AC response of a polyphase filter

[yxo]

filtering an ac constant current source

Dear colleagues,
I am struggling with a polyphase filter. Everything almost ok (or at least it seems to me so), but my AC response is not symmetrical. I don´t know if it is correct or something going wrong. What the source if a such response? Thank you in advance.

 

[yxo]

ac response filter

The filer was made using gm-c technique. Transconductors are ideal.

 

[sutapanaki]

ac response gain from current source

I didn't really understand what you mean by the ac response being not symmetrical. Then, is it normal to have dc gain of almost -80db?

 

[FvM]

AC response of a filter

I guess, if you show the circuit, the reason for the residual DC gain will be obvious.

 

[yxo]

Re: AC response of a filter

sutapanaki, "Not symmetrical" for me is attenuation from the low frequency not the same as from the high frequency.
FvM, I attached the basic cell. There are five similar ones with a different gm and capacitors.

 

[FvM]

Re: AC response of a filter

I guess, if you show the circuit, the reason for the residual DC gain will be obvious.

Yes, it's as expected. The filter has a DC (respectively low frequency) path to the output, DC/LF suppression is only achieved by feedback. Due to the finite amplifier gain, a resisual DC/LF transmission exists.

 

[mengcy]

AC response of a filter

I think you'd better check your transfer function.
what is the Q value of this filter?
If Q>1, the peak would appear.

 

[yxo]

AC response of a filter

FvM, my OTA are just voltage controled current source and a resistor 1000G to ground.
mengcy, this filter has butterworth response. A peak will appeared if Q>0.707. Also, Q is determined for every complex pole in a filter, not for a whole filter.

 

[yxo]

Re: AC response of a filter

I guess, if you show the circuit, the reason for the residual DC gain will be obvious.

Yes, it's as expected. The filter has a DC (respectively low frequency) path to the output, DC/LF suppression is only achieved by feedback. Due to the finite amplifier gain, a resisual DC/LF transmission exists.

I would appriciate it, if you suggest any methodology how to decrease the residual DC gain. Thank you

 

[JoannesPaulus]

AC response of a filter

what about AC-coupling the input...

 

[FvM]

AC response of a filter

what about AC-coupling the input...

Yes, I think, that's the obvious way with finite gain components. As an additional comment, I don't think that filter gain values below -120 or -140 dB have much real meaning, so the request for symmetry can be somewhat relaxed.

 

[LvW]

Re: AC response of a filter

I would appriciate it, if you suggest any methodology how to decrease the residual DC gain. Thank you

In most cases the observed residual dc gain is a result of inproper part and gain matching between both branches.

 

[LvW]

Re: AC response of a filter

FvM, my OTA are just voltage controled current source and a resistor 1000G to ground.

Is this ("1000G to ground") a joke or a typing error ?

 

[yxo]

Re: AC response of a filter

FvM, my OTA are just voltage controled current source and a resistor 1000G to ground.

Is this ("1000G to ground") a joke or a typing error ?

I was talking that I used ideal model (macromodel) which included ideal controlled current source and ideal infinite output impedance due to I made the design with transconductance amplifiers.

 

[LvW]

Re: AC response of a filter

FvM, my OTA are just voltage controled current source and a resistor 1000G to ground.

Is this ("1000G to ground") a joke or a typing error ?

I was talking that I used ideal model (macromodel) which included ideal controlled current source and ideal infinite output impedance due to I made the design with transconductance amplifiers.

OK, I see your point. Forget this resistance.
However, as far as your question at the beginning is concerned:
Remember how the polyphase filter is developped: A hypothetical lowpass (for pos. and neg. frequencies) is somewhat shifted as a result of a frequency transformation.
Because this lowpass approaches zero in the neg. frequency region only at minus infinity, you cannot ecpect that the resulting polyphase bandpass approaches zero at f=0. Instead, there must be a finite dc value. That is no error, but a normal behaviour as expected. And therefore, there is a certain filter symmetry only in a limited region around the mid frequency. Try to redesign the filter based on a larger frequency shift - and you will see that the residual dc gain is smaller.
Regards
LvW

Added after 5 minutes:

Correction: The resulting bandpass has a fully symmetrical response when you extend the frequency range into the neg. region. But, that´s only theoretical.
In practice, the left and the right branch of the transfer curve around the mid frequency are symmetrical only in a band which is twice the frequency shift.