9
\$\begingroup\$

I have a question regarding the stability of a circuit I plan on building. This is a voltage-controlled current source, with the IN-AMP being used to sense the current through Rsns and provide feedback to the op-amp. I'm trying to use a programmable instrumentation amplifier and turns out most of the ones that fit my requirements are chopper amplifiers.

However, as I understand it this means there will be some delay from when the current through Rsns changes to when the capacitors in the chopper charge and discharge, and then the output of the in-amp is changed. Am I correct in assuming this delay will lead to oscillation? (I don't have the parts yet or I would just build it up). Is it in general a bad idea to introduce delay elements into a feedback loop, or is there a way to use them without instability? Thanks!

Circuit

UPDATE: For those that would like an update: I built this circuit with a vanilla op-amp and instrumentation amplifier, with the instrumentation amplifier having a G=100, Vin= 1Vpp sinewave at 60Hz, Rsns=1R, and ZL=22R, and I see my 60Hz signal "amplitude modulated", if you will, at an oscillation frequency of 133kHz. Here is the oscilloscope trace across ZL. oscilloscope trace

\$\endgroup\$
  • \$\begingroup\$ If I could upvote twice, one for interesting question, one for clear schematic, I would. \$\endgroup\$ – Neil_UK Sep 14 '16 at 7:57
  • \$\begingroup\$ @Neil_UK I'll help you with that. \$\endgroup\$ – Arsenal Sep 14 '16 at 8:40
3
\$\begingroup\$

Yes, stability will likely be a problem, and the internal construction has little to do with it. Most (modern) chopper amps have many MHz bandwidth and behave similarly to normal op-amps or in-amps aside from really nasty spikes out the inputs and some small noise near the modulation frequency.

However, you are introducing a lag and more gain into the feedback loop and both of those will tend to result in less phase margin and thus potential instability. By keeping the gain of the in-amp low and perhaps introducing some compensation you should be able to make this concept work.

\$\endgroup\$
  • \$\begingroup\$ Generally, I'm under the impression that you can (usually) fix this by just making sure the UGBW between the two amps is a factor of 5 from each other. Do you have experience with this? \$\endgroup\$ – Andrew Spott Sep 14 '16 at 16:21
  • \$\begingroup\$ If they aren't you can slow one down so that there is one only dominant lag. \$\endgroup\$ – Spehro Pefhany Sep 14 '16 at 16:43
  • \$\begingroup\$ So if there is a sinewave at the input, am I correct in assuming the "importance" of the delay in the gain loop in relation to oscillation is dependent on that frequency? Or will the circuit oscillate independent of a signal at the input? \$\endgroup\$ – Jordan Sep 14 '16 at 19:08
  • \$\begingroup\$ In real life there is usually enough noise to get things going. If you're interested in simulation (and you ought to be) then feed the input a square wave and look at the decay of the overshoot (if any) to see what kind of damping you are getting. The closer you are to oscillation the longer it will take for the ringing to die out, until the ringing gets bigger each cycle and you have an oscillator. Check under varying load conditions including some parasitics too. \$\endgroup\$ – Spehro Pefhany Sep 14 '16 at 19:14
  • 1
    \$\begingroup\$ OK, I've been looking for an excuse to learn to design SPICE models. I'll get on that and post the results when I finish. Thank you for all your help. \$\endgroup\$ – Jordan Sep 14 '16 at 19:36
1
\$\begingroup\$

As long as you pick the right device there shouldn't be any problem.

The term chopper amplifier is not well defined and used for a number of different topologies. However, the datasheet should give enough information to get an idea what topology or method was used for a particular device.

For continuous time signal processing a chopper (stabilized) amplifier usually consists of two amplifiers. The main amplifier which is in the signal path and a nulling amplifier which is there to deal with its own offset and the offset of the main amplifier.

The principle is shown below: enter image description here

The circuit operates in two phases, in one phase the nulling amplifier measures its own offset and stores it in capacitor A. This voltage is fed back to the nulling amplifier and is used by the amplifier to correct its own offset. In a second phase the now almost offset free nulling amplifier measures the offset of the main amplifier and again stores a voltage in a second capacitor B which corrects the offset of the main amplifier.

The offset correction is done using a modified input stage, that has a low gain at the compensation input.

Ideally this method works transparently and is invisible from the outside. In practice the switching frequency can be seen at the output but the amplitude is usually very low. Sometimes spread spectrum methods are used to distribute the spectral components over a wider range.

This is just one principle but the other methods are often similar to this one.

\$\endgroup\$
  • \$\begingroup\$ I was looking at the AD8557, if I'm not mistaken the most relevant performance measure here is settling time, which is given to be 8 microseconds. \$\endgroup\$ – Jordan Sep 14 '16 at 19:16
  • \$\begingroup\$ Settling time gives an indication. In order to ensure stability a circuit simulation should be done. \$\endgroup\$ – Mario Sep 14 '16 at 19:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.