Post by thread:[OT:] Temp sensor recommendation please

I need to measure temperature inside a circuit enclosure box. I have digital and analog 0 - 5V inputs, serial port or SPI available, and would welcome pulse width measurments too. Temp sensing speed is not critical. I would like something that is easily obtainable, not nearing obsolesence and under $2.00 US. Resolution of 1 degree F would be fine. I have been working with thermisters but the hysterisis is unacceptable. Any recommendations? Thank you.

In message <001c01c52dca$a7f62fa0$09313fd1@xero> "John Pearson" <spam_OUTxeroTakeThisOuTcmc.net> wrote: > I need to measure temperature inside a circuit enclosure box. > > Any recommendations? Dallas DS18S20? 1-Wire bus, -55 to +125 deg. C measurement range, 0.5% accuracy between -10 and +85 C, 9-bit resolution, 750mS measurement time, 3V to 5.5V power. The datasheet is on <http://www.maxim-ic.com> somewhere. I've got some code that does 1-Wire comms on a 16F PIC - ask me nicely and I'll send you a copy :) Later. -- Phil. | Acorn Risc PC600 Mk3, SA202, 64MB, 6GB, .....philpemKILLspam@spam@philpem.me.uk | ViewFinder, 10BaseT Ethernet, 2-slice, http://www.philpem.me.uk/ | 48xCD, ARCINv6c IDE, SCSI ... Complaints? Write them here legibly [] <-

>Temp sensing speed is not critical. I would like something that is >easily obtainable, not nearing obsolesence and under $2.00 US. > >Resolution of 1 degree F would be fine. I don't know what the price is, but I would have thought a likely starting point would be the LM35 series sensor from National Semiconductor. They do a Fahrenheit version (LM37?) which is 10mV/F, unless you are specifically looking for something with digital output. >I have been working with thermisters but the hysterisis is unacceptable. What hysteresis? This suggests to me that you have a large thermal resistance between the sensor and the temp source which gives the appearance of hysteresis.

>>I have been working with thermisters but the hysterisis is >>unacceptable. > > What hysteresis? This suggests to me that you have a large thermal > resistance between the sensor and the temp source which gives the > appearance > of hysteresis. ie A temperature stabilised thermistor has NO hysteresis at all (for practical purposes). It's resistance is a function only of temperature. If you are measuring before it has stabilised then the result will differ ascending and descending. This is not hysteresis in the thermistor but thermal time constant. A possible method if you have uP A2D available is to use a standard silicon diode and measure the voltage drop at two known currents. The currents can be eg varied by a PIC pin. The differential voltage is an extremely accurate function of temperature AND is device independent so needs no calibration. (Depends only on nature of diode (silicon, germanium etc)). RM

From: "John Pearson" <xeroKILLspamcmc.net> > I need to measure temperature inside a circuit enclosure box. You don't specify the range, but I assume its <125C. > I have digital and analog 0 - 5V inputs, serial port or SPI available, and > would welcome pulse width measurments too. > Temp sensing speed is not critical. I would like something that is > easily obtainable, not nearing obsolesence and under $2.00 US. LM34 or LM35 offer analog outputs of 10mV/degree (C or F depending upon sensor model) Dallas DS1621 or DS1721 have I2C (2-wire or SMBUS) outputs giving a reading directly in degrees C. DS1821 is a 1-wire version of the same. LM75(also LM76 and LM77) are I2C. LM74 is SPI. > Resolution of 1 degree F would be fine. All of the above are higher resolution than that. Most are least .5C resolution, some are .0625C resolution. Accuracy is usually around 1C. > I have been working with thermisters but the hysterisis is unacceptable. > Any recommendations?

On 21 Mar 2005 at 7:56, michael brown wrote: {Quote hidden}> From: "John Pearson" <.....xeroKILLspam.....cmc.net> > > > > I need to measure temperature inside a circuit enclosure box. > > You don't specify the range, but I assume its <125C. > > > I have digital and analog 0 - 5V inputs, serial port or SPI available, > and > > would welcome pulse width measurments too. > > > Temp sensing speed is not critical. I would like something that is > > easily obtainable, not nearing obsolesence and under $2.00 US. > > LM34 or LM35 offer analog outputs of 10mV/degree (C or F depending upon > sensor model) Dallas DS1621 or DS1721 have I2C (2-wire or SMBUS) > outputs giving a reading directly in degrees C. DS1821 is a 1-wire > version of the same. LM75(also LM76 and LM77) are I2C. LM74 is SPI. > > > Resolution of 1 degree F would be fine. > > All of the above are higher resolution than that. Most are least .5C > resolution, some are .0625C resolution. Accuracy is usually around 1C. > > > I have been working with thermisters but the hysterisis is > unacceptable. > > > Any recommendations? > I used one PC motherboard thermistor as in this design: http://elm-chan.org/works/temp4/report.html It worked very well at .1C resolution for controlling my home air conditioning system. Mark Jordan

