1
\$\begingroup\$

Here's my requirement:

I need an analogue switch that can do SPST/DPST/DPDT with very low On-Resistance (<10mOhms). There would be current flowing through this switch, albeit in the range of a <10mA. This current DOES need to flow across, and hence my understanding is that I can't use any buffered(?) switches. I want the voltage drop across the switch to be absolutely minimal - very critical. Effectively I want the same voltage seen on the input to be present in the output, while allowing a small current to flow through.

The requirement for the very low On-resistance is an effort to keep the voltage errors below 0.1 ppm seen at the output. This is for a precision instrumentation circuit where the switches select various range resistor networks. Currently, these switches are mechanical and the contacts are gold plated. But I wanted to see if I can use solid-state switches in an effort to allow for the instrument to be more automated.

I understand this seems a difficult thing and it might be near impossible, but I wanted to see if anyone had any thoughts.

Here's my question/problem:

A MOSFET Mux seems that ideal candidate and is - except the lowest R-ON I could find was around 250 mOhms. This introduces errors in the range of 2-3 ppm.

And hence the first thing I was looking at was making my own MOSFET Mux using Discrete MOSFETs. The issue I'm having is that I can't seem to find a MOSFET that is 4 pin with a dedicated body-pin. My understanding is that to construct a MOSFET Switch/Mux, you need the body pin of the N-MOS to be connected to VSS and of the P-MOS to be connected to VDD.

I also did have a look at using series JFETs as switches, but they exhibited significant voltage drop across the FET.

  1. Is there a way to construct a MOSFET Mux/Switch with conventional MOSFETs that have their body connected to the source?
  2. Is anyone aware of 4-pin MOSFETs (with source and body disconnected) and if there was any specific terminology I might use to find them?
  3. Is there any other type of FET/BJT that can be used to accomplish what I want.
  4. Is anyone aware of low R-On Mechanical Relays?

Thank you for all your help. Let me know if this is a hopeless endeavor :)

The circuit for a MOSFET Mux/Switch with separate body pin The circuit for a MOSFET Mux/Switch with separate body pin.

Series JFET switch Series JFET switch

\$\endgroup\$
  • \$\begingroup\$ Lets assume you want 0.1ppm out of 1 volt. That is 0.1uV or 100 nanovolt. Yet temperature gradients across the metal-metal connection of two copper wires will be 5uV/degree Centigrade. From what I've read about Seebeck Effects. en.wikipedia.org/wiki/Seebeck_coefficient#/media/… \$\endgroup\$ – analogsystemsrf Aug 7 '19 at 5:21
  • \$\begingroup\$ "This current DOES need to flow across", why? - It sounds like you have a compensation problem. \$\endgroup\$ – Harry Svensson Aug 7 '19 at 5:48
  • \$\begingroup\$ Celcius or Kelvin. There is no such thing as centigrade. \$\endgroup\$ – winny Aug 7 '19 at 6:59
  • \$\begingroup\$ Considering that 2 ppm/deg C resistors cost several dollars each, I'd be interested in why you think an absolute error of 0.1 ppm is at all relevant to your design. \$\endgroup\$ – Andy aka Aug 7 '19 at 7:05
  • \$\begingroup\$ It is part of a resistor network - acting as a voltage divider. And hence for the effective resistance of the network to be adjusted, the current has to flow between the two parts. \$\endgroup\$ – leviathan Aug 7 '19 at 8:57
0
\$\begingroup\$

Here are two NMOS based analog muxes. I don't know how your resistor network looks like schematic wise, but I suppose you know how to glue it all together and make it work. So I'll leave that problem to you.


enter image description here
Link to interactive simulation

You can make a transmission gate (TG) by tying 2x NMOS's source and gate together.

If the input lives between -5 V and 5 V then you need to pull the gate above 5 V + the gate threshold voltage which in this case is 1.5 V. But the higher the gate is above the gate threshold voltage, the less Rd(on) you'll have. I chose 12 V because it will turn on the TG hard.

To turn it off you simply pull the gate around the floor of the input voltage, -3.5 V will make the TG to start turning off. So I chose -5 V.

Then there's some other NMOS around to act as boolean algebra + logic level converters to make it into a functional analog mux.

Concerns with this circuit that you should know about:

  • The gate to source voltage will be flying all over the place, all from 17 (12+5) V to -10 (-5-5) V. It might be hard to find a NMOS with those voltage ratings. I haven't checked, but thought I'd mention it.
  • I have absolutely no idea if it will introduce less than 1 ppm errors, I like designing circuits, so I did it because it was fun. If I'd want to be professional about it I'd demand your resistor network schematic. Maybe there's an entirely other way of solving your problem that is much more realistic. Who knows.
  • NMOS and all other MOSFETs capacitances, so you are introducing quite a lot of capacitance to your signal, maybe your signal is quite DC, maybe it's AC, who knows. Maybe you can find a NMOS with low capacitances.

enter image description here
Link to interactive simulation

This version uses two op-amps to set the gate to source voltage. This way you won't see 17 V and -10 V between the gate and the source. It's easier to find these kinds of NMOS. Now you can buy logic level NMOS. The problem with logic level NMOS are that they usually have very low maximum gate to source voltage and drain to source voltage.

The 100 nF is only there for simulation, in the real world you'll maybe have around 500 pF which is a part of the NMOS which you can't remove.

The resistor going from the op-amp to the gates should be there so you turn it into a low-pass filter. You could add the 100 nF in, in case the op-amps are too slow, or is it too fast? I think there might be oscillations if there's no low-pass filter. Not 100% certain.

Concerns with this circuit that you should know about:

  • If your analog signal has high frequencies then the op-amps might have problems to follow it and will mess up the voltage at the gate. So make sure that your op-amps bandwidth is above whatever frequencies you are working with.
  • When you set the input to 0 V, meaning that you want 0 V across a gate to source, then it won't get to 0 V because the TG deactivates itself when the gate to source is below the threshold voltage. You can see this in the image above where I set it to 0.5 V and get 1.384 V instead. It turned off itself at 1.5 V and then floated randomly downwards.

If anyone else sees any obvious mistakes or problems in my circuits, tell me about it. I'd love to hear them.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.