usb-tc01 cold junction noise?

[Blokk]

Hello,

I am using USB-TC01 thermocouple devices a lot from NI, and I have a practical question.

The scenario: I "Peltier-control" an electrical cabinet's (1 cubicmeter) temperature, I measure the temperature with the above sensor. I feed this value into a PID VI, which is driving a current source connected to Peltier-modules.

Sampling rate (also for PID) is 1 second. I can reach pretty nice temperature stability, between 0.02 - 0.04 Celsius standard deviation calculated from 600 samples. I have tuned my PID using the Ziegler-Nichols rules.

So basically I am satisfied, but one thing I notice on the measured USB-TC01 signal. It sometimes, randomly "jumps" up and down, with an amplitude of ~0.05 Celsius. Is this the effect of the cold junction compensator? I have attached a figure to show what I mean.

Sometimes, for certain intervals, which can be even like 10 minutes, I do not see these "jumps". In these intervals, the stability reaches +-0.02 Celsius.

Including the noisy parts, the stability drops with a factor of 2 (+- 0.04 Celsius).

I guess this does not mean real jumps in the real temperature, so this is an artifact. My question is that, do I need to filter these jumps somehow to have an even better PID, or am I already at the practical limit using this sensor? I tried to put a FIR filter before the PID VI, but I did not notice a significant improvement...

Thanks for advice!

[johnsold]

My first thought is to wonder if you are getting power line frequency interference. Because the sampling rate of the TC01 is so low, you cannot even tell whether that is the problem by any measurement you can make with the device.

Does the noise correlate with anything in the system or nearby which gets turned on or off, such as a fan, motor, heater, or other high power device?

Also note that the Input noise specification is 2 uVpp which roughly correlates to 0.05 degrees. This would not explain why the noise is intermittent.

Lynn

[Blokk]

well, this is a quite complex experiment, there are two vacuum pumps, 12 GPIB Keithley devices (4 of them power source), 4 NI USB-TC01, air mixing fans like 14 separate units, etc... 🙂 But all the fans for example, continously running, so no on/off switching happens in high power devices here...

But. As I remember, I always saw such behaviour of this NI device before (I have installed like 6 in the last 2 years), even in a much less (electrical) noisy environment. So maybe this is an internal behaviour...?

What I can do, tomorrow I will take one sensor to a shielded room, where I will run only a PC and the USB sensor attached. I will check whether I see this kind of jumpy noise.

[johnsold]

That sounds like a good check.

Lynn

[Blokk]

well, i have managed to test another sensor (the sensor which gave the posted data is recently used and running, cannot take it) in a less noise environment. I see continuous noise, with a pp amplitude of 0.05 degrees, what is a value specified for the sensor as you mentioned previously. Maybe the first sensor is defect in a way, it sometimes has less baseline noise? 🙂

Actually it is an interesting question: NI produces a lot of these sensors, and I guess they state the "worst possible" noise value in the spec. papers. What about if a certain number of the sensors are better? 🙂 Well, I have no idea if this makes any sense... 😄

when I have the chance, I will take the first sensor to the same less noisy location, and I will compare them. I am very interested what I will see...

[johnsold]

I have no experience with those devices. I have seen excess noise effects in sensitive amplifiers which have been damaged by transient overvoltages on their inputs. They usually are not as consistent in amplitude as you are seeing. Certainly any device which can measure uV signals is more likely to be damaged by out of rang inputs than less sensitive devices.

Because of the wide range of possible applications, NI (or any other manufacturer) can only test and specify their devices under a limited set of reproducible conditions. NI generally tries to specify near worst case conditions so that most customers will experience better performance.

Are the thermocouples grounded or shielded?

Lynn

[Blokk]

regarding to the spec, there are two types of TCs:

(http://sine.ni.com/ds/app/doc/p/id/ds-215/lang/en)

"Choose from two thermocouple options: an exposed-tip junction or a grounded probe. The exposed-tip junction thermocouple offers the best response time, which is ideal for measuring rapid temperature changes. The grounded probe thermocouple is enclosed by a metal sheath and is ideal for measuring corrosive gases and liquids."

I use both types, but I only used the "grounded probe" type in the above tests. So I guess they are grounded to answare to your question. But I do not know the wiring structure, I never cut it...

Edit: Actually the peaktopeak noise amplitude and the resolution of the CJC is very close:

"Cold-junction compensation sensor resolution: 0.0625 °C typical"...

[johnsold]

Unfortunately the thermocouple terminology is misleading.

The grounded probe thermocouple has the metal sheath connected to the thermocouple junction inside. However, the TC01 only has a two-conductor connection to the thermocouple. At that point in the circuit, neither conductor is at ground potential. If the sheath is "grounded" to the chamber and leakage currents from some of the equipment in or near the cabinet flow through it, you could have problems. The USB ground is connected to the computer ground (if any). The specs look pretty good at rejecting line frequency noise, but this could still be an issue. 

If the sheath is connected to a grounded portion of the cabinet, try running the system with the sheath pulled away from any electrically conductive connection. Of course the temperature measurement will not be as accurate, but you are looking for the noise mechanism. If that solves the problem, you will need to get a thermocouple which has a sheath which is isolated from the thermocouple wires. Omega and most other thermocouple manufacturers make such devices.

Lynn

[Blokk]

My sensor is "hanging in the air", in the middle of the chamber, so measuring the air temperature, which I want to control (there is no galvanic connection). Actually I am very satisfied with the recent performance of this USB TC01 sensor, I just like to learn and see a bit deeper how these sensors work, so thanks for the replies!

The sensor does what is stated in the spec, I was surprised to see how easy I got the ultimate stability with a quick PID tuning (okey, I have quite robust Peltier units, each of them has ~300 Watts heat pumping performance, of course depending on the temperature differences).

[johnsold]

Good.

I like it when people want to understand how the components of their systems work as well as just getting them to work.

The biggest issue with getting stable thermal systems is often the time delay between an input of energy and the change of the output sensor temperature. Long delays can make it very difficult to stabilize the systems.

You are probably close to the limit of what you can expect from air temperature control in a chamber of that size without going to extraordinary efforts.

Lynn