I don't have 6 pre drivers, those positions A, B and C, were alternatives. I'd either use pair A, pair B or pair C
I don't actually have an exact schematic as it's in pieces between the simulator (which is using virtual constant current sources) and the schematic in the other blameless thread. Really though, all I've done is add in 3 extra pairs of output transistors and added in 0.1 ohm emitter resistors. In the PCB layout I've provided I'm using 4 paralleled resistors for RE.
Mmm I managed to get all that sorted with the other amp. I did have one ground loop when I added in more then one channel, this was between the signal grounds, but I managed to solve that. I'm not expecting to have any real trouble with regards to grounding and wiring the thing up, the only bit that could perhaps cause issues is the fact that I will be dealing with significantly higher currents and hence the chance for coupling or injection is higher.
I am thinking of using RC filtering between the output stage and the input/VAS though, as this might help. It's easy enough to add in later though if I'm having trouble.
One other thing that comes to mind when mounting PCBs. A company I worked for back in the early nineties had me prototype a power amp for a subwoofer design they were working on. I was assembling a number of these power amps and testing them. Back then we had another tech make our boards for us. He did a small low run service out of his home. One of the board had mounting holes slightly offset from center. This particular amp measured the distortion a magnitude higher than the rest, which was puzzling to me. I dismounted the board to check the soldering and everything was fine. I put the board back on the standoffs only this time without tightening the nuts down and the distortion measured as it should. I tightened the nuts down and again and the distortion was a magnitude higher. It took me awhile to realize what was going on. The traces on the board were being flexed and that created a strain gauge effect on the traces causing the distortion to rise. It was an easy fix, just drill the mounting holes on the board a bit larger to relieve the strain.
If you unfamiliar with the strain gauge effect, strain gauge effect is, when a conductor is flexed under tension or stretched under tension, the resistance of the conductor increases. The principle is used in instrumentation a lot for measuring weight. Strain gauges are commonly used at intersection to detect the presence of a vehicle stopped at a traffic light.
The point being, make sure there is no stress on the PCB when mounting the boards.
If you unfamiliar with the strain gauge effect, strain gauge effect is, when a conductor is flexed under tension or stretched under tension, the resistance of the conductor increases. The principle is used in instrumentation a lot for measuring weight. Strain gauges are commonly used at intersection to detect the presence of a vehicle stopped at a traffic light.
The point being, make sure there is no stress on the PCB when mounting the boards.
I like CFPs to, here's a link to a few ways of doing a CFP with multiple output transistors, one is obviously more efficient than the other as i'm sure you'll see
The output is a triple, removing the pre driver will obviously make it not so. As you have said previously "sorry for the low tech approach"...Bests, Mark.
Ah, so I'm not the only one who's seeing this in simulation
I really should buy a proper function generator so I can create nice square waves. So far I've been using the computer to generate all my test signals. For sines, it's great, but for squares...as you'd expect, it leaves a lot to be desired.
I like CFPs to, here's a link to a few ways of doing a CFP with multiple output transistors, one is obviously more efficient than the other as i'm sure you'll see
Bests, Mark.
Low tech is just as good as any modern approach as long as it gets the job done
For now I think I'll be sticking with a single set of drivers. 8 output transistors per channel would = 8 drivers I'd rather not have to deal with that if I can help it.
"I'm curious how well the split-pole compensation worked on this circuit in practice. I've been simming it and getting excellent low THD but very poor square wave performance with overshoots of nearly 3V. How well does the real one do ?"
Have you tried putting a filter on the input to the amp (sorry, I don't have your circuit in front of me).
This might help (methodology works for 2 pole and normal single pole compensation):-
Step1: compensate the amp so that the phase margin (open loop!) is in the region of 60degrees minimum. Use LTP emitter degen (typically 100 Ohms to 220 Ohms) and Cdom (typically 50pf to 100pf) to do this.
Step 2: close the loop and look at the unity gain frequncy. Typically it will be between 3 to 10MHz. This is too high and will cause the overshoot you are seeing in your sims
Step3: place a low pass filter on the input of the amp. Adjust the filter to get a unity gain frequency in the closed loop condition of between 1MHz and 2Mhz.
Your amp will now be compensated.
Pointer 1: Use an output inductor of between 1uH and 2uH with a parallel 2.2 Ohm resistor - you need this!
Pointer 2: Use a Zobel network (10 Ohms and 0.1uF). connect from junction of emmitter resistors straight back to power supply capacity start ground.
Pointer 3: note that when you do step 1 you MUST not have any input filters connected - this will screw your phase marging readings up - you must remove the input filter for Step 1.
Good luck.
Have you tried putting a filter on the input to the amp (sorry, I don't have your circuit in front of me).
This might help (methodology works for 2 pole and normal single pole compensation):-
Step1: compensate the amp so that the phase margin (open loop!) is in the region of 60degrees minimum. Use LTP emitter degen (typically 100 Ohms to 220 Ohms) and Cdom (typically 50pf to 100pf) to do this.
Step 2: close the loop and look at the unity gain frequncy. Typically it will be between 3 to 10MHz. This is too high and will cause the overshoot you are seeing in your sims
Step3: place a low pass filter on the input of the amp. Adjust the filter to get a unity gain frequency in the closed loop condition of between 1MHz and 2Mhz.
Your amp will now be compensated.
Pointer 1: Use an output inductor of between 1uH and 2uH with a parallel 2.2 Ohm resistor - you need this!
Pointer 2: Use a Zobel network (10 Ohms and 0.1uF). connect from junction of emmitter resistors straight back to power supply capacity start ground.
Pointer 3: note that when you do step 1 you MUST not have any input filters connected - this will screw your phase marging readings up - you must remove the input filter for Step 1.
