Help needed understanding a switched cap filter

[stickjam]

I'm trying to understand how (and just as importantly, why) this lowpass filter works in an existing design. 

It appears to be a switched capacitor filter, however, I thought the switches on either side of the cap need to open/close alternately?  In this case, they are all controlled by the same phase of digital clock which swings from +5v to -5v which powers the CD4066 bilateral switch chip. 

So what would the filter's corner frequency be related to clock frequency?

Thanks

-Bob

 

[JohnRoberts]

This is not a classic "switched capacitor" filter. What is actually going on is the resistor values are being duty cycle modulated by the on/off ratio.

The switching frequency doesn't matter but should be high enough to not interfere with the passband. When the switches are closed continuously the Rs look like 1.5k. When the switches are connecting the Rs for less than 100% duty cycle, the effective resistance is increased proportionally. 50% closed they look like 3k, 25% closed 6k, etc..

I once considered using a variant of this to make a tracking anti-alias filter for a BBD delay so at long delay with low sampling frequency I could automatically optimize the filters. I was going to use a fixed one shot for the on time, and the varying BBD clock period for the off time. This would give me a filter that changes in the correct direction. As I recall I abandoned this approach without thoroughly exploring it because it wasn't worth the complexity, for my application.

JR

Note: If there is significant bias current in the opamp used it will cause a DC shift with effective R changes. Also the opamps need to have good gain bandwidth to deal with HF switching components.

 

[PRR]

> what would the filter's corner frequency be related to clock frequency?

Wrong Thinking.

Simple 2-pole active filter. Like Sallen-Key, though when unity-gain may be under another name. Resistors are varied by clock Duty Cycle, not clock frequency.

If the switch is closed 100% of time, the "1.5K" IS 1.5K and R-C corners are near 10KHz.

If the switch is closed 1% of time, the "1.5K" acts-like 150K and R-C corners are near 100Hz.

Clock must be much higher than upper bandpass. The low-pass nature will tend to smooth-out the abrupt on/off action of the switches. How much? I dunno, but you say the design is "existing", so somebody already figured it out and felt it was good-enuff to get paid.

 

[stickjam]

A-ha!  I knew you guys could make sense of it!  I had a passing thought if duty cycle had something to do with it, but I was somehow fixated on its resemblence to a switched cap scenario, figuring it has to be frequency modulated.

That answer also changes my whole paradigm about all the digital crap that produces Fc; that part makes sense now too.

Actually, JR. it IS an anti-aliasing filter.  ;D

-Bob