1

for a phase sensitive measurement, I would like to have the DC offset of a signal cancelled. The AC-signal will be amplified and then the phase of it is measured in reference to another AC-signal. My attempt was using a Sallen-Key-Highpass filter with a cutoff frequency of 4 Hz which first worked fine. I recognised, that the phase of the filtered signal shifts with temperature which I want to avoid. I am using C0G ceramic capacitors (220 nF) and resistors with a relatively high resistance of 180 kOhms.

I read about the concept of the DC servo filter which I could also try to implement, but is it promising for what I want to achieve?

Schematic of the filter

T.K.
  • 11
  • 2
  • Comments are not for extended discussion; this conversation has been [moved to chat](https://chat.stackexchange.com/rooms/120463/discussion-on-question-by-t-k-low-frequency-high-pass-filter-drifts-with-temper). – Voltage Spike Mar 05 '21 at 18:45

3 Answers3

2

If you must still use an analog filter then it is important to use good caps like metal film or polyprop .Never use those nasty ceramics here ,they are best for decoupling .

Autistic
  • 14,235
  • 2
  • 27
  • 65
1

An obvious way to avoid temperature effects is to use a digital filter instead. The filter coefficients will not vary with temperature, and the only effect you might see is a small change in the processor clock frequency over temperature.

Elliot Alderson
  • 31,192
  • 5
  • 29
  • 67
  • Wouldn't I need to do this after the ADC? I have a signal with of some 100 mV slowly varying DC offset and an AC amplitude of maximum 3 mV. This is a quite high dynamic range. Or do you mean the switched-cap filter as Pete mentioned? – T.K. Mar 03 '21 at 17:28
  • No, I mean digitize the signal with an ADC and do all of the processing digitally. At such a low frequency you should be able to get very high resolution, I would think, and your "quite high dynamic range" would actually be a fairly small dynamic range. – Elliot Alderson Mar 03 '21 at 19:16
  • @T.K. -- If it is all going to ADC, then this is the way to go. You could subtract a DC signal to relax the demands on the ADC design, if you really need to – Pete W Mar 03 '21 at 20:05
  • DC offset will limit my amplification which I will definitely need for higher frequencies. At low frequencies 300 mV DC offset would only be a factor 100, comparable to 400 LSB when taking a 12 bit ADC. I think I will need the possbility to cancel DC before digitizing the signal. – T.K. Mar 03 '21 at 21:35
  • You could use a 24 bit sigma delta ADC, or even a 192kHz soundcard with the DC blocking capacitors hacked out so it goes to DC... – bobflux Mar 03 '21 at 21:51
0

It is known that all positive-gain Sallen-Key topologies have relatively large passive sensitivity figures. In particular when (as in your case) rather low pole frequencies are to be realized. Perhaps you should consider another filter topology?

  • Negative-gain Sallen-Key filters have better passive sensitivities;

  • Multi-feedback filters also have better passive sensitivities (large active sensitivity is not a problem due to the low pole frequency).

  • The most promising filter structure (in this respect) is a GIC-based highpass. (Keywords: GIC filters, Fliege filter). These filters offer a great flexibility for selecting proper capacitor values.

LvW
  • 24,857
  • 2
  • 23
  • 52
  • I am trying to understand why the GIC-highpass would be more stable here. Do you have an example with a negative-gain Sallen-Key-Filter? – T.K. Mar 05 '21 at 16:36
  • The advantage of a GIC-filter is the following: You have a maximum of freedom for chosing low cap values because they appear to gether with 4 resistors in the equations for the pole data (pole frequency and pole-Q). – LvW Mar 06 '21 at 09:18