Analog Multiplexer for Thermocouples

[BuffaloEngineer]

Will it distort the measurement of a thermocouple to use an analog mux on the input (before the signal reaches the amplifier)? If it was a differential analog mux that added the same resistance to both inputs (±2 or 3Ω) to both inputs, is that going to force me to come up with a new equation to linearize the amplified signal?

 

[Diver300]

The resistance itself won't make much difference to the thermocouple readings, as there is very little current. It doesn't make much difference if the resistance is balanced or not.

The multiplexer might add a small offset voltage, which would make much more difference to the temperature readings.

What temperature range are you reading? Thermocouples only give out a tiny signal and for a small temperature ranges there are much more sensitive devices.

 

[BuffaloEngineer]

We don't need ultra high precision, maybe ±1C centered around 150C. Offset voltage is not a specified parameter of the parts I'm looking at, but I'm sure it's an issue. I feel like it's a great economic decision but with a high risk that small offsets will find their way through to the output.

 

[MikeMl]

We don't need ultra high precision, maybe ±1C centered around 150C. Offset voltage is not a specified parameter of the parts I'm looking at, but I'm sure it's an issue. I feel like it's a great economic decision but with a high risk that small offsets will find their way through to the output.

Look at the Omega web site for guidance. Based on my experience, 150C is too low to use thermocouples. I would be looking at Pt RTDs

 

[BuffaloEngineer]

Look at the Omega web site for guidance. Based on my experience, 150C is too low to use thermocouples. I would be looking at Pt RTDs

Why too low? They cover this range, and as far as I know most folks are using K and T type in the field. An RTD will be more accurate, but more expensive, which is why I think thermocouples are so prevalent.

 

[MikeMl]

Why too low? ...

SNR sucks. Small offsets cause huge error... You need cold junction compensation with thermocouples.

 

[BuffaloEngineer]

I guess the question is what is a huge error? I was thinking of using something like the MAX31856 to linearize the input (make it easy for me) - I can always skip the analog mux and just use a few of these thermocouple ASICs. I have been trying to figure out a good way to define the accuracy - most sites list a K type thermocouple at ±2.2C, but that does not account for offsets in the electronics. It's hard to define what the accuracy might be without building it and testing it.

https://datasheets.maximintegrated.com/en/ds/MAX31856.pdf

 

[BobW]

The biggest drawback to using thermocouples where I worked (industrial processing plants) was the requirement for long runs of expensive TC wire back to the main control panel, often several hundred feet of wire per TC. That combined with the other drawbacks of TCs made RTDs much more appealing, and cheaper in the long run. However, if you have only a very short run from the TC to the amplifier then using a TC may not be a bad choice.

 

[MrAl]

Hi,

K type to ADC box should be ok. Any long runs can then be digital or a higher level analog.
K type should be ok for general 300 degree F use. Calibrate if higher accuracy is needed.

 

[KeepItSimpleStupid]

The biggest drawback to using thermocouples where I worked (industrial processing plants) was the requirement for long runs of expensive TC wire back to the main control panel, often several hundred feet of wire per TC. That combined with the other drawbacks of TCs made RTDs much more appealing, and cheaper in the long run. However, if you have only a very short run from the TC to the amplifier then using a TC may not be a bad choice.

Then you did it wrong. You use "Extension wire" to bring the TC back to the "cold junction" or use a temperature transmitter that uses Fieldbus or 4-20 mA.

 

[BobW]

Extension wire *is* thermocouple wire. And yes I know there are lots of ways to get those horrible inaccurate TC signals back to the control panel. All of them involve extra cost and messing around which these companies didn't want to do. In some cases where they had to use TCs because of the high process temperatures involved they often decided to use regular copper instrumentation cable and not even bother with cold junction compensation, because it didn't make the accuracy much worse than it already was.

TCs do have their applications, but not in any of the industrial plants that I've worked in in the past 40 years.

 

[Diver300]

Extension wire, or compensation cable, is probably thermocouple wire for type K. For type R, the are much cheaper compensation cables that work well as long as the joints between the thermocouple and the compensation cable is at a moderate temperature.

Also when using a compensation cable that isn't identical to the thermocouple wire, the + and - joints must be at the same temperature as each other. So where the compensation cable joins the thermocouple, the + and - must be close together. I think that there are junction boxes made with that in mind. You can easily get the situation where a 1 ° difference between the + and - interconnects results in an error that is far more that 1 °.

Running long runs of thermocouple wire or compensation cable was considered outmoded when I was in the process control industry 30 years ago. It is far, far better to put a transmitter on the back of the thermocouple pocket and send back a more robust signal. It was 4 - 20 mA when I was doing it. I guess it is multiplexed now.

 

[Diver300]

I guess the question is what is a huge error? I was thinking of using something like the MAX31856 to linearize the input (make it easy for me) - I can always skip the analog mux and just use a few of these thermocouple ASICs. I have been trying to figure out a good way to define the accuracy - most sites list a K type thermocouple at ±2.2C, but that does not account for offsets in the electronics. It's hard to define what the accuracy might be without building it and testing it.

https://datasheets.maximintegrated.com/en/ds/MAX31856.pdf

Just use several of those. It is a far better way of doing it than multiplexing the analog signal. Your cold junction compensation will be done separately in each MAX31856. If your application is price sensitive, you could look at multiplexing, but then you would have to factor in the time taken to measure how much the accuracy is degraded by a multiplexer, and you could make a lot of units using separate MAX31856 converters for the cost of checking your multiplexer, at various temperatures of both the multiplexer and what you are measuring.

