Use analog switches or MOSFETs controlled via MCU GPIOs instead of mechanical switches (S1–S8), for automatic control.
Among Analog Switch ICs, you can choose CD4051 8:1 mux, 74HC4051 or ADG708. Or use N-MOSFETs/P-MOSFETs (with proper gate drivers). You have to use 3.3V as ADC reference to match the top of your divider. Add a small capacitor (e.g. 0.1 µF) between ADC input and GND to filter noise. Following are some designs that you may use for references.
https://www.hackster.io/Costalegre/auto-ranging-ohmmeter-6fa4af
https://www.pcbway.com/project/shar...e_Meter_without_microcontroller_a98f1825.html
https://simple-circuit.com/arduino-auto-ranging-ohmmeter-lcd/

Thank you. Although my initial post was based on voltage dividers, I currently plan to use a constant current source. Still, I appreciate the post, and I'll take the opportunity to analyze these analog switch ICs.

 

You need to understand that THERE IS NO SUCH THING AS A PERFECT ANALOG SWITCH!! All of them have some amount of charge injection and internal resistance. Of course there are ways to work around those effects, but they do not happen by themselves.

I see no answer to the questions posed in my post #52. Especially important is the purpose of the test. Without knowing the purpose of the test it is not possible to create any reasonably cost effective testing scheme. All we have is suggestions of circuits that may, or not, be applicable.
Common purposes of production testing include:
1. Verify that a part is within some defined specification at a single point.
2. Verify that the part response is within some defined specification though out it's entire operating range.
3. Create a set of part responses to allow a system calibration for that specific part.
4. Measure the part variation within specifications for the purpose of part production process control.

The hardware and software to carryout the four different types of tests are quite different, as are the times for the tests and the cost of both the tester and the testing.
So until the thread starter provides an answer as to the purpose of the test there is no reason to continue.

I agree. Since I plan to test four thermistors simultaneously, I plan to use four constant-current circuits to test each thermistor and connect the output of each of these circuits to an analog input of an ADC. This way, I'll avoid using switches and multiplexers.

 

OK, and that is why I did not even mention those variables.
The single IMPORTANT question is what the actual purpose of the testing . That really does matter a great deal.

I have asked that question three times, #31, #52, and again in post #57. We do get a sort of description about the type of test but still not one word about the actual purpose.
Is that omission possibly a translation problem???

I will try to be quicker in my responses, out of respect for the help you are giving me, but, as I mentioned before, I don't have any free time while I'm at work and, currently, I only have a few hours available.

 

I apologize for the delay in responding. My work schedule is a bit hectic, and I don't have time to access and respond while I'm working.

This thermistor reports, in real time, the internal battery temperature to the aircraft's electronic systems. As you may know, this information is very important for flight safety. For safety reasons, there's also a thermostat. The thermostat also needs to be tested, but since it only displays open or closed states, it's easier to test.
The purpose of this test is to ensure that the thermistor still functions according to the manufacturer's specifications. I understand that, to an electronics specialist, the test data reported in the manual may seem incomplete or inaccurate, but aircraft maintenance, overhaul, and repair manuals are written for aircraft technicians. For an aircraft maintenance technician, the temperature vs. resistance curve of a thermistor doesn't matter, for example. The only thing that matters to him is knowing that, within a certain temperature specified in the manual, the thermistor must exhibit a certain resistance, also specified in the manual.
The test will be considered successful if the resistance is within the specified range. Any value outside this range results in a failed test, in which case the thermistor is destroyed and a new one is installed. The tolerance is already within the range specified in the manual, so no values outside the range are tolerated.

OK, that can certainly be done with a single fixed current source per thermistor installation tested. The system would monitor the voltage across each thermistor at each specified temperature test point. Using a one milliamp current makes the system checking process and verification simpler and faster. It also removes a lot of hardware and software from the test system. My design of the CC source uses only two components. The downside is that it requires using true "Kelvin" connections. But that is actually not a big deal.
BUT cooling and heating the battery assemblies does seem like a big effort. And it seems to me that two or three points should be able to verify that it fits the curve. But if the system is fully automated then it could produce detailed test reports with more measurement points. The down side is that the testing will take a few hours.
The place for the analog switches will be in the tester system self check portion. So it can verify correct calibration before and after each test sequence.
Hopefully there can be a battery performance test happening at the same time. It has been many years since I was involved with a battery performance tester, and that was for audit use, not all production.

 

OK, that can certainly be done with a single fixed current source per thermistor installation tested. The system would monitor the voltage across each thermistor at each specified temperature test point. Using a one milliamp current makes the system checking process and verification simpler and faster. It also removes a lot of hardware and software from the test system. My design of the CC source uses only two components. The downside is that it requires using true "Kelvin" connections. But that is actually not a big deal.
BUT cooling and heating the battery assemblies does seem like a big effort. And it seems to me that two or three points should be able to verify that it fits the curve. But if the system is fully automated then it could produce detailed test reports with more measurement points. The down side is that the testing will take a few hours.
The place for the analog switches will be in the tester system self check portion. So it can verify correct calibration before and after each test sequence.
Hopefully there can be a battery performance test happening at the same time. It has been many years since I was involved with a battery performance tester, and that was for audit use, not all production.

Of course, I prefer the device to maintain accuracy across the entire temperature range, but since the test must be 100% accurate to the manual, the final report will need to report the resistance readings for only the two required points [0°C ±1.7°C (+32°F ± 3°F)] and [+71°C ± 1.7°C (+160°F ± 3°F)].

I'm only discussing this test with you, but since you mentioned it, the manual requires us to perform a series of other steps on the battery. The normal flow of operations, in very brief terms, consists of the initial inspection, complete initial discharge of the battery (including the use of dissipation resistors to short-circuit the battery after the "traditional" discharge), disassembly, washing of the cells, insulators, case, etc., sensor testing, reassembly, intermediate inspection, recharge, capacity test (which consists of another discharge measuring the time it takes the battery to reach 20V), and, if the battery passes the capacity test, the final recharge and final inspection. Remember that, regarding the complete discharge, the battery is a Ni-Cd.

 

That seems like a very extensive and labor intensive validation sequence. Certainly the testing would benefit a great deal from being automated. The inspection portion should include a detailed description of pass and fail.
Twenty years ago I would have messaged you to send contact info so that we could quote the project of producing the machine. But for the controls it would have used an industrial PC.

 

Certainly the overall complexity of the sequence of tests demands much more than a small processor module can provide. And certainly such tester software will require adequate verification. So in addition to everything else, this tester system will require a separate verifiable checking standard.
This is by no means a hobby-class design project.