The accent isn't a problem but he's a tedious presenter, even if he is an expert EE.
Couldn't stand any more after the first half so I finally did the derivation myself.
I found my mistake and now understand the role of the unity gain frequency.
That will be set by the GB product for a simple first order compensated op-amp, as seemed reasonable.
A bit more complicated if more than first order, now I better understand the AD797 compensation recommendations.
Very educational to work this out so thanks for the push.
Best wishes
David
I find a lot of folks don't use the distortion reducing capacitor simply because they can't roll op-amps into the same socket.
The notes in the AD797 data sheet (p.15 rev. K) say the distortion reduction capacitor is beneficial if the gain is more than 10, by implication not useful for an inverter.
So is it useful for the inverter or integrators?
I would like to do a better job than the typical "op-amp roller"
Best wishes
David
Last edited:
Distortion is reduced no matter what (noise) gain. What they wanted to say is that at low (noise) gains distortion is very low anyway, and not easily measured. That is a relative judgment, and our context probably warrants a different point of view.
What a careful read of the datasheet, and/or a bit of practical experimentation, will reveal is that the AD797 is not unity-gain stable if the inverting input sees a low impedance at HF. IIRC not mentioned in the datasheet is that the tendency of instability is higher with the distortion cancellation capacitor installed.
In an integrator, exactly this condition is present. So be prepared to come up with a suitable fix.
Samuel
That was my expectation, just wanted to have it confirmed and check if there were any subtleties I missed.
Yes, I note the requirement in the data sheet for 100 ohms of series resistance, preferably by-passed with 33pF.
That was the bit I understood better after I studied it a bit.
My first idea would be for more capacitance, to land the phase correction directly on top of the 16MHz unity gain frequency.
That's only first order approximation, do you have any experimental data on this?
Best wishes
David
its still my belief that the 797 "error correction" connection of the compensation capacitor is, like HEC, a feedback technique - not Black's 1st feedforward error correction
http://www.diyaudio.com/forums/soli...terview-error-correction-337.html#post1513928
http://www.diyaudio.com/forums/soli...-blowtorch-preamplifier-1770.html#post1809194
http://www.diyaudio.com/forums/solid-state/123613-class-biasing-ad797-6.html#post1525757
and like HEC it can be viewed as a local positive feedback boosting local loop gain around the output stage or as canceling the internal Ccomp roll off of gain V for that local loop
which will of course have stability consequences - try "measuring" in sim the loop gain around the output, cutting the Ccomp cancellation too
http://www.diyaudio.com/forums/soli...terview-error-correction-337.html#post1513928
http://www.diyaudio.com/forums/soli...-blowtorch-preamplifier-1770.html#post1809194
http://www.diyaudio.com/forums/solid-state/123613-class-biasing-ad797-6.html#post1525757
and like HEC it can be viewed as a local positive feedback boosting local loop gain around the output stage or as canceling the internal Ccomp roll off of gain V for that local loop
which will of course have stability consequences - try "measuring" in sim the loop gain around the output, cutting the Ccomp cancellation too
Last edited:
Crossover distortion is a function of output current. Unless the load is known, the output current can't be derived from the input voltage of the output stage. Thus I think feedforward control is in no way applicable to reduce crossover distortion (unless I'm missing something, which is not unlikely).
Samuel
Just reinforcing your thoughts! I have surely observed the effect on stability in transistor-level simulations of this topology, and IIRC it prevents the use of the technique for power amplifiers.
Are you sure a 33 pF by-pass capacitor would improve stability..? I think having a resistive HF impedance is what helps.
It's been a while since I've used the AD797, but IIRC with the distortion cancellation capacitor in place, 150r is needed. Also consider a low-Q inductor/ferrite bead instead of a resistor (lower noise in the audio band), a Zobel network at the output or a feedback network arrangement that increases noise gain at HF (series RC from inverting input to ground).
Samuel
Are you sure a 33 pF by-pass capacitor would improve stability..? I think having a resistive HF impedance is what helps.
It's been a while since I've used the AD797, but IIRC with the distortion cancellation capacitor in place, 150r is needed. Also consider a low-Q inductor/ferrite bead instead of a resistor (lower noise in the audio band), a Zobel network at the output or a feedback network arrangement that increases noise gain at HF (series RC from inverting input to ground).
Samuel
Yes, IIRC Cherry tried it. He was rather optimistic about the results, though.
Whether justified or not, many high-end audiophile designers shun the use of ferrite beads. Nevertheless, they are a handy tool for stabilizing many circuits. If they were not shunned, they might usefully substitute for base stopper resistors on power transistors. I'm not sure if anyone has tried to measure distortion created by a ferrite bead with the level of detail and sensitivity that has been applied to investigations of capacitor distortion. Bruce Hofer would probably have some really good observations in this regard.
Cheers,
Bob
I have used ferrites a lot. There are many formulations. Some will saturate more easily than others. But if kept within bounds they are fairly benign. When you need to filter at low freqs you might run into issues due to iron and nickel being popular mix in the alloy for low freqs (like ac power line). But some exotic materials I have measured for speaker crossovers have nothing to worry about even at high currents. So it just depends on your choice of materials used in the ferrite for a particular app. Which materials are you interested in measuring... I still have a stash of various ferrite types.
THx-RNMarsh
Last edited:
Not sure, but the data sheet mentions it and it should put in a Bode step close to the unity gain frequency and improve the phase.
So it seems plausible.
This is very relevant, I wasn't sure if the omission in the data sheet was because the cancellation capacitor was just unnecessary for most applications or actually problematic.
Like Bob, I am wary of the ferrite but not obsessive.
More ideas to consider, thanks.
I initially expected this too but I posted previously why most of the contenders looked problematic in one area or another.
I am open to specific recommendations but so far only the OPA1611 looks even equivalent and I don't have independent data to confirm this (a hint for Samuel
).Scott says the main market driver is lower power consumption so that explains some of the reason why newer is not better.
Presumably a better cost-no-object op-amp could now be made but insufficient market, if it's really needed then it's possible to use a composite, as you have remarked.
Best wishes
David
Last edited:
Intelligent comments, you have to understand that there is a very limited demand for this at this time. What could be done is nice but who is going to pay for it.
Composites or what about a discrete circuit. Like the SWOPA? The thd instruments I have (3) all use discrete front ends. Why cant we design a discrete? Also wont have to worry about a low volume opamp being cut from the line-up. Easier to do a composite though, I suppose.
THx-RNMarsh
THx-RNMarsh
Last edited: