Monitor temps

[faultman]

I want to monitor the temperatures of up to 100 different locations and record the temps once or twice a day. I can easily hard wire thermisters from each location to a central point and I can dedicate a PC for data acquisition to this project. Is there an off-the-shelf device that could monitor that many inputs? Does anyone have an idea of how it could be done? Thanks,<p>Frank

[bodgy]

Don't know about off the shelf products with that many nodes, but it wouldn't be that hard to impliment.<p>Basically with that many inputs, you'd be really talking about a uP to log and convert the temperatures or as you've got a PC a multplexor or in other words a cascaded X to 8 (if using the LPT on the PC), and then the PC would query each unit for the temperature.<p>This could be made a lot easier (at cost) if you changed from thermistors to dedicated addressable temperature sensors such as the Dallas/Maxim range.<p>Then all your software would need to do is send out the address of the sensor to be read, perform the calculations, store/display and then go on to the next sensor. You could also do somehting like this with the thermistors, but the interface circuitry may or may not be cost effective -I haven't looked into that.<p>If the locations are far away like 10ft or more, it may be better to impliment the sensors as a current loop instead of voltage. You could also perhaps use AM of FM ready made modules in conjucnction with something like a Holtek encoder IC to transmit the data. In fact you could use one of those even if you had everything hardwired and bingo you have your addressable sensors.<p>Just some ideas.<p>Colin

[Chris Foley]

There are a number of ways to do this, Frank. But I'm not sure that thermistors are the way to go here, because a thermistor is a resistor whose resistance is dependent on temperature. If you use a standard two-wire connection to the thermistor, you have to take into account the resistance of the wires. Copper wire resistance is also temperature-dependent, which adds a problem if you've got 500 feet of it, and degrades your precision. To do it accurately, you would need a kelvin (4-wire) connection to the thermistor, and you would have noise problems with the voltage sensing lines. I'm assuming the 100 sources are at some distance (say, up to a couple of hundred yards). Of course, if you already have the 100 precision thermistors, or just want to do it that way, it's a different story.<p>Two methods for measuring temperature at a distance are common -- the thermocouple, and the silicon temperature sensor. Thermocouples are widely available and relatively inexpensive. They produce a very small (microvolt-to-millivolt level) voltage which is dependent on temperature, and can measure from well below zero to over 2000C. This voltage can be read and interpreted (the voltage is very non-linear for all T/C types). When applying thermocouples, you have to be very careful to take into account the effects of using dissimilar metals (this creates competing thermocouples), which means you need to use thermocouple extension wire for all your wiring (costs add up here) and special connectors which have the same type of metals as the two wires in the T/C (more $$). This is starting to look like a fairly high cost solution here.<p>The second common method of measuring temperature at distance is to use silicon temperature sensors. A number of manufacturers make ICs with an output whose magnitude is linearly dependent on temperature. These work from about 0 to 100 or 125C. Of these, one good possibility is the AD590, an IC made by Analog Devices as well as a number of second sources. The IC outputs a current instead of a voltage, so voltage drops over long lines don't make any difference.<p>Unless you want to have 100 voltmeters, the way to read all the inputs is with a signal multiplexer, or using relays to switch the signals in to the meter. The thermocouple method has an extra layer of expense here. You have to have an isothermal block at the point where you go from thermocouple wire to copper wire and board traces (that means having the terminals guaranteed to be at the same temperature), and use precision reed instrumentation relays with guaranteed low thermal offset. You also need to measure the temperature at the isothermal block, and take that into account when calculating temperature if you want any accuracy at all. Coto makes these relays, and they're about $10 a pop. You go from there to the voltmeter, make the reading, and interpret it. Or you can buy the multiplexing/measuring system off the shelf. The AD590 solution can use standard relays. You would multiplex all the signals, using standard reed relays, to a precision resistor, and measure the voltage across that. Calculation would be easy, as output current is directly proportional to temperature.<p>Below are some liks that might give you a hand. If you want a thermocouple solution and decide to do the A/D conversion route, you'll need to have the conversion equations for the thermocouple you're using. Ususally they're eighth or ninth order polynomials which achieve "best fit" for the thermocouple curve (note that many of the equations are only good for a portion of the range of the thermocouple). The NIST link can direct you to the equations you need. The second link is for Omega Engineering, which has everything you need off the shelf. Their free Temperature Handbook is a great source of technical information in itself, apart from being a candy store of tech toys. If you decide to go the AD590 home-brew route, you might want to look at the AD data sheet and an AD app note on how to use the AD590 for remote temperature measurement.<p> http://srdata.nist.gov/its90/main/
http://www.omega.com/
http://www.analog.com/UploadedFiles/Dat ... D590_b.pdf
http://www.analog.com/UploadedFiles/App ... 6an273.pdf<p>Before getting into specifics, you need advice on some hard questions: What's your budget, what degree of accuracy/precision/temperature range do you require, are there environmental constraints (e.g. water, corrosives, shock/vibration, &c), what equipment/materials do you already have available? If you need more help on this, feel free to e-mail me.<p>[ March 30, 2003: Message edited by: Chris Foley ]<p>[ March 30, 2003: Message edited by: Chris Foley ]</p>

[bodgy]

Just going on from what Chris said, standard silicon diodes are quite good at temperature sensing and are cheap!<p>They are sometimes used as a temp probe in some industrial ovens - if the they get too how and the glass or connections melt then they are cheap and easy to replace.<p>Colin

[Dimbulb]

A pic timer program counts to 100 and recycles every twelve hours. It uses the pic serial port to send sequential resistance samplings to an adc each count the pic advances the sequencing of 100 switching diodes that are connected to 100 thermistors. The ADC thinks it is one thermistor
The pc data file is generated giving a table of 100 sequential resistance measurements.<p>If you use 101 then the last one can be zero

[hlreed]

Here is a good way to hook up 128 analog wires.<p>Use 4066 analog switches in a tree structure. Two inputs to the 4066 go to one output. So each column of switches divides the input number by 2.
The 4066 has a control for every data bit.<p>4066 64 32 16 8 4 2 = 126 switches= 32 4066 ics
data 128 64 32 16 8 4 2 1 to ADC
control bits needed are one each column or 7
bits. 7 inverters are needed. I have not worked out the control sequence yet, but it should be doable. Select half of half of half of, etc.<p>Parts are temp. sensors, switches, inverters, ADC, MC, display or recorder. ADC and MC can be the same unit that provides input control.
If you like, I will do the control system for you.
You have 7 bits whose value will choose one out of 128 inputs. 2^7 is 128, so this will work.<p>[ April 01, 2003: Message edited by: Harold ]</p>

[Will]

I think your best bet is to use silicon diodes. They have a temperature coefficient of about minus 2.0 mV per degree C. You might want to check calibrate individual diodes if you want reasonable accuracy - that isn't too difficult, just tie groups of them together, immerse them in ice water and boilinf water, stick about 1.0 mA through them, measure individual diode voltages and then label and identify diodes. You can then store the calibration figures in your micro. The switching arrangement which Harold proposed would work OK but you might want to take account of the volts drop across the 4066's as well. You could prpobably do all of the above by setting up the micro and then using it to calibrate the overall system taking account of both lead resistance and 4066 volts drops.
Don't go near thermocouples they will eat into your bank account - even if you made your own couples which is fairly simple. The real problem is that whilst you don't really need connection terminals made of t/couple materials (Unless the terminal blocks are so large that you can't assume even temperatures across whole devices) but the required cold-junction compensation would give you difficulries because either your multiplexing device would have to be constructed from t/couple materials (Because you probably couldn't assume uniform temperatures across the whole device) or you would have to effect cold-junction compensation for each of your 100 channels - Which you would not like !
Cold-junction compensation is required because, where you terminate the thermocouple leads (Which also have to be manufactured from thermocouple or similar materials) you are effectively connecting copper or brass etc into each channel, which, while the copper or brass does not present a problem, means you have essentially connected into the circuit an additional thermocouple which measures the temperature of the instrument terminals, injects an equivalent voltage in opposition to the measuring voltage at the hot junction and will therefore reduce the (hot) signal by approximately the terminal temperature equivalent. The cold-junction compensation consists of adding the lost volts signal back into the circuit. That also is fairly simple, but not if you have to do it 100 times.
Hope this helps you Will

[haklesup]

For off the shelf, two companies immedietly come to mind. Omega and National Instruments.<p>Omega is big in the temp sensing field. A quick look at one of their excessively thick catalogs shows a model OMB-DAQ-56 can be configured with options for 120 inputs for around $3k. It's USB, comes with SW and is very expandable (to 8000)and flexible (other types of inputs also).<p>National Instruments is big in PC based data acquisition. They have many offerings also. The AMUX-64T can be cascaded for up to 128 inputs and works with a DAQ card and the labview software which can be integrated with all sorts of other instruments for about the same price.<p>Just about anything will require some user configuration, that's the nature of the data acquisition business.

[Bob Scott]

Frank,<p>Yes there is an off-the-shelf solution. We use a device called a Watlow Anafaze in our ESS (Environmental Stress Screening) lab. It has facilities for recording the output of 32 J-type thermocouples over just about any time period. The thermocouples are inexpensive easy to place "peel-and-stick" type. It comes with a Windows PC interface. You can easily transfer your data recordings to Excel data.<p>We use these in conjunction with programmable environmental test chambers that cycle electronics under test to various temperatures over a period of time.<p>Bob

[Will]

That sounds like a reasonable solution Bob Scott - What is the model # of the Anafaze devices and how much do they cost ? Are they all furnished with Automatic Cold Junction Compensation ?
Will

[Chris Foley]

