NEEEED help for switched capacitors!!

[lemin_zhong]

Hi all,

I am doing my project for a switched capacitor bandpass fitler now, there is a problem stop my flow.

as the figure shown, i modified those resistor to switches & capacitors, but the R3 which is connnected to ground was not work correct(after my test)

that means if I use resistor for R3, others use switches capacitors, the ckt was worked ok, use SC instead of R3 will make it fail

can anybody help me ? thx alot

 

[Audioguru]

you do not have a switched capacitor filter. Instead you have a Multiple Feedback Bandpass Filter. Instructions and examples can be found in Google.

Your circuit does not show its dual polarity supply.

The input must be fed from a very low inpedance like another opamp.

 

[LvW]

I understand, that you have replaced each resistor in the shown circuit by a switched-capacitor devive, right ?
But that is NOT the correct way of designing an SC filter.
The reason is as follows: Each resistor which is to be simulated by a switched capacitor arrangement must be voltage driven! That means, both ends must be connected to a voltage source (input voltage, opamp output) or ground potential (real ground, zero volts or virtual ground from neg. opamp input). In some cases such a voltage source may be a capacitor which is much larger than the switched cap.
To summarize, you cannot simply convert a multi feedback topology to an SC filter.
The preferred solution is based on a filter topology which consists of integrating stages.

Added after 41 minutes:

Addendum/Correction:
The mentioned restriction (voltage driven nodes) does not exist if the bilinear approximation (instead of EULER) is used to simulate the resistor action. (But his approach has some other disadvantages).

 

[lemin_zhong]

you do not have a switched capacitor filter. Instead you have a Multiple Feedback Bandpass Filter. Instructions and examples can be found in Google.

Your circuit does not show its dual polarity supply.

The input must be fed from a very low inpedance like another opamp.

sry guy ,here is my schematice in S_Edit, I use VCVS for a ideal opamp first, and use a sine wave DC sourece to determine each freq input get different output to calculate the gain of the filter than plot in excel to get the frequency response.

the problem is, when I use SCs(in red) instead resistor it doesn't work correct(that means this resistor which connected to ground changed to SCs got wrong)

thx very much

 

[lemin_zhong]

I understand, that you have replaced each resistor in the shown circuit by a switched-capacitor devive, right ?
But that is NOT the correct way of designing an SC filter.
The reason is as follows: Each resistor which is to be simulated by a switched capacitor arrangement must be voltage driven! That means, both ends must be connected to a voltage source (input voltage, opamp output) or ground potential (real ground, zero volts or virtual ground from neg. opamp input). In some cases such a voltage source may be a capacitor which is much larger than the switched cap.
To summarize, you cannot simply convert a multi feedback topology to an SC filter.
The preferred solution is based on a filter topology which consists of integrating stages.

Added after 41 minutes:

Addendum/Correction:
The mentioned restriction (voltage driven nodes) does not exist if the bilinear approximation (instead of EULER) is used to simulate the resistor action. (But his approach has some other disadvantages).

yup, I convert the Axtive MFB filter to SCs, this is the way tutor suggested me.

so, in ur explain, I cannot simply convert an active filter to a integrating filter.

but here, is it the problem is the ends of SCs?
as u said, ends must connect to voltage or real ground, all of SCs were correspond with this principle(first I thought that maybe SCs cannot ends to ground, but seems not?)

and I don't know ur addendum's meaning...could explain more clear?(thxxxxx aloooot) -_-

so, I ask for a suggestion if it's possible to build a SC MFB filter... or I will return to the first...MY GOD - -

 

[LvW]

..............
and I don't know ur addendum's meaning...could explain more clear?

* For replacing a resistor by a switched capacitor you have THREE alternatives based on two EULER approximations (backward, forward) or on the bilinear approximation.
All three alternatives look different!

*Unfortunately, I cannot see which method you are going to use as your picture shows not enough details. Please, add a new picture which is larger and - if possible - with a white background.

 

[lemin_zhong]

..............
and I don't know ur addendum's meaning...could explain more clear?

* For replacing a resistor by a switched capacitor you have THREE alternatives based on two EULER approximations (backward, forward) or on the bilinear approximation.
All three alternatives look different!

*Unfortunately, I cannot see which method you are going to use as your picture shows not enough details. Please, add a new picture which is larger and - if possible - with a white background.

sry, here is the schematic the element at left top is clock generator for have 2 cross wave to product wave to control switches.

 

[LvW]

Sorry, couldn't answer earlier.
The attachement (pdf file) shows the 3 alternatives to simulate a resistor.
As you can see, up to now your approach is identical to the ER-approximation (capacitor switched through and, then, deloaded to zero), which does not work for the MFB filter topology.

Recommendation: Use the bilinear form (BIL) for all resistors. You can do it very easily with your circuit by modification of the pulse phases in the correct way.
Question: Has your VCVS a gain of at least 10k ?

 

[LvW]

But do not expect, that the SC circuit behaves identical to the RC bandpass.
It is an approximation following the rules of digital signal processing (with a zero at 50% of the clock frequency) .

 

[lemin_zhong]

Sorry, couldn't answer earlier.
The attachement (pdf file) shows the 3 alternatives to simulate a resistor.
As you can see, up to now your approach is identical to the ER-approximation (capacitor switched through and, then, deloaded to zero), which does not work for the MFB filter topology.

Recommendation: Use the bilinear form (BIL) for all resistors. You can do it very easily with your circuit by modification of the pulse phases in the correct way.
Question: Has your VCVS a gain of at least 10k ?

yep, so i need to try BI type of SC for that resistor(because I sure that other SCs were worked ok)?

and sure that my VCVS has very high^_^

thx alot for help, and SCs filter is the integrated purpose, so it don't need to work as well as RC bandpass~~

 

[Audioguru]

Look at the detailed datasheet of the LMF100 switched capacitor filter IC from National Semi.
It does lowpass, highpass, bandpass, notch and allpass filters.

 

[lemin_zhong]

Look at the detailed datasheet of the LMF100 switched capacitor filter IC from National Semi.
It does lowpass, highpass, bandpass, notch and allpass filters.

yep, thx alooooot, I am very exciting now!!wow, I use BI type of SC, it works!

but why it ? and seems like what I've use were non of the three type u gave in that pdf, is is good? it is modified from EV type.

 

[LvW]

Hello lemin_zhong !

Congratulation! When it works now - fine!

Your first approach was according to the ER approximation which does NOT work for resistors which have at least one floating node.

Now - with the bilinear form - floating nodes are allowed. That`s the reason it works now. Perhaps you can show us some simulation results.

 

[lemin_zhong]

Hello lemin_zhong !

Congratulation! When it works now - fine!

Your first approach was according to the ER approximation which does NOT work for resistors which have at least one floating node.

Now - with the bilinear form - floating nodes are allowed. That`s the reason it works now. Perhaps you can show us some simulation results.

ummm, seems like what happened was as u said, but i am not sure that 'floating node' meaning, the node ends to ground?

here is a graph of my simulation(the filter has passband from 40k to 60k), its from a 50k sine wave(1v phase & it is the centre frequency), the gain was not reach to 0 dB, so i think that i should increase the gain of circuit, and later maybe upload the frequency response plot by excel tomorrow^^

ah, 1 am now~~~i am going to sleep now, hope u have a good night, and i think i will have a sweet dream tonight, hoho