An opamp, driving a class A buffer in the loop will get you to 100 ppb - so about 15x better than an AP272 . . .
It is worth triying, I'll make a test tomorrow
I'll try it too, but I think it will stop oscillations: the final effect is comparable to the feedforward cap I tried earlier
I don't know of any opamp capable of <10e-7 at over 100KHz
All the capacitors are solid tantalum, they combine very low ESL with moderate ESR, the best mix for damping.
I assembled an opa1611 driving a lme49600 on Friday and did some testing yesterday.
Squarewaves good from 1kHz to 100kHz.
But at 10Hz there is some slope (~20% of signal) on the "flat" sections. I think this indicates roll off of ~-2.4dB @ 10Hz.
Yet there is no input filtering. It is DC coupled through out !
^
Looks like o'scope prob (AC-coupled, mabye AC/GND/DC switch is broken). Does the generator signal look OK?
-----:-----
A sub-ppm buffer at 100kHz.... how do you plan to verify this once it's debugged? Just curious....
Looks like o'scope prob (AC-coupled, mabye AC/GND/DC switch is broken). Does the generator signal look OK?
-----:-----
A sub-ppm buffer at 100kHz.... how do you plan to verify this once it's debugged? Just curious....
There is a sort of in-built mechanism for checking, through intermodulation (which is basically the same as THD, just generalized to any frequency).
If I detect IM products downstream, it will mean the buffer is not up to the task.
It seems to work, as I have already had to rule out all current high perf opamps I could get hold of.
Think of PIM measurements: the test instruments need to be OK for sure, but then the measurement floor can be much lower.
I habitually use AC coupled input.
I'll check again on DC input.
I'll just accept the lme49600 datasheet info: 0.00003% (300ppb)
It's only a headphone driver/buffer.
OK, I have found the time to make some tests
But this negative opened up new possibilities; in fact only one is left: the cascode stage. I disabled Q5, and bingo, oscillations were gone!
Since this type of cascoding has nothing unusual and works properly most of the time, it is probably the combination cascode + CFP which has such a weird effect, so weird that the sim didn't catch it.
With this knowledge, I concentrated my efforts on Q5, but the problem seems exceptionally persistent: I tried base and emitter stoppers, and even with a combination resistor + ferrite bead, it wouldn't go away. If anything, the remedies seem to make things worst.
I'll have to think about it overnight, try to figure out what's going on. Very strange.
This one had no effect, nothing positive anyway.
Quite surprisingly, this one had no effect either. This was completely unexpected, as I imagined the current stasis loop played a big role in the instabilities.
But this negative opened up new possibilities; in fact only one is left: the cascode stage. I disabled Q5, and bingo, oscillations were gone!
Since this type of cascoding has nothing unusual and works properly most of the time, it is probably the combination cascode + CFP which has such a weird effect, so weird that the sim didn't catch it.
With this knowledge, I concentrated my efforts on Q5, but the problem seems exceptionally persistent: I tried base and emitter stoppers, and even with a combination resistor + ferrite bead, it wouldn't go away. If anything, the remedies seem to make things worst.
I'll have to think about it overnight, try to figure out what's going on. Very strange.
Here's another experiment that might further the divide and conquer strategy.
With the stasis loop disabled as described in my earlier email, you report that the oscillation remains. You also report that eliminating the cascode transistor
Q5 stops the oscillation.
Perhaps you've done essentially this already with the base and emitter stoppers...but, what happens if you remove C2 while the cascode transistor is still in place?
With the stasis loop disabled as described in my earlier email, you report that the oscillation remains. You also report that eliminating the cascode transistor
Q5 stops the oscillation.
Perhaps you've done essentially this already with the base and emitter stoppers...but, what happens if you remove C2 while the cascode transistor is still in place?
Member
Joined 2009
Paid Member
Perhaps, but the original circuit looked stable too. Meddling with the current loop without modifying the cascode will only have minor effects, if any
This will essentially leave the 6 diode's dynamic resistance as base stopper.
I suppose that at some level, it will manage to kill the oscillations, but at the cost of the HF performance, which is important in my case
6.8Ω + 270nH ferrite bead
That's true...it will affect the high freqs, but if it succeeds (or fails) in stopping the oscillation, it may give insight into a better way to find an answer that you can live with.
Any changes with capacitance between Q2 base and Q4 collector?
Q5 base capacitance about 20pF to gnd shows instability.
Q5 base capacitance about 20pF to gnd shows instability.
Last edited:
Voltwide had a good observation. I repeated his test and found:
1. disabling the cascode helped phase margin a little, but it was still kind of iffy.
2. Adding 100 pF from collector to base of Q1 gets pretty good phase margin.
Of course, if the sim and reality agree...that remains to be seen!
1. disabling the cascode helped phase margin a little, but it was still kind of iffy.
2. Adding 100 pF from collector to base of Q1 gets pretty good phase margin.
Of course, if the sim and reality agree...that remains to be seen!
The input transistor Q2 acts as voltage amplifier with collector-base capacitance producing miller effect:
-Input impedance is more of a capacitor, i.e. drops linearly with signal frequency
-loop bandwith is limited by base series resistance R12 and varies with actual source impedance Rg.
For maximum Bandwith R12+Rgen < 10Ohm!
But this is not what I expect from a voltage buffer amp🙄
-Input impedance is more of a capacitor, i.e. drops linearly with signal frequency
-loop bandwith is limited by base series resistance R12 and varies with actual source impedance Rg.
For maximum Bandwith R12+Rgen < 10Ohm!
But this is not what I expect from a voltage buffer amp🙄