Common mode choke for Instrumentation amplifier input

[ricko_uk]

Hi all,
with reference to the attached schematic with the common mode choke at the instrumentation amplifier input.... The instrumentation amplifier input impedance is extremely high (so the currents are extremely low) and the voltages at the CM choke inputs are also extremely low making the currents even more microscopic (pico amps or even femto amps??). So does the CM choke work at all? Would it not need some "decent" currents to create a magnetic field so the noise cancel itself?

Thank you

[unitedatoms]

Impedance is frequency dependent. There is input capacitance in pF range. So the current is not femtoamperes in AC, but rather nanoamperes or more (million fold for MHz interference).

[ricko_uk]

Thank you Unitedatoms
would it still work at removing noise between 50KHz and 200KHz?

How do I choose it correctly? What parameters do I need to check in the datasheet to choose the most appropriate CM choke?

Thank you again

[T3sl4co1l]

Right, to make a filter, some impedance is required.  A filter for infinite impedance is meaningless.

Tim

[unitedatoms]

Perhaps the chokes have some kind of specs for common mode loss, blocking. I saw example of choke in HP LCR meters, where coaxes from BNC ports go through large rings, two or three turns. I'd look at there part numbers and make conclusions.

The correctness of choice is possibly debatable. May be the most common chokes are most correct. The most common ones are this lossy ferrite rings for power supplies and VGA cables, can be found as pair of semi rings with plastic clips to mount on cable.

That kind of chokes is designed for DC DC converters frequencies in high kHz range.

Edit: @Teslacoil. Well, I can not follow that logic that countermeasures is meaningless. For receiver with infinite differential and common mode impedances and non zero sensitivity to common mode signal, the required filter simply has to have common mode impedance greater than infinity. The practicality of such filter is debatable.

[T3sl4co1l]

Perhaps the chokes have some kind of specs for common mode loss, blocking. I saw example of choke in HP LCR meters, where coaxes from BNC ports go through large rings, two or three turns. I'd look at there part numbers and make conclusions.

The correctness of choice is possibly debatable. May be the most common chokes are most correct. The most common ones are this lossy ferrite rings for power supplies and VGA cables, can be found as pair of semi rings with plastic clips to mount on cable.

Right, those have a substantial impedance, namely the ground-ground impedance.  So the choke is quite effective -- the ground impedance might be a few ohms while the ferrite is a hundred, or thousands for multi-turn beads or CMCs.

Edit: @Teslacoil. Well, I can not follow that logic that countermeasures is meaningless. For receiver with infinite differential and common mode impedances and non zero sensitivity to common mode signal, the required filter simply has to have common mode impedance greater than infinity. The practicality of such filter is debatable.

Into an instrumentation amplifier as shown, there is no load impedance -- well, a few pF from the amp pins and stray wiring, that's it.

The CMC introduces a series impedance, perhaps some kohms.  Which does absolutely nothing when the load impedance is say Mohms.

The only solution is to give in and permit some parallel loading, or to use an instrumentation amp with common mode range and bandwidth adequate to cover the expected interference -- which may be 10s of V in an industrial setting.  (Bandwidth is probably a good idea, so as to avoid phase mismatch near cutoff, and RF rectification.  MOS/FET types recommended.)

Tim

[EEEnthusiast]

In order to improve the EMIRR, most instrumentation amplifiers recommend to add a few 100pF of cap right at the input pins to ground. With the caps added at the input terminals, the common mode choke becomes even more effective in reducing the common mode signal. Without those, the effectiveness of the choke is questionable as it depends on the current to create a magnetic field. With an open circuits (opamp input) in loop, the currents would be minimal to have any impact.
Please refer pg 18 of the below datasheet for some more info.
https://www.analog.com/media/en/technical-documentation/data-sheets/AD8429.pdf

[David Hess]

In order to improve the EMIRR, most instrumentation amplifiers recommend to add a few 100pF of cap right at the input pins to ground.

Just be careful about mismatch in the common mode capacitors converting common mode signals into differential mode signals.  If this is a problem, then a larger third differential capacitor can be added to swamp the mismatch.  Or sometimes in old circuits, you will find that the common mode capacitors are trimmable.

Do *not* control instrumentation amplifier bandwidth with input filtering; instead filter the output.  AC coupling the input is also usually a mistake so instead apply the integrated output signal to the reference input.

[EEEnthusiast]

Reducing the input bandwidth helps improve the DC offset drift. RF energy could be rectified by the non linearity of the OPAMPs and end up as DC offset at the output of the opamps. This can never be filtered at the outputs. In such cases, limiting the input bandwidth is really needed.

[David Hess]

EEEnthusiast, controlling the instrumentation amplifier bandwidth and EMI filtering are two different things.  Controlling bandwidth at the inputs introduces more problems than it solves because differences in the input filters will convert common mode interference into a differential signal.

[T3sl4co1l]

But, if CM-D conversion occurs within the amp's well-behaved range (say, below GBW), and you filter that down later to the desired signal BW (assuming it's lower), you cut out the CMRR and avoid spooky effects (rectification, etc.).

If you need high BW and can't afford excess GBW, you have more limited options.

Worth noting most precision amps claim EMI immunity, though they never go into detail what that constitutes, or how effective it is.  As far as I know, it's usually an RC filter at the input pin, so there is some impact on e_n and Cin, and some opportunity for imbalance anyway; but again, who knows.

Tim