Pull Down resistor

[Nikos A.]

Hi everyone,

I design a circuit that will support 8 analog sensors and the MCU will read each analog value through a multiplexer. After debugging the circuit I realized that the empty header sockets (no sensors connected) are actually floating inputs... I am thinking of placing a pull-down resistor on each of the multiplexer inputs (15 in total) to avoid this situation. Is there a better solution instead of placing fifteen pull-down resistors? This is the initial circuit.

[David Hess]

That depends on the application and on how the multiplexer is constructed but in general, unused inputs should be tied to some potential within the supply range.  Otherwise current injected into the substrate through the unused input can cause error in other channels.  If impedances and accuracy requirements are low, then this may not be an issue.

Multiplexers built on dielectrically isolated processes may not suffer from this.

[Nikos A.]

That depends on the application and on how the multiplexer is constructed but in general, unused inputs should be tied to some potential within the supply range. 

Thank you for your answer. I will tie the pins to the ground but I was wondering if there is any better solution besides connecting a single pull-down resistor to each of the MUX io pins.

[GeorgeOfTheJungle]

The signals TEMP and HUM are digital or analog?

[Wimberleytech]

Why not use a single pull down resistor on the downstream side of the mux?

[Nikos A.]

Thank you for your answers guys!!

The signals TEMP and HUM are digital or analog?

Are analog signals

Why not use a single pull down resistor on the downstream side of the mux?

You mean a pull-down resistor on SIG pin? Yes, I suppose that this could be an ideal solution for me. Since I control pins C0-C15 with a 4-bit sequence (S0-S3), by pulling down SIG pin is the same as pulling down C0-C15 individually isn't it?

[David Hess]

Why not use a single pull down resistor on the downstream side of the mux?

The problem is with floating inputs on the upstream side of the multiplexer.  Depending on the multiplexer design, it may tolerate inputs floating past its supplies or it may not leading to problems like leakage between channels or improper operation.

[GeorgeOfTheJungle]

Thank you for your answers guys!!

The signals TEMP and HUM are digital or analog?

Are analog signals

Watch out then, because the resistance of those analog switches isn't a constant, IIRC it's an ƒ(V) (or an ƒ(I)?). There's a chapter in TAoE that explains that very well (again, IIRC).

Why not use a single pull down resistor on the downstream side of the mux?

You mean a pull-down resistor on SIG pin? Yes, I suppose that this could be an ideal solution for me. Since I control pins C0-C15 with a 4-bit sequence (S0-S3), by pulling down SIG pin is the same as pulling down C0-C15 individually isn't it?

Yes! You can do with only one pulldown.

[Wimberleytech]

Why not use a single pull down resistor on the downstream side of the mux?

The problem is with floating inputs on the upstream side of the multiplexer.  Depending on the multiplexer design, it may tolerate inputs floating past its supplies or it may not leading to problems like leakage between channels or improper operation.

I looked at the datasheet before answering.  The mux consists of a conventional CMOS transfer gate.  While no details are given about input protection the datasheet specifies HBM/MM/CDM performance which would indicate serious diode junctions of both flavors on all inputs.  Thus I suspect it unlikely that a truly floating node would bump up against a supply and cause crosstalk via power rails.  Crosstalk voltage levels between the rail potentials is whatever it will be regardless of whether the signal is floating or forced.

[Wimberleytech]

You mean a pull-down resistor on SIG pin? Yes, I suppose that this could be an ideal solution for me. Since I control pins C0-C15 with a 4-bit sequence (S0-S3), by pulling down SIG pin is the same as pulling down C0-C15 individually isn't it?

Yes!

[David Hess]

I looked at the datasheet before answering.  The mux consists of a conventional CMOS transfer gate.  While no details are given about input protection the datasheet specifies HBM/MM/CDM performance which would indicate serious diode junctions of both flavors on all inputs.  Thus I suspect it unlikely that a truly floating node would bump up against a supply and cause crosstalk via power rails.  Crosstalk voltage levels between the rail potentials is whatever it will be regardless of whether the signal is floating or forced.

It is not crosstalk via the power pins.

The problem comes about with junction isolated IC processes where the substrate is connected to the negative supply.  Negative going signals are clamped by the diode to the negative supply but still inject carriers into the substrate.  That is how junction isolated CMOS processes can die from SCR latchup even though they have input protection diodes.

It is more of a problem with multiplexers because the substrate current can leak into other channels.  For this reason, some multiplexers are deliberately built on dielectric isolated processes making them immune to this problem.  JFET multiplexers may be immune also although I am not sure.  Another way I have occasionally seen used with very old junction isolated multiplexers is to bring the substrate connection out separately and require it to be tied to a more negative voltage.

Is it a real problem?  At high resolutions it sure is.  For 3.5 digits, leakage is unlikely to be an issue.

[Wimberleytech]

The problem comes about with junction isolated IC processes where the substrate is connected to the negative supply.  Negative going signals are clamped by the diode to the negative supply but still inject carriers into the substrate.  That is how junction isolated CMOS processes can die from SCR latchup even though they have input protection diodes.

A floating node (a high-impedance to any source) will not initiate latchup in a properly designed integrated circuit (junction-isolated or otherwise).

[David Hess]

A floating node (a high-impedance to any source) will not initiate latchup in a properly designed integrated circuit (junction-isolated or otherwise).

That is right but an analog multiplexer is sensitive to other failure modes from substrate currents like non-destructive leakage between channels which a digital IC would ignore.

[Wimberleytech]

A floating node (a high-impedance to any source) will not initiate latchup in a properly designed integrated circuit (junction-isolated or otherwise).

That is right but an analog multiplexer is sensitive to other failure modes from substrate currents like non-destructive leakage between channels which a digital IC would ignore.

For high-res analog circuits, I would never leave a floating node.  In general, I do not like any unknown node potential, but in this application it is just not an issue.  I spent my life designing precision analog/mixed-signal ICs--in that world you certainly worry about everything.