I am trying to design an instrumentation amplifier, having 2 bio-potential sources as its differential inputs, each having peaks in the order of 10's of uV and a frequency band of 3-50 Hz. One of the main characteristics of the input signals, however, is a DC offset and depending on different conditions, it is about 600-800 mV.

This limits the gain of the amplifier, as it will also shift the output signal up about (Gain * DC-offset) and this in turn, will limit the gain and for a small signal, this means a noisy output (from different sources). One obvious solution is to put a DC blocking capacitor prior to the positive input into the InAmp. This, to me seems not to be a problem, but I am not really familiar with noise sources in capacitors (like Johnson noise in resistors), but i know that some ceramic caps experience that microphone effect, and most of the surface mount caps I have on hand are SMD MLCCs.

Would this be a good solution, or would it cause strange behaviour?

  • \$\begingroup\$ (1st sentence) They are not "differentiating" inputs; they are differential inputs. \$\endgroup\$
    – Andy aka
    Commented Aug 8, 2020 at 19:03
  • \$\begingroup\$ @Andyaka noted. thanks \$\endgroup\$
    – NeuroEng
    Commented Aug 9, 2020 at 8:50

2 Answers 2


I would have agreed with the below Cap material choices, NP0/C0G or Film without thinking about it. But after consideration for EMG, you don't need more than 10nF * 10M but for ECG or EEG, you may want 1M*1uF=T. (EMG for muscle signals, ECG for cardio-heart and EEG for brainwaves.)

Choosing 0.1uF or more might lead you to expensive or non-existent solutions in NP0 or film. The value can be small if the value is 1nF for higher frequency muscle signals as this material are low density (Dk) so the values for inexpensive caps are limited yet input impedances to bias might be small by choosing a large bias resistor. e.g. 10M+1uF More about this at the end.

Thus the next choice is the low-frequency cutoff. Due to galvanic motional skin voltage there is a chemical offset modulated by electrode pressure and motion that can also be a nuisance as noise. So you can determine what is your ideal Bandwidth for SNR and compute the breakpoint for ω=1/RC for the high pass filter, HPF.

But in the end, the most important design criteria is reduction of stray line noise which is a large high impedance V/m stray electric field. This creates a Common Mode noise to both signals and returns cables and the INA IC's are already designed for you with -120 dB CMRR rejection ratio with laser-trimmed resistor matching to 1 part per million. So do yourself a favour and choose an INstrument Amplifier (INA) IC. Some also do multi-channel and allow selection of common-mode inputs to invert for providing biometric 0V return signal shared by all other electrodes so you get the best CMRR from > 50V/m E-fields from nearby line voltage. This is also called Right Leg Drive (RLD) signal = 0V return.

What's all this stuff about Cap. Noise?

I thought it was only resistors that make noise. Well-known are the microphonic effects of ceramic capacitors when you hit the board. But what if the ceramic was ruggedized? Well, it turns out you can do that. SMD's will pick up on the audio resonance frequencies of stiff circuit boards together with the ceramic electrode noise.

But it turns outs, Capacitors on their own do not make noise, but it is also the Electrostrictive effects of ceramic capacitor current in addition to external vibration, that current creates a force to make ceramic crystals resonate from the excitation frequency voltage. The external ruggedized 1uF ceramic cap is available and looks like this from Murata in the KRM series.
enter image description here

They also make an internal ruggedized ceramic in the GJ8 series, but not this large value of 0.1uF, also neither in NP0/C0G, the thermally stable series. Murata's general-purpose (GP) ceramic is called the GRM series.

But what if the cap never sees any large excitation currents? Then it won't create any noise. True except for the microphonic effects, which you can minimize by using an epoxy leaded part that doesn't resonate audio with the board. Yes, let's go for this leaded ceramic cap. or the Murata KRM series SMD part (~$0.5 to $1 1pc). Now, these leaded parts are rapidly going expensive $2 (1pc) before they get obsolete, but your lab may have plenty of stock. The SMD


Ceramic Cap's do not generate thermal noise but rather conduct noise by;

  1. Electrostrictive effects from excitation current and frequency,
  2. Mechanical microphonic effects in vibrational bandwidth together with PCB resonance
  3. Galvanic skin response from electrode motion creates low-frequency noise and then we have DC input offset to deal with high gain amplifier requirements.

But fortunately for most biometric sensing, we do not need a DC response. So I'm saying for low current uV level signals you can use your lab stock of general purpose 0.1 to 1uF ceramic X7R thru-hole caps. enter image description here

  • \$\begingroup\$ I see, I used to think that leaded caps would pick up humming environmental noise, but every day is a school day! What a great answer, my understanding of this specific issue is way better just because of your answer, so thank you for your valuable input. \$\endgroup\$
    – NeuroEng
    Commented Aug 9, 2020 at 8:53

NP0 ceramic capacitors are okay, however larger values tend to be a bit large and expensive. Film capacitors are good, but SMT versions are not so great. Tantalums are not a noise problem but are leaky (and polarized), which might be an issue in your application.

X7R and similar class 2 capacitors should probably be avoided- they can exhibit mV level noise. Here are some informal tests where they whacked them hard and got ~100mV with what looks like mechanical resonances in the several mV range.

I would suggest leaded film capacitors or NP0 ceramic. Don't forget that you need to provide a DC path for your amplifier input bias current so you'll need large-valued resistors to ground (or some other reference voltage).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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