Relay / fet matrices for low current / voltage switching?

[alanambrose]

Hi,

I have three related questions:

(a) What are your go-to switching matrices for switching with better than 10uV thermal EMF?

(b) Does anyone have any leads on the documented behaviour of relay contacts / contact materials when switching low voltages / currents?

(c) Failing off-the-shelf solutions, are there any go-to designs for low voltage switching (i.e. uV with max 5V) into high-ish impedances (i.e. 10M ohms)?

I'm looking for 25, 50 and 64 channel matrices - but only the 64 channel is critical. All leads welcome

TIA, Alan

[Zucca]

What are your low current spec?

[alanambrose]

>>> What are your low current spec?

Good question and we need do to some tests - this is an opt101 photo-diode but with v. low light levels on particular IR wavelengths.

Alan

[Alex Nikitin]

https://www.pickeringrelay.com/applications/low-thermal-emf/

Cheers

Alex

[alanambrose]

Thanks Alex.

FYI, for anyone else looking at this in the future, there's a doc here which gives a bit more analysis of relay contacts re low thermal:

https://www.mouser.com/pdfDocs/switching-low-offset-voltages-training.pdf

Alan

[Kleinstein]

With so many channels an a rather low voltage, I would consider CMOS switches in IC form (like chips for a 8:1 MUX or even 16:1 MUX). The big down-side is possible leakage current.   Due to usually very low power consumption and a compact design temperature differences can be kept low.

Switching between photodiodes usually is not that critical with respect to voltage, it's more like a current sensitive case.

[Echo88]

Thats the most comprehensive document on low-thermal-relay-construction ive ever seen and it even states the recommended manufacturer test procedure. Thanks. 

[Alex Nikitin]

>>> What are your low current spec?

Good question and we need do to some tests - this is an opt101 photo-diode but with v. low light levels on particular IR wavelengths.

Alan

From the datasheet that photo-diode + amplifier combination is rather noisy, with 300uV RMS noise in 20kHz BW (for a comparison the 1M feedback resistor thermal noise in the same bandwidth is ~18uV at room temperature). If you switch the amp outputs you will have more problems with the amps noise, offset and drift than with thermal voltages IMHO.

Cheers

Alex

[rhb]

Has anyone documented the thermal EMF of the 44421A and 44422A relay boards from the 3497A?

[martinr33]

The quad CMOS switches that seem to populate precision meters might help:
https://www.maximintegrated.com/en/products/analog/analog-switches-multiplexers/MAX333A.html

I'd be more worried about thermals around the connections.

Keithley scanners /meters with scan options are reasonably inexpensive. Might be possible to switch out the relays with CMOS switches (maybe a little patch PCB). Then, you'll have a manageable package. 10uV isn't too hard to deal with.

There are not so many of them on eBay right now, but they do show up.

The bigger scanners have a lot more channels. The thermal offset is kind of in range as well.

https://www.conres.com/test-equipment/wp-content/uploads/sites/4/2014/04/selector_guide_7001_7002.pdf

[feedback.loop]

How about Agilent 3499A/B/C + some multiplexing module like 44470A?
3499B is a small unit with two slots for modules. 3499A is a bigger unit with five slots.
They can be found for under $100. Modules are also cheap enough.
Typical thermal EMF of 44470A relay module is specified as < 3uV
Be warned that many modules are sold on ebay with no terminal blocks.

[alanambrose]

As ever, some quality insight and experience many thanks to everyone who gave advice.

For speed, I think we'll try to pull in some off-the-shelf matrices and then develop our own custom ones as and when we've been able to characterise a bit better.

Probably 2 off 34972A plus 4 off 34908A 40-channel (3uV) or something similar.

Good point Alex about whether we're being too ambitious on the scanning given these particular photodiodes.

I'll keep you posted.

Alan



 

[splin]

The quad CMOS switches that seem to populate precision meters might help:
https://www.maximintegrated.com/en/products/analog/analog-switches-multiplexers/MAX333A.html

That won't cut the mustard - the spec shows a typical ON leakage of 200pA; combined with the 10M source impedance requirement that gives a 2mV offset! And that will vary drastically with temperature.

The best such part I know of is the TI MUX36S08 with a typical ON leakage of 8pA. Still too high for this application but the leakage is a function of the input voltage relative to the supply voltages so actual performance may be better in this case. The scales of the leakage v temperature diagram in the datasheet are not helpful for room temperature applications but a more expanded graph is shown in figure 5 here:

http://e2e.ti.com/blogs_/archives/b/precisionhub/archive/2016/02/19/does-a-low-leakage-multiplexer-really-matter-in-a-high-impedance-plc-system

The 2N/PN/SST4117A Jfet (< 1pA leakage) might just do it but they are expensive, hard to get and require a lot of support circuitry. The MBM4117 is cheap but has a worse leakage spec - but possibly only because it isn't production tested.

[feedback.loop]

Check this out:
https://www.ebay.com/itm/Agilent-3499A-Switch-Control-System-with-N2261A-40-Channel-Relay-Plugin-Card-DIM/352018520458

[EmmanuelFaure]

The best such part I know of is the TI MUX36S08 with a typical ON leakage of 8pA.

MAX326/327 : Max leakage 10pA, typical < 1pA.

[Kleinstein]

The max326/327 are low leakage, but also relatively high resistance. It dependents on the application if this is ok.  The mux36S08 looks good - low leakage and much lower on resistance.