frequency compensation method keeping gain and bandwidth

[anhnha]

Hi.

My op amp is unstable. However, I don't want to reduce gain and bandwidth. Is there a frequency compensation method that only change phase without affect to gain?

 

[FvM]

OPs with compensation are usually implementing minimum phase systems https://en.wikipedia.org/wiki/Minimum_phase

As a consequence, phase and amplitude response have a fixed relation. You can't raise the phase without reducing the gain. But the conclusion is only valid for compensation networks cascaded with the shown amplifier response. It may be still possible to change the internal amplifier compensation scheme, you have to analyze the circuit in detail.

 

[LvW]

Hi.
My op amp is unstable. However, I don't want to reduce gain and bandwidth. Is there a frequency compensation method that only change phase without affect to gain?
]

Are you sure about instability? Which functions do your diagrams show?
More than that, as you don´t want to "reduce gain and bandwidth": Do you speak about open or closed-loop gain?
There are method to realize stable closed-loop gains - even in case the opamp is not universal-compensated.
Please note that an opamp cannot be unstable - it is only feedback circuit that can cause instability.

 

[anhnha]

Thank you, FvM and LvW.

Are you sure about instability? Which functions do your diagrams show?

Yes, I plotted Vout/Vin of op amp in closed loop.
Here is the op amp and simulation schematic.

More than that, as you don´t want to "reduce gain and bandwidth": Do you speak about open or closed-loop gain?

I meant closed loop bandwidth.

There are method to realize stable closed-loop gains - even in case the opamp is not universal-compensated.

Could you suggest some methods?

 

[LvW]

Yes, I plotted Vout/Vin of op amp in closed loop.
Could you suggest some methods?

1) Stability check is based on LOOP GAIN response. However, your closed-loop response looks good (in particular, the phase function shows no abnormal behaviour)
2.) These methods lead to a smaller closed-loop bandwidth because the LOOP GAIN is reduced (without change of closed-loop gain value).

 

[FvM]

The phase is apparently wrongly annotated, it should start at -180°.

 

[anhnha]

1) Stability check is based on LOOP GAIN response. However, your closed-loop response looks good (in particular, the phase function shows no abnormal behaviour)

Could you tell me why it is normal?
From the picture, it is clear that while phase shift frpm 0 to 180 degree, gain is still big.
At 180 degree, Av = 39.41dB.

The phase is apparently wrongly annotated, it should start at -180°.

Could you explain why?
I don't think it is wrong. I simulated it in cadence.

 

[Audioguru]

Are you designing a slow opamp? Then make it faster with a higher bandwidth.
Or you could add "feed-forward frequency compensation" explained by Texas Instruments here:
https://www.ti.com/lit/an/snoa646/snoa646.pdf

 

[anhnha]

Thank you. I am designing an opamp. If I make it faster, I am afraid that it will become a comparator.

 

[LvW]

Could you tell me why it is normal?
From the picture, it is clear that while phase shift frpm 0 to 180 degree, gain is still big.
At 180 degree, Av = 39.41dB.
Could you explain why?
I don't think it is wrong. I simulated it in cadence.

Anhnha,

May I tell you something general?
From our different responses you see that something in our answers is correct and something is not OK. Can you guess why?
Because in your first post you didn´t tell us - frankly spoken - nothing ("my opamp is unstable").
We did not know if you are designing an opamp or if you were using a commercial opamp with feedback.
We didn´t know what functions are displayed in the BODE diagram.
We didn´t know if you are using split or single supply.
Now you told us that you are designing not only an opamp with resitive feedback but with a phase inverting FET amplifier in the feedback loop.

Now we are in post#8 and are able to slowly approach some suitable answers.

What I can tell you now is the following:
1.) As mentioned already: Stability check (determination of stability margins) requires the BODE diagram for the LOOP GAIN - not for the closed-loop gain.
2.) If the phase response of the closed-loop function does exhibit an abnormal behaviour (sharply rising phase in the region where the loop gain is app. zero) we can expect that the circuit is unstable. The magnitude of the closed-loop gain may still look OK - even for unstable circuits. But this surprising phase response does not tell anything about the degree of instability.
3.) Your signal input is at the inverting opamp input. Hence, the closed-loop phase must start with -180deg (see FvM`s comment).
4.) From your diagram it seems that you are using single supply. Did you check the operating point? Did you perform a transient analysis? This is more confident than an ac analysis.

 

[anhnha]

Thank you.
I think my schematic is a little confusing.
Vin is the voltage feedbacked from voltage divider. It is at the non-inverting input.
The inverting input is reference voltage. It is 0.8V.

As you mentioned about loop gain. I simulated it and now the circuit is unstable.
The loop gain = Tout/Tin
Tin is a sinusoidal voltage.
Not sure why my picture for gain and phase of loop gain is damaged.
The result is that at for loop gain, phase starts at 180 degree.
As gain falls to 0dB the phase is -22.4 degree. That is unstable.

 

[crutschow]

The op amp internal schematic shows no compensation, so naturally it will be unstable. An op amp without compensation is basically a comparator. If you want the highest bandwidth with minimum compensation, then you could use a current-feedback type op amp design.

 

[LvW]

The op amp internal schematic shows no compensation, so naturally it will be unstable. An op amp without compensation is basically a comparator. /QUOTE]

Perhaps it is helpful to complete the sentence:.....it will be unstable for feedback factors above a certain limit (equivalent to closed-loop gains below a certain limit).
An opamp as a stand-alone unit never can be "unstable".

 

[FvM]

Some more points about the present circuit and simulation problems

The loop gain = Tout/Tin
Tin is a sinusoidal voltage.
Not sure why my picture for gain and phase of loop gain is damaged.
The result is that at for loop gain, phase starts at 180 degree.
As gain falls to 0dB the phase is -22.4 degree. That is unstable.

You'll "measure" the gain in small signal analysis. It uses an AC, not a sine voltage source. To get correct results, you have to assure a correct bias point. That's not necessarily achieved by applying 0 V input differential voltage to an OP, at least you need to check the DC output level. Better use DC feedback to set the bias point.

The negative phase margin in post #1 is even larger, rather -100 than -22 degree. The result seems realistic for an uncompensated amplifier with four gain stages, three internal to the OP and one external transistor.

To design a compensation for stable closed loop operation (which surely won't keep the original bandwidth), the location of individual poles must be known. The OP topology seems not well suited for standard compensation schemes because second and third stage are both inverting. You can perform pole splitting between 2nd and third stage by miller compensation of the third stage, but the first stage pole will be kept unchanged. So the compensation must be strong to act as a dominant pole for the OP open loop gain.

The external transistor stage adds loop gain and a fourth pole and will also need compensation.

 

[crutschow]

The op amp internal schematic shows no compensation, so naturally it will be unstable. An op amp without compensation is basically a comparator. /QUOTE]

Perhaps it is helpful to complete the sentence:.....it will be unstable for feedback factors above a certain limit (equivalent to closed-loop gains below a certain limit).
An opamp as a stand-alone unit never can be "unstable".

Of course, that's true. I was referring to the ops use of the opamp in his posted circuit showing a feedback loop.