Author Topic: Temperature measurement  (Read 3057 times)

0 Members and 1 Guest are viewing this topic.

Online BradC

  • Super Contributor
  • ***
  • Posts: 1291
  • Country: au
Temperature measurement
« on: November 20, 2016, 10:45:14 pm »
while true ; do echo "I am not a volt nut" ; done

As part of the Australian traveling voltage standards group, I'm first up. I've been measuring standards sent to me on my uncalibrated HP3457a and the best room in the house still has fairly average thermal stability. I'm logging voltage using a homebrewed GPIB interface, but I really need to be logging temperature at the same time.

How accurate does it have to be if it's to be useful? I was thinking of just adding a couple of DS18B20's as I have a drawer full of them. I already have code to use the high resolution mode, but how accurate does it have to be? The 18B20 is speced at about +/- 0.5C over the temperature range I'm likely to be seeing, and I'd have thought what I should be looking at is more a trend when superimposing the temperature and voltage graphs, but I'm not really sure.

Thus my question. If you were measuring a group of uncalibrated standards against an uncalibrated meter what else do you want to know? Provided things are kept reasonably stable, is humidity important?

As it is, it seems to take about a day for the meter and reference to really get stable, now whether that is the meter achieving internal thermal stability, the reference, both or black magic, I'm really not sure, but after about a day things stop really moving around.

 

Offline guenthert

  • Regular Contributor
  • *
  • Posts: 166
  • Country: us
Re: Temperature measurement
« Reply #1 on: November 21, 2016, 03:04:47 am »
> How accurate does it have to be if it's to be useful?
Useful for what?  What is the target accuracy of the measurements of the voltage standards?

DMMs have a temperature coefficient, but that is fairly small and might not even be specified (might be thrown into the general uncertainty) for a 'comfort zone' (a few Kelvin somewhere around 23 degree C, should be specified).

Voltage standards will also be optimized for a small TC in such a comfort zone (ideally overlapping with the comfort zone of the DMM) or (should) have a built-in oven.

If you observe large temperature related swings (more than a few ppm/K), then I would have a closer look at the contacts - they might need some cleaning.

Achieving accuracy for temperature readings better than 0.5 degree C will get tricky (I'm sure there are temperature-nuts out there ...), but might not be needed here.  Better precision can be had using a NTC or RTD; that at least will then clarify whether there is a significant TC in your voltage measurements.
 

Offline saturation

  • Super Contributor
  • ***
  • Posts: 4784
  • Country: us
  • Doveryai, no proveryai
    • NIST
Re: Temperature measurement
« Reply #2 on: November 21, 2016, 03:25:45 am »
Envrionmental conditions in a metrology lab are ISO 17025 standard.  The bigger problem is trying to meet them.  If your measurement conditions are not stable, any variation on your references cannot be ruled out as due to the environment.

The good news, temperature is more a factor than humidity, it depends on the overall precision you are achieving, the more precision the more control you need over other variables.

http://us.flukecal.com/literature/articles-and-education/general-calibration-metrology-topics/papers-articles/temperature-h

http://what-when-how.com/metrology/the-role-of-metrology-laboratory/

I personally do not know what the current values are, so the above are just examples.

Best Wishes,

 Saturation
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 3350
  • Country: tw
  • Volt-bolt
    • xDevs.com
Re: Temperature measurement
« Reply #3 on: November 21, 2016, 03:06:21 pm »
Given the mentioned level of instruments (3457A), it's safe to say that you don't need absolute accuracy, all you need is relative temperature change.
So 18B20 will be well enough for your purpose, to keep track of temperature changes and calculating impact on your measurements.
For same reason you don't have to worry about humidity as well.

It's very probable situation , that absolute error on your measurements would be much much larger than any effect from the temperature&humidity changes. If you have access to calibrated meter, it would be best to run comparison to estimate this error, before going temperature-nuts  :-+
xDevs.com YouTube | Have test gear documentation to share? Upload here! No size limits, firmware dumps and teardown photos welcome.
 

Offline Assafl

  • Frequent Contributor
  • **
  • Posts: 593
Re: Temperature measurement
« Reply #4 on: November 22, 2016, 07:36:59 pm »
Preface: I am a noob to making precise measurements at home. So everything I state here is probably obvious to everyone.

Oddly enough I am just running into temperature related issues trying to measure the cell voltage of a Muirhead cell I salvaged from a local scrapyard. It serves a mental exercise to see if I can deduce from measurements if the cell is "good" or "not good". (I don't really need a 1ppm voltage reference at home - but it is a nice exercise to see if I can measure something with a predictable outcome).
 
I was getting 3-4uV swings in reading (which - if true - meant the cell should be scrapped).


I was pretty careful to ensure clean copper connecting to the cell (knurled nut) - but I didn't want exposed banana plugs getting shorted by accident so I used a Hirschmann safety banana (nickel over brass). These are large and therefore have a thermal mass (way more than I had given them credit). I underestimated Seebeck effect....

