What type of temp sensor do you recommend...

29 posts / 0 new
Author
Message

Hi Everyone,

Looking at Digikey I can't believe how many different types there are!

My goal is trying to control some fans with PWM to keep a specific temperature. I probably need to study up on some PID stuff for this.

On the sensing side, the temperature sensor will be around 12" away and off board. Do you guys often use a voltage output type of sensor and ADC mostly, or something else?

Thanks,

Alan

Building one, or one thousand?

And what is the absolute accuracy, and the precision with which you need to make the measurements and maintain the environmental temperature?

"Smart" sensors, such as DS18B20 and Sensirion's sensors, (e.g. SHT1x & SHT7x), do the conversion and output a digital signal. Easy to go a foot off the PCB, and no noise issues with the distance between the sensor and the micro.

NTCs however are very inexpensive. For a norrow temperature range one can ignore the Steinhart-Hart equation.

There are other sensors better suited for high temperature environments.

Low cost IR sensors are yet another high tech approach, and sometimes useful.

JC

Hi JC,

Just a one off. Accuracy isn't as important and consistency. Mostly measuring a metal surface around 50 deg C.

Thanks,

Alan

Then a thermistor is an easy choice. Typically use it in series with a resistor equal to its resistance at 25C.

To keep the self-heating down, make the thermistor power under a few milliwatts. With a 5V supply and the arrangement just described, the thermistor voltage will be 2.5V. Since P = V*V/R, this means a thermistor resistance of at least 3K.

You read the divider with the ADC and can set the trip point in software as you wish.

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

My preference is a voltage output type sensor such as an LM335. It's output is proportional to temperature - 10 milli-volts per degree C. Measure the output with an ADC input, then do the simple math to convert the voltage to temperature.

Thermistors are not a bad choice either, but you have to deal with the non-linearity issues in the math if the expected temperature range is "extensive".

I use a forward biased silicon diode fed through a resistor to set the current in the range of around 10 to 100 uA, but it is not critical. The forward voltage drop changes approximately -2mv per degree C.

What do you guys think of this:

http://www.digikey.com/product-s...

Do you think it could be double stick taped to a metal surface?

Thanks,

Alan

Do you think the sticky tape will be a good thermal conductor? You can get thermistors with a screw tag for bolting to a heatsink.

Kartman wrote:
Do you think the sticky tape will be a good thermal conductor? You can get thermistors with a screw tag for bolting to a heatsink.

I wondered that, but tested it some with my fluke temp probe and it seemed to be within one degree F. The double stick tape I have is very thin. Accuracy isn't my big concern for this application as much as consistency is.

alank2 wrote:
What do you guys think of this:

Looks interesting. I think my next Digi Key order will include several of these to play with.

Don

It looks like a good choice for your project.

I don't like the double sticky tape idea, however.

The tape is a thermal insulator, and the ones with foam conduct heat very poorly and SLOWLY.

As you develop your control software you will need to consider the "thermal" time constant of the system. If the sensor reading is very slow and delayed, it significantly impacts your control software methodology.

Also, the tape will dry out with time, and not stick.

If you can drill a small hole in the metal then you can insert a sheet metal screw with a small metal strap to press the sensor against the metal. Or two screws with a small band (strap) between them, squishing the sensor against the metal.

A drop of thermal paste, used for PC processors and heat sinks, would be great, but maybe overkill if you don't have any handy.

I did not read the data sheet, just the overview. With an analog sensor and one foot leads you probably want a small cap on the micro's ADC input pin. Whether it is just a cap or a small RC depends on the output driver for the sensor. Or at least add some pads to your PCB for this, in case you need it. Depending upon what else is in the area, you might want to use a shielded cable for the sensor signal.

If you really don't want to use a screw and strap then think about using epoxy or some other bonding agent for metal and plastic. My builder's supply store and Ace hardware have 20 different glues these days, good for prototyping. I'm sure there are expensive, special purpose glues available.

Remember to post a photo in the AVR Freaks Projects Thread when you have it up and running!

JC

Hi JC,

Thanks for the tips. I've got to types of double stick tape, both have a very thin liner. One is polyester and the other is a white PVC. I would like a better solution, but the metal box is a lpro rubidium oscillator and is mostly a smooth metal box except for the mounting holes.

My unit is similar to this one on eBay, but I only paid \$100 for it.

http://www.ebay.com/itm/LUCENT-R...

It has a goldish looking casing that I'm not sure is anodized aluminum but possibly it is copper? The rubidium module fastens securely to the bottom inside of this giant heat sink with 6 flush screws. Do you think it conducts well enough that I could mount the sensor very near the rubidium, but instead drill a hole in the case?

I'm going to build a single case that has a u-blox Neo 6M GPS board in it, an AVR to convert the Neo 6M data format to the Motorola Oncore format that the rubidium unit needs, a TTL serial to RS422 conversion, and the fan controller I've been asking about in this thread. I'm going to mount two 60mm fans on the backside of the unit. Add some power switches and a real 24V desktop adapter instead of using my bench supply and it should be good to go for a GPSDO.

Thanks,

Alan

The casing is unlikely to be Cu, Cu is too soft.

Can you post a photo of the rub unit sitting in the case?

I'm absolutely the last person on this Forum that should be giving thermal advice.

That said, if there is room adjacent the Rub unit you could put the temp sensor on top of it, with a strap holding it down. The strap anchors to the case adjacent the rub unit.

If the screws have enough thread length, or you can find some slightly longer ones and saw them to length, another option might be to put a small rectangular sheet of copper under the Rub unit, like a big flat washer. The copper has holes for the original mounting screws. The copper piece sticks out a centimeter or so from one side, and to that you attach your temp sensor.

You would like the temp sensor as close to the rub unit as possible. Such a large case, with fans, will have quite a temperature gradient across its surface.

And that said, if the unit is steady-state thermally, then the exact location of the temp sensor won't matter as much, but it will be measuring a different temp that the rub unit itself, if it is not attached to the rub unit.

See what others have to say, also.

I'm thermally challenged.

JC

I'm also artistically challenged, especially with a schematic drawing package, (I.e. no bolts, nuts, etc...), but this is the concept:

Attachment(s):

Hi JC,

The next time I have it apart I'll take a picture! The RB sits in the rear right corner from the front view and the bottom of it is held to the bottom of the unit flush and tight with six screws.

I like your copper sheet idea with a tab hanging off to one side of some sort to mount the sensor. I'll see if I can locate some thin copper somewhere.

Thanks,

Alan

The temperature sensors arrived today and are quite sensitive to noise. I tried to test it on my frequency counter breadboard (with a 150 MHz oscillator being measured on it) and it will throw off the output voltage of the sensor by 0.3V which is 15 deg C.

I measure about 70mV rms on the VCC line with the 150MHz oscillator connected. I won't have an oscillator like this present, but it does seem like this sensor is well pretty sensitive to VCC noise. What is the best way to filter out this noise? A low pass filter like they are suggesting?

The datasheet mentions:

4.3 Layout Considerations
The MCP9700/9700A and MCP9701/9701A family
does not require any additional components to operate.
However, it is recommended that a decoupling
capacitor of 0.1 Î¼F to 1 Î¼F be used between the VDD
and GND pins. In high-noise applications, connect the
power supply voltage to the VDD pin using a 200Î©
resistor with a 1 Î¼F decoupling capacitor. A high
frequency ceramic capacitor is recommended. It is
necessary for the capacitor to be located as close as
possible to the VDD and GND pins in order to provide
effective noise protection. In addition, avoid tracing
digital lines in close proximity to the sensor.

What about the idea of using a linear regulator and supply a lower voltage to the sensor?

Well, I got the sensor installed in the unit. I couldn't do the underneath plan you suggested DocJC because the unit height has to do with aligning the interface pins to the main board outside of the unit.

I was all ready to drill the outer shell of the Rb module when I figured out I can just make a small aluminum strap like your picture above and use the interface hex screws to hold it down. They don't mate to anything in the main board and worked perfect for this.

So far the sensor is working very well. As the unit warms, the sensor is maintaining millivolt precision which equates to 0.05128 deg C...

Thanks,

Alan

alank2:

You wrote:

"So far the sensor is working very well. As the unit warms, the sensor is maintaining millivolt precision which equates to 0.05128 deg C..."

From which I assume you mean the sensor's output has a resolution of 0.05128 deg C per millivolt. Is that assumption correct?

Are you saying that you are getting a reading from the sensor that remains steady with no variation of more than 1 millivolt? And therefore the temperature of the RB is remaining steady within 0.05xxxx deg C?

Hi Chuck,

No, the Rb it isn't stable, it is always moving up or down, what I'm saying is that the sensor doesn't jump around, but moves from one millivolt to the next smoothly. I guess I should say it isn't noisy. Should be nice to bring in on an ADC and average/smooth.

Thanks,

Alan

Omega has these really small thermocouples and potting compound to adhere the sensor to surfaces.

Another option would be to use an IR temp sensor like this one; http://www.pololu.com/catalog/product/1061

But sure, they are an order of magnitude more expensive.

fleemy wrote:
Another option would be to use an IR temp sensor like this one; http://www.pololu.com/catalog/product/1061

But sure, they are an order of magnitude more expensive.

...and you need to know the emissivity of the surface that you are measuring, but at least knowing it, and if it is a constant, you should be OK. It would not be all that good for general "any surface" temperature measurements because of the range of emissivities encountered.

Cheers,

Ross

Ross McKenzie ValuSoft Melbourne Australia

Can you expand on this Ross? I've noticed that some surfaces seem to want to take a decent IR temp reading and others do not...

fleemy wrote:
Another option would be to use an IR temp sensor like this one; ...
World's first single-chip digital IR MEMS temperature sensor (Texas Instruments, TMP006)

"Dare to be naïve." - Buckminster Fuller

alank2,

So, it seems you are not currently reading the temp sensor ouput with the ADC, but a DVM?

In "staedy-state" operation of the Rb what is the temperature range you are seeing with this sensor the way it is currently mounted? I'd expect less than one degree-C or two.

alank2 wrote:
Can you expand on this Ross? I've noticed that some surfaces seem to want to take a decent IR temp reading and others do not...

Alan, it is probably best to have "someone else" explain it better than me. My college physics is rusty, but the important "take away message" is that two different materials/surfaces will register different ir temperatures depending upon their emissivity values. Watch this youtube explanation for a real world example.

Cheers,

Ross

Ross McKenzie ValuSoft Melbourne Australia

Chuck-Rowst wrote:
So, it seems you are not currently reading the temp sensor ouput with the ADC, but a DVM?

I started with the DVM when I first hooked it up, but it is ADC now with an AVR PI control loop controlling the fan PWM.

Chuck-Rowst wrote:

In "staedy-state" operation of the Rb what is the temperature range you are seeing with this sensor the way it is currently mounted? I'd expect less than one degree-C or two.

What do you mean? With the same fan RPM?

Thanks,

Alan

alank2,

Yes, with the fan running and the unit situtated as it would be in actual use, encosures all butoned up,same mounting orientation as it will beused in, etc. In other words, typical operating environment. In this condition with the unit warmed up, say 30 minutes, what temprature swing are you reading on your sensor over a period of say 10 minutes? Does the temperature swing y 1 degC, 2 degC, etc. Does there seem to be any repetitive characteristics to the swing, or does it seem random?

You might want to take the same measurements without the fan, assuming the Rb will not run up to an out-of-spec temperature.