differentiator

[farshaddd]

Hi every one
I'm trying the differentiator block of vhdl lib in cadence but it does not work. The error massage is:
No convergence achieved with the minimum time step specified. Last acceptable solution computed at 0 s.

I also wrote my own verilog A code for differentiator in cadence but is has the same error massage in simulation. We know that the differentiator transfer function in s domain is Vo/Vin = s

what is the problem.

Thanks a lot.

 

[FvM]

Ideal differentiator can't be implemented in time discrete circuit. Use DT1 with small T1 instead
H(s) = s / (1 + T1s)

 

[farshaddd]

Ideal differentiator can't be implemented in time discrete circuit. Use DT1 with small T1 instead
H(s) = s / (1 + T1s)

Thank you for your reply.
I used T1 of for example 1e-9.
The simulation is done.
The phase difference between input and output is 90 degrees and that is correct. However the gain is not 1. Actually, the gain depends on the input frequency!!!!!!!!

 

[KlausST]

Actually, the gain depends on the input frequency!!!!!!!!

This is what one expects form a differentiator!

In doubt: Read any differentiator documentation.

Klaus

 

[LvW]

I also wrote my own verilog A code for differentiator in cadence but is has the same error massage in simulation. We know that the differentiator transfer function in s domain is Vo/Vin = s

Please note that the function H(s)=s is pure mathematics (like 1/s for integrating).
However, an electrical circuit which is intended to differentiate an incoming signal must have a realizable (unitless) transfer function H(s)=s/(1+sT).
That means: It is not possible to realize an IDEAL differentiator. We only can realize a first-oder highpass function which can approximate the desired function as good as possible (large corner frequency with T as small as possible)