Switched capacitor filters

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm curious if anybody has tried using switched capacitor filters for audio like the MF10? I'm in progress of making an audio system consisting of these speakers:
4 x 3 way (left/right front/rear) and 2 subwoofers (I have a soundblaster live with 4 outputs for surround sound).
All speakers are active active (yes, 3 monoamplifiers per speaker). The preamplifier is designed to filter all audio signals for the amplifiers and is controlled via an atmel microprocessor.
I just want to know if anybody has any tips to share with me?

btw: if you're interested in the schematics for this project: mail me and I will send them to you. If too many people want them, I will post them.


I just worked with the MF10 for a school project two months or so ago. The filter works with the "State var" filter type

After trying different setups of the device, I tried to make nutch filter (Band stop). It worked nicely, but the clockfreq of the system was very noticable on the output.

I would not use it for audio, build your active filters using opamps.

Ok, then what about this: The MF10's clock ratio can be set from 50:1 to 100:1. Suppose I want a low pass filter at 150Hz with clock ratio set to 1:100. That means the clock runs at 15Khz. If I then use an opamp filter at the output stage with it's -3dB point at let's say 1Khz, then where is the problem? The clock can't get through because it runs at a frequency 15x higher than the analogue lowpass filter. Next I use another 1Khz lowpass filter (for ease) at the input to ensure the input signal's frequency can't rise beyond half the samplerate to avoid aliasing.
Now I have a very good lowpass filter that is clock tunable without clock feedthrough or aliasing effects.
Am I right? Or am I missing something? This was the original setup I have for the amplifiers...
Multiply or divide

Let's say you are using an LTC1063 -- for a 200 Hz cutoff point you have to clock the filter at 20kHz. There's about 50uV of 20KHz clock feedthrough, however, so you would need to filter this at 100 * 20kHz -- you could cascade a couple decade counters to generate both frequencies, i.e. start with a xtal oscillator at 2MHz, divide down by 10 for 200kZ and 10 again for 20kHz.

Let's say you just have a 20kHz source, however, if you use a divide by 10 counter in a 4046 CMOS phase-lock-loop chip you multiply the output of the 4046 by 10.

Besides the clocking noise, there's other noise associated with these chips, but I find it a great device to work with, and as always, Linear's tech support is excellent.
Nice aproach, but if you say that the switched cap filters have more noise than traditional opamp filters, I going for the first aproach to get as little noise as possible.
But the ltc1063 is a butterworth filter? I was taught that the only filters suitable for audio are the bessel filters because of it's constant delay? What about that? Does lineartech have clock tunable filters that you think are suited for audio use?
Or do you have another approach for my setup? I just want to be able to change cutoff frequencies with a microcontroller. Switched cap filters where the first I thaught of, but maybe there is more?

Analog and Digital Filters

If you want go analog, do it with a microprocessor cotrolling the digital potentiometers from Dallas (now Maxim), Analog Devices, Microchip etc. I was thinking to gang together several and writing a program to change the resistors on the fly. Bessel, Butterworth, Chebyshev -- whatever you want, knock yourself out. Figure that a 10K digital pot has about 39.1 ohms resolving.

The noise of some (not all) of these devices really gets down to the basement of the measuring equipment I happen to own.
OK, Jens

Analog has 1k pots with 1/(8^2) resolving capacity, put in series a pair of 10K pots and you can go anywhere, do anything. The math is trivial.

Not enough time to figure the passband and stopband ripple given the resistance and statistical variation on a pair of 10k pots (that's about 40 ohms on 20K) is at the extrema.

I've tried switched cap filters and was unhappy with them.

Althought I have not tried it, your idea of an additional filter at 1K would probably work just fine.

What are you going to do about the high pass section? Your idea won't work here.

But I'm curious. Why not select the correct filter frequencies in the first place? What's the attraction of a microprocessor controlled filter?
Switchable filter

I used a switchable filter with a microprocessor controlled signal generator -- cuts down the harmonics -- but the switched cap filters have this 50uV of clocking noise feedthrough -- I spent an entire weekend breadboarding active filters, and indeed, matching of components is quite necessary, so it's a point well taken.

My audio system is biamped, or rather tri-amped if you include the subwoofer, and I am happy not to twiddle with anything on it. I guess I got started in this about 10 years ago when I saw a switchable xovr at the local music store ... that's how a lot of this stuff gets started. Unfortunately, for reasons of domestic tranquility I don't get to listen to the system too much, save Sunday evening when the XYL heads for the sack and I can close the door and listen full crank.
Well, the advantage of a microprocessor filer is that it's completely universal. It can be used with any kind of speakers in any room. The listening room greatly affects the audio spectrum of the speakers you have (reflections...). And since I don't have a clue how to calculate something like that, I want to be able to change my settings. Why a microprocessor: simple.. I can store different settings that way. Listening to a dvd requires a very different setup than listening to audio.
And about the highpass section: why souldn't it work? cut off freq is to be set at 5-8Khz for the tweeters. Ratio is set to 50:1 so clock freq varies between 196-314Khz (Mode 3 highpass, see formulas on page 10 of the maxim MF10 datasheet...). All I need to do is add an analog lowpass filter after the switched cap filter with a cut off freq at 30Khz or something.. Right :D ?

I'm talking about the high pass at 150Hz. I'm assuming you will be having one.

I know a great deal about room interactions. I can state without fear of contradiction (Hah!) that you should NOT tailor a speakers crossover network in an attempt to correct room problems. Use an equalizer, if you must, but better to pay attention to speaker and listener placement then treat the room for remaining problems.
Well, the 150Hz highpass could be a problem if it wasn't a midrange speaker... The mid has a highpass at 150Hz and a lowpass at 5-8Khz. Clock for the 150Hz can be set to 100:1, so I need a freq. of 15Khz for the highpass wich is still higher than the lowpass at 5-8Khz :D
And about the room interactions: I know you could use an eq if the filters are tuned for a certain set of speakers. But I don't see how it's possible to determine the correct x-over frequencies for a certain speaker. Say the tweeter can start from 5Khz up and your mid is able to go up to 8Khz. Where do you choose the x-over point? At 6Khz or 7Khz, or maybe 6.5? That's why I want to make it variable. I want to experiment with that to see what's the difference is sound. And if I want to make another speakerset for me, I don't need to do some dull calculations to try and predecit what the set will sound like. I can just try it out. Sound is very subjective...

you can use CMOS counters and divide by anything.

Let us say that you don't like the clock-through noise, or 0.01% distortion of the LTC1063 -- and you decide to go with a filter made of op-amps, caps and resistors. <p> It's not difficult(but it itsn't simple) to take the equations and factors and write an algorithm to allow you to change Q, f3 etc. on the fly by changing the values on several digital pots, and using FET's to range switch with capacitors. Well, it's a little hairy using integer math but you get my point. I think that this method would be a lot easier and less expensive than ganging pots together, or making discrete plug-ins (which I have done -- using 14 pin DIP plugs). The cost for a DIY'r to make a mcu controlled 3 way crossover with an LCD display would probably be less than $100 -- assuming you make your own pcb's know how to program a PIC or AVR.
Versatile filter

Hi Peter,

I think you have to think grand here. Check out http://dias.umist.ac.uk/PAG/signalwizard.htm, a 2-channel dig filter unit, complete with graphics software, to implement simultaneously x-over filters AND room-correction equalisers, store filters and switch on the fly, very nice. If you can afford it..

Jan Didden

Edit: I think you have to go through http://dias.umist.ac.uk/PAG/ and then click on signalwizard.
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.