Two Pole, Low Pass Active Filter with Differential Inputs

[gnuarm]

I found this circuit in a four year old thread here.  It is a Multiple FeedBack (MFB) topology if I'm not mistaken.  It looks like exactly what I want.  I've never seen this filter in a differential input arrangement.  I will need to dig a bit to find info on how to set all the values.

The bit that concerns me, is the fact that in my design, I need to use it both as a differential input circuit, and as a single ended input, in different operating modes.  When a single ended input is applied, I don't have a way to ground the unused input.  I expect the signal will need to be applied to the inverting input, but that still leaves the non-inverting input flopping around.  Maybe it's better to say the non-inverting input would be dragged around by C9, the 470 pF cap. 

This design is very tight for space, very tight, so no room for extra logic or amps!  Anyone know if this circuit will still function effectively if C9 is replaced by a pair of 1 uF caps to ground?  Or does that impact the differential aspect of the filter? 

The input operates in several modes, including 50 ohm terminated, single ended; 600 ohm, differential; and "high" impedance, differential.  "High" impedance doesn't currently have a spec number on it.  I expect 5k would be adequate, preferably 10k.  If each input leg had a 4.7k resistor to ground, that should serve as a ground for the non-inverting input when in single ended mode, in place of splitting the C9 cap, no?

(ignore the op amp part number, I'll be using something else)
[attachimg=1 width=1200]

[srb1954]

I found this circuit in a four year old thread here.  It is a Multiple FeedBack (MFB) topology if I'm not mistaken.  It looks like exactly what I want.  I've never seen this filter in a differential input arrangement.  I will need to dig a bit to find info on how to set all the values.

The bit that concerns me, is the fact that in my design, I need to use it both as a differential input circuit, and as a single ended input, in different operating modes.  When a single ended input is applied, I don't have a way to ground the unused input.  I expect the signal will need to be applied to the inverting input, but that still leaves the non-inverting input flopping around.  Maybe it's better to say the non-inverting input would be dragged around by C9, the 470 pF cap. 

This design is very tight for space, very tight, so no room for extra logic or amps!  Anyone know if this circuit will still function effectively if C9 is replaced by a pair of 1 uF caps to ground?  Or does that impact the differential aspect of the filter? 

How does substituting two 1uF caps for one 470nF cap help save space?

If you use two caps the circuit will still work, but not as well. Any mis-match in the values of the two caps will degrade the circuit's common-mode rejection ratio and might allow higher frequency common-mode signals to pass through to the output.

[Terry Bites]

Yes, grounding the inverting or non-inverting input will be fine. It won't have any effect on roll off or passband gain.
I would be preferable to ground the non-inverting input.

[gnuarm]

I found this circuit in a four year old thread here.  It is a Multiple FeedBack (MFB) topology if I'm not mistaken.  It looks like exactly what I want.  I've never seen this filter in a differential input arrangement.  I will need to dig a bit to find info on how to set all the values.

The bit that concerns me, is the fact that in my design, I need to use it both as a differential input circuit, and as a single ended input, in different operating modes.  When a single ended input is applied, I don't have a way to ground the unused input.  I expect the signal will need to be applied to the inverting input, but that still leaves the non-inverting input flopping around.  Maybe it's better to say the non-inverting input would be dragged around by C9, the 470 pF cap. 

This design is very tight for space, very tight, so no room for extra logic or amps!  Anyone know if this circuit will still function effectively if C9 is replaced by a pair of 1 uF caps to ground?  Or does that impact the differential aspect of the filter? 

How does substituting two 1uF caps for one 470nF cap help save space?

If you use two caps the circuit will still work, but not as well. Any mis-match in the values of the two caps will degrade the circuit's common-mode rejection ratio and might allow higher frequency common-mode signals to pass through to the output.

Every part in the circuit has an equivalent part in the other half.  If none of them are perfectly matched, why would these two need to be matched?

They are two, in place of one, so that each is grounded, rather than being tied across the input.  Tying across the input means a single cap only works differentially.  Grounding the two caps means they will filter both differentially and in common mode.  I've tried this and it works better than the single cap.  With the single cap, the stop band reaches a minimum and turns back up producing significant ripple in the stop band.  With the two caps to ground, the low pass response continues to very, very low numbers. 

The circuit does not achieve filtering through matching of the two signal paths.  The filtering is an inherent part of the operation of the RC circuits.

[gnuarm]

Yes, grounding the inverting or non-inverting input will be fine. It won't have any effect on roll off or passband gain.
I would be preferable to ground the non-inverting input.

The non-inverting input does not need grounding.  It has a DC set point from the circuit itself.  If the inverting input is left floating, it disrupts the operation of the inverting feedback, and the circuit becomes a unity gain amplifier within the pass band.

[srb1954]

How does substituting two 1uF caps for one 470nF cap help save space?

If you use two caps the circuit will still work, but not as well. Any mis-match in the values of the two caps will degrade the circuit's common-mode rejection ratio and might allow higher frequency common-mode signals to pass through to the output.

Every part in the circuit has an equivalent part in the other half.  If none of them are perfectly matched, why would these two need to be matched?

They are two, in place of one, so that each is grounded, rather than being tied across the input.  Tying across the input means a single cap only works differentially.  Grounding the two caps means they will filter both differentially and in common mode.  I've tried this and it works better than the single cap.  With the single cap, the stop band reaches a minimum and turns back up producing significant ripple in the stop band.  With the two caps to ground, the low pass response continues to very, very low numbers. 

The circuit does not achieve filtering through matching of the two signal paths.  The filtering is an inherent part of the operation of the RC circuits.

I didn't mention anything about filtering. I was commenting on the ability of the circuit to reject common-mode signals, which is the main reason for using a differential input circuit such as the one you posted.

Common-mode rejection is quite independent of the filtering function of the circuit and applies to signals from DC upwards, not just for frequencies above the stop-band of the filter. To achieve good common-mode rejection it is critical to have all the corresponding components in the inverting and non-inverting paths closely matched.

[gnuarm]

How does substituting two 1uF caps for one 470nF cap help save space?

If you use two caps the circuit will still work, but not as well. Any mis-match in the values of the two caps will degrade the circuit's common-mode rejection ratio and might allow higher frequency common-mode signals to pass through to the output.

Every part in the circuit has an equivalent part in the other half.  If none of them are perfectly matched, why would these two need to be matched?

They are two, in place of one, so that each is grounded, rather than being tied across the input.  Tying across the input means a single cap only works differentially.  Grounding the two caps means they will filter both differentially and in common mode.  I've tried this and it works better than the single cap.  With the single cap, the stop band reaches a minimum and turns back up producing significant ripple in the stop band.  With the two caps to ground, the low pass response continues to very, very low numbers. 

The circuit does not achieve filtering through matching of the two signal paths.  The filtering is an inherent part of the operation of the RC circuits.

I didn't mention anything about filtering. I was commenting on the ability of the circuit to reject common-mode signals, which is the main reason for using a differential input circuit such as the one you posted.

Common-mode rejection is quite independent of the filtering function of the circuit and applies to signals from DC upwards, not just for frequencies above the stop-band of the filter. To achieve good common-mode rejection it is critical to have all the corresponding components in the inverting and non-inverting paths closely matched.

I will try simulating this.  I don't have requirements for common mode rejection.  I do have  a requirement for single ended operation which the single cap version does poorly. 

[ejeffrey]

Yes dividing that cap and grounding it will help with the response when input is unused at the expense of probably worse common mode rejection if they are unmatched.  It's not really ideal, you still have a high impedance floating node that can pick up stray signals at low frequency but if your really don't have any way to ground the unused input it might work.

If you truly don't have any requirement on common mode rejection  can you just use a single ended filter and ignore the anti phase signal even when it's present?  That would save components and avoid unwanted signals coupling in on the inverting input.

[gnuarm]

Yes dividing that cap and grounding it will help with the response when input is unused at the expense of probably worse common mode rejection if they are unmatched.  It's not really ideal, you still have a high impedance floating node that can pick up stray signals at low frequency but if your really don't have any way to ground the unused input it might work.

If you truly don't have any requirement on common mode rejection  can you just use a single ended filter and ignore the anti phase signal even when it's present?  That would save components and avoid unwanted signals coupling in on the inverting input.

I suppose that could work.  But it would provide half the signal level.  There would also be a problem of reference voltage.  I don't know that the inputs are ground referenced.  Actually, I have little info on the equipment this will be connected to.  it all sucks in many ways, but this is not a new design.  It's an update to an old design that seems to work well enough.  So I'm not going to make such major changes as only using one input of a "balanced pair".