Transfert function of complex polyphase filter: simulation gain issue

[Chr's]

Hi,
iḿ currently work on plyphase filter.
I arrive to trace the correct frequency curve, where the cut-off frequency fit with the theory, but the gain.
I've a difference of 3 dB (more) with the theory.

So my question is:
Like 3dB=√2, when the calculator does the modul of I+jQ, the magnitude of I and this od Q are equal so the addition give twice the magnitude, so when de module is computed may be i've to divide by 2.
Like |I+jQ| =√((I²+Q²)/2)
= √(I²+Q²)/√2
= 20log(√(I²+Q²))-20log(√2)
= 20log(√(I²+Q²))-3

=> and I get my 3dB less and fit with the theory.

If I'm wring, anyone have an idea on the reason of 3dB more compare to the theory ?

Thanks in advance for your help.

 

[crutschow]

How is the filter implemented?
Post a circuit diagram.

 

[Chr's]

Hi,
It's a Tow thomas polyphase filter diffrential with current input, so my gain is dB(V/mA).

 

[Chr's]

Hi,
Find in attached file the structure of the filter.
I've to redign it in new technology but I didn't find any information about it.
In fact, it's a fully differential band pass polyphase filter with center frequency of -1MHz but normaly a simulation of I (or Q) only give me an low pass with f-3dB=1MHz with the same DC gain (DC gain of low pass= Gain in band of band pass).
The input signal is a current.
My question is, i beleive that's a Tow-Thomas structure but I'm not sur. I'm trying to compute the transfert function but the result found doesn't fit with the simulation.
Regarding the value: C1=2*C2=22.6pF
The gain triming are on R1, R2 and R3.
The frequency triming are on C1 and C2.
If someone can help to find a paper or give me some explanations.
Thanks in advance.



- - - Updated - - -

I made a error, it's not Vin but Iin.
And in the implementation, like it'a fully differential structure, the last inverter is not, the investion is done by P-N before R2

 

[Chr's]

Hi.
So I've progressed in the comprhension of this filter. Following computing of transfert function I'm getting results corresponding to it. I've a new question. When the interconnect between I and Q to do the polyphase filter, the signal is introduced throw resistor and the theory saying that the pole Ra*C1 does the 1st frequency shift for the 1st stage and the pole Rb*C2 does the 2nd frequency shift for the 2nd stage (in attached file). If I change the resistors value I change the frequency shift so that fit with the theory.
But add to this frequency shift I've a gain variation. But in my Gain formula those resistors are presnt in. So I don't inderstand why i'm getting this gain variation. If someone can help me, he's welcome.

 

[Chr's]

Following error about included picture



- - - Updated - - -

I found by simulation that the Gain is not impacted by Ra and Rb if the poles (Ra*C1) and (Rb*C2) are equal.