I think that the 1/128 °C resolution of the MAX31856 is a bit over the top, but you can ignore the last few bits.

When I had to use thermocouples, I used similar temperature converters. It was only for two inputs, but there was no way that I was going to mess around with the thermocouple signals. I just wired the + and - straight to the board as near to the inputs of the converters as possible.

 

[KeepItSimpleStupid]

The biggest issue is cold junction compensation. Most companies do it wrong Measurement Computing just puts a sensor anywhere nearby. It's affected by air currents etc. Totally useless. This https://www.microdaq.com/measurement-computing-usb-temp-daq.php is NOT the way to do it.

Omega used to distribute a large termination point with a bunch of screw terminals and a buffered temperature readout of cold junction. I ccan't find it now.

I used a relay based multiplexor at one time. Not for computer control. It just paused a few seconds at each point selectable.

Analog Devices made a 4 TC compensator to place on the edge cards of one of their products.

So, it's mostly in the mass of the screw terminals and the need to be in close proximity with a sensor if you want to handle multiple types.

 

[Reloadron]

The biggest issue is cold junction compensation. Most companies do it wrong Measurement Computing just puts a sensor anywhere nearby. It's affected by air currents etc. Totally useless. This https://www.microdaq.com/measurement-computing-usb-temp-daq.php is NOT the way to do it.

Omega used to distribute a large termination point with a bunch of screw terminals and a buffered temperature readout of cold junction. I ccan't find it now.

I used a relay based multiplexor at one time. Not for computer control. It just paused a few seconds at each point selectable.

Analog Devices made a 4 TC compensator to place on the edge cards of one of their products.

So, it's mostly in the mass of the screw terminals and the need to be in close proximity with a sensor if you want to handle multiple types.

That right there. Where the TC meets the MUX you will need CJC (Cold Junction Compensation). Mike also mentioned it. CJC aside if you want +/- 1.0 Deg. C. uncertainty you will need a certified calibrated TC. For a low temperature the use of Type T would come down to:

Type T:

MAXIMUM TEMPERATURE RANGE
Thermocouple Grade
– 328 to 662°F
– 200 to 350°C
Extension Grade
– 76 to 212°F
– 60 to 100°C

LIMITS OF ERROR
(Whichever is greater)
Standard: 1.0°C or 0.75% Above 0°C
1.0°C or 1.5% Below 0°C
Special: 0.5°C or 0.4%

As Mike mentioned the Omega Engineering Website contains a wealth of information on temperature measurement in general using just about every sensor you could imagine.

How many TCs did you want to MUX? Back to CJC how close to standardized will the junctions be? Like a room temperature of 70 F. give or take? You may also want to give this a read as it gets into the CJC thing pretty good, thermocouples and their uncertainty.

Ron

 

[atferrari]

I recall doing basic maintenance in engine rooms fitted with control systems RACAL DECCA. Their ISIS models varied from each other but in all, every sensor had their own xmtr to the respective board. 4-20 mA loops used profusely. Made easy to say who failed: sensor, connectors, wires or the associated electronics. No muxes.

Noise, temperature and VIBRATION. Fast marine engines.

 

[Reloadron]

I recall doing basic maintenance in engine rooms fitted with control systems RACAL DECCA. Their ISIS models varied from each other but in all, every sensor had their own xmtr to the respective board. 4-20 mA loops used profusely. Made easy to say who failed: sensor, connectors, wires or the associated electronics. No muxes.

Noise, temperature and VIBRATION. Fast marine engines.

I used mostly Inor temperature transmitters, the newer ones were nice in that you could program them.

They can also be used in applications beyond temperature.

Ron

 

[KeepItSimpleStupid]

Reloadron The transmitters look pretty cool. Not sure if the speed would matter in some cases.

 

[Reloadron]

Reloadron The transmitters look pretty cool. Not sure if the speed would matter in some cases.

If you ever have need for one let me know. I did have the programming software but have since changed computers. I also had some old non-programmable transmitters which were Type J and I think 35 to 700 F. Getting old and have stuff piled all over the house I will never use in a lifetime, not my lifetime anyway.

Ron

 

[BuffaloEngineer]

That right there. Where the TC meets the MUX you will need CJC (Cold Junction Compensation). Mike also mentioned it. CJC aside if you want +/- 1.0 Deg. C. uncertainty you will need a certified calibrated TC. For a low temperature the use of Type T would come down to:



As Mike mentioned the Omega Engineering Website contains a wealth of information on temperature measurement in general using just about every sensor you could imagine.

How many TCs did you want to MUX? Back to CJC how close to standardized will the junctions be? Like a room temperature of 70 F. give or take? You may also want to give this a read as it gets into the CJC thing pretty good, thermocouples and their uncertainty.

Ron

Here's the point I get stuck on for calibration - if the thermocouple is ±2.2C, then you can calibrate for better results. But what if they change thermocouples? You would have to ship it and say it's ±1C as long as you never mix up your thermocouples, correct? Is that realistic?

Thanks for the reading material! I'm just getting started on this so I'm doing a bit of reading on it all!