Behavioral model using VerilogA

Dear TUKA,

TUKA said:

I wish to build a behavioral model for a transimpedance amplifier (TIA) using verilogA. 

However, I also wish to see gain peaking in my response as well, as is the case with a practical TIA.

As I am still not into circuit level implementation, I am missing out the poles and zeroes that the parasitic inductors and capacitors would contribute.

How could I use verilogA to model gain peaking in my frequency response??

I am not sure of all the constraints you intend to include in your model (i.e., gain, number of poles/zeroes, gain flatness, etc.) nor the type of simulation you are trying to perform. Hence, my comment may not be helpful.

However, a simple model for AC simulations whose frequency response will show a rolloff of -40 dB/decade and can include a peak in its transfer characteristic might be a second order lowpass filter. The parameters for your model might include the Q and wo frequency as a minimum. The Q will determine the amount of transfer function peaking. The transfer function for a second order lowpass filter as function of frequency, w, is well known and may be found in many references.

The following Forum post provides a methodology for including the transfer function in a veriloga model.

https://community.cadence.com/cadence_technology_forums/f/mixed-signal-design/48127/how-to-model-a-low-pass-filter-using-veriloga

Shawn

Dear TUKA,

TUKA said:

I wish to build a behavioral model for a transimpedance amplifier (TIA) using verilogA. 

However, I also wish to see gain peaking in my response as well, as is the case with a practical TIA.

As I am still not into circuit level implementation, I am missing out the poles and zeroes that the parasitic inductors and capacitors would contribute.

How could I use verilogA to model gain peaking in my frequency response??

I am not sure of all the constraints you intend to include in your model (i.e., gain, number of poles/zeroes, gain flatness, etc.) nor the type of simulation you are trying to perform. Hence, my comment may not be helpful.

However, a simple model for AC simulations whose frequency response will show a rolloff of -40 dB/decade and can include a peak in its transfer characteristic might be a second order lowpass filter. The parameters for your model might include the Q and wo frequency as a minimum. The Q will determine the amount of transfer function peaking. The transfer function for a second order lowpass filter as function of frequency, w, is well known and may be found in many references.

The following Forum post provides a methodology for including the transfer function in a veriloga model.

https://community.cadence.com/cadence_technology_forums/f/mixed-signal-design/48127/how-to-model-a-low-pass-filter-using-veriloga

Shawn

Dear Shawn,

Thank you for getting back to me.

I have considered my TIA as a black box. I have only 3 constraints in my model :

1. Photodiode capacitance

2. gain of the amplifier

3. Feedback resistance (approx. equal to transimpedance gain)

Sure, as you suggested, I will model it using a second order Low pass filter and play around with parameters such as Q and wo to see a peaking in the gain.

However, I found that an AC simulation might not work in my case, because it only gives the voltage gain. What we expect from a TIA is the transimpedance gain (Output voltage / Input current).

Thus, I am doing the same using xf or hb analysis -> Results -> Direct plot -> Main form -> Transimpedance -> dB20.

I am getting a flat curve, and the rolloff of -20 dB/decade at the dominant pole. Now, I believe a second order LPF might provide another -20 dB/decade rolloff.

Thanks once again for your helpful feedback!

TUKA said:

However, I found that an AC simulation might not work in my case, because it only gives the voltage gain.

One small (maybe rather pedantic) point. An AC analysis doesn't actually give you gain (voltage or current gain); you apply a small-signal input (which can be a voltage or a current), and then you can observe the signal at any node in the circuit (or any branch current) which arises from that input - assuming that you are simulating the small-signal linearisation of the circuit. 

Commonly people apply a 1V input, and then the output in volts is indeed the voltage gain as you've normalised the signal levels. You could equally apply a 1A AC current input and then the output voltage you observe would be the transimpedance.

However, an xf analysis is more direct and is really computing a gain. 

hb would only be important if you're trying to look at nonlinearities in the circuit; if it's linear, then hb will just be a slower way to simulate it (not much for a behavioural circuit though).

Andrew

Dear Andrew,

Thank you for giving me proper insights about the different type of simulations- AC, xf and hb.

I will do an AC simulation and check by applying 1A current input and see the output voltage, which would be my transimpedance gain.