Unstable feedback loop.

[king.oslo]

Hello there,

I am working to stabilize have this 500KHz opamp with a feedback loop. I cannot get it stable. I have been trying all sorts of things. High pass / low pass filters, AC-coupling. But I have given up.

I am sure this is very simple. What method do I need to stabilize it?

Attached is my schematic and LTSpice file.

Thanks.

Kind regards,
Marius

[Bored@Work]

Are you serious?

[king.oslo]

Yes.M

[amspire]

The use of the non-inverting input is correct, as the transistor inverts the opamp output.

The circuit is close, but there are  few errors.

First in my sketch there was a 100k resistor from the opamp input to the 10V. This is important for the compenstaion to work at all.

Secondly, it will be really hard to stabilize this circuit with such a big MOSFET with a large drain capacitance. It is a 8A mosfet, and and your maximum current is under 10mA. Can you find a transistor or MOSFET rated at about 100mA that can take the voltage? The thing is if you can find a device with a collector/drain capacitance of just a few pF, it will boost the -3db corner of the transistors well above the frequencies that would cause the opamp to be unstable.

I will modify your spice file and post it back.

Richard.

[amspire]

Here is an updated Spice file.

I removed Q2 as it seems that Q1 by itself can do OK without any help.

Q1 has to be a high voltage low current transistor like the ones used in almost every CRT driver circuit.  Perhaps a MPSA42 would work well. The transistor I used in the simulation was a 150V one which was the best I could find in the LTSpice library.

The 200pF cap on the output is just to simulate test lead capacitance - you do not need to add it to the circuit. I added it to check that extra capacitance on the output would not make it unstable.

Richard

[amspire]

By the way, a much better opamp for the job would be the LTC1052 chopper stabilized opamp. with the LTC1052, you can use a 220pf compensation cap, and the opamp will have a lot more accuracy (extremely low bias current and offset).

The one you picked has a fairly large bias current which causes significant (for your meter  ) voltage drops across the 100K resistors.

The LTC1052 is 16V maximum supply, so you will want to have a 15V regulated supply for the opamp and regulator.

Richard.

[king.oslo]

Thank you.

Is this a high-pass filter? (When the cap is on the positive side of the output, right? Or not?)

And which formula is used to work out the phase shift?

Thank you for your time.

Kind regards,
Marius

[amspire]

Thank you.

Is this a high-pass filter? (When the cap is on the positive side of the output, right? Or not?)

And which formula is used to work out the phase shift?

Thank you for your time.

Kind regards,
Marius

Don't get too concerned about the capacitor on the output, but you just want to test that the supply is still stable with a capacitor on the output.  I haven't worked out any phase shifts. I have built similar circuits before, so I just used seat of the pants reasoning. The transistor circuit has a gain of 22, and I know the opamp is stable when the total loop gain is unity. So the first thing is to say at high frequencies, I want the total gain to be one, so I want the opamp to have a gain of less then 1/22.  So I set the resistor in the compensation circuit for a gain of 2k2/100K = 1/45. This is a lot less then 1/22, so I am happy with that.

Next the capacitor in the compensation circuity sets the frequency the gain of the op-amp circuit wants to fall off. The bigger the capacitor, the slower the feedback circuit will work, so the idea is to make it as small as possible.  The transistor circuit will start to add extra phaseshift somewhere near 1MHz, so you want the opamp to be at a gain less then 1/22 by then. 220pF sounds about right, but if you are not sure. make it a bit bigger. try not to let it get above 1nF or it will degrade the regulation.

Does that make sense? It is probably easier to understand if replace the dc voltage source with a 100mV AC source with a 10V DC offset and do an AC analysis. You can then see what is happening with phase shifts around the circuit.

Richard.

[king.oslo]

Thank you.

Is this a high-pass filter? (When the cap is on the positive side of the output, right? Or not?)

And which formula is used to work out the phase shift?

Thank you for your time.

Kind regards,
Marius

Don't get too concerned about the capacitor on the output, but you just want to test that the supply is still stable with a capacitor on the output.  I haven't worked out any phase shifts. I have built similar circuits before, so I just used seat of the pants reasoning. The transistor circuit has a gain of 22, and I know the opamp is stable when the total loop gain is unity. So the first thing is to say at high frequencies, I want the total gain to be one, so I want the opamp to have a gain of less then 1/22.  So I set the resistor in the compensation circuit for a gain of 2k2/100K = 1/45. This is a lot less then 1/22, so I am happy with that.

Next the capacitor in the compensation circuity sets the frequency the gain of the op-amp circuit wants to fall off. The bigger the capacitor, the slower the feedback circuit will work, so the idea is to make it as small as possible.  The transistor circuit will start to add extra phaseshift somewhere near 1MHz, so you want the opamp to be at a gain less then 1/22 by then. 220pF sounds about right, but if you are not sure. make it a bit bigger. try not to let it get above 1nF or it will degrade the regulation.

Does that make sense? It is probably easier to understand if replace the dc voltage source with a 100mV AC source with a 10V DC offset and do an AC analysis. You can then see what is happening with phase shifts around the circuit.

Richard.

I think I am about to understand. I have to play with it a bit more.

Do you give this transistor the nod?

https://www1.elfa.se/data1/wwwroot/assets/datasheets/07105356.pdf

Kind regards,
Marius

[amspire]

Do you give this transistor the nod?

https://www1.elfa.se/data1/wwwroot/assets/datasheets/07105356.pdf

Kind regards,
Marius

A lower capacitance and faster transistor is better. It does not need to be a power transistor. Say the DC supply is 200V. The maximum power that the transistor has to handle is half this (100V), and the current would be 3.7mA, so that is a power dissipation of 370mW. The one I suggested, the MPSA42 would be good. If the data sheets do not mention the collector capacitance, look at the F_T (basically the maximum frequency the transistor can actually still amplify). For the TIP47, it is 10MHz at 200mA. For the MPSA42, it is 50MHz at 10mA. Speed is improved when the current is increased, so at the working current of about 3mA, the TIP47 may have a F_T of only 150KHz, compared to about 16MHz for the MPSA42.

My concern with the TIP47 is that it speed is slow enough to really mess with the stability of the circuit.

Richard.