BB OPA627 as naked buffer with little gain and got this:
Sorry, I didn’t talk about THD+N.
I looked at just one filter bin of 1Hz and mentioned that this will set the theoretical limit for THD distortion measurement at -174dB when having 0.5nV/rtHz at the amps input.
Your plot of the OPA627 with 5nV/rtHz seems to fully confirm this calculation.

Hans
 
-250 dB? Arithmetic on ideal models maybe but no real parts come close. Even more pointless than the hopeless misunderstanding of phase margin in feedback.
I know, for me it is something like a intellectual challenge.

All resistors will be matched as good as it gets, starting from 0.01% parts. At the moment everything important is just 0.01% out of the bag without checking..
I have LS318 dual transistors removed from equipment, all of them have hfe of 421 on both sides and no temperature drift when touching them. The manufacturer seems to have made a very strict selection. Also a quartet of nos MAT 01 with color dots on the top, both sides perfect on all of them.
So, I´m working on it.
 
I understand your point. Did similar things with the development of a class AB-latFET-amp. After several months of simulation I built a prototype and afterwards I spent much time on my test equipment to verify the results until theory and praxis matched somehow, in the ballpark of THD=1ppm. Theoretical improvements beyond any physical limits, outlined by Hans Polak, do not make much sense - because there is no way for us mortals to verify these.
 
Don't know about your schema, but Samuel Groner made a measured -180dB THD amp as posted by Jan Didden in Samuel Groner's super opamp thread.
Currents in an opamp are rather low so the effect of coupling via leads / wires is manageable. In a class A amp it's similar but AB, don't think so. My schema was just an effort to see how far you can go with the cheap MJH6284_ON darlington and the now obsolete LM49860. IOW w/o ppower supply, cost below ~ $10.
 
Don't know about your schema, but Samuel Groner made a measured -180dB THD amp as posted by Jan Didden in Samuel Groner's super opamp thread.
So let's do the calculation again. Input noise was at 0.39nV/rtHz with input short circuited.
Measured with 10V output and a gain of 1000, output noise per Hz was 20log(10V/0.39uV) = -148dB being the theoretical measurable limit for THD.
THD may be below this noise level, but then you can't measure it.
So where did you find the -180dB ?

Hans
 
Last edited: