Two Pole, Low Pass Active Filter with Differential Inputs

Sorry about that.

You don't seem to understand the question.

1) I can't ground an input, as there is no circuitry to do that. There is no room on the board to add a switch. One input will have to float when used single ended. 2) This will run on a single power supply voltage, but you can consider the terminology of "ground" to refer to a reference voltage that is half the power supply. 3) The inputs will be coupled to the signal source through capacitors, so no DC path. 4) The concern is that when using the circuit with a single ended input, and one input floating, that C9 will cause signal to drive the floating input, impacting both the gain, and the filter response. Actually, the filter response might not be so important. The single ended mode does not use the full 20Hz to 20kHz bandwidth. It's a 1 kHz sine wave with 100 Hz amplitude modulation.

Does this make more sense?

It's too bizarre to think about.

Hmmm... I'm not sure you understand. If the non-inverting input is grounded, what you say will be useful. The non-inverting input will be held at an AC constant value and the circuit will work as intended for the single ended input. But, leave the non-inverting input open, it will be dragged about by C9, the input will not be an AC constant value, and the resulting output will not be the same as if the non-inverting input were AC grounded.

When running with a single ended input, there is no DC common mode rejection, because it's single ended and the concept of common mode does not apply. DC is not an issue anyway, as the source is coupled to the circuit input through capacitors.

I'm more interested in finding a similar circuit that will work effectively with differential inputs, as well as with a single ended input. That's why I asked about replacing C9, with a pair of capacitors, each to AC ground Then there's no direct path through the capacitors to the other half of the circuit. The amplifier + input will be connected to the DC reference through the network and only the inverting half of the circuit will operate, as in this drawing.

I'm not sure connecting two capacitors in place of one, will not degrade any performance in the two use cases. I am currently drawing up the circuit to simulate. I'm pretty convinced this will work fine, I just thought it might be interesting to discuss. I had to do a lot of searching to find any low pass, active filter, that was designed for a differential input and single ended output.

It's funny that Larkin says the circuit is "bizarre". But then he's never claimed to be well versed in filter design, has he?

I'd donate a nice circuit if he wasn't such a consistent jerk.

So you do know filters? What is so "bizarre" about a MFB filter designed for differential inputs?

I've already simulated this and the performance seems fine. What's the squawking about?

But thanks for your offer.

Wouldn't it be easier to connect unused input R16 to the -V rail for single ended operation making a potential divider against R13 that has the effect of setting the DC bias of the + input to be almost zero.

That doesn't disturb any of the other filter properties apart from common mode rejection (which is obviously inferior when unbalanced).

Not sure what you are getting at about bias. In this schematic, ground is the appropriate bias point, center of the two power rails. Why would you want to tie R16 to V-??? Doesn't matter. This design is severely space constrained, so adding something to tie R16 to anything is space prohibitive.

As it turns out, the appropriate circuit has C9 split into two Cs to ground. I'm not sure there is any advantage to using a single capacitor, other than using a single capacitor. When used in single ended mode with two capacitors to ground, leaving R16 floating, the circuit works perfectly. When the single capacitor is used, with R16 floating, the stop band maxes out (or minimizes out?) around -80 dB. That's no significant issue, but why use an inferior solution?

This may be a moot point. The input section of this design has selectable termination along with attenuation for different signal strengths. That circuitry is hard to combine with this circuit, unless the input impedances are rather large. That makes the capacitors small enough that circuit parasitic elements can impact the filter details.

Still, it's an interesting filter design.

I finally found a reference of the differential input version of the MFB filter, in an old Cirrus Logic app note, appropriately titled, "DESIGN NOTES FOR A 2-POLE FILTER WITH DIFFERENTIAL INPUT".

The indicated design process is to design the filter you require, using the single ended version of the topology. Then add the same components to the non-inverting portion of the circuit. Simple, and effective.