bandpass switched capacitor filter design problem

HI,guys, I am a new comer for help.
I am working on designing a bandpass SC filter. The specification are as follows:
center frequency is 120K; passband 10K-15K(the bandwidth), passband gain 0-3dB;stopband 70K-80K(the bandwidth), and stopband attenuation >-60dB;

I designed this filter totally according to the chapter 9 of Allen's book<CMOS Analog Circuit Design,2nd>, taking the lowpass-to- bandpass normalization method of cascade biquads . I took the order of 6 and chebyshev filter.

After work out all the required capacitors' value, I use Hspice to do an transient simulation (such a pity that the SC circuit cannot be simulated in AC sweep) The input is a sine wave with 120K of interest, which should be output without attenuation. However, after a relatively long settling time, the output settled to a smaller value than input (rougly -15dB). Because the quality factor of this filter is rather high, so the capacitance spread is rather high, which I guess is the reason why the input is attenuated.

I do follow the design procedure on Allen's notes, so I doubt whether this way has the practical application value. Can anyone give me some advice ? Thanks in advance.

PS: The OPAMP I used has a bandwidth of 10M with 10p capacitive load (two stage), DC gain 85dB.

Have yoiu taken any special precausion measures for the switch capacitor circuitry?
- Is it fully differential?
- Have you negate clock feedthrough?
- Have you buffered the switches?
- Have you negate the dc offset voltage of the OPamps? ( this is to be taken care at t=0+1).

I believe that you loose on these points, but as I am not an expert on filters, it mey be another cause as well.

D.

dkace said:

Have yoiu taken any special precausion measures for the switch capacitor circuitry?
- Is it fully differential?
- Have you negate clock feedthrough?
- Have you buffered the switches?
- Have you negate the dc offset voltage of the OPamps? ( this is to be taken care at t=0+1).

I believe that you loose on these points, but as I am not an expert on filters, it mey be another cause as well.

D.

Click to expand...

Thanks for the reply!
I used single-ended OPAMP,no differential output. I do agree the clock feedthrough and OPAMP's offset have bad effects on filter's performance, but I don't know whethey they mainly contribute to the attenuation of output with the frequency of interest. And I don't know what's effect if the switch is not buffered.

I have a little doubt on the designing way. because in the biquad , the integrator capacitor have the biggest value, so there is no gain of the integrator, I think that's why my output is reduced. And I made a test and found that the attenuation degree is relation of the ratio of integrator capacitor to the smallest one in the biquad .

thanks !

I undestand your point. I have almost no knowledge on filters so I would appreciate if you could answer me this:

DC offset of the OPAMPs is feed through the SC network to the output. At the same time, charges that are finding there way through the network due to the switching of the MOS are also accumulating to the capacitors.

Now I understand that in a filter like that the SC network is working in an intergrator configuration. If the Capacitors charge is differing than the ideal charge, don't you have problems with attenuation at the output?

Why don't you try a fully differential topology? You will copy everything and you will just name a Common mode voltage. I guess it will be quite easy to see if these effect your design...Sorry I can't help on the filter each self- it is not my specialty ( but I want it to be, that's why I ask you)!

Thanks,
D.

dkace, thansk for sharing your idea with me ! Now I realized that the DC offset of my OPAMP may be a big problem, because in order to improve capacitive drive ability, I increased the its bias current from original 10uA to 100uA, and the unity gain bandwidth increased a lot while the DC offset also increased from 160uV to almost 600uV. I may find the direction which I should make effort to.

By the way , fully differential topology is really a good choice, but I have none available in hand now, so I have to design one by myself . CMFB is another challenge and I am sure I will meet more problems, and I am sure I will come back!

Added after 2 minutes:

Here is a link to a good article, which explains the non-linearity factors in SC-filter design, enjoy it!
**broken link removed**