flexible connection matrix

[OM222O]

I'm designing a very high precision DMM (I know everyone's going to point out that I'm a fool for trying this, but I've made great progress so far and I'm pretty happy with the results) and I need a way to change connections between the modules. for example there is a wide dynamic range shunt (1 ohm to 100Kohm in 6 steps, each being an order of magnitude), the ADC, the probes, etc.

I do not wish to manually change the probe to different positions, or reconfigure stuff by hand (for example change different ranges) for ease of use.

The simple solution would be relays but they're mechanical parts and prone to much earlier breakdown compared to solid state solutions. For somethings like the ADC, analog muxes are an obvious choice, but for things like the probes and the shunt, high currents (a few amps) can be involved so using a mux is out of the question.

N channel fets would be the next best solution but it's not possible to use them everywhere due to high gate voltages required.

Please let me know if there's anything besides relays that can do this reliably.

[Zero999]

What do you mean by flexible connection matrix?

Are you talking about an anlalogue multiplexer? What voltage are the signals? Most analogue multiplexters are limited to a maximum voltage of +/15V, with +/-5 being more typical, in the case of the 74HC4051.
http://www.farnell.com/datasheets/2024221.pdf?_ga=2.34285609.966527979.1596465371-2127120132.1578504493&_gac=1.242411062.1596465371.EAIaIQobChMI5vKp8qD_6gIVF-vtCh0-TwHxEAAYAyAAEgKBy_D_BwE

[OM222O]

Yes, I would like to multiplex connections, for example, when going from voltage measurement to current measurement, the probe should obviously be switched to a different part of the circuit (voltage requires high impedance and current requires low impedance). The reason for the title was having used FPGAs where you can connect any pin to any point of the circuit you want (connection matrix) and I don't know anything similar.

I'm more limited by the current rather than voltage as this is a precision DMM for low power electronics (aka no mains voltages, 10s of amps, etc).
The chip you recommended for example, has a current rating of only 20mA. That's practically useless for a switching the probe, as it limits the current measurement to 20mA.

Edit: I've looked into how those analog switch ICs work and it seems like they use a P channel and N channel fet , but they require connections to the body of the fet for biasing which I don't think is possible with discrete fets. So far relays look to be the only real option.

[razvan784]

Nevermind advising you to not attempt designing your own meter; it's generally a very bad idea to allow software-controlled switching between voltage and current measurement. There are commercial meters that go to great lengths to alert you of a potential misconnection, e.g. probe detected in Amps jack while Volts mode selected. Also, commercial meters have complex, multi-layered protection circuitry built in, that comes with several compromises that you should be aware of. But maybe your application needs none of that - in that case I suggest researching PhotoMOS solid-state relays.

[OM222O]

That is exactly why I want software control. when you switch modes, it automatically changes the input, so you can't mess up the jack connection even if you tried to, plus the continence that I mentioned earlier. Also like I said, no high voltages or energies, so it should be fine.

[ejeffrey]

For current up to a couple hundred mA, reed relays are the best.  They are sealed so the contacts don't get contaminated and they can last for millions of cycles. 

Depending on your leakage requirement you may not have any option besides relays. The leakage current of high current FET switches will show up as an error on the low current ranges.  Relays can be plenty reliable if used properly.

I second razvan784 that muxing current and voltage leads on the same jack is generally a bad idea.  It can make sense in an automated test setup but isn't really needed or desirable for a bench instrument.  You still may need relays to select between different current shunts however, so the question is still relevant.

[tooki]

I'm designing a very high precision DMM (I know everyone's going to point out that I'm a fool for trying this, but I've made great progress so far and I'm pretty happy with the results) and I need a way to change connections between the modules. for example there is a wide dynamic range shunt (1 ohm to 100Kohm in 6 steps, each being an order of magnitude), the ADC, the probes, etc.

I do not wish to manually change the probe to different positions, or reconfigure stuff by hand (for example change different ranges) for ease of use.

The simple solution would be relays but they're mechanical parts and prone to much earlier breakdown compared to solid state solutions. For somethings like the ADC, analog muxes are an obvious choice, but for things like the probes and the shunt, high currents (a few amps) can be involved so using a mux is out of the question.

N channel fets would be the next best solution but it's not possible to use them everywhere due to high gate voltages required.

Please let me know if there's anything besides relays that can do this reliably.

Maybe… look at how existing DMMs solve this?

What makes you think relays aren't suitable to this task? Pretty much every bench DMM (and some handheld ones!) uses relays for various things.


Why would you switch the high-current current measurement input at all, if as you say later, you'll have separate current jacks anyway? Just have the high-current input go through a shunt and switch the voltage output from that. If your voltage detection circuitry is sensitive enough you can still have tremendous dynamic range. (I mean, my Keithley 2015 can measure up to 3 amps with 10µA resolution or 1A with 1µA resolution.) Also, 1Ω-100kΩ seems like an odd choice of shunts anyway — why would you use such crazy high values? There are good reasons why current measurement shunts are typically very low values. A typical shunt in a DMM is around 0.1Ω for a 10A range. My K2015 looks like it uses 15Ω for the 10mA range (10nA resolution) and 0.3Ω for all the rest. Even 15Ω is enough to be troublesome in some circuits. A 100K current shunt is… preposterous? What kind of circuit could tolerate that?!?

That is exactly why I want software control. when you switch modes, it automatically changes the input, so you can't mess up the jack connection even if you tried to

But above you say you don't want to have to move the probes. So which way is it: separate current jacks or not?

[OM222O]

Nope, I never mentioned wanting separate jacks. Those shunt values are also for low currents as well as being used for a current source, used for 4 wire measurments. there are milliohm shunts for higher currents but again, they need to be switched in / out.

Only problem with relays: bulky and unreliable compared to solid state solutions. there are small ish relays but they're still larger than a few SMD fets.

[tooki]

Nope, I never mentioned wanting separate jacks.

You suggested it when you said:

That is exactly why I want software control. when you switch modes, it automatically changes the input, so you can't mess up the jack connection even if you tried to

But if the same jacks are used for both current and voltage, then they are BY DEFINITION capable of being "messed up". The only way to not mess it up is to have them separate.

Only problem with relays: bulky and unreliable compared to solid state solutions. there are small ish relays but they're still larger than a few SMD fets.

Bulkier? Sure. Unreliable? Again, they're used in practically every bench DMM. If they had meaningful reliability problems, we'd all know. But while not unheard of, relay failure in DMMs is rare. (And the FETs, if anything, seem to fail more.)

[OM222O]

I will go with the slim relays then. They seem to come with a version that doesn't need active driving to stay on either, so just pulsing it on and off is good enough and saves a lot of power too.

Regarding messing up the measurement types: if your software does the job correctly and switch to the correct nodes, then no, you can't mess it up! protection features like over voltage and over current also stop you from blowing up the DMM or the circuit.

[tooki]

If you tell it to switch to amps while connected to a voltage source, you ARE creating a problem. The software can't know this.

[RJSV]

WOW, that's exactly what I've been thinking, at least the connection parts! My project has cascaded elements, i.e. 'modules' where the serial order is dialed in but later an expanded version will need software control and a large switching array.
  Reminds me of efforts made, in early telephone switch technology, developments AKA 1895. In my case each separate circuit module interfaces using voltage in range of zero through about four volts. My switch choice is a two-tier 1 of 6 rotary. Pict shows how the rotary and SpDt switch are for a 1 of 12 selection, (each module selects the input it wants).
  Could you browse over to read my post, in this PROJECTS section? Meanwhile, thanks, I want to keep reading your thoughts, there, you can see "Multiple Module Connection Box".
  Your work results can directly help me, trying to switch vast numbers of signals and network concepts for doing so.  RJHayward

[OM222O]

If your circuit doesn't require high currents like mine, then your best bet are the analog mux ICs that were mentioned earlier, otherwise it seems relays are the best option. Maybe have a separate daughter board with the relays attached to it and have the modules on another PCB on top.