2
\$\begingroup\$

Can an auto-zero or chopper amplifier be used in a low-pass filter to remove a high frequency carrier? I am trying to design an active low-pass filter with a 1kHz cutoff and very low DC drift, offset, 1/f noise. The noise needs to be less than 20μVrms in 1kHz BW and -80dB attenuation is needed at 100kHz. I intend to use a Sallen-key 2nd order Butterworth with a single op-amp.

My concern is whether I need to worry about my 100kHz carrier (that I am trying to extinguish) beating against the chopping amplifier modulator to produce more low-frequency artifacts in my 1kHz pass-band, or that higher frequency intermodulation products could alias when I digitize. Would I be better off using a passive filter and dealing with any residual aliasing of the 100kHz carrier digitally?

My power supply will be ±5V and the output will drive a single ended ƩΔ ADC.

\$\endgroup\$
2
\$\begingroup\$

Can an auto-zero or chopper amplifier be used in a low-pass filter to remove a high frequency carrier?

Yes, if you really are using it as a low pass filter, the fact that its a chopper doesn't interfere with the signal unless but only if the auto zero/chopper has a unity gain bandwidth of greater than 100kHz (or whatever your frequency of interest is that might conflict with the chopping frequency ). The reason I say yes, is because most all chopper/auto zero amplifiers have a unity gain bandwidth in the Mhz range. The chopping frequency is higher than that of the unity gain crossover point.

The worst case scenario is the chopper/auto zero amplifier is chopping at 100kHz (or a harmonic). Because these amplifiers modulate and then demodulate signals, if the demodulation happens at the same frequency as the input bad things happen. Usually the chopping frequency is stated in the datasheet, or can be seen in the noise diagram of the amplifier.

I have had issues with this, I had an auto zero amplifier that was picking up RF in the range of 70-700Mhz and it was causing a shift (ever so small) in the amplifiers output. I put an RF low pass filter on the front end and the problem went away, later on analog devices released this part with the EMI filter built in:

enter image description here

So if you are having problems with a chopper op amp, it may be useful to use a passive low pass filter followed by a buffer amplifier at unity gain (or thereabouts).

Would I be better off using a passive filter and dealing with any residual aliasing of the 100kHz carrier digitally?

There are two options, an active filter and making sure the any high frequencies that are to be filtered are within the bandwidth of the amplifier (and thus well away from the chopping frequency)

A passive filter followed by an impedance buffer (this would be the only way to go if you had a frequency on the input that was at the chopping frequency)

\$\endgroup\$
  • \$\begingroup\$ Exactly what I was looking for. It seems like most chopper amps have sufficient GBW so shouldn't be a problem. Do you recommend the ADA4522 series for this kind of application? This is my first real experience dealing with nV signals. \$\endgroup\$ – Mike Sep 6 '18 at 1:21
  • \$\begingroup\$ The ADA4522 works well, so yes I do recommend it \$\endgroup\$ – Voltage Spike Sep 6 '18 at 1:37
  • \$\begingroup\$ Is it okay to cascade a pair of ADA4522 amplifiers in an active filter like Tony's answer, or do I risk inter-modulation problems? I would suppose it is okay if the amps share a chop clock. \$\endgroup\$ – Mike Sep 8 '18 at 3:06
  • \$\begingroup\$ That doesn't matter, the only information that would pass between amps is a noise spike around the chopping frequency. This spike occurs because the 1/f noise in a chopper is modulated to higher frequencies (this is how they rid themselves of 1/f noise. So there is a small spike at the chopping frequency point. If you have any kind of cutoff then this will be filtered out. \$\endgroup\$ – Voltage Spike Sep 8 '18 at 4:53
2
\$\begingroup\$

A chopper is basically a mixer that accepts the entire spectrum of the input above and/or below the chopper frequency including harmonics of the chopper rate to produce intermodulation products to produce an exact replica of the signal. A variation of this is the Sample & Hold (S&H) mixer which does the same intermodulation.

correction

I would suggest a 3rd order 0.1dB to 1kHz or 2kHz with -120dB rejection @ 100kHZ (not -170) so that the chopper signal is well below your ADC resolution and you can provide 60 dB gain or as needed. But your phase distortion spec is unknown, so this can be easily tailored to suit the application. A passive RF filter is always a good idea as the output impedance rises with frequency.

enter image description here

Anecdotal

I recall in the late 70's, in Aerospace R&D, designing a S/H mixer in this way for the purpose of doing Doppler ranging measurements. I used a 10MHz OCXO and fast switches to create a 100kHz Sawtooth signal with the fall time of 10ns in a 10us cycle or 0.1% This had harmonics over 300MHz and was the signal I was sampling with a 10nS pulse created by the zero crossing of a 100kHz sine wave pilot tone from a similar 10MHz OCXO and a PLL to create my pilot tone that was broadcast from the rocket with all the other channels of FM data. All I needed was 1Hz of signal Bandwidth which I multiplied to 300MHz with a zero-cross-switch (ZCS) 1-shot. The result was up to 1Hz Doppler shift of the 100kHz pilot tone equivalent to x km per beat frequency cycle at speed of light while the rocket was going up to Mach 7.

  • The interesting part was the exact shape and any wrinkle of the 100kHz sawtooth was replicated at 0.001 Hz or less when the ground station OCXO was tuned to null prior to launch.
  • At lift-off, the sawtooth was DC then ramped up to 1Hz or more after lift-off representing Doppler Range (distance from the antenna) then became null at Apogee and reversed slope of the sawtooth coming down.
  • This was a low-cost tracking system to replace WWII trailer full of electronics and 2 techs used at the time in Churchill Research Range with NRC-funded Black Brandt and NASA-funded sounding rockets. It was used to locate rockets for possible pickup by chopper later after they travel over 500 miles up and then down. If there wasn't a parachute they would get buried in the ice ground like a dart. The OCXO was redesigned and tested to be stable << 1e-10 with 50g acceleration and 100g shock and 15g vibration.
\$\endgroup\$
  • \$\begingroup\$ Version 2 had two ground station mixers with the RF signals from 2 separated quad Helix steerable antenna to get azimuth phase angle along with range using nanosecond phase angle of the 100kHz pilot tone sub-carrier which had very low phase noise. The VHF Tx carrier was not stable enough to use for this with AT type Xtal. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Sep 5 '18 at 4:24
  • 1
    \$\begingroup\$ Fascinating anecdote Tony, now I have to go back through all of your answers looking for gems like this. I had been avoiding a 2-amp filter, but the -170dB looks worthwhile. And thanks for pointing out this TI filter designer, much better than the Analog.com filter design wizard. \$\endgroup\$ – Mike Sep 6 '18 at 1:18
  • \$\begingroup\$ Well, I've got my nights planned out for a while! FYI, link should be electronics.stackexchange.com/search?q=user%3A17574+Anecdotal \$\endgroup\$ – Mike Sep 6 '18 at 2:52
  • \$\begingroup\$ @TonyEErocketscientist Shouldn't it be -120dB for a 3rd order filter? I thought it was -40dB/decade/pole for a total of -120dB at 100kHz? I actually only get -110dB in FilterPro. \$\endgroup\$ – Mike Sep 8 '18 at 3:01

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.