0
\$\begingroup\$

We need to design a HPF filter to AC couple a bipolar signal to a high gain amplifier. Any substantial DC left would saturate the amplifier, and thus it must be removed with the high-pass filter(HPF).

But the bandwidth of interest extends down to 0.1Hz and therefore the cut-off for the HPF should be way below 1Hz.

A simple first order solution would use a capacitor in series with a resistor.

Then C needs to be large to achieve a very low cut-off freq., but electrolytic capacitors can not be used as the signal is bipolar. Further, tolerances are poor for electrolytic caps, which would not be acceptable for our precision amp.

What are the best designs for this HPF which avoid electrolytic capacitors?

A possible solution seems to be to implement a second order Butterworth with a Sallen-Key architecture. The (two) Capacitors needed in this design (according to TI's Webench filter design tool) are smaller than the C called for for a first order HPF.

Is this a good way forward (increase order and use multiple stages to avoid large capacitors)?

Which are the established approaches to this problem?

\$\endgroup\$
  • \$\begingroup\$ It's often possible to arrange a DC bias across an electrolytic, utless you need to preserve the DC value. They still aren't terribly accurate or low distortion, but they do work. \$\endgroup\$ – Brian Drummond May 14 '16 at 14:46
  • \$\begingroup\$ What kind of amplifier do you plan to use? What is the desired gain? \$\endgroup\$ – Master May 14 '16 at 15:00
  • \$\begingroup\$ Signals are in the uV range with offsets in the mV. So gains in the 1000-5000 are usually needed and AC coupling is a must to avoid saturation. \$\endgroup\$ – Eric T May 14 '16 at 15:05
  • \$\begingroup\$ What is the highest frequency you need to amplify? \$\endgroup\$ – Master May 14 '16 at 15:06
  • \$\begingroup\$ Do you use operational amplifiers to do the job? \$\endgroup\$ – Master May 14 '16 at 15:07
2
\$\begingroup\$

A common way to deal with this is to use a "DC servo" circuit to cancel out the DC component. Build a low-pass filter to isolate the DC and then invert it and add it to the original signal. The advantage is that you can use high values of resistance and relatively low values of capacitance in the low-pass filter.

\$\endgroup\$
  • \$\begingroup\$ I should have mentioned that low noise is important in our application since the input (bio)signals are in the uV range. Therefore add/substract prior to amplification is usually ruled out. BUT, I would like to confirm I fully understood your suggestion. Why implement a separate LPF and then substract? You are avoiding the electrolytic by using a very large resistor and a smallish capacitor which could have been done in the first place in the HPF. Correct? \$\endgroup\$ – Eric T May 14 '16 at 15:16
  • \$\begingroup\$ Yes it is exactly the same design I suggest. The correct implementation does not create any problems with noise. \$\endgroup\$ – Master May 14 '16 at 15:39
  • \$\begingroup\$ @EricT: Yes, correct. However, there are often reasons (bias currents, noise, etc.) that you can't have high-value resistors directly in the signal path. The servo circuit puts those components in the feedback loop -- a "side channel", if you will -- where those concerns don't exist. The inverting opamp in the feedback loop has no gain in the frequency band of interest, so noise should not be a particular problem, as Master says. \$\endgroup\$ – Dave Tweed May 14 '16 at 16:08
  • \$\begingroup\$ Thanks for your comments/solution. The DC servo solution is certainly worth exploring. And for completeness, possibly it would also be interesting to look at an RC HPF with a >1Mohm resistor (and low-cap ceramic capacitor) to get a large tau>1sec. Then check noise and effect of bias current of the downstream opamp. Thanks again. \$\endgroup\$ – Eric T May 14 '16 at 17:28

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.