So I plotted V over 30 minutes or so:



Clearly it took over 10 minutes for the thermal effects to stabilize, and even then it would swiftly swing at any breeze. Covering it with cloth helped, but I am certain that unless I can control that junction temp (or replace the conductors with copper) - I will not be able to measure this with sufficient precision to answer the postulated question.

So to answer the OP question, to me it seems temperature is important in order to compensate for temp effects. I use it in my calculations. But as it turns out -  even more important is to maintain thermal equilibrium at all costs. Cover all exposed connections, use highest gauge wire, keep all connectors together and at the same temp, if possible in a draft free box.

Oh - and ditch large brass and nickel contacts: As it turns out the Nickel Plated brass banana plugs can cause a drift of over 10uV given a nice temperature change as evidenced by using a cooler spray or hair dryer on the exposed areas of the banana....

Or - another option is to ignore the uV digits (if possible).

 
« Last Edit: November 22, 2016, 07:43:52 pm by Assafl »
 

Offline Conrad Hoffman

  • Frequent Contributor
  • **
  • Posts: 825
  • Country: us
    • The Messy Basement
Re: Temperature measurement
« Reply #5 on: November 23, 2016, 12:11:54 am »
At that level all manner of things will affect the readings. I use plain untinned wire pulled from old phone wiring for my interconnections with good success. No bananas or other connectors, just fasten them under the screw-down part. The stuff known as "bell wire" here also works. I'd think you'd want about 0.1 degree C resolution and at least calibrate the sensor in an ice bath and maybe boiling water. I heat with wood and have similar environmental issues!
 
The following users thanked this post: saturation, Assafl

Offline Assafl

  • Frequent Contributor
  • **
  • Posts: 593
Re: Temperature measurement
« Reply #6 on: November 23, 2016, 01:29:45 am »
At that level all manner of things will affect the readings. I use plain untinned wire pulled from old phone wiring for my interconnections with good success. No bananas or other connectors, just fasten them under the screw-down part. The stuff known as "bell wire" here also works. I'd think you'd want about 0.1 degree C resolution and at least calibrate the sensor in an ice bath and maybe boiling water. I heat with wood and have similar environmental issues!

How do you hook it on the DMM side? (I ask because that is the source of my issues - I use untinned wire on the cell side which has copper posts).
 

Offline Conrad Hoffman

  • Frequent Contributor
  • **
  • Posts: 825
  • Country: us
    • The Messy Basement
Re: Temperature measurement
« Reply #7 on: November 23, 2016, 06:14:31 am »
In my case it's easy because my DMM is an old HP 3455A with exposed screw down banana sockets. If you have a modern recessed banana, some kind of plug will be needed. I've never tried it, but I'd be temped to turn some copper plugs on the lathe. How anybody gets through life without a lathe is a complete mystery to me. Otherwise you might have to search out some low thermal emf ones or maybe experiment with a couple different types.
 

Online Gyro

  • Super Contributor
  • ***
  • Posts: 3150
  • Country: gb
Re: Temperature measurement
« Reply #8 on: November 23, 2016, 06:29:05 am »
Using Cotton Buds for retaining bare wires in 4mm sockets has been mentioned a couple of time in the last week or so.
Chris

"Victor Meldrew, the Crimson Avenger!"
 

Offline Assafl

  • Frequent Contributor
  • **
  • Posts: 593
Re: Temperature measurement
« Reply #9 on: November 23, 2016, 07:32:15 am »
Those low thermal EMF things are spectacularly expensive. I would have expected Pomona to make one of their dual banana to binding post adapters in copper - but alas - no.

Maybe it is possible to convert one of those to copper? Mate with a Mueller or Pomona Te/Cu binding posts?

I tried the cable stuffed in the banana socket using a ear swab and it worked. But it didn't  inspire confidence for anything but a one-time use.

I guess it is time for experimentation. And to buy a lathe.
 

Online BradC

  • Super Contributor
  • ***
  • Posts: 1291
  • Country: au
Re: Temperature measurement
« Reply #10 on: November 23, 2016, 08:04:55 am »
And to buy a lathe.

Every workshop needs a lathe. Turning unadulterated copper is a challenge however as if your tools aren't razor sharp and geometrically correct it just smears everywhere.

Thanks for the input everyone. I'll start fiddling and see what I come up with.
Resolution better than 0.1C is easily achievable with the old DS18B20, but accuracy is more of a challenge.
 

