3

I don't mind if it's a band-pass or low pass filter (although I think low-pass will give a better result). Input operating frequency range is 100Hz to 1kHz. Input amplitude will be about 1Vp-p and although it will be a bit noisy and not very square-wave-ish it'll be good enough to feed into a comparator to get a decent square-wave should that be required.

Basically I need a design that will take a noisy square-wave output from a strain gauge amplifier, determine the frequency of said square-wave and filter out the high frequency noise and harmonics leaving a sinewave.

EDIT - the end game is to measure the amplitude of the squarewave accurately - by filtering-in the fundamental frequency it becomes easier to measure because the squarewave amplitude is very accurately linked to the sine amplitude (4/Pi from memory)

This is for a test box where the strain gauge (connected to the amplifier) is supplemented with a FET switching a large value resistor connected across the strain gauge - this simulates a small but predictable amount of strain. The output from the amplifier needs to be converted to a sinewave for measurement purposes. I'm aware that this signal could be taken into a PC running a Fourier analysis on it but it's not always convenient to have a PC handy.

I suspect, for this tracking filter a switched-cap filter like the MF10 (or based on similar) will be required and a phase-locked-loop to track the actual input frequency to generate a clock at a much higher rate.

Currently, I have a 12th order low-pass filter that is good for a small range of frequencies around 600Hz (with minor adjustments) but this is no good for lower or higher frequencies.

Please let me know if there is information missing that would help answers. I am looking for quite high precision and if it needs a select-on-test components that's not a problem.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • 3
    Read up on the [lock-in amplifier](http://en.wikipedia.org/wiki/Lock-in_amplifier), which sounds like exactly what you need. – Dave Tweed May 23 '13 at 13:48
  • 1
    have you peeked at the VFC32?http://www.ti.com/lit/ds/sbvs015/sbvs015.pdf – Scott Seidman May 23 '13 at 14:01
  • @DaveTweed interesting observation - I have the signal that inputs to the amp under test and this could work - OK I'd need to generate 0º and 90º reference signals to properly do the job. Good call. – Andy aka May 23 '13 at 14:01
  • @ScottSeidman are you suggesting using the VFC32 as some kind of signal demodulator? – Andy aka May 23 '13 at 14:08
  • Just sort of a rough idea that popped into my head. The f to v functionality should let you get a voltage proportional to the freq, and then you could either use that to drive a sine oscillator, or use it as an input for a lock-in amplifier (which needs to know what frequency it's looking for!) – Scott Seidman May 23 '13 at 15:09
  • Sounds like you want a PLL that can track the input square wave frequency from 100~1000Hz and which controls a sine wave oscillator. The range 1:10 is a bit large for an oscillator built from discretes though, maybe use a DDS (digital synthesizer)? That way you don't have to worry about tracking filters, a simple low pass filter slightly above the maximum expected frequency should be sufficient. – jippie May 23 '13 at 19:02
  • @jippie before asking this question I had in mind a PLL locked to the frequency but running maybe 50 times greater controlling switched-cap filter. I'm not so sure now but also I'm not sure how your PLL controlling an oscillator gets me a solution so please elucidate. It's definately not an easy application. – Andy aka May 23 '13 at 20:10
  • You want to get rid of the harmonics, right? So if you manage to use an oscillator that as no harmonics (sine wave) then you don't have to filter them out. Your challenge is to get that sine wave oscillator in sync with the original frequency, hence the PLL. – jippie May 23 '13 at 20:13
  • @jippie, so you get a decent sinwave in sync with the noisy squarewave and multiply them? Remember I'm trying to accurately measure the noisy squarewave amplitude and if that isn't clear in my question I immediately apologize and I'll have a rethink about the question content. You guys - I really appreciate giving this some thought. – Andy aka May 23 '13 at 20:16
  • 1
    @Andyaka Since you're considering the MF10, look at the [LTC1068-25](http://cds.linear.com/docs/en/datasheet/1068fc.pdf) as an option: 4 matched 2nd order filters, clock to corner ratio 25, can be cascaded for 8th order low-pass. Use a zero crossing or low pass filter plus comparator on the source to get a clock, divide it by 24 (not 25), and feed that clock in, for a sharp cutoff just above the source signal's primary frequency, that will track as the source frequency changes. That 24 --> 25 margin will give the filter the breathing room you might need. Would that work for you? – Anindo Ghosh May 25 '13 at 10:00
  • @AnindoGhosh this would probably work just fine - when you say "divide" I assume you mean "multiply" and use a PLL for generating 25 x F? – Andy aka May 26 '13 at 14:27
  • @AnindoGhosh - I'm not sure now! Just had a quick read of the pdf and it seems the clock ratio is nominally 25 times higher than actual filter frequency. Are you sure you haven't got this the wrong way round? – Andy aka May 26 '13 at 16:07
  • @Andyaka Oh, you are right: The ratio is 25:1, not 1:25. I've just proved the "*If it's too simple to be true, it isn't true*" tenet! You're back to PLLs, unfortunately. – Anindo Ghosh May 26 '13 at 16:21
  • @AnindoGhosh I don't mind using a PLL of course and the device you unconvered has some really high performance filters that look attractive. Another company I know in the same business use this technique but MF10s and they don't performa that good from memory. – Andy aka May 26 '13 at 16:53
  • @Andyaka No, the MF10 got superceded by the LMF100, I think... Much as I try to stick with Texas Instruments parts, this is one area where LTC and Maxim both have some very impressive ICs, so I would drift from TI options. – Anindo Ghosh May 26 '13 at 17:15
  • Beware of switched capacitor filters they are very noisey I have featured them in www.badbeetles.com a website that I got set up for people to share problems with ICs – Autistic Jul 23 '15 at 02:48
  • @Autistic I think you should add a proper hyperlink that works. – Andy aka Jul 23 '15 at 07:19

0 Answers0