That would increase loop phase shift at HF so is probably the wrong direction, but you could still try it anyway. In my experience conducting only experiments that are supported by theory is very limiting. What works always makes sense in the end, but what makes sense in the beginning often doesn't work in the end. I would love to test each stage of an amp with a good network analyzer one day and see how close they are to simulation, because I suspect there are a lot of surprises despite my best effort.
A scope may not have enough resolution to provide accurate phase information. But, as you point out, it's also possible that the test setup mucked things up a bit. That's why network analyzers and bode plot analyzers will have a calibration function that allows you to subtract the gain/phase variation of the test setup from that of the DUT.I used a python to drive my function generation and scope to get the data for the plot.
Over all it worked well. However I don't think the phase margin was completely accurate possibly do to the bandwidth of the op-amp.
I'd be tempted to try something like the OPA859 in that application.
Tom
I actually did the same thing. The software was written in such a way that you could ask it to calibrate then it would create a csv file that would be use to tweak both the gain and the phase. Obviously the DUT wasn't connected during this test. It was basically testing itself out to 10Mhz from memory and the a file was create to flat line everything. You could then run a regular test once again with the DUT not connected and verify that you had a flat response.That's why network analyzers and bode plot analyzers will have a calibration function that allows you to subtract the gain/phase variation of the test setup from that of the DUT.
Unfortunately I still didn't think that the phase was being measured accurately as it wasn't showing the numbers that I was seeing in the sim. Close but not quite there. Maybe it was right but unfortunately I did not work out a way to verify.
This amp UT#2 started out stable, is stable closed loop, and shows no sign of ever blowing up. One theory is
that when they blow up with rail to rail oscillation, maybe a cap or some part fails that is missed on repair. But
the previous owner reported that they never blew up. And the driver board that I put in had never been
powered up, all new old stock transistors.
Running it "open loop" with the 20K feedback resistor for DC it oscillates at 15 to 30 KHz just by touching the
input hot. Obviously, with no feedback there should be no way for it to oscillate from the global loop. I also
tried it without the cap in the feedback path, making it a feedback amp just set for much higher gain, oscillates
in a similar way. And by the way, the other amp UT#1 that oscillates closed loop does it at 1-3 MHz so
something very different is going on.
I was going to write here that I'm stumped and, by the way, I've tried many things to fix the grounding, adding
caps for stability etc. and nothing fixes this oscillation. But on further thought I believe that with the long wires
and perhaps, still not the best grounding, there's just no way to avoid induced or radiated feedback paths.
Grounding the input makes it stable and that's how I'm going to try to measure the loop gain from now on.
I'm open to other suggestions/theories.
It is so strange that in the Tigersaurus video he also sees tens of KHz oscillations IN ONLY ONE OF TWO SAMPLES
and he did not open the loop. His solution was to change the CFP output stage to EF - so strange and quite
the puzzler. Some feedback path through the power supply?
that when they blow up with rail to rail oscillation, maybe a cap or some part fails that is missed on repair. But
the previous owner reported that they never blew up. And the driver board that I put in had never been
powered up, all new old stock transistors.
Running it "open loop" with the 20K feedback resistor for DC it oscillates at 15 to 30 KHz just by touching the
input hot. Obviously, with no feedback there should be no way for it to oscillate from the global loop. I also
tried it without the cap in the feedback path, making it a feedback amp just set for much higher gain, oscillates
in a similar way. And by the way, the other amp UT#1 that oscillates closed loop does it at 1-3 MHz so
something very different is going on.
I was going to write here that I'm stumped and, by the way, I've tried many things to fix the grounding, adding
caps for stability etc. and nothing fixes this oscillation. But on further thought I believe that with the long wires
and perhaps, still not the best grounding, there's just no way to avoid induced or radiated feedback paths.
Grounding the input makes it stable and that's how I'm going to try to measure the loop gain from now on.
I'm open to other suggestions/theories.
It is so strange that in the Tigersaurus video he also sees tens of KHz oscillations IN ONLY ONE OF TWO SAMPLES
and he did not open the loop. His solution was to change the CFP output stage to EF - so strange and quite
the puzzler. Some feedback path through the power supply?
Correction to the above post, it is true that UT#2 never blew up but it was UT#1 where I replaced
the driver board with one with all new parts.
the driver board with one with all new parts.
I'm going to put UT#2 to the side for now since it only has stability problems when run open loop
and will keep thinking about the 15 KHz oscillation. Also, might learn something while working on
the other amps.
Cleaning up UT#3 and powering it up for the first time showed no oscillation at all, it is the "brother"
of UT#2, they were used as a stereo pair and probably bought at the same time. The behavior is the
same as UT#2 and my theory is that the parts are from a common batch. Probably a semiconductor
difference. I believe that the output devices have date codes and I will make note of them soon.
I tried the usual, driving it with a square wave trying an 8 ohm load, no load, clipping and all looks
clean, no overshoot at all. The amp already has .1uF across the output, so adding 1000pF, .01uF and
even .1uF makes very little difference. 1 uF causes slight ringing and I'm looking before the output
inductor. I forgot to remove the 220pF input filter and I'm going to just try to put 75 ohms in parallel
with the input resistor to significantly increase the rise time since that cap seems to play a part in the
stability of the amp.
and will keep thinking about the 15 KHz oscillation. Also, might learn something while working on
the other amps.
Cleaning up UT#3 and powering it up for the first time showed no oscillation at all, it is the "brother"
of UT#2, they were used as a stereo pair and probably bought at the same time. The behavior is the
same as UT#2 and my theory is that the parts are from a common batch. Probably a semiconductor
difference. I believe that the output devices have date codes and I will make note of them soon.
I tried the usual, driving it with a square wave trying an 8 ohm load, no load, clipping and all looks
clean, no overshoot at all. The amp already has .1uF across the output, so adding 1000pF, .01uF and
even .1uF makes very little difference. 1 uF causes slight ringing and I'm looking before the output
inductor. I forgot to remove the 220pF input filter and I'm going to just try to put 75 ohms in parallel
with the input resistor to significantly increase the rise time since that cap seems to play a part in the
stability of the amp.
Last edited:
- Home
- Amplifiers
- Solid State
- Measuring Loop Gain on the Bench - Best Way?