Recomendations for Temperature Sensor for Custom Thermostat?

"Norm Dresner" wrote in news:sfZib.1215$ snipped-for-privacy@bgtnsc05-news.ops.worldnet.att.net:

I'd use thermistors with a Cygnal 8051 w/ ADCs from

The present "cute game" for temperature sensors is to use a diode or diode-connected transistor and drive it with a square-wave of current with minimum value 1*Io and maximum value 10*Io (your choice of Io); thus the square-wave voltage (AC-term) that results is kT/q*ln(10) (T on °K)

Thus you don't have to calibrate the device ;-)

...Jim Thompson

I am impressed.

I like the Analog Devices AD590. It is a current device so voltage drop when used remotely is not a problem. Go to their website for info and check out application note AN-273. Incidently, the resistor values in the app note for the voltage divider appear to be wrong.

Doug

use

predicting

use

predicting

I've used planar epitaxial diodes as temperature sensors with good results. They have a nearly linear voltage vs. temperature coefficient over a reasonably wide temperature range. The only problem with them is you would need some amplification to get the voltage up to a useable range and they should be driven by a constant current source.

You're not alone. I like "cute" in electronics.

While I like the method that Jim mentioned for more sloppy no-cal temperature control, I'd hesitate to recommend it for HVAC systems because it's not quite accurate enough. It also has a fairly low output voltage of a couple of hundred uV/K which requires a bunch of compenents unless you're designing an ASIC. Jim:- one manufacturer claims *automatic* compensation for rb (?) the base resistance, which as I understand it is the most important error term. Do you know how they are doing that?

Instead, I suggest using a precision 1% thermistor. It will be much more accurate than most other types of cheap sensors at room temperature (which is pretty much the only temperature that counts in the HVAC field). It's also cheap and low system component count if you ever decide to make a zillion of them. It requires no absolute voltage reference, only a reference resistor.

P.S. This is the type of sensor you'll find in virtually every piece of consumer electronics with a temperature display, from indoor/outdoor thermometers to fever thermometers. The technology was refined by the Japanese to create all those markets.

Best regards, Spehro Pefhany

Maxim has some nice new chips (MAX6642 and MAX6665) that can use an external diode as the sensor. They have an SPI interface.

Leon

-- Leon Heller, G1HSM Tel: +44 1424 423947 Email: aqzf13 at dsl dot pipex dot com WWW:

You might want to consider the Maxim (Dallas) 1-Wire temperature sensor, which can simplify wiring of multiple remote sensors:

I've used the National LM34 series for about 20 years(ugh, time flies) in hundreds of remote energy control systems. Have never had a failure yet. Those Dallas 1-wire parts look nice,providing you have a micro( my system is discete 4000 series CMOS) Be aware of RF,EMI,etc as possible 'noise' sources. hth jay