[Edited to give more details]

I'm trying to build an 6th order Butterworth active op-amp low pass filter with programmable cutoff frequencies of 1/5/10/50/100/250/500/1000/2000Hz.

Another limitation is that it must fit in a very small footprint, about 10mm^2.

I've come up with a design using a three circuit op-amp with a six resistor digital pot with a max resistance of 100kOhms. However, it's only 256 taps.

With those specs, I can only get a f_cutoff of about 2.5Hz to a little over 500Hz.

Can I use a multiplying DAC instead of the digital pot?

I'm going to baseline the noise versus a switched capacitor filter to see which is better. I've been told it's preferable to not use the latter.

  • \$\begingroup\$ You can configure programmable filters using either. There are lots of digi-pots available, probably more than MDACs. So you'll have to be more specific about what characteristics you need from your control element (not ' ... I can't find the characteristics ... ') to get any meaningful help. \$\endgroup\$
    – Neil_UK
    Oct 18, 2016 at 8:16
  • \$\begingroup\$ My paint is the wrong colour, can you advise me the right colour? \$\endgroup\$
    – Andy aka
    Oct 18, 2016 at 8:18
  • \$\begingroup\$ How about a clock tunable 6th order switched capacitor butterworth filter All you need to tune it is a clock at 50 (or 100) times the desired cutoff frequency. Lots of companies make them, that one from TI was just the first that google spit back at me. \$\endgroup\$
    – JRE
    Oct 18, 2016 at 8:33
  • \$\begingroup\$ Hi JRE, I'm comparing it to a traditional active op-amp filter w.r.t. noise. \$\endgroup\$
    – sjlee001
    Oct 18, 2016 at 8:36
  • \$\begingroup\$ Sounds like you need to switch the caps to get two sub-ranges (I know that will cost you area, but life's a beach). MDACs usually get very sad right down in the LSBs, even though they may have more bits than your digipots. \$\endgroup\$
    – Neil_UK
    Oct 18, 2016 at 8:36

2 Answers 2


The potential problem you are going to see when using using digipots in a sallen key low pass filter is capacitive feed-through ruining the high frequency rejection response. If you examine the data sheet (of say the AD5206), it states the capacitance on each node is about 50 pF.

50pF and 100k ohm have the same impedance at about 32 kHz so this could severely impact the filter's rejection characteristic. This needs considering.

Temperature also needs considering; when used as a rheostat, a digipot has a very poor temperature stability (usually). Again, looking at the data sheet for the AD5206, the tempco is 700 ppm/degC i.e. terrible (or not too bad) depending on your target performance requirements! This could swing your filter's response around quite a bit so please do consider it.

So now, you've been told about digital pots and now you must consider whether the problems I've pointed out are a worser evil than the noise produced by a switched capacitor filter. Do some research on this and try to understand whether the noise is likely to degrade your performance expectations.

Switched capacitor filters also can have significant pass-band ripple (maybe +/- 0.5 dB) so it's another thing to note.

Multiplying DACs can be very good but trying to find several of these to fit in such a small room might be the biggest hurdle. INL (integral non-linearity) performance could be a showstopper in that the straight-line linearity of some DACs may not be as accurately predictable as you think they should be.

I've just been looking at a similar job and I would consider this device: -

enter image description here

  • \$\begingroup\$ Great answer Andy! Quid pro quo: blue. \$\endgroup\$
    – sjlee001
    Oct 18, 2016 at 9:02
  • \$\begingroup\$ @sjlee001 switched-capacitor filters may contribute noise. Look carefully at your noise budget to see if you have enough room. Seems to be a great solution otherwise. You may still need a clean-up post-filter (LPF) to clean up chopping remnants. \$\endgroup\$
    – glen_geek
    Oct 18, 2016 at 14:30

1 ... 2000Hz.

If you try to cover this range merely by varying resistances, you'll need high resolution (12-bit at least) MDACs.

These have relatively low ON-resistance switches compared to 8-bit MDACs, which translate to large area FETs and, as Andy points out, large switch capacitances.

I believe to cover this range you need to think about two ranges, each covering about sqrt(total range), switching in additional capacitors for the LF range. As sqrt(2000) is about 44, maybe 50:1 in capacitance range.

Then choose between digipots or 8-bit MDACs, and you'll need one FET switch per capacitor for the range change.

  • \$\begingroup\$ you read my mind! \$\endgroup\$
    – sjlee001
    Oct 18, 2016 at 16:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.