Dummy Load II

[jocanon]

...I'm not sure how much attention we paid to the grounding. That could be a problem we should take care with in this one.

ronv, for the sake of my education, what do you mean exactly by that? Everything in the schematic that says it should be connected to ground is through the ground pin which plugs into my household plug...but I assume you mean something more than just this.

 

[jocanon]

Okay, just measured the voltage on the DUT with my normal VM from Harbor Freight and it is stable, it didn't oscillate at all while the little voltmeter attached to the dummy load was still oscillating. I even opened up the case and touched the leads of the normal voltmeter to the same exact location where the other one is and there was no oscillation whatsoever, it read a constant 24.1v under full load. I wonder if the "normal" voltmeter (i.e. not the one that is attached to the dummy load that oscillates) has some caps and resistors in there to reduce noise whereas that little one on the dummy load is just a bare voltmeter, susceptible to all the noise. I even got my hands on another little voltmeter like the one on the dummy load just to see if it would oscillate too and it did. So normal VM from Harbor Frieght doesn't oscillate little VM from China does.

 

[ronv]

It's only a guess, but it is probably the switching noise from the big power supplies getting into the little meters and yes most do have some high frequency filtering to prevent it. If you have some spare parts you can try putting about 100 ohms in series with the positive lead of the volt meter (value is not to important) then one of your .1 ufd caps from the meter side of the resistor to the negitive lead of the meter to see if it gets better. We will use a bigger cap in the real one so we are just looking for an improvement.
For the grounds we want the negative lead from your PSU and the 12 volt power supply ground tied to the same point where the big ground ties into the big buss bar. Then a wire from there to all the op amp grounds and small circuit grounds. (logic ground) The only other things going to the big ground are then the power resistors. (power ground)
It is probably already this way, but just to make sure.

 

[jocanon]

Sounds good Ron, I'll try the resistor and cap you mentioned with the little china VM when I get home tonight to see if its an improvement.

As far as ground, yes I do believe everything is tied to the same ground as you described.

 

[jocanon]

Maybe I should check and see if I still get that noise with the Chinese VM when my PSU is powering the charger instead of being connected to the dummy load. That could pin point it down to the possible noise from high frequency switching you mentioned versus the feedback loop in the dummy load.

 

[ronv]

Maybe, but the charger will be a reactive load so may give somewhat different results. You can also try just tieing the +/- of the voltmeter together and holding them close to the power leads.

 

[jocanon]

You can also try just tieing the +/- of the voltmeter together and holding them close to the power leads.

I'm not sure I know quite what mean by that?

 

[ronv]

If the noise from the supply is high enough it will couple into the measurement leads even though they are shorted together. Kind of like an antenna. May work, may not.

 

[jocanon]

The little voltmeters are 3 wire, the pos (red) and negative (black) to power the LED screen on, then a blue wire to read voltage. In normal operation I connect the black to the neg of the power supply and the blue and red to the positive. So are you saying to tie the blue, black, and red all theee together and then just hold them close to the neg and positive power leads of the PSU, but not actually touching them and it may pick up a signal?

 

[ronv]

No, the red and black where they ususally go then another wire from blue to black. This is like shorting out the leads to your voltmeter but it may still pick up a small voltage from the radiated noise.

 

[jocanon]

I see, I knew what I said didn't make sense...in fact almost laughable...that is why I needed clarification, and you provided it, thx

 

[Mr RB]

I don't think so but maybe Mr. RB could recheck my math. It is very close if the over temperature kicks in, so since it runs so cool we should probably lower where that shuts it off. Since you are going to add more for higher power anyway we can throw in a few extras to increase the margin. It's not like we are selling a million of them. I just don't like to add a lot of parts since everyone has a failure rate. At some point it becomes self defeating.

All looks good to me, and you're way better at opamp hardware than me anyway.

Since stability is a concern, I would make a couple of small changes, like increase the value of the current sense resistors so they give a higher voltage output signal for the same amps.

Also on my dummy loads I use quite a bit of RC integration on the drive to the power transistors, this gives massive stability and there's no need for the dummy load to be able to change current very quickly, it's adjusted by a pot!

On your circuit you could put a series resistor plus a sizable cap (10uF tantalum?) from the FETs gate-source pins, so that the FET drive is only able to change slowly. Since you have no time constants in your voltage reference or current feedback voltage, it will integrate very nicely and be as stable as the Pyramids.

 

[ronv]

Since stability is a concern, I would make a couple of small changes, like increase the value of the current sense resistors so they give a higher voltage output signal for the same amps.

Yep, we could do that now that we are changing to 10 watt resistors. Maybe move it up to 3 watts or so in each one. We were trying to cool them down as they were hotter that an overworked pistol.

