Temperature Sensors for Touch Sensor Pad Application

[Lagomorph]

Hi,
I'm trying to make a touch sensor pad using a grid of temperature sensors that would be covered with a thin, heat-conducting rigid sheet. I need these sensors to be able to pick up the temperatures on the sheet above them when the sheet is touched with the fingers and give the readout to the microcontrollers that will be polling them. This will mostly be used in room remperature conditions.

Do any of you know of a good temperature sensor for this kind of application?

Thanks in advance.

 

[Danger]

How discriminatory does this thing have to be? Most touch pads that I know of work by capacitance.

 

[Lagomorph]

I want this thing to accept multiple inputs, so I don't think I can use capacitance.

What do you mean by discriminatory? As in how fine do I want the grid of coordinates that it outputs? I figure that multiple sensors should detect a temperature change around the area on the top sheet that was touched. Then, I should be able to find the weighted average between all the readings and find a coordinate anywhere between the sensor positions, as precisely as I want/can.

If you mean what kind of resolution I'm looking for in the sensors, I'm not entirely sure. I don't know how much heat a touch can produce. I'm a bit worried that the sensors would have trouble picking up quick touches by cold hands. Anyway, I figure it would need to pick up pretty small fluctuations in temperature pretty precisely, so I'm not sure. I know I don't need a large range, but I don't know any specifics. Hence, I'm seeing if anyone has any suggestions.

 

[brewnog]

Why does it need to use temperature as a signal?

 

[Danger]

I was primarily asking about 'resolution'. Even with weighted averages of multiple sensors, I doubt that you could get much more accurate than 1/32" or so. The other problem that I can see is that it would have to measure a temperature change, not just an increase. It's entirely possible (not likely, perhaps, but possible) that the ambient temperature might be higher than that provided by a touch. If you're using it outdoors in Las Vegas, for instance, you would have to detect the coolness of the finger.

 

[FredGarvin]

The easiest, I think, would be thermocouples. They are simple and relatively inexpensive, plus the circuitry is pretty well understodd and widely available. Bare wire TC's are pretty responsive and could be tack welded to the bottom of your rigid plate. It will not be an instant reading, bit I can't think of any temperature sensor that operates that fast.

 

[Cyrus]

This sounds very tricky. You are going to need to find something that has a high thermal diffusivity. This is the ability of a material to respond quickly to changs in thermal environment. If you're not mindful of this, when you touch the foil it is going to take a long time before the temperature changes. And afterwards, its going to take a long time for the temperature to become that of the free stream environment once again.

If it's something you want to give a quick touch-touch-touch, your going to need very low time constants on your sensor.

 

[chroot]

The heat capacity of even small bare-silicon thermal sensors is very large; coupled with the relatively low thermal conductivity of human tissue and plastics, you're looking at large time constants. This concept is flawed. I suggest you look into capactive touch sensors instead.

- Warren

 

[Cyrus]

I was primarily asking about 'resolution'. Even with weighted averages of multiple sensors, I doubt that you could get much more accurate than 1/32" or so. The other problem that I can see is that it would have to measure a temperature change, not just an increase. It's entirely possible (not likely, perhaps, but possible) that the ambient temperature might be higher than that provided by a touch. If you're using it outdoors in Las Vegas, for instance, you would have to detect the coolness of the finger.

This is another very good point, appart from the time constant, resolution is going to be a significant issue. If you want to detect a rapid touch, you are going to have to measure a very very small change in temperature. I.e., small local change in temperature where the sensor is placed, and thus high resolution with a low threshold.

This was a very nice post Danger.

 

[FredGarvin]

Where in the OP did it mention anything about time response as a design criteria?

 

[Danger]

This was a very nice post Danger.

Thank you, cyrus; I consider that an extreme compliment coming from someone with your knowledge.

Fred, the response time wasn't mentioned. It seemed to be implied that it should be quite rapid (within a second or so). Nobody wants to stand there with his finger on a particular spot for a couple of minutes while the system sorts things out.

 

[Cyrus]

Thank you, cyrus; I consider that an extreme compliment coming from someone with your knowledge.

Fred, the response time wasn't mentioned. It seemed to be implied that it should be quite rapid (within a second or so). Nobody wants to stand there with his finger on a particular spot for a couple of minutes while the system sorts things out.

Your brain is a dictionary of scientific terms, you know that? You always seem to use the precise terminology. It is impressive. I myself need to work on that skill.

 

[Danger]

You do just fine, yourself. Don't sweat it.
I simply try to use the most applicable terms to cut down on confusion. Sometimes the terms that I use are in conflict with professional usage, which has resulted in misunderstandings a couple of times. I always try (not always successfully) to remember the error and not repeat it. I guess it's just a hold-over from my professional science/science fiction writing days. A newspaper or book audience doesn't have the luxury of requesting clarification, so I tried to minimize ambiguity.
It can have some really weird consequences. When W and I were at the hospital a few weeks ago when our niece's baby died, the nurse/counsellor thought that I was a doctor because of the questions that I asked. She was a bit weirded out to learn that I never finished high-school.

 

[NoTime]

Just a thought.
You might want to consider an IR sensor like those used in motion detectors.
IIRC, the sensor is an array with a lens system to focus the image of the covered area on it.

 

[Danger]

That's not a bad idea at all, NoTime. I can still see problems with it, but not as many as other techniques would have.

 

[NoTime]

Thanks, Danger.
Looks like a good reserch problem.
While not conclusive, I don't know of any off the shelf solution here.

 

[FredGarvin]

Another approach could be a set up that uses multiple Peltier circuits. In the "untouched" mode the controller regulates the temperature of the board, call it the main reference temperature, by applying a specific current to each junction. Then, once the plate is touched, the controller measures the current differential required at those junctions being warmed by the hand to try to keep the plate at the reference temperature. The increased current at each location would then correlate to temperature rise.

 

[Danger]

Another good idea, Fred.
The main thing that's really bugging me about this is the thermal conductivity of the cover plate. You need high conductivity to get the heat to the sensor, but that same feature will spread the heat laterally enough to blur things severely. I can think of only two ways around this.
One would be to have the cover made of insulative material with a 'well' full of high-conductivity material over each sensor element. (Maybe heat-sink paste in an aluminum capsule?)
The other way, and I'm pretty sure it doesn't exist, would be a sheet of 'thermally-polarized' material. By that I mean something that conducts heat in one direction, but not perpendicular to it. Is there such a material?

 

[NoTime]

I've seen aligned metal wires in a plastic matrix.
Wasn't intended for heat transfer(too thick), but sounds like what you are thinking of.

 

[Danger]

Yeah... that's something like it. All that comes to mind right now for making one at home would be to get a bunch of thin (1/2mm or so) aluminum disks, perhaps cut from a 1mm wire. Set them in a pan with something like soft wax on the bottom to keep them in position, then carefully pour 1/2mm of liquid plastic such as Casolite around them. They'd need to have surface imperfections around the sides, though, to give the plastic something to bite to so they won't be popping out all over the place. I think that it would be quite a nightmare to produce, and probably wouldn't work all that well anyhow.

 

[Lagomorph]

I've been away for a bit and I'm really thrilled to see this much discussion.

It was really an idea I just started thinking about. At first I looked into light sensors and most information that I found suggested that those might have too many problems with ambient noise. All the pressure sensors that I've managed to find are too expensive. So I figured why not try heat?

I don't know what kind of temperature the human hand tends to be at and what the difference between cold hands and warm hands is. In other words, I don't really know what kind of temperature changes this thing would need to detect. I suppose if the ambient temperature caused the heat-conductive sheet to be at the same temperature as that of the user's hand, then there would be an extremely small reading. Also, I've considered that heat changes are not as instantaneous as light or pressure, so I don't know if it is even possible to find a combination of material and sensor to pick up quick touches that would definitely be necessary for a pad like this.

Essentially, I don't know there are sensors or materials that would allow for a thermal multi-input touch pad like this. If there is another type of sensor that could do this, I am completely open to suggestions. Perhaps I'm too attached to this grid of sensors idea. Maybe there is another way to simulate the same type of input.

 

[NoTime]

There are touch pad sensors imbeded in keyboards and available as separate units.
They handle touch as well as position.

So why won't those work for you?

 

[Danger]

Again, those are capacitance-measuring units. Very effective. If, for some reason, Lagomorph still can't use those, another idea just crossed my mind. Again, it would be a bugger to construct. The advantages would be that the resolution would be far higher and it should be immune to extraneous signals.
What I'm thinking is to embed pairs of fibre-optic strands in a matrix, like a toothbrush, but fully encased in the plastic except for a tiny hole over each pair. Strand 'A' of every pair goes to a common IR LED. Every Strand 'B' goes to an individual photodetector. When a finger gets within contact range, the IR from Strand 'A' is reflected into Strand 'B'.

 

[Darryl]

i would use two solid state temperature sensors in a "nulling" circuit, one as your sensor the other the reference. as long as ambient temp was not very very close to body temp. a circuit like this with a comparator would have a very quick response LM335Z, scaled to body temp range with a single resistor. TO-3 transistor backage, or metel case. connected to a multimeter will show you their response is almost instant when held between your fingers.

mounted on a small low thermal mass metal plate, i think would meet your needs.