filter design problem

[hebu]

scf biquad

I designed a third order switched capacitor filter, cascaded with first-order highpass
and a lowpass biquad, but I got a problem.

1)I'm wondering where do I put the first-order highpass filter preceding biquad
or following biquad?

2)My simulation told me this is different, but why? I don't have any
idea about this.

3)I got good results of two seperated filter(seperating the highpass and low pass)
but I got a strange result while I make them cascaded.
Anything I missed?Anybody has the idea?

Thanks,

 

[flatulent]

low outband noise rc filter

One problem to avoid is the out of band switching noise. The order you cascade the filters has an effect on the entire circuit output noise.

One way to reduce this noise is to use an analog RC filter on the output.

 

[willyboy19]

A few things to check:

1) Do you apply a sample-and-hold input to your filter or your filter is continuous-time? The SCF should handle S/H inputs.

2) The cascade implementation of SCF is inherently sensitive to component mismatch (especially high-Q filters). A more robust implementation will be ladder filter based on R-L-C prototype, though the design and debugging is more complex.

3) If I were to implement this 3rd-order BPF using cascade structure, I would put the lowpass biquad as the first stage, followed by the 1st-order highpass. This will prevent high frequency noise from getting into the filter stages.

4) Theoretically (or mathematically), the order of this 1st-order HPF and 2nd-order biquad should not matter. Maybe your implementation has some problems since you did observe difference. You can post either your implementation or your simulated frequency response so that we can see what might potentially go wrong.

 

[hebu]

One problem to avoid is the out of band switching noise. The order you cascade the filters has an effect on the entire circuit output noise.

One way to reduce this noise is to use an analog RC filter on the output.

Sorry, I don't really understand, do you mean I need to insert a RC filter between
the biquad and first-order high pass filter?
What's the "out of band switching noise" in your statement? Does this noise affect
transfer function of sc filter?
How I judge how many order of cascade will be affected by the output nosie?

Thank you very much,

 

[sunking]

pay attention to you clock frequency.
the output prestage and the input nextstage pay attention to no equal R

 

[flatulent]

My advice about the switching noise is from a board design I did using two Linear Technology switching filter ICs. The output switching noise was different depending on which chip was first and second. I had to put a RC low pass filter on the final output to get rid of most of the switching noise.

In your case, first look at the output noise on each one when it has a resistor on the input that is the same as the impedance of the circuit preceding it.

You can then decide to put a RC low pass or high pass filter after it to get rid of the out of band noise. Then try the filters together in both combinations and select the sequence with the lowest output noise.

 

[hebu]

pay attention to you clock frequency.
the output prestage and the input nextstage pay attention to no equal R

Thanks for your reply,
What do you mean "no equal R?"

For the sampling clock, in my knowledge, they should be referred to the same
clock phase. If we define the sampling phase is phase I for first stage, and the
second one must sample in the same clock phase.

Added after 31 minutes:

I'm wondering, I have taken the loading effect of the second stage into
consideration, So, I increase the OPamp bandwidth of first stage to prevent the loading to kill the bandwidth and reduce the settling time.

By my simulation, I exchange the position of HPF and LPF alternatively, the transfer function is ok for the first stage, but it of the second stage goes wrong.
For example, if the HPF is placed at the first stage, the transfer function is ok, but
it's wrong if it is placed at second stage. If the LPF in the first, the LPF transfer
function is ok, but it's wrong if LPF is in second.

So, by this, I'm wondering is this result is because of the change of source
impedance? In the individual simulation, I put a 50 ohms to be the source impedance. The situation is different from the cascaded case.

I'm not sure the analysis is correct or not, any comment on this?

Added after 2 hours 48 minutes:

My advice about the switching noise is from a board design I did using two Linear Technology switching filter ICs. The output switching noise was different depending on which chip was first and second. I had to put a RC low pass filter on the final output to get rid of most of the switching noise.

In your case, first look at the output noise on each one when it has a resistor on the input that is the same as the impedance of the circuit preceding it.

You can then decide to put a RC low pass or high pass filter after it to get rid of the out of band noise. Then try the filters together in both combinations and select the sequence with the lowest output noise.

I can't understand why the switch noise of first stage will affect the transfer function of second stage? Would you explain it in more detail?

I have tried to make the HPF be the first and LPF be the second. The transfer
function is looked better than the case LPF be the first. But it's still a little different
from the characteristic of LPF when it's alone.

In the cascaded filter design, I didn't see any textbook or reference with the
precaution of this. But it seems to happen often.