Also on my dummy loads I use quite a bit of RC integration on the drive to the power transistors, this gives massive stability and there's no need for the dummy load to be able to change current very quickly, it's adjusted by a pot!

On your circuit you could put a series resistor plus a sizable cap (10uF tantalum?) from the FETs gate-source pins, so that the FET drive is only able to change slowly. Since you have no time constants in your voltage reference or current feedback voltage, it will integrate very nicely and be as stable as the Pyramids.

I think you could do that on your load because it is open loop. But with the closed loop it needs to be pretty fast or it will oscillate. If we were using BJT we could run it open loop with just some emitter resistors to even up the Vbe drop of the parallel transistors, but with the wide range of threshold voltage using FETs we can't use the same method.

 

[jocanon]

It's only a guess, but it is probably the switching noise from the big power supplies getting into the little meters and yes most do have some high frequency filtering to prevent it. If you have some spare parts you can try putting about 100 ohms in series with the positive lead of the volt meter (value is not to important) then one of your .1 ufd caps from the meter side of the resistor to the negitive lead of the meter to see if it gets better. We will use a bigger cap in the real one so we are just looking for an improvement...

Ok, here's the result, the noise almost complete went away, it fluctuated one decimal point instead of a whole volt, but the read out was 1 whole volt too low. It said 23.1 volts instead of about 24 when under full load, maybe this was expected because of the resistor, but how will we compensate for that in the dummy load?

 

[jocanon]

If the noise from the supply is high enough it will couple into the measurement leads even though they are shorted together. Kind of like an antenna. May work, may not.

Tried this too, but didn't have any reading in the voltmeter.

 

[ronv]

What size resistor did you use. I think the meter is 100000 ohms/volt. I'll go back and check.

 

[()blivion]

Also on my dummy loads I use quite a bit of RC integration on the drive to the power transistors, this gives massive stability and there's no need for the dummy load to be able to change current very quickly, it's adjusted by a pot!

On your circuit you could put a series resistor plus a sizable cap (10uF tantalum?) from the FETs gate-source pins, so that the FET drive is only able to change slowly. Since you have no time constants in your voltage reference or current feedback voltage, it will integrate very nicely and be as stable as the Pyramids.

This is exactly what I was first thinking too.

To me, one stops oscillation with a low-pass filter (slow things down, because the opposite of AC is of course DC.) But I understand the implications of what ronv is saying and it makes perfect sense too. However, what he says is in direct contradiction to the above idea. The more delay in a closed loop, the more it should tend toward oscillation.

I think I understand, but I'm not really sure

During simulation I can only get our FET/Op-Amps circuit to oscillate if there is some significant inductance in series the gate/output wire.



But with the resistor and capacitor integrator it hasn't produced any negative results, which is not in accordance with ronv's take on this, as this configuration does add a whole lot of delay. More than the above inductor + cap does.



It looks like the RC version has delay, but is not phase delay, and that is making all the difference. If I had known that, I would have finished my original thought and suggested we use a RC low-pass filter on the FET gate like Mr RB has done.

Ronv: ... try putting about 100 ohms in series with the positive lead of the volt meter (value is not to important) then one of your .1 ufd caps from the meter side of the resistor to the negitive lead of the meter to see if it gets better. ...

...

jocanon: ... the noise almost complete went away ...

Good, now we are getting somewhere.

This could indicate that the circuit is in fact genuinely oscillating and needs to be worked on. But it could still also be a misbehaving power supply. We can't do anything about the power supply really, but it is fairly easy to fix oscillations.

My simulations above show that the system will oscillate if there is an inductance (the wire) in series with a capacitance (the FET gate) if there is little to no resistance in line with it. My sim also showed that if you put a resistance in the circuit, it will stop it. This is about the only thing I could see causing oscillations.

 

[ronv]

(), Try it in spice with a little pulse on the reference voltage and you will see the ringing on the voltage on the sense resistor.

Back a few posts his good meter shows no variation.

J, They don't say what the input impedance is on the meter, but it sounds like a smaller resistor is in order.

 

[()blivion]

(), Try it in spice with a little pulse on the reference voltage and you will see the ringing on the voltage on the sense resistor.

Hummm.... Looks like you are more or less right, here are the results.



Though I'm not getting a full and steady oscillation out of it, just the ringing. We are seeing a steady oscillation in the real circuit, which is what I am trying to reproduce.

I only just started playing with it though. Still want to try some more things. . .

 

[jocanon]

(), Try it in spice with a little pulse on the reference voltage and you will see the ringing on the voltage on the sense resistor.

Back a few posts his good meter shows no variation.

J, They don't say what the input impedance is on the meter, but it sounds like a smaller resistor is in order.

OK, I will have to try that out, maybe tomorrow night. Orders are coming in that I have to fill now... I was using a 100 ohm resistor.