Multiplexed DMM using LTC2400

[Christopher]

Making a multiplexed DMM with one of these chips, a bunch of relays and a 32-bit pic.

Hacked up an input stage in about half an hour earlier:

Questions:
- 100nF cap across the input, how's that? Will I also need cap across the ADC input?
- Will be potting down the voltage using the resistors on the input of the relay. (Only looking for 12+ useful bits, but will use tight tolerance resistors)
- SPI lines on the LTC2400 will use optocouplers
- Differential measurements. I am using one of those potted 12V isolated board mount converters to power the LTC2400 and ref. Is it safe to float the circuit with respect to the return of whatever I'll be measuring?? Could be dangerous?
- No diodes across coils currently cos im lazy and just want to get something on paper!

Relay cross-conduction shouldn't be a problem, if I add delays in the code? I've seen some ICs which control relays with cross-conduction protection, but I think that's a bit overkill

That's all for now. Will look forward to sharing my finished 12-channel version soon

Current BOM including chips: £100. The relays are a huge part of this cost, they're the same as the relays used in the Agilent dataloggers. Rather stick with something the pros use

[tszaboo]

That is not a DMM. It is a data acquisition system. It cannot measure current, and Ohm and stuff. Also, there is no input protection, which makes it dangerous, or not usable for any serious work.

[Christopher]

That is not a DMM. It is a data acquisition system. It cannot measure current, and Ohm and stuff. Also, there is no input protection, which makes it dangerous, or not usable for any serious work.

Sorry for not confirming my project requirements. This will be a data logger system, for ATE testing of power supplies.

It'll only be used to measure divided voltage and current (Voltage across a shunt resistor which isn't part of this).

Good point regarding input protection. MOV across the input should do?

[tszaboo]

TVS diode, resettable fuse and some sort of diode clamping to the power supply after the resistor dividers, and relays. MOV is useless for non 230V (or 110 or 220 or 240 or anything mains power supply)systems.

[Andreas]

- 100nF cap across the input, how's that? Will I also need cap across the ADC input?

Relay cross-conduction shouldn't be a problem, if I add delays in the code? I've seen some ICs which control relays with cross-conduction protection, but I think that's a bit overkill

Any cap at the input together with the input impedance will generate non-linearities for the ADC.
Ok at 12 bits that does not mind.
After a voltage divider you may probably need a voltage follower to keep the input impedance low for the ADC.
One single voltage divider after the relays should be sufficient. (I would use one low pass capacitor between voltage divider + voltage follower).

Cross conduction could be avoided when stacking the relays. (You would even save money since for 4 inputs you will need only 3 relays).

With best regards

Andreas

[Christopher]

OK i will not bother with any capacitance. I will add a voltage follower using a high-spec chopper amp after the relay stage, right before the ADC

[dannyf]

Any cap at the input together with the input impedance will generate non-linearities for the ADC.

I am not sure what "nonlinearities" you are talking about but in general you do not want to have reactance on the front end as it potentially alters the circuit the probe is on.

The usual way is to have an impedance translater (a follower for example) and then you send it to the adc.

I would put a serial resistor on the input, with clamps to Vcc and gound.

[Marco]

Why use relays? Sub ohm analogue muxes are under a buck.

BTW, if you use a high ohmic input divider do remember the leakage current out of the ADC.

PS. if you only need 12 bit I guess leakage current or mux resistance isn't relevant ... you can just use a 74HC or CD4000 mux. Of course you could just use something like a CS5528 which has 8 inputs to begin with.

[Christopher]

Only reason i'm using relays is because the Agient datalogger we're currently using (and replacing with this) uses relays and has 60 channels.

The reason I'm using them is because its what it uses, and I can add as many channels as I need to. (1 reading every 10 seconds or more).

I am not sure what "nonlinearities" you are talking about but in general you do not want to have reactance on the front end as it potentially alters the circuit the probe is on.

The usual way is to have an impedance translater (a follower for example) and then you send it to the adc.

I would put a serial resistor on the input, with clamps to Vcc and gound.

OK. Thanks. Why does it alter the circuit later on? I can see it making a low-pass type filter?

What kind of value series resistor? I will add diode clams to just over the ref voltage after the potential divider, and before the follower. Bare in mind I have the potential divider.

I was going to possibly add some kind of high-frequency filter on the ADC input...

[dannyf]

If you use BAT54 clamping diodes (pretty common) here, those are 100ma If diodes so something like a 1k serial resistor will go a long way (100vdc) in protecting the input.

[awallin]

- SPI lines on the LTC2400 will use optocouplers

Did you look at the AD71xx converters too? I guess they are a bit more complex to drive/program, but could be lower noise.

For 24-bit precision over longer time the voltage reference becomes critical. What are you planning to use?
For many transducer measurements it can be useful to get the voltage-ref as an output for the user too. i.e. drive a transducer bridge with the ref-voltage (buffered) or put a current proportional to the ref-voltage through a resistive sensor.

I'm putting together a similar system for pt100 sensors with many channels and an Agilent 34970A. I want sub mV stability with a range of +/ 10 V, and it looks like the only way to get this is to do a measurement with two polarities, e.g.:
- put +500uA through the pt100 sensor, measure the amplified voltage (4-wire connection) across it
- put -500uA through the pt100 and do the same
- subtract the results.
This should subtract out any offset-voltages (opamp offsets, thermovoltages, etc) that offset the measurement in the same direction.
So you might want a digital or analog output also from your system, to do this "lock-in" technique.

For the input-stage I'd try to make it with programmable gain/offset or at least easy (a few compontents to solder, or jumpers) to change. Some applications may want +/-10V signals others 0-5V or 0-1V or whatever.

If you split it into a DVM board and one or many relay-boards it can be modular like the Agilent. If you just put the optoisolated ADC + input-stage + voltage-ref on the DVM board with SPI + a few DIO lines interface then everyone can interface it to their favourite uC or SBC. Not everyone likes PIC, and if I mention something else someone will always hate/love it (arduino, beagle, olinuxino, whatever)

Useful project for slow but precise datalogging - keep us posted!
AW