Good luck.
I should have added above that the final step is to inject a squarewave input to the amp to get about 2V pk to pk on the output - check that there is no overshoot. Adjust the input filter for minimum or no overshoot. You need to work on this because you don't want to be too heavy handed - just the right amount to get the job done.
the open-loop gain (OLG) is kind of a worst case indicator for stability.
it is approximating the loop gain for a unity gain configuration of your amp. (e.g. output short circuited to inverting-input)
just leaving the feedback path open circuit is not the ideal way of determining the OLG, as the effect of feedbacknetwork itsself on the amp is not taken into account.
to measure OLG you can setup your amp as you would use it (with R_load e.g. 4ohm, feedbackpath in place, sourceimpedance, etc.).
place a unity gain voltage controlled voltage source sensing the differential voltage from non-inverting to inverting input. (or just calculate it as the difference of the two node voltages. why, see below)
now perform an AC analysis. put an ac source to the input like you were feeding your audiosignal to the amp.
plot the Voltage at the output divided by the differential voltage from inverting to non-inverting input.
now you see the OLG of your amp.
But still this is not telling you what you need to know when you want to check if the amp is stable.
therefore you need to take a look at the loop gain (note that its not "open") and check this as usual for the nyquist critereon.
one way to get to the loop gain is the "middlebrook method". this is a bit complicated to explain here, so I have to leave the research up to you.
here's a site on middlebrook I found quite helpfull: Simulating Loop Gain - Spring 1997
if you find any better resources, feel free to let me and the rest of us know
EDIT:
one more link: http://sites.google.com/site/frankwiedmann/loopgain
it is approximating the loop gain for a unity gain configuration of your amp. (e.g. output short circuited to inverting-input)
just leaving the feedback path open circuit is not the ideal way of determining the OLG, as the effect of feedbacknetwork itsself on the amp is not taken into account.
to measure OLG you can setup your amp as you would use it (with R_load e.g. 4ohm, feedbackpath in place, sourceimpedance, etc.).
place a unity gain voltage controlled voltage source sensing the differential voltage from non-inverting to inverting input. (or just calculate it as the difference of the two node voltages. why, see below)
now perform an AC analysis. put an ac source to the input like you were feeding your audiosignal to the amp.
plot the Voltage at the output divided by the differential voltage from inverting to non-inverting input.
now you see the OLG of your amp.
But still this is not telling you what you need to know when you want to check if the amp is stable.
therefore you need to take a look at the loop gain (note that its not "open") and check this as usual for the nyquist critereon.
one way to get to the loop gain is the "middlebrook method". this is a bit complicated to explain here, so I have to leave the research up to you.
here's a site on middlebrook I found quite helpfull: Simulating Loop Gain - Spring 1997
if you find any better resources, feel free to let me and the rest of us know
EDIT:
one more link: http://sites.google.com/site/frankwiedmann/loopgain
Last edited:
Krachkiste,
I do not want to hijack 5th Elements thread, so I'll answer your points and then propose that AndrewT opens up a separate thread on 'Compensating an Amplifier'. Lets keep in mind that there is more than one way to skin a cat, so not everyone will approach this issue in the same way. I have used my methodology and it works. The problem with compensation, especially if you are an experimenter, is that without a simple practical approach, you can end up chasing your tail with a less than optimal end result. The problem has to be approached in a step by step manner.
1. This is about practical audio power amplifier compensation - so using a simulator to help understand what is going on as you compensate the amp is perfectly valid in my view.
2. If the feedback network consists of only a resistor (so no other frequency shaping componenets) then making this resistor 'open' or a very high value is a valid near enough assumption to gauge the OLG and phase margin (in open loop condition).
3. I do not know why you would use open loop gain to acertain the loop gain in a unity gain situation - most power amps have gain of 20-30db so why would you want to know this anyway?
4. Yes, Middlebrook and others have accurate ways of measuring loop gain, but for a simple 2nd order system (which most power amps are), this is not necessary.
5. Its quite easy for the OLG in an amp to vary by quite a bit in a practical amp - so what? Knowing the loop gain accurately is not a requirement for stabilizing an audio power amplifer.
I do not want to hijack 5th Elements thread, so I'll answer your points and then propose that AndrewT opens up a separate thread on 'Compensating an Amplifier'. Lets keep in mind that there is more than one way to skin a cat, so not everyone will approach this issue in the same way. I have used my methodology and it works. The problem with compensation, especially if you are an experimenter, is that without a simple practical approach, you can end up chasing your tail with a less than optimal end result. The problem has to be approached in a step by step manner.
1. This is about practical audio power amplifier compensation - so using a simulator to help understand what is going on as you compensate the amp is perfectly valid in my view.
2. If the feedback network consists of only a resistor (so no other frequency shaping componenets) then making this resistor 'open' or a very high value is a valid near enough assumption to gauge the OLG and phase margin (in open loop condition).
3. I do not know why you would use open loop gain to acertain the loop gain in a unity gain situation - most power amps have gain of 20-30db so why would you want to know this anyway?
4. Yes, Middlebrook and others have accurate ways of measuring loop gain, but for a simple 2nd order system (which most power amps are), this is not necessary.
5. Its quite easy for the OLG in an amp to vary by quite a bit in a practical amp - so what? Knowing the loop gain accurately is not a requirement for stabilizing an audio power amplifer.
Last edited:
RE that RC network. Are we talking a resistor and cap, in series, connected between the base and the collector of the drivers?
If not please correct me
And what values are we talking about?I'm making the final touches to the PCB and I figured I might as well add in some extra options, just in case I have problems with stability.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.