At 08:22 AM 4/23/2013, microsoftwarecontrol wrote: >max6675, TC to SPI, > >best solution says that the Maxim max6675 costs around $14 in singles, dropping to around $8 in hundreds. I would be needing to use them in hundreds or thousands, so lets use that number. You would need one of these for each thermocouple input. Contrast that with the LT1013 - its about $1 in hundreds (from TI). The extra components (two resistors, two capacitors) add another few cents. Then add in the cost of the cold-junction compensation network: either 3- 1N4148 diodes and a resistor *or* a 2n4401 transistor and 3 resistors - this cold-junction circuit is common to all thermocouple inputs on the card. That means the total cost to do one thermocouple input is perhaps $1.10 if you are doing only a single input - but the cost drops to about $0.60 per channel if you are doing two or more thermocouples on the c= ard. But lets be honest here: I don't need extreme precision for my applications. I'm perfectly willing to settle for accuracy of a couple of percent in exchange for low cost. I mentioned above that I use either 3- Si diodes in series *or* a BJT and 3 resistors for cold-junction compensation. Using a BJT with a two-resistor divider network feeding the base of the transistor allows me to tailor the slope of the temperature compensation so that one LSB of the 10-bit a/d equals 1 degree C. It just makes for easier math, that's all. You still have to read and save the offset into eeprom to get an accurate result. I do that using room temperature (which I know and control) at first power-up. dwayne >----- Original Message ----- >From: "Dwayne Reid" >To: "Microcontroller discussion list - Public." >Sent: Monday, April 22, 2013 3:17 PM >Subject: Re: [EE] How to measure higher temperatures? ( was: Besttemperatu= re >sensor ) > > > I'm still fond of thermocouples for measuring temperatures higher > > than semiconductor sensors can tolerate. Circuits are simple - I > > have two standard circuits that I make use of. > > > > One project (group of projects, actually) that I did uses a single > > op-amp stage to amplify the thermocouple to a level that a PIC a/d > > could make use of. Cold-junction compensation is done with a chain > > of 3- 1n4148 diodes in series and fed into another a/d input. Very > > simple and very effective. > > > > The other thermocouple amplifier that I use is actually a > > thermocouple transmitter - it takes the signal from a "K" type > > thermocouple and converts it to a current loop signal. Cold-junction > > compensation is derived from a single 1n4148 diode right on the > > board. I've built several thousands of these over the past couple of > > decades - they are used in our Catalytic Heater-based Industrial Oven > > control systems. The earliest version of these transmitters used a > > 4-20mA loop but all are now configured for a 1-5mA loop > > instead. There are just too many transmitters in a large system and > > the preamp needs only a few hundred uA to operate - it didn't seem > > reasonable to waste all that electrical energy as heat. > > > > But I digress. > > > > If you want to build a simple TC preamp, use an op-amp with low > > Vos. I used a LT1013 running from my unregulated rail of 15V (the > > LT1013 isn't R-R). The op-amp is configured as an inverting > > amplifier with a gain of 200. The (+) input is grounded, the input > > resistor on the (-) input is 1k0; feedback resistor is 200k with a > > 100n cap across the resistor (we want a slow response). The > > thermocouple is connected with (Y) lead grounded, (R) lead feeds the > > 1k0 input resistor. Note that means that the TC is providing a > > negative signal which is then inverted by the amplifier stage. There > > is also another 100n noise suppression cap at the junction of the TC > > (R) lead and the 1k0 input resistor. > > > > I chose the LT1013 for many reasons: its low cost if purchased from > > TI (the Linear Technology version is stupidly expensive, the TI > > version is darned reasonable); the input includes Gnd within its > > common-mode range; it has a very low Vos; it doesn't run hot. > > > > The LT1013 has two identical op-amps - you can process 2 separate > > thermocouples if you wish. For me - the thermocouples are used in a > > catalytic heater system and I used the > > extra op-amp stage as a comparitor as part of the hardware > > safety-shutoff system. > > > > Hope this helps! > > > > dwayne > > > > > > At 01:45 AM 4/22/2013, KPL wrote: > >>Could this (diode junction) be (cheapest | easiest) way to measure > >>temperatures up to about 200C? Looks like PT100 and similar need about > >>the same amount of analog circuitry to work, but those are more > >>expensive than diodes. > >> > >>I want to add more advanced regulator to my laminator, so I could use > >>it for PCB toner transfer, but could still use it for it's original > >>function as well. > >>Great precision is not required, but temperatures can reach close to > >>200 degrees celsium. > >> > >>I have no experience with any of these, as usually ds18b20 were good > >>enough until now. > >> > >> > > >> > Diode bandgap measurement does most of what you want and cost can b= e > >> > lowish if you work at it. > >> > > >> > Wikipedia covers basics: > >> > http://en.wikipedia.org/wiki/Silicon_bandgap_temperature_sensor > >> > > >> > If you sequentially apply two currents to a silicon diode the > >> > delta-voltage is a function of only temperature and the two currents > >> > and a constant. > >> > You can notionally swap sensors using any silicon diode and get the > >> > same result without recalibration - but using the same type of diode > >> > is better in practice. > >> > This is a substantially more accurate and generally superior method = to > >> > just measuring diode forward voltage drop. > >> > > >> > The method and the formula are not "hard" - just perhaps unexpected. > >> > > >> > The sensor is in theory ANY siliicon diode, swappable with out > >> > calibration or recalibration, but using say 1N4148 as standard will > >> > improve ease of getting good results. > >> > > >> > Any diode material can be used if due allowances made > >> > Si is good for up to about 200C with SiC being better above 200C > >> > > >> > Delta_Vdiode =3D KT/q * ln ( i2 / i1 ) > >> > > >> > or > >> > > >> > T =3D Delta_Vdiode / (k * ln (i1/i2)) > >> > > >> > ln may be annoying but is not intractable. > >> > > >> > Where > >> > K =3D Boltzmann's constant > >> > T =3D temperature in degrees K > >> > q =3D electron charge > >> > i1 =3D current 1 > >> > i2 =3D current 2 > >> > k =3D K/Q > >> > > >> > > Accuracy is .5C + or - .25 C > >> >> Range is - 20C to 75C > >> >> Long term stability > >> > > >> > Achievable, yes, yes. > >> > > >> >> simple interface such as i2c, SPI , voltage , current , resistance > >> >> ..... > >> > > >> > Needs two switched precision currents and suitably accurate voltage > >> > measurement. > >> > OR two non precisely produced currents which can be precisely measur= ed. > >> > > >> >> simpler is better > >> >> Low price > >> > > >> > Sensor is a silicon diode. Cost is a diode for the sensor plus a > >> > suitably accurate ADC (assuming digital) for voltage and two accurat= e > >> > currents OR a means of accurately measuring currents. > >> > eg imagine that you inject I1 then I2 into a series combination of D= 1 > >> > + R1 to ground. > >> > Measure VR1 and then V(D1 + R1) > >> > If R1 is stable and known then this allows measurement of 2 x curren= t > >> > and Vdiode with a single switched ADC input or two 'unswitched' ADC > >> > inputs. > >> > In this case current switching could be achieved with eg applicaton = of > >> > a dfigital high or low to a resistor network using relatively low > >> > precision resistors. > >> > With D and R in series the relative voltage drops of R and D can be > >> > adjusted to best use ADC range. > >> > > >> >> no external parts > >> > > >> > Deep ends what you mean by "external". > >> > Needs above parts. Can be integrated into whatever as required. > >> > > >> >> no calibration > >> > > >> > Yes. Neither additional or for sensor swap. > >> > > >> >> Can be slow ..... 30 second update .... can be as large as golf b= all > >> > > >> > Easily met. > >> > > >> > > >> > Russell > >> > -- > >> > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive > >> > View/change your membership options at > >> > http://mailman.mit.edu/mailman/listinfo/piclist > >> > >> > >> > >>-- > >>KPL > >>-- > >>http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive > >>View/change your membership options at > >>http://mailman.mit.edu/mailman/listinfo/piclist > > > > > > -- > > Dwayne Reid > > Trinity Electronics Systems Ltd Edmonton, AB, CANADA > > (780) 489-3199 voice (780) 487-6397 fax > > www.trinity-electronics.com > > Custom Electronics Design and Manufacturing > > > > -- > > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive > > View/change your membership options at > > http://mailman.mit.edu/mailman/listinfo/piclist > >-- >http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive >View/change your membership options at >http://mailman.mit.edu/mailman/listinfo/piclist -- Dwayne Reid Trinity Electronics Systems Ltd Edmonton, AB, CANADA (780) 489-3199 voice (780) 487-6397 fax www.trinity-electronics.com Custom Electronics Design and Manufacturing -- http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .