0
\$\begingroup\$

In our project, we developed and tested a auto-zero chopping op amps coupling with Sigma-delta ADC for low-frequency voltage signal measurement. Below is the FFT spectrum from the 1-bit data stream of the ADC. Now, I am designing a decimation filter to convert 1-bit to high-precision data word. I noticed that there are several noise peaks around our chopping frequency. I wonder if I have to design a notch filter to take care of those peaks or just a regular sinc3 filter would suppress the noise and do the conversion efficiently.

Thanks in advance. enter image description here

\$\endgroup\$
4
  • \$\begingroup\$ What delta sigma converter and auto zero amp? need a part numbers. What is your signal bandwidth? \$\endgroup\$ – Voltage Spike Oct 16 '19 at 15:23
  • \$\begingroup\$ Does the chopper frequency change with temperature or from opamp to opamp? If it comes from an on-chip oscillator, it is likely to be quite variable, which is not compatible with a notch filter... \$\endgroup\$ – bobflux Oct 16 '19 at 15:57
  • \$\begingroup\$ Hi, this is a customize ASIC which is not available on the market and we don't have the data sheet for it yet. The chopping frequency is generated by external circuit using a phase-locked-loop \$\endgroup\$ – Dat Tran Oct 16 '19 at 17:46
  • \$\begingroup\$ worked with some people who developed AutoZero opamps. The onchip surge currents, of analog switches and switch drivers and FlipFlops and oscillator, transitioning in a few nanoseconds (even with reduced VDD) would make the GND and VDD rails ring, and that ringing comes thru the large output analog FETs to induce the ringing and spikes into the analog output. At about the 10 millivolt level. Thus you need to attenuate that 10 milliVolts down to << 1 microvolt? \$\endgroup\$ – analogsystemsrf Oct 17 '19 at 3:28
0
\$\begingroup\$

When you're designing a decimation filter, the purpose of the stopband is to reduce that part of the input signal spectrum below the level that would cause you problems if aliased down to baseband.

If your approach was only to design sufficient stopband to suppress the noise-shaping rise, then you would be short of attenuation for the noise peaks around your chopping frequency. Whether you go for additional attenuation by means of notch filters, or simply a more aggressive general stopband, is up to you. Note that if the chopping frequency were to change, perhaps due to substituting a different amplifier, then you would have to redesign the notch filter to cope.

My general strategy is to do limited sinc decimation to get the data rate down, followed by further polyphase FIR decimation to flatten the passband.

\$\endgroup\$
1
  • \$\begingroup\$ Thank you for your detailed answer. It helped me to confirm my thought. I was thinking the same too to flatten the passband and give more attenuation for the noise peaks around my chopping frequency. \$\endgroup\$ – Dat Tran Oct 16 '19 at 16:27
0
\$\begingroup\$

Standard practice with analog clocked filters chopper amps is to analog filter to take out the clock frequency (don't forget that Nyquist criteria apply to these filters and amps!)

Standard practice for decimation is to digital filter below the new Nyquist frequency.

\$\endgroup\$

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.