RTD help?

Huh?? Lead resistances are virtually equal; they add to the voltage on each terminal at the op amp which gives a common mode voltage that any decent op amp will safely ignore. On a 3-wire RTD, that all makes sense and your observation does not track..

Hi Bill,

Analog make this, something similar perhaps.

But does it need to be AC (reversing) excitation? I suppose that does remove the huge common mode component that would otherwise exist.

--

John Devereux

This project is becoming a lot more than i bargained for. All i want is a simple self-contained temperature "probe" that approaches 1C accuracy with maybe 0.1C repeatability. Would be nicer if the readout was direct, like thermometers aer. Already have thermocouple-type thermometers, a glass thermometer and an LM35 which are in the rough ballpark but emphasis is on "rough". They all indicate something different; max differences 2-3C. Then there is the thermocouple connected to the Omega controller i bought ages ago from Omega..NO direct readout. Set temperature desired with dial. Calibration makes it "correct" at one temperature, with error increasing as one departs from that point; error looks parabolic and is a PITA (being so large as on gets away from cal point).

The Crystal parts had a bad habit of disappearing. And spiking up around $100 for a while before. I'm not afraid do doing the math, and having access to the raw data allows sanity checking.

The analog mux scheme can be easily configured to reverse the RTD leads, if you really wanted to. But measuring the RTD voltage with zero excitation current is just as good, and allows you to hard-ground the low side sense lead, better for EMI.

I used 2-wire 1K thinfilm RTDs for my cabin automation system, and just took out the small wire resistance in the math. The longest run is about 40 feet, to the outdoor sensor.

ftp://jjlarkin.lmi.net/Auto_wired.jpg

ftp://jjlarkin.lmi.net/RTD.jpg

ftp://jjlarkin.lmi.net/CABIN.TXT

The 1K resistors are Susumu thinfilms. All the math is ratiometric. Good enough for the purpose.

You should design and build something now and then. It's fun.

John

only

ards

n e

umu

o

Companies do do that kind of thing shortly before they get taken over.

You want Kelvin - four-wire - excitation for maximum accuracy, and anyway you'd be better off screening the all four leads. Having current circulating through a nominally grounded lead isn't a good idea.

Halving the power dissipated in your sensor increases the root mean square Johnson noise by 42%, and halves the signal you actually measure.

It isn't going to matter in your application, but it isn't "as good" as doing the job properly.

And how much resistance was that? Copper has about the same temperature coefficient as platinum, so that wouldn't have been a good idea if you were aiming for any kind of sensitivity or accuracy at the resistance thermometer.

Of course.

That counts as design? You are easily amused.

-- Bill Sloman, Nijmegen

ly

ds

u

nd

Calibrate your LM35 in a well stirred ice bath, ideally using ice made from distilled - or at least deionised - water. As a set point it is good to about +/-0.003K.

The device has already been trimmed for gain and linearity.

National guarantee 10.00 +/-0.1mV per K, which could be as high as

+/-1K at 100C.

If you want to do better, you have to find another predictable temperature.

The steam point is pretty good, if you know the local atmospheric pressure

You need to place your sensor above a pot of boiling water in a place where it enveloped in steam, and water is condensing on it and dripping back down into the boiling water.

A conical flask with it's neck loosely blocked with twist of porous rag works pretty well.

Make sure that the - super-heated - water boiling in the flask isn't going to splash up onto the sensor, and make sure that the water in the flask is boiling fast enough that there is steam condensing in the air above the plug - you want the atmosphere inside the flask - and around your sensor - to be 100% water vapour.

-- Bill Sloman, Nijmegen

The feedback current is a linear function of the out-of-balance signal that you measure, which is to say it generates a parabolic correction by increasing the out-of-balance signal that is being fed back.

Trust me - and Honeywell. It works. It is positive feeback, but nowhere near enough to make the system unstable - as one idiot engineer tried to tell me. Note that it also increases the self- heating in your sensor as the sensor gets hotter, and decreases it as it get cooler, but you can compensate for that - to some extent - by moving away from using a load resistor exactly equal to your Pt- resistor value at the centre of your range.

-- Bill Sloman, Nijmegen

Do the math. 40 feet of RG174. The errors are tiny.

It was fun. I laid out the lower PCB myself, all thru-hole for easy maintenance. A netbook computer talks serially to the upper data acq board, computes temperatures, runs the software thermostat, and bounces files through my FTP site. The software is in PowerBasic.

Before we leave the city, I can turn on the heat so it's all toasty when we arrive. My wife loves it.

I looked into a bunch of standard home automation products and didn't especially like any of them.

Webcams next.

John

Linear works because the RTD resistance is itself increasing with temperature.

In fact, you can drive an RTD with just a resistor from a voltage source that is DC plus a bit of the RTD voltage itself, and get a linear on temperature output. You can do a zeroed, calibrated, 3-wire, linear RTD signal conditioner with a dual opamp and a few resistors.

John

t f

or

a

page

s -

e more

ach

t

it was in a different configuration and I explained it bad.

the connection to the rtd was done balanced into a diff amp because otherwise electric noise would mess up the signal.

current source didn't work because then it is not really balanced, one wire see low impedance to ground, the other see a high impedance current source

-Lasse

Am i missing something here?

Maybe older timers disease is setting in!

Jamie.

I'm making a guess as to the design date of the instrument in question from your description of the design.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

only

hazards

I've found a Belden spec for their RG174 which includes the resistance of the inner - 97R per 1000 feet - and the braid - 10.7R per 1000 feet,

40 feet of their RG174 would introduce 4.31R of extra resistance. Small compared with 100R, but not tiny. The inner is copper-coated steel wire, and represents 3.8R of that resistance. Iron's temperature coefficient of resistance is half again higher than that of copper but whatever steel is copper-coated to form the inner may have a different temperature coefficient of resistance again.

-- Bill Sloman, Nijmegen

Yeah; seems that the current sources driving the RTD legs need to be balanced better than 0.05%.. Anyone know of a design that would guarantee this tight matching?

Don't leave us in suspense..open the envelope and name the winner...

On a sunny day (Tue, 14 Dec 2010 16:55:01 -0800 (PST)) it happened " snipped-for-privacy@fonz.dk" wrote in :

To be more precise: The 'other wire' sees the RTD impedance to ground.

only

hazards

Susumu

Coppefclad steel? Mine is more like 40 milliohms per foot.

But I'm using 1K RTDs.

And it's just an outdoor temperature measurement. It's tweaked to be dead on at ice point, which is compulsive for an outdoor thermometer, but why not?

Most of the coax runs through an unheated garage, so ice point should be pretty accurate.

but not tiny. The inner is copper-coated

So, you didn't do the math.

John

:

will only

art to

me hazards

uses an

balance

th Susumu

ing is

t to

ugh, do

RTD

out

o the

r A/

e

df

nd

.

und

n d e

So whose RG174 are you using?

Sorry. You have mentioned that and I should have remembered.

Huh? You complain that Stevenson boxes don't neasure the "right" temperature, but tolerate the contribution from an unheated garage. A

4-wire connection would have eliminated the problem.

Actually, you didn't provide the information. Most of us know that there is more than one source of RG174, but you didn't specify which source you used - and you still haven't.

Working out which steel Belden used as the core of their copper-coated inner would have taken quite a while, always assuming that Belden has actually specified this somewhere, and it would have been a complete waste of time because you aren't using Belden RG174.

There's not a lot of point in doing math on arbitrary numbers.

-- Bill Sloman, Nijmegen

only

hazards

Susumu

it was for a 1K rtd, so 4.31R is ~ 1C @ 0C remove that with calibration and you are left with ~0.4% of ~1C from the temp coeff of copper wire

I'd think that's good enough for an outdoor thermometer ;)

-Lasse