How to design this Chebyshev filter??

[wjxcom]

Hi all: I am a fresh man to design filter! Now I am designing a Chebyshev filter, this filter can be found in the book "CMOS Analog Circuit Design (2th edition)" writed by Allen, i.e. the Example 9.7-5.

After the AC simulation, I found the magnitude response and the phase response of the filter is not right!

I do not why!

At the same time, if the periods of the CLOCK is 0.8u, what's the value of the GAIN and GBW of the OPA? and why??

Because the filter includes some circuit, so I put this circuit in the file "filter.rar", at the same time, the library file "csmc.lib" and the netlist file can be found in "filter.rar" too.

Help me please, thanx!!!

 

[skythunder]

Hi,
I am also working on designing a bandpass SC filter, so I have read the chapter you mentioned in ALLEN's book. You should note that Allen also mentioned that this example's magnitude is affected by sin(x)/x, which leads to the passband gain is not satisfied and it needs prewarping technique.

On the other hand, designing a SC filter needs to consider many factors. For example, the effects brought by switches, charge injection, clock feedthrough. You need to take an early-late clock phasing, just turn off the switch connected to the input of opamp a little earlier, which is also mentioned by Allen in section 9.3 SC integrator. This may improve the magnitude response.

As to the requirements of OPA, the larger the gain is , the better. while GBW is at least ten times the sampling rate (for we need the output settle to final value within far less than T/2).

All above are observed when I do my own design. However, I still can't not get good result of my design. I can observe that my design can perform bandpass function right, but the passband gain get attenuated much.

maybe we can keep discussing this problem.

 

[wjxcom]

Hi skythunder， I'm glad to discuss this problem with you!

I am a fresh man to design filter, so the most thing I cared is the function of the filter, rather than the performance of the filter!

I think but from the switches, charge injection, clock feedthrough and so on can infulence the performance of the filter, but they can not infulenct the function of the filter.

you said " GBW is at least ten times the sampling rate ", I think you are right! may be I didn't notice this so I can get a good function of the filter!

thanx you!

 

[skythunder]

hehe, yes, I also focus on the correct function of the filter.
And I have downloaded your .sp file, and can you make sure that it can perform AC analysis of SC filter ? Because I always only do trans simulation , which consumed me too much time. If so , I will try your way to analysis my design.

regards !

 

[wjxcom]

From the book writed by Allen, in fig 9.7-7, there has a sentence:"VIN 1 0 DC 0 AC 1", so I think I add a AC source to the input of the filter and perform AC analysis then I can get the awave about magnitude response and phase response.

 

[wjxcom]

Hi skythunder, have you already simulate my circuit?

regards!

Added after 4 hours 43 minutes:

Hi everyone: no body can help me??

 

[Yanhui]

From the book writed by Allen, in fig 9.7-7, there has a sentence:"VIN 1 0 DC 0 AC 1", so I think I add a AC source to the input of the filter and perform AC analysis then I can get the awave about magnitude response and phase response.

It's not a continues circuit! so u cann't use AC analysis.
If u have spectre, u can use PSS & PAC to simulate.

give a pulse source, get enough tran analysis data. sample and FFT at MATLAB.
so u can get the freq-character about this filter.

 

[wjxcom]

Hi: though It's not a continues circuit, I use this method to perform AC analysis:

1. I perform TRAN analysis and save the operating point at 100us
2.then I load the operating point and perform AC analysis.

in fact, I think use PSS & PAC to simulate circuit is the method that I mentioned above.

At the same, I use Hspice rather than Spectre, so I can't perform PSS & PAC analysis?

regard!

Added after 35 minutes:

Hi Yanhui:

give a pulse source, get enough tran analysis data. sample and FFT at MATLAB.
so u can get the freq-character about this filter.

would you give me the detailed process to get the freq-character about this filter?

thanx!

 

[neoaspilet11]

IF you can digitize your signal, then you can easily program a DSP chip to implement a digital Chebeshev filter. Its far more convenient, flexible, and less complicated. DSP chips are getting cheaper now adays.

 

[andre_luis]

Why don't you generate your filter using this software ?

**broken link removed**

It's the best I saw.
Using it, you can create several circuit topologies, active, passive, switched capacitors, FIR, IIR, etc...


+++

 

[Yanhui]

Hi: though It's not a continues circuit, I use this method to perform AC analysis:

1. I perform TRAN analysis and save the operating point at 100us
2.then I load the operating point and perform AC analysis.

in fact, I think use PSS & PAC to simulate circuit is the method that I mentioned above.

At the same, I use Hspice rather than Spectre, so I can't perform PSS & PAC analysis?

regard!

Added after 35 minutes:

Hi Yanhui:

give a pulse source, get enough tran analysis data. sample and FFT at MATLAB.
so u can get the freq-character about this filter.

would you give me the detailed process to get the freq-character about this filter?

thanx!

a pulse source includ all range frequency. after this filter, signal in band will be lose. so, u can get the freq-domain character.
the most simply method is get the output data of a 'tran' analysis with a pulse source.
sample use the clock frequency. then do FFT analysis to those data after sample.

 

[wjxcom]

Hi all: I have already perform the TRAN simulation, and I think the wave of TRAN simulation is right!

but the result of AC is not right!

 

[pipida]

I rarely simulate filters in time domain. Why don't you try varying frequency.

 

[wjxcom]

Hi pipida: would you explain you meaning?