Offline guenthert

  • Regular Contributor
  • *
  • Posts: 166
  • Country: us
Re: Temperature measurement
« Reply #11 on: November 23, 2016, 11:23:31 am »
[..] I'd think you'd want about 0.1 degree C resolution and at least calibrate the sensor in an ice bath and maybe boiling water. [..]
Well, better measure the atmospheric pressure then when determining the boiling point.  Now, how do you measure that pressure accurately ...  ;D


Finally found what I was looking for (boiling points of water vs. atmospheric pressure):
http://www.hi-tm.com/Documents/Calib-boil.html
« Last Edit: November 23, 2016, 11:27:54 am by guenthert »
 

Online BradC

  • Super Contributor
  • ***
  • Posts: 1291
  • Country: au
Re: Temperature measurement
« Reply #12 on: November 23, 2016, 12:11:33 pm »
[..] I'd think you'd want about 0.1 degree C resolution and at least calibrate the sensor in an ice bath and maybe boiling water. [..]
Well, better measure the atmospheric pressure then when determining the boiling point.  Now, how do you measure that pressure accurately ...  ;D

Easy. I take my maritime barometer to the local met bureau and check its cal, then take it home and put it in the kitchen. A couple of pascals here or there is not going to make much of a difference. Does that make me a pressure-nut?
 

Offline Conrad Hoffman

  • Frequent Contributor
  • **
  • Posts: 825
  • Country: us
    • The Messy Basement
Re: Temperature measurement
« Reply #13 on: November 23, 2016, 03:59:04 pm »
Some time back Scientific American mag had a nice article for making a DIY triple point cell. Alas, the world is full of far more interesting things than I'll ever have time to pursue.
 

Offline Assafl

  • Frequent Contributor
  • **
  • Posts: 593
Re: Temperature measurement
« Reply #14 on: November 23, 2016, 08:01:51 pm »
Again - I am a noob, so everything here is probably considered obvious and trivial by others. So I apologize....

As I am adding another measurement point, the post by BradC is bugging me quite a bit I admit.... I think what bugs me is my being completely befuddled about "why to measure"? The rest of the WH questions: "How to measure", "what to measure" and "where to measure" - or even "when to measure" becomes secondary once an understanding of "how it will be used" displaces the cerebral fog...


As an example, let's take the lab which calibrated the HP34401A:
The lab measured the ambient temp and RH. Why? What purpose does it serve? (I admit it is a rhetorical question - they have a process they follow, standards and regulations they follow and their customers demand it). But I can find no practical use for it. It is a single measurement at a single point in time.

I admit to being ambivalent about the single point in time measurement since it makes the report look good (so I feel good about it) - but it may be for show (not feeling as good anymore). It is one point that shows me that it is within HP/Agilent/Keysights recommended envelope for calibration. In the mind's theater somebody somewhere read the manual.... Does it attest that the rest is up to scrutiny? Wishful thinking yearns for it to be so - but I wish I knew better....


A better example:
I am not as ambivalent about measuring the temp of the Standard Cell (as accurately as I can) if only because of the 64uV/C tempco of the cell. Were I not able to compensate for it - my readings would be fluctuating even more. 64uV/C actually make it very worthwhile to figure out where I am (or should be) measuring the well temp and how.

In the "assigning blame" game, cell temp uncertainty is a good candidate (perhaps even more than the brass bananas) to "blame" for the measurement fluctuations. Now - Surely there is a lag between the cell temp and the well temp. As an example, the relative position of the PT1000 in the well affects the measurement since the bottom of the well touches the heater (whose temperature mildly oscillates). My current hypothesis is that the actual "thermal sink" cell temp changes considerably less than the copper well temp.
 

So what about the why's of ambient temp and humidity?
I don't know if a single point is actually worthwhile - except to validate that your DMM is calibrated at the lab at the temp and humidity recommended by the manufacturer. Finding "practical" uses for the ambient temperature measurement (e.g. as a compensation for Seebeck and other effects) I believe is nearly impossible.

However, registering the temp and humidity over time can perhaps help you find seasonal correlations between room conditions and drift (albeit one has to figure out if it is the DUT, the measuring device, or more likely - both - to blame).

For a travelling box of references my guess would be that the highest ROI for the exercise would be to set the measurement device (+ all the necessary conditioning electronics for it - so the samples can be compared) at a specific location relative to the references as part of the travelling setup. Minimizing the variability of the setup (e.g. fixing the setup in a "whisky" wooden box) will help trying to weed out temp (and humidity) related trends. (but this is a guess by a clueless noob...)

BTW - regarding humidity - I don't think humidity is important for measurement except for very high resistance or sensitivity measurements where humidity can create alternate current paths... I think wet/dry bulb measurement is far more important (since it changes the air's capacity for energy - hence for a given stable ambient temperature - the temperature rise of the air surrounding the devices with the tempco will indeed be significantly different).

I think humidity is measured because it is far easier to measure if not as accurate than to use a real psychrometer - especially the sling types.         

« Last Edit: November 23, 2016, 08:13:46 pm by Assafl »
 

Offline Echo88

  • Frequent Contributor
  • **
  • Posts: 327
  • Country: de
Re: Temperature measurement
« Reply #15 on: November 24, 2016, 03:41:50 am »
As Conrad mentioned: the triple point cell article from scientific american: https://www.scientificamerican.com/article/tackling-the-triple-point/
Interesting  :-+
 
The following users thanked this post: Assafl, guenthert


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf