temp sensor

[zkt]

Whats the usual way to sense temps in the 150-200C range? All I find at the usual parts suppliers have a tempMax<=150C.
thanks

[Teo]

Hello,use thermocuple

[janni]

Easiest would be Pt1000. Comparatively easy to measure (two wire method should be accurate enough), easier to linearize than thermocouple.

[LGR]

Thermocouples are a nightmare to cold-junction compensate. Pt RTDs are somewhat expensive. National semiconductor has a huge variety of temperature sensors, some of which can go that high, and are much cheaper. Here's one, but it may not be the best choice for your application. http://www.national.com/pf/LM/LM95235.html I recomment searching National's site for more options.

[zkt]

Thanks guys Ive deceided to go with a type J thermocouple from Omega engineering because of the lower cost. Accuracy isnt all that important either. Now I suppose I will need a PIC with a A/D converter. By the way, this is simply an oven control. According to the spec sheet, the TC voltage in the range of operation is approx 8-11mv. Have never implemented an A/D so I guess its time to start reading up. The PIC F87 seems a good choice.

[LGR]

I hope you understand how thewrmocouples work. They don't measure temperature, they only measure differential temperature. You have to use special wiring, and you have to do cold-junction compensation, or they're worthless. And J is particularly bad, because one of the metals is iron, which rusts easily.

Thermocouples can be made to work, but don't come back here in a couple of weeks, asking why it didn't work when you just connected the TC wires straight to the millivolt amp (which you will also need; there isn't enough voltage to drive an A/D directly). It's not as simple as it looks.

FYI - J thermocouples are typically used in furnaces, where temperatures are typically in the 500-1000 C range. In the 100-200 range, the signal is rather weak.

[zkt]

Thanks LGR. I thot I had read up on the subject but apparently missed a few things. Can you recommend a tutorial, link or whatever on the subject?

[FRM]

A moot point, but you could utilize one of several 'all in one' t/c signal conditioning packages, which most generally have a digital interface. Some have SPI, cold-junction comp and more besides.
Off the top of my head - a MAX6675 from www.maxim-ic.com uses type k as input. Not the cheapest, mind. There are others that may be more suitable and cheaper. Worth an educational look if nothing more.

[LGR]

I can't recommend anything on the net about this, because I've been working with this for decades. There are tutorials about thermocouples, that explain how the metals have to be coordinated, and how the cold juntion has to be compensated with a thermistor, but really, thermocouples are only suitable at this point for very high temperatures, such as in furnaces or in refineries. The RTD option is easier to work with, and more accurate as well, and roughly the same cost, but you still have to build a bridge. IIRC, there's specialized silicon out there for both, and if you're lucky, you can find a chip that will bridge an RTD and do the A/D conversion all in one chip.

Depending on what you're doing, the National sensor may be a better option. If it's an industrial process where you need to put a probe in a thermowell, use an RTD. If it's an OEM application, use the integrated sensor. You're going to have to make that call, because I don't know what you're trying to do.

[janni]

The cheapest sensor is KTY84-xxx. Works up to 300 degrees Celcius. You would have to calibrate it, though, as the initial accuracy is not high.

[LGR]

Janni - that's a good choice if you're doing an OEM application where that kind of package can be used. That's why I say it's very much dependent on the nature of the application. Calibrating individual temperature sensors in a production item may end up costing more than using Pt RTDs (which don't need to be calibrated). And that item may also be hard to mount.

Another consideration is transient response. If fast response is needed, that's a good sensor, provided it's directly exposed. But ultra-low mass sensors are better. You really can't properly select a sensor without looking at the application in it's entirety.

[janni]

Hi LGR,

You're right. I've just passed some info and zkt will have to do the choosing. That's the way forums work. Quality of the answers depends on the question. That is also why I didn't go into details - not enough data to work with .

[LGR]

As long as it's not under pressure (and thus needs to be sealed), I'd use the diode Janni suggested, mounted bare, so that response time isn't more than necessary. The longer the response time, the greater the temperature swings. If the distance between the sensor and the PIC is significant, you probably want to mount the amplifier near the diode. You also may want to shield the wiring, and/or use a dsPIC to filter the 50 hZ and harmonics if that gets to be a problem. Otherwise, that sounds very doable. If the temperature swing is excessive, it's reladively easy to do PID control of an electric heater using PWM or phase angle control, but you don't want that complication is it can be avoided.

If you have access to a mechanical pilot setup, you can answer all of these questions experimantally. If not, I don't know what to tell you. It should work well enough with on-off control, but no guarantees.

[zkt]

Here is the concept. A small sample of a material is placed on a small brass screen about a quarter inch in diameter. A resistance wire heats the air under rhe screen. A temp sensor above the screen monitors the temp of the convection air passing thru the sample . When a user defined temp in the range 300-400 F is reached a led is activated and the PIC switches the heater on-off holding the target temp. The sample screen is surounded by a copper tube which insulates the sample from the ambient air convection curents and helps retain heat. An accuracy of + or- 10 degrees F is good enough. The target temp only has to be held for a few seconds. Simply cutting the power to the heater would probably good enough but needs to be determined experimentially. Therefor the transient response is not all that critical. This is not a precision device.
Hope this helps.

[LGR]

OK. As long as you have some significant mass in the system (effectively thermal capacitance), that should work great. Your first description sounded like there was flow directly from the heater to the sensor, with no thermal capacitance between. That would have cycled like mad. What you're doing should work just fine.