0
\$\begingroup\$

Reaching out as I am in the initial stages (read: crude block diagram) of circuit design for a project aiming to automate resistance checks between various connectors. We'll call these connections A and B. Each connector has 48 lines. These connectors interface with a PCB. This PCB mostly consists of an adapter from A to B but also contains passive components which I'd like to also test resistance (these have a specification defined below).

Testing for short circuits between each pin of B is required due to the quality control nature of this device.

As of right now, I've gone with (6) 8x 1:1 analog switches for each connector and (2) 8x 1:1 analog switches behind each for measurement bus selection (dependent on if the measurement is from Connector A->B, A->A, B->B). This seems like a bit much to me, with the number of analog switches reaching sixteen.

Resistance measurements must be taken from various lines belonging to A and B. Of note, the connections between A and B are not straight through. The measurements taken will require to not only go from each connector but between lines on each connector themselves (shown below):

A --> B
A --> A
B --> B

Additionally, the expected resistance measurements fall within a range of acceptable standards dependent on which pair is under test:

less than 1.0 ohm
between 300-360 ohms
above 150k ohms

The outputs from the (2) 8x 1:1 analog switches will then be presented to constant current ohmmeter circuitry which in turn is fed to an ADC for usage by the main MCU.

For this application, what is the best topology for the analog switches? Or is there another solution I am overlooking?

I hope I've provided enough information. In the fear of none of the above making sense, I've included the crude block diagram of what I have thus far. Any suggestions for optimization would be much appreciated, I've found myself a bit blocked up.

Block diagram

Thanks!

\$\endgroup\$
14
  • \$\begingroup\$ this is a Q&A site, not a discussion forum ... please ask a specific, answerable question \$\endgroup\$
    – jsotola
    Commented Jan 24, 2022 at 5:11
  • \$\begingroup\$ Start by listing your specs or expectations of test results , for interconnections, voltage , impedance, isolation , speed, and user interface for errors or states. Define the overall purpose then after that think of a better question \$\endgroup\$
    – D.A.S.
    Commented Jan 24, 2022 at 5:17
  • \$\begingroup\$ @TonyStewartEE75 Thanks for stopping by, I'm not looking to get down into the weeds of component specification. Only looking for any suggestions regarding topology optimization concerning the analog switches. The "measurement" section of the device is simply a constant current ohmmeter fed into an ADC. Speeds are not important for this application. \$\endgroup\$
    – ski
    Commented Jan 24, 2022 at 5:26
  • \$\begingroup\$ "This seems like a bit much to me"- Why? \$\endgroup\$ Commented Jan 24, 2022 at 5:28
  • \$\begingroup\$ @BruceAbbott Mostly due to the relative complexity and elevated BOM cost, but I suppose that comes with the territory. This is the solution I arrived at, was wondering if others would find it acceptable. Thanks! \$\endgroup\$
    – ski
    Commented Jan 24, 2022 at 5:31

1 Answer 1

0
\$\begingroup\$

One mux chip can only have one output enabled, so with the schematic in the question, you can't test for shorts between two pins of the A connector if they are on the same mux chip. Same for the B connector. But you can test A->B.

Measuring below 1 ohm will be difficult if the switches are in series with the resistance being measured. The lowest RdsON 8:1 mux on digikey has resistance between 0.5 ohm typical and 0.75 ohm max. Since different switches may have different resistance, and it depends on temperature... you could use calibration, but... this is going to be annoying. Unless you relax the requirement to something like "below 2-3 ohms", in which case this switch will be fine.

If you want accurate measurements for low resistance values to spot a bad contact, you'll need 4-wire measurement. Fortunately, on the measurement side you can use any RdsON switch including the very cheap HC/HCT4051, on the current injection side you can use a bit higher RdsON switches which are cheaper than the 750 mOhm ones, so in the end, it shouldn't matter much.

Assuming you only want to measure resistance on A->B, and short circuits on A->A and B->B, then you'd need 7x 8:1 current injection muxes and 7x 8:1 measurement muxes on each side.

Measuring for shorts: nested loop, for each pin, inject current, then for each pin, measure voltage. This will need a pulldown on each pin. The pin where current is injected should be near VCC, all the others should stay at 0V.

Measuring resistance: for each pin on A, inject current, set the mux on the other connector to GND, then measure voltage across the cable with the measurement muxes. Measured voltage should be below (maximum allowed resistance) x current.

The mux that connects to ground could be replaced with a shift register.

Personally I'd put the connectors on small boards connected to the main board with 0.1" headers, to change them easily when they get damaged. Or put the connector on the board, but use a male-female connector extension that gets the wear of plugging/unplugging.

Then I'd do two identical boards, one per connector. That should result in a smaller cheaper board, and again easier to replace if things go wrong. The microcontroller should be populated only one one board. They only need to be linked with supply, ground, and mux control signal, via a small mezzanine connector.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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