Is it good to equalize input impedances in cascade active opamp filters?

HI Electronx,

In general, when selecting the filter passive components, the designer has to scale the passive components considering the circuit frequency response, while also accounting for the intrinsic noise performance of the circuit, as well as the circuit stability. 

In general, I try to reduce the resistance values while ensuring the filter remains stable.  Another factor to consider, the input filter impedance becomes the load to the prior stage or sensor. Therefore, making the impedances low reduces the resistor noise contribution, but reducing the resistor exceedingly could change the filter response shape as the prior stage or sensor could have trouble driving low impedances.  In addition, relatively low bandwidth / low power, precision amplifiers such as the OPAx189 can drive loads at the ~1kΩ - ~2kΩ range without issues, so I typically scale the resistors around this range. Choosing to scale the resistances high, which will scale the capacitors values down, helps in some cases with the circuit stability, but also comes with a trade off, increasing the resistor broadband noise (resistor Johnson noise).

Regarding the OPA189, this is a zero-drift or chopper amplifier, offer outstanding DC performance with ultra-low offset and low drift.  Zero-drift amplifiers such as the OPAx189 use switching on the inputs to correct for the input offset and drift of the op-amp. The chopping frequency of the device occurs around ~200-kHz which is close to your 100-kHz frequency bandwidth requirement.  This type of amplifier is typically used on applications that require a very high level of DC accuracy. On chopper amplifiers, the charge injection from the integrated switches on the inputs can introduce extremely short current transients in the input bias current of the amplifier. These input bias currents cause no issues when using relatively low source impedances at the inverting and non-inverting op-amp inputs.  However, chopper amplifiers can be sensitive to high source impedances, or relatively high impedance mismatch at the inverting and non-inverting terminals of the op-amp. In general, we recommend keeping the equivalent input impedances of the OPAx189 at the inverting and non-inverting inputs of the op-amp relatively low, around the 1kΩ for best performance.  

Since this circuit is required for an AC coupled application, bandpass filter application, from 0.2Hz to 100kHz, perhaps you do not require a chopper architecture amplifier and we could also look at other precision linear (non-chopper), low noise amplifiers, that still offer very good DC performance, low noise, and relatively low drift without sensitivities to the impedance mismatch.

Do you have a strict requirement for a 250kΩ input impedance? i.e. is the IEPE sensor preceding the Sallen-Key high, pass filter require a high input impedance? If you require the high input impedance while keeping noise low, we could also consider buffering the Sallen-Key filter stage.

Does the filter have a intrinsic noise performance requirement? What filter response characteristic is required, for example, Butterworth (maximally flat)?

Do you have a schematic for these filters?  We could help verify the circuit with TINA simulation. as well as make suggestions.

Thank you and Best Regards,

Luis  

Hey Luis thanks for the information ,this is really useful.

What I want is that when a 12 Volt offset 10V(sinus) peak to peak ((this offset is a typical feature of the iepe sensor because the signal and supply line are the same, the dc offset is deleted through the high pass filter capacitor at the input and after the dc offset is deleted, the signal oscillation is +-10 volts(sinus) )) input signal between 1-20khz is applied, I want the amplitude to reach the output without changing, with the gain being 1. For this, I chose the cut-off frequencies of 0.1 Hz for HPF and 110 khz for LPF. Normally, it would be enough to set the cut-off frequency at 24 khz, but in this case, the amplitude at 20 khz causes attenuation. However, I do not increase the filter order because the number of opamps will increase and the cost will increase. In this case, I increased the bandwidth of the filter's cut-off frequencies to 0.1 Hz and 110 khz and ensured that the gain was = 1 at 20 khz and 1 Hz.(So ı use butterworth  for no ripple  flat gain)

I use +-15 volt dual supply ,The goal  tolerance for me is to see +-1mV on my filter output.

I am writing these informations  because if there is any mistake in my information, please correct me.

When I apply a bode plot graphic with ac sweep, I get a result like this, which is good.

Bandpass filter filter ac sweep local simulation results is looking okay

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

HPF ac sweep local simulation results is looking okay

----------------------------------------------------------------------------------------------------------------------------------------------------

HPF  Transient analsys local simulation results is looking okay 20 khz applied to input

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------

LP filter ac sweep local simulation results is looking okay

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------

LPF  Transient analsys local simulation results is looking okay 20 khz applied to input connected differential probe to output

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

But when I connect the filters sequentially and apply a sine wave with 20khz 12volt offset and 10v peak to peak and connected differential probe to output

auto convergence is opened

I get such a bad result, the offset has increased in an interesting way. 

Where am I missing? Are the references correct?

ı also tried in Tina TI

Transient analysis is looking good

but when ı do a steady state analysis

I get an unwanted result with a shifted offset, similar to the one in pspice.

Hi Electronx,

Regarding the circuit, I am a little concerned about using the OPAx189 zero-drift, chopper amplifier.  One concern is that you could see some chopper noise:  The OPA189 chopping frequency is around ~200kHz but it is possible you may see a noise artifact at around ~100-kHz. The filter cutoff is just at 100kHz, and maybe the circuit above with the OPA189 could work but I believe there is no benefit on using a chopper amplifier in this application.  Please note, the chopper behavior/noise artifacts are not incorporated in the simplified/behavioral SPICE models.

Below is a very similar circuit with the OPAx192 (Linear amplifier, non-chopper), 10-MHz.  The circuit provides just about the same frequency response. I checked the OPAx192 and THP210 circuits for stability and phase-margin as well,  Both circuits are stable. The 20-ohm resistors inside the feedback may help with the loop-gain response/stability, but may not be absolutely required.  The total output noise is about 11.18uVRMS.  Let me know what load or circuit is driven by the THP210. The R-C-R filter at the output is optional, could be useful if you are driving an ADC.

Regarding the original OPA189-THP210 simulations, I attempted transient, stability and steady-state analysis using TINA-TI.  It worked well for me, and was not able to reproduce.  The issue may be related to convergence or an interaction between the two models.  If issues persist please share the simulation files or project file so we can review.

Thank you and Regards,

Luis

Hey Luis, really thanks for your reply. The two products  OPAx192 and  TLV197 look very similar. Could TLV97 also be an option? You wrote that OPA189 chopping frequency is around ~200kHz. Which parameter can I see this in the datasheet?

The chopper frequency is sometimes reported in the spec tables, often not. But, you can sometimes see where it is in the spot noise plot, this one does not spike as much as some do, but it is there,