Transimpedance amplifier. Gain and Phase margin

[Alexandra1055]

i am designing TIA with the variable resistive feedback gain. I have a problem with simulating gain and phase margin. For the test input signal I am using AC current source and perform AC simulation. When I plot the Magnitude, I get a transimpedance gain which is equivalent to the value of my feedback resistance. But for the phase plot I get a strage data, starting from -180 degree and decreasing further.
Can you please tell me what can be wrong and how to build a testbech for transimpedance amplifiers?

 

[FvM]

Probaly normal behaviour. Transimpedance amplifier is inverting, thus 180° phase shift observed. Phase decreases according to the poles in frequency characteristic.

 

[LvW]

Alexandra - it seems that you are simulating the closed-loop system, correct? However, because you were speaking about stability margins - are you aware that stability margin simulations require loop gain analyses?

 

[Alexandra1055]

Alexandra - it seems that you are simulating the closed-loop system, correct? However, because you were speaking about stability margins - are you aware that stability margin simulations require loop gain analyses?

Yes, I am simulating TIA in a closed loop with the variable resistive gain and feedback capacitance. So I would like to see if my TIA is stable in a closed loop.

Can you please give me more detailes about loop gain analysis? I don't know exactly how to set my testbench...

 

[Dominik Przyborowski]

If You are using Spectre simulator, it provides stb analysis based on Middlebrook method to obtain in easy way the open loop gain and phase.
You have to insert only an iprobe element into your loop.

On the other hand, for classic AC analysis You have to use at least 4 replicas of your TIA and look for the feedback current.

 

[LvW]

On the other hand, for classic AC analysis You have to use at least 4 replicas of your TIA and look for the feedback current.

Dominik, do you agree that "4 replica" are necessary only in case you want an error in the order 0f 0.01% or less?
I have determined the loop gain for countless opamp based circuits based on one single voltage injection - with an accuracy of 99% or even better.

- - - Updated - - -

Yes, I am simulating TIA in a closed loop with the variable resistive gain and feedback capacitance. So I would like to see if my TIA is stable in a closed loop.
Can you please give me more detailes about loop gain analysis? I don't know exactly how to set my testbench...

* Yes - you can perform a stability test for the closed-loop sysytem - however, in the time domain only (step response, TRAN analysis). The step response and the amount of overshoot/ringing gives you a rough information about stability properties.
* But if you want to determine the stability margin you have to simulate the LOOP GAIN (that is the gain of the complete feedback loop if it is OPEN.)
There are several methods for performing such an analysis (small-signal ac analysis). Try a search engine for "Loop gain analysis" or "phase margin determination".

 

[Dominik Przyborowski]

Dominik, do you agree that "4 replica" are necessary only in case you want an error in the order 0f 0.01% or less?
I have determined the loop gain for countless opamp based circuits based on one single voltage injection - with an accuracy of 99% or even better.

For opamp working as voltage amplifier is not a problem to obtain almost correct PM using two replicas, but for amplifier working in transimpedance mode I'm not sure to believe in that simulation. But maybe I'm supersensitive for this kind of circuits ;-)

 

[LvW]

For opamp working as voltage amplifier is not a problem to obtain almost correct PM using two replicas, but for amplifier working in transimpedance mode I'm not sure to believe in that simulation. But maybe I'm supersensitive for this kind of circuits ;-)

According to my experience - when the output resistance of the active device is sufficiently low (in comparison to the disconnetred load) it is really sufficient to place the exciting voltage source for loop gain ac analyses BETWEEN opamp output and load (this correponds to Middlebrooks method without additional current injection). No replica necessary.

 

[Alexandra1055]

Dominik, do you agree that "4 replica" are necessary only in case you want an error in the order 0f 0.01% or less?
I have determined the loop gain for countless opamp based circuits based on one single voltage injection - with an accuracy of 99% or even better.

- - - Updated - - -



* Yes - you can perform a stability test for the closed-loop sysytem - however, in the time domain only (step response, TRAN analysis). The step response and the amount of overshoot/ringing gives you a rough information about stability properties.
* But if you want to determine the stability margin you have to simulate the LOOP GAIN (that is the gain of the complete feedback loop if it is OPEN.)
There are several methods for performing such an analysis (small-signal ac analysis). Try a search engine for "Loop gain analysis" or "phase margin determination".

Thanks a lot, the problem is solved

- - - Updated - - -

I performed the step response and did stb analysis with the iprobe inserted between the output of my amplifier and the feedback path. When the PM is sufficient (was ~75 degrees in my case and is defined by feedback capacitance) there is not overshoot.

 

[LvW]

Thanks a lot, the problem is solved

I performed the step response and did stb analysis with the iprobe inserted between the output of my amplifier and the feedback path. When the PM is sufficient (was ~75 degrees in my case and is defined by feedback capacitance) there is not overshoot.

Between? Are you sure? I rather think, an i-probe has to be placed at one single node only (in closed-loop conditions, other than a voltage probe).