The multiple-input scanner is a good idea, and gives you the best of both worlds, especially if you can bunch your inputs up in groups geographically. Unfortunately, Anafaze doesn't make a 32-input scanner anymore. Their 16-point scanner, the CAS200, goes for around $110 a point, with the base unit (RS-485 comm, terminal blocks, you supply the +24V power) setting you back about $1750. Here's the spec sheet on the scanner. Its basic accuracy is +/- 1.2 degrees C for J-type T/C. It does have cold junction compensation, but you should keep a thermal baffle around the rear terminal blocks for greatest accuracy.<p>http://www.anafaze.com/literature/specs ... as0502.pdf<p>
$100 a point is, unfortunately, pretty much a standard minimum for off-the-shelf thermocouple scanners. And faultman hasn't even bought any thermocouples, extension wire, or connectors (you need a connector of the same metal types to go from the thermocouple to the extension wire, unless you want to have straight runs from the measurement points to the scanners). If faultman can live with input temperatures between -25C and 125C, the temperature-to-current IC (AD590) could be a more economical way to go. These are based on the temperature-dependence of the forward voltage drop of a silicon diode, but have additional circuitry to compensate for non-linearities and hysteresis, and a circuit to convert the voltage to a current (1 microamp per degree K). These go for between $7 and $100 each in single quantities. You can approximate the degree of accuracy and precision of the higher-priced spread by doing a software calibration for each of the $7 sensors, and then fitting the curve.<p>One of the neat things about this setup is that it's fundamentally hacker-friendly, and it has a low installed cost for the accuracy you get. I think that, with 3-point software calibration of each sensor, you could be looking at less than $700 for 100 sensors with calibrated accuracy of +/- 1.5C. Now, get a 24 reed relay I/O board for your PC (available from several sources in N&V), a 4-2/3 digit DMM with RS-232 comm, a +12V linear open frame P.S. with OVP, a precision 1K wirewound resistor(0.05% preferred, 0.1% O.K.), a handful of C-J 11-point terminal blocks, and a mess of 11- or 12-conductor wire along with a mess of 2-conductor, and you're in business.<p>Your 12VDC Linear supply is going to power all of these sensors, one at a time. You're going to use 20 of the relays as a multiplexer to do this. Put a 22 ohm 1/4W carbon film resistor in series with the + of the PS, and spread that to relay 11-20 common. Relay 1-10 common goes to the 1K ohm precision resistor. The other end of the precision resistor goes to the P.S. common. Now each of your 11-conductor cables is going to have a + feed from the power supply (one of RY 11-20), which will go to all 10 sensors hooked up to that cable, and one - wire for each of the 10 sensors (RY 1-10). Use two C-J terminal blocks at each junction to spread the signals from the 11-conductor wire to 10 two-conductor wires. So, at the computer, you should provide terminals so each of the + outputs goes to its own 11-cond. cable, and so that all negative wires #1 are connected together and go to relay 1, all negative wires #2 go to relay 2, &c up to relay 10. This multiplexing method (it's easier to draw than to write) allows you to switch 100 inputs with 20 relays. Just measure the voltage across the precision 1K resistor, and convert to temp (1mV = 1 degree K). Scroll thru the relays, and you have your 100 measurements. Wait several seconds after switching relay contact for the DVM measurement to stabilize (but since you're only measuring twice a day, time is on your side here).<p>A couple of details. Use small pieces of aluminum heatsink (at least 3/16" thick) to mount the metal-can IC package. Place a small resistor (47 ohms) in series with the IC, and a small ceramic cap (around 220 pF) across the IC to help with transients. More complex protection schemes (e.g. bridge rectifier, low-leakage crowbar) can be used, but watch for leakage currents. Use small gauge wire at the IC connect to reduce heat flow to the IC from the wires, and thermally insulate the back of the IC. Keep everything electrically isolated from ground to avoid ground loops. For the 2-conductor wire, use something distinctive, so a maintenance person won't think it's a phone line or something. If you have one of those $169 handheld meters, you should use an RS-232 isolator (the RS-232 on those isn't isolated from the measurement -- avoid ground loops, microamps count), and use one of your spare relay contacts to switch the battery power on for the meter, so it's only on when you need it. Use the software provided with your meter to write the data collection program in your language (if you don't need bells or whistles, you could have this running in an afternoon, along with a reasonable display and data collection to file on HD or floppy). If you want to spend more, there's a number of commercial packages available to gather data from meters.<p>That's one of many possible hacker solutions. The advantages are limited cost, fair precision, and no need for special hardware. The biggest advantage of this setup is that the signal is robust, and can handle plain reed relay contacts, the voltage drop of 500 feet of 22AWG wire, thermoelectric potential differences from dissimilar metals, and a fairly wide range of input power. The main limitation of this solution is that you should probably beef up the protection for each individual AD590 (doable with a small circuit at each sensor/heat sink). The AD590 is rated to operate to 30V, but transients happen. A second limitation is that this isn't a "store-bought" solution, and in many work environments, that's the only acceptable one. If you need to just buy something, replace it if it goes bad, and call in the vendor tech rep if you can't get it going, this solution isn't for you. Your total hardware cost for this hack (excepting wire) is probably close to one of those controllers alone, less than 2K. And you can use plain-Jane copper wire instead of T/C wire for the runs to the scanners. If you've got a well-stocked junk box or have any of these parts handy, probably less. Again, if you need help, feel free to e-mail me. Happy hunting.<p>[ April 04, 2003: Message edited by: Chris Foley ]</p>