I am having a hard time figuring out what is going on with the "anti-aliasing filter" for this FDA (pictured below). I have attached my LTSpice drawing and the model used for the specific FDA.


For reference:
Pin 1 - Non-Inverting IN
Pin 2 - Inverting IN
Pin 3 - V+
Pin 4 - V-
Pin 5 - Out+
Pin 5b - Out-
Pin 17 - Vocm


From working with the actual circuit, it works. The ADC on the output seems to have no issue with the differential signals coming out.
From playing with the simulation I can tell that the circuit doesn't seem to work without the zener diodes AND 2.5V bias on the input signal.

I have looked at the FDA document by TI that discusses anti-aliasing using their FDAs but this does not look anything like a 2nd or 3rd order Filter. It looks like a failed mash-up of different order filter designs. I have even used the 2nd and 3rd order filters shown in the document with my own values and they appear to perform similarly.

I would appreciate any insight into the filtering on the output of this FDA.

 

Thre is a low pass filter in the feedback path of each differential channel with C1, R7 and C2, R10. There is also a low pass filter on the output, R9, R10, C3. What don't you understand?

 

I am having a hard time figuring out what is going on with the "anti-aliasing filter" for this FDA (pictured below). I have attached my LTSpice drawing and the model used for the specific FDA.

A balanced FDA is essentially an op amp with two feedback paths. Each half of the FDA has the standard negative feedback op amp topology: an output of one polarity feeds back to an input of the opposite polarity. The FDA's differential gain is set by the ratio of the feedback resistor to the input resistor. And -- just like a regular op amp -- a capacitor in parallel with the feedback resistor forms a LPF, as the capacitor gives high frequency content a low-impedance path through the feedback loop, allowing more of it to be "cancelled out" by the negative feedback.

So, as KeithWalker said, you have an active first-order LPF formed by R7and C1 (and on the other side R8/C2), with a cutoff frequency of about 159 Hz. Technically, you also have a passive first-order LPF formed by R9/R10 and C3, but its really functioning as a "charge bucket" to help quickly settle the ADC's sample and hold capacitors. For that reason, I'd say that you have a first-order anti-alias filter.

As for the circuit itself, it's kind of weird and doesn't seem optimal. The dc operating point of the FDA is 2.5 V, and so the currents through R3 and R4 are wasting power. Typically you'd also want to ac couple the input, which likely has its own dc operating point. The THS4131 is a pretty terrible FDA, with dreadful performance near the rails. A symptom of this is your having to use +/- 7V rails for a 5V peak-to-peak signal. A good replacement would be the LTC6362.

 

This might be of interest,
https://www.analog.com/en/technical...tial-filters-for-communications-systems.html#

Towards the bottom, they talk about how to design a differential filter.
Basically you design a standard single ended filter, then reflect it,

 

And -- just like a regular op amp -- a capacitor in parallel with the feedback resistor forms a LPF, as the capacitor gives high frequency content a low-impedance path through the feedback loop, allowing more of it to be "cancelled out" by the negative feedback.

Which resistor/cap combo are you suggesting are connected in parallel? Correct me if I am wrong, but R7/8 and C1/2 are not in parallel to my understanding -- Part of what is tripping me up.

So, as KeithWalker said, you have an active first-order LPF formed by R7and C1 (and on the other side R8/C2), with a cutoff frequency of about 159 Hz. Technically, you also have a passive first-order LPF formed by R9/R10 and C3, but its really functioning as a "charge bucket" to help quickly settle the ADC's sample and hold capacitors. For that reason, I'd say that you have a first-order anti-alias filter.

Thank you. My main issue was that I had not seen this configuration for an anti-aliasing filter (granted I haven't been in the industry long). It doesn't look 'intuitive' to me as any of the FDA filter configurations in the .pdf document attached to my post. But this was a very helpful response.

As for the circuit itself, it's kind of weird and doesn't seem optimal. The dc operating point of the FDA is 2.5 V, and so the currents through R3 and R4 are wasting power. Typically you'd also want to ac couple the input, which likely has its own dc operating point. The THS4131 is a pretty terrible FDA, with dreadful performance near the rails. A symptom of this is your having to use +/- 7V rails for a 5V peak-to-peak signal. A good replacement would be the LTC6362.

The input signal pictured is more just for demonstration purposes, but the 2.5V bias is true to form. I am not the original designer but it is my responsibility now so I appreciate the insight on the component!

 

Which resistor/cap combo are you suggesting are connected in parallel? Correct me if I am wrong, but R7/8 and C1/2 are not in parallel to my understanding -- Part of what is tripping me up.

You are correct, my mistake. I saw what I'm used to seeing and not what's actually drawn. R7 and C1 are in series in the feedback path, which makes the amp act as an integrator. Any dc at the input will saturate the output. Are you sure this is the correct topology?

If it is, I would re-draw the schematic to make it plainly clear what's going on, i.e., use the standard triangle symbol for the amp and put R7 and C1 (R8/C2) on the same horizontal line to emphasize that they are in series. While an integrator has low-pass behavior, it is definitely not a standard anti-alias filter.