At 00.07 2005.03.22 +1200, you wrote: >>>I have been working with thermisters but the hysterisis is unacceptable. >> >>What hysteresis? This suggests to me that you have a large thermal >>resistance between the sensor and the temp source which gives the appearance >>of hysteresis. > >ie A temperature stabilised thermistor has NO hysteresis at all (for practical purposes). It's resistance is a function only of temperature. If you are measuring before it has stabilised then the result will differ ascending and descending. This is not hysteresis in the thermistor but thermal time constant. > >A possible method if you have uP A2D available is to use a standard silicon diode and measure the voltage drop at two known currents. The currents can be eg varied by a PIC pin. The differential voltage is an extremely accurate function of temperature AND is device independent so needs no calibration. Whoha.. any pointers to more data on this fascinating subject? >(Depends only on nature of diode (silicon, germanium etc)). > > > > > RM > >-

> A possible method if you have uP A2D available is to use a > standard silicon diode and measure the voltage drop at two > known currents. The currents can be eg varied by a PIC pin. > The differential voltage is an extremely accurate function of > temperature AND is device independent so needs no calibration. The only problems are that both the current through the diode and the ferenence voltage used to measure the drop must be accurate enough. Wouter van Ooijen -- ------------------------------------------- Van Ooijen Technische Informatica: http://www.voti.nl consultancy, development, PICmicro products docent Hogeschool van Utrecht: http://www.voti.nl/hvu

On Monday 21 March 2005 06:07 am, Russell McMahon scribbled: > A possible method if you have uP A2D available is to use a standard > silicon diode and measure the voltage drop at two known currents. The > currents can be eg varied by a PIC pin. The differential voltage is an > extremely accurate function of temperature AND is device independent > so needs no calibration. (Depends only on nature of diode (silicon, > germanium etc)). Hmmm... interesting. And is that "extremely accurate function" linear? Any pointers to more info on doing this? Thanks, -Neil.

On Monday 21 March 2005 07:56 am, michael brown scribbled: > From: "John Pearson" <EraseMExerospam_OUTTakeThisOuTcmc.net> > ... > LM34 or LM35 offer analog outputs of 10mV/degree (C or F depending upon > sensor model) Dallas DS1621 or DS1721 have I2C (2-wire or SMBUS) > outputs giving a reading directly in degrees C. DS1821 is a 1-wire > version of the same. LM75(also LM76 and LM77) are I2C. LM74 is SPI. > > > Resolution of 1 degree F would be fine. > > All of the above are higher resolution than that. Most are least .5C > resolution, some are .0625C resolution. Accuracy is usually around 1C. How do you define resolution for analog-output sensors such as the LM34, LM35? Cheers, -Neil.

----- Original Message ----- From: "PicDude" <picdude2spam_OUTnarwani.org> > On Monday 21 March 2005 07:56 am, michael brown scribbled: > > From: "John Pearson" <@spam@xeroKILLspamcmc.net> > > ... > > LM34 or LM35 offer analog outputs of 10mV/degree (C or F depending upon > > sensor model) Dallas DS1621 or DS1721 have I2C (2-wire or SMBUS) > > outputs giving a reading directly in degrees C. DS1821 is a 1-wire > > version of the same. LM75(also LM76 and LM77) are I2C. LM74 is SPI. > > > > > Resolution of 1 degree F would be fine. > > > > All of the above are higher resolution than that. Most are least .5C > > resolution, some are .0625C resolution. Accuracy is usually around 1C. > > > How do you define resolution for analog-output sensors such as the LM34, LM35? Just like the datasheet does: 10mV/degree

michael brown wrote: >> How do you define resolution for analog-output sensors such as the >> LM34, LM35? > > Just like the datasheet does: 10mV/degree That's not resolution, more like scale. The resolution of an analog sensor is in theory infinite, although added noise will bound the useful range. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com

On Tuesday 22 March 2005 04:47 pm, michael brown scribbled: > > How do you define resolution for analog-output sensors such as the > > LM34, LM35? > > Just like the datasheet does: 10mV/degree I read that as an output function, rather than resolution. They could've also stated 1V per 100 degrees to mean the same thing. I've used the LM34/35/335, etc and you can get more resolution than that. Depends on the accuracy of the A/D, vref, etc. Cheers, -Neil.

In another life and long ago I worked in a research envrionment. That was before factory made thermocouples were common place. We made ours up from thermocouple wire. Chromel-Alumel was favored but Iron-Constantan had higher output. to form the bead at the business end, we used a small torch. Or we welded them with a big capacitor/knife switch arrangement. I believe they worked from 0degreesK until they melted. If you went to the trouble of properly rferencing them to a cold junction(melting ice), calibrated them against a standard, and read them with a balanced bridge there was no more accurate measurement to be had. Or cheaper! John Ferrell http://DixieNC.US {Original Message removed}