Author Topic: GBP and the reduction of Gain  (Read 644 times)

0 Members and 1 Guest are viewing this topic.

Offline LoveLaikaTopic starter

  • Frequent Contributor
  • **
  • Posts: 558
  • Country: us
GBP and the reduction of Gain
« on: September 15, 2021, 06:54:42 pm »
This may be rudimentary, but I wanted to ask for my own peace of mind. For voltage-feedback op-amps, GBP is a quick way to see how the bandwidth decreases as you increase gain. Does it work the opposite way as well? (i.e., if you decrease the gain for whatever reason, you also increase the bandwidth)

Also somewhat related, if you decrease the gain of an op-amp, is it still recommended to include a feedback capacitor as well or is it better to remove it if you see undesirable behavior (like oscillations) as a result of reducing gain?
 

Offline Manul

  • Super Contributor
  • ***
  • Posts: 1105
  • Country: lt
Re: GBP and the reduction of Gain
« Reply #1 on: September 15, 2021, 09:20:21 pm »
Does it work the opposite way as well? (i.e., if you decrease the gain for whatever reason, you also increase the bandwidth)

All other things beeing equal, yes.

Also somewhat related, if you decrease the gain of an op-amp, is it still recommended to include a feedback capacitor as well or is it better to remove it if you see undesirable behavior (like oscillations) as a result of reducing gain?

Generally we could say that capacitor in feedback is needed even more at low gains then high, all other things beeing equal. That is because at lower gain overall bandwidth is higher. Capacitor gives frequency dependent negative feedback, so high frequencies are amplified less, essentially acting as low pass filter that reduces bandwidth.

But the roots of op amp stability is more complicated subject. That is called compensation and you can find a source to read about that. It is about op amp output driving loads which are (partly) capacitive, thus introducing phase shifts at higher frequencies and thus affecting gain. At some frequencies your negative feedback loop may even get positive and op amp will oscillate. Series resistor at op amp output may also be used to compensate for phase shifts. Also op amps are not equal by their performance, some have more or less internal compensation.
 

Offline Terry Bites

  • Super Contributor
  • ***
  • Posts: 2357
  • Country: gb
  • Recovering Electrical Engineer
Re: GBP and the reduction of Gain
« Reply #2 on: September 15, 2021, 10:26:21 pm »
https://www.google.co.uk/url?sa=i&url=https%3A%2F%2Felectrosome.com%2Fopamp%2F&psig=AOvVaw39vwtoJGbs6s9C1Otceayk&ust=1631829668372000&source=images&cd=vfe&ved=0CAoQjhxqFwoTCKCZnrD9gfMCFQAAAAAdAAAAABAU

Draw a hoizontal line across the chart at the gain you want to set (in dB). Where it hits the sloping line draw a vertical lline down the the x axis. The will give you a fair aproximation of the frequncy response.
Its a little more complex in reality becasuse the opamp's abilty to generate output current in to a load or feedback network also falls off. AKA full power bandwidth. It makes sense to roll off the repsonse with a feedback capcitor to reduce noise a higher frequencies. It may introduce unexpected phase shift which may be undesirable. The casue of BW limitation is the finite GBW of the transistors inside the opamp and it's internal compensation scheme. The opamp designer has done their best to make a stable amplifier. You only need a again of 1 to make an oscillator! The opamps internal gain is in the millions. See the attched guts of a 741 and that little 25p stabilty cap that stops the whole thing going crazy.

Or you can set the gain to some arbritary value, say 10, plug in a signal generator and monitor the output on a scope.
 You can also simulate this setup on the many platforms available like LT spice.
 

Offline LoveLaikaTopic starter

  • Frequent Contributor
  • **
  • Posts: 558
  • Country: us
Re: GBP and the reduction of Gain
« Reply #3 on: September 15, 2021, 10:30:45 pm »
Thanks for replying. I was simulating an OPA818, and I saw that at -1x inverting-unity-gain, 0.4 pF was stable; however, as I reduced the gain to -0.5x, the capacitance had to be changed, as somehow there was a really large oscillation. That was what brought this question to mind.

Since you mentioned frequency-dependent feedback, it brought to mind a thought. Given GBWP and an RC-negative-feedback network, the smaller cutoff frequency will be dominant, correct? For example, say an op-amp has a GBWP of 200 MHz; however, the negative feedback network consists of 1 nF and 10k-ohm resistors (using extreme values to make a point). Looking at just the negative feedback components, it forms an RC filter with a cutoff of 15 kHz. So, the op-amp will only be stable up to 15 kHz?
 

Offline Manul

  • Super Contributor
  • ***
  • Posts: 1105
  • Country: lt
Re: GBP and the reduction of Gain
« Reply #4 on: September 15, 2021, 10:47:52 pm »
Given GBWP and an RC-negative-feedback network, the smaller cutoff frequency will be dominant, correct? For example, say an op-amp has a GBWP of 200 MHz; however, the negative feedback network consists of 1 nF and 10k-ohm resistors (using extreme values to make a point). Looking at just the negative feedback components, it forms an RC filter with a cutoff of 15 kHz. So, the op-amp will only be stable up to 15 kHz?

Stability and bandwidth (or cutoff frequency) are different topics, but yes, high RC constant in feedback loop will limit amplifier bandwidth and it can be way lower then GBWP and gain would allow otherwise.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21612
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: GBP and the reduction of Gain
« Reply #5 on: September 15, 2021, 11:54:06 pm »
Note that such a connection makes an integrator, of some limited DC gain (due to the resistor), and some cutoff frequency (due to the capacitor).

That cutoff (GBW), can go up to a limiting value of the amp's GBW, when C = 0.  Or put another way, we can calculate an effective capacitance from the amp's GBW, and say that, even for in-circuit C = 0, there is some minimum actual value of C always present, and we can only connect more capacitance in parallel with it to further lower GBW.

Note that this is really only true at low frequencies, or large capacitance, so that the opamp still has high loop gain at frequencies of interest.  For frequencies near GBW, there's all the non-idealities of the circuit: the input and output impedances rise, phase shifts, there's even a feed-forward path directly through the capacitor so gain actually goes to zero and reverses (goes positive) past some point!

And yeah, as for stability, that depends on more subtle things; most importantly, how much phase shift is incurred around fT.  Note that, especially as you go up in bandwidth, you may find a lot of "decompensated" amps, that are only stable down to a certain noise gain -- that is, a measure of how much negative feedback is applied around the amp.  These will NOT be stable with a unity-gain configuration (note that, asymptotically speaking, an integrator requires a unity-gain-stable amp: at high enough frequencies, the capacitor looks like a short circuit, so the amp looks like a voltage follower with +in = GND).  If you must, you can sometimes improve stability, by reducing noise gain at high frequencies -- a typical way of doing this is placing a cap between input pins (+in to -in).  This effects a lower input voltage difference, at frequencies where the amp's gain is low and therefore the input voltage would otherwise be rising; which in turn reduces the amount of negative feedback applied, which also *drumroll* increases the noise level, hence the term "noise gain".  (By definition, noise gain is the expected output noise or offset voltage, given the amp's input voltage offset or noise level.  Note that input-referred noise is nothing more than an AC offset voltage.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline LoveLaikaTopic starter

  • Frequent Contributor
  • **
  • Posts: 558
  • Country: us
Re: GBP and the reduction of Gain
« Reply #6 on: September 16, 2021, 02:08:42 am »
Funny how you mentioned a decompensated op-amp. The datasheet referred to it as such. Thanks for the info. This gives me some ideas on how to proceed with my next design.

Given what I learned from version 1, I dont think a decompensated amp is what I need if I need to achieve unity gain somewhere. I'm gonna have to research more.
« Last Edit: September 16, 2021, 02:12:37 am by LoveLaika »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf