Fluke 540B Reversal Error

[essele]

A quick question about the significance of the thermocouple reversal error on the Fluke 540B...

I've been using the 540B to calibrate an AC calibrator from a calibrated DC calibrator and it all seems to be working out ok, it's entirely repeatable but I don't have anything accurate enough to actually measure it to 0.01% tolerances (I'm just coming to terms with how rubbish my 34470A really is for this stuff!)

I had noticed a delta on the reversal, so I've just tested it using the method in the manual, and cross-checked with my earlier measurements during calibration and I'm seeing a consistent 0.05% where it should be 0.01% or better. Most stuff I've read suggests that this is a result of overloading and the thermocouple would need to be replaced, which clearly isn't an option (unless someone has a handy source?)

Is it really that important? Doesn't the method of averaging the two results remove the significance of this error? Or is there something about it that would impact the overall accuracy?

Thanks,

Lee.

[Bill158]

Is it really that important? Doesn't the method of averaging the two results remove the significance of this error? Or is there something about it that would impact the overall accuracy?

Thanks,

Lee.

Lee
The answer is yes and no depending upon how accurate you want your AC Calibrator to be.  If you can accept +/- 0.5 % then the "DC Reversal error" will not be significant.  If you expect the AC Calibrator to be +/- 0.01 % then the "DC Reversal error" is very significant (i.e. a FLUKE 5200A).  All AC/DC Thermal Converters have a reversal error.  Some larger than others and it can depend upon variations during manufacturing, mainly due to the exact placement of the placement of the glass bead insulator on the heater element.  Read "Thermal Transfer Stds.pdf" from "Calibration Philosophy in Practice 1st ed." that I have attached.  Another document would be section 5 of "jresv48n2p121_A1b.pdf" by Hermach from NIST https://nvlpubs.nist.gov/nistpubs/jres/048/jresv48n2p121_A1b.pdf.  These may help to answer your question.
Bill

[essele]

Thanks Bill ... that's an interesting paper. I can't pretend to have understood all aspects, but my short summary is:

If the reversal error is due to thermoelectric effects then this will be effectively nulled out by the averaging process, but any other cause, most significantly the metal used in the heater and the positioning of the hot junction, will cause a transfer error ... so I'm somewhat confused as to how the latter two can change over time, or through abuse of the thermocouple?

Through a bit more tinkering I have discovered that the polarity switch is a bit questionable ... measuring voltage at the shunt terminal when applying a DC voltage as normal (or just resistance when not) I can see a very repeatable reading when the switch is pressed in, however when switching it 'out' there's quite a lot of variation each time and I can cause a reasonable movement just by wiggling the button. I've tried cleaning it, and it's better, but still not right.

I don't believe it's responsible for the whole error, but it's certainly contributing ... I can easily see a 20uV delta on the 0.45v reading just by wiggling, so I'm now not really comfortable believing the numbers until I sort that out ... even if it were consistent a difference in resistance between the two sets of contacts would manifest itself as a reversal error.

Not sure how I'm going to fix it though ... I don't think I can replace the switch with a simple DPDT, it seems to run through an intermediate state when a 100R resistor is across the input, presumably not to leave it open circuit at any point.

EDIT: Potentially a make-before-break rotary switch with three positions might do the job ... I think I'll spend some more time cleaning before I resort to this though.

[Bill158]

Lee:
The Hermach paper isn't for the average person.  The basic part of this is that there IS reversal error due to a number of issues.  Just accept that it happens.  The 540B Reversing switch is a REAL problem.  They wear out after many many uses.  I tried cleaning mine with all types of cleaning including DeoxIt from Caig Labs.  That worked for a while but the problem always came back.  The switch is just worn out.  I tried a few different ways to replace it with a rotary switch but not enough space in that area.  So since I now use a Fluke 5440B as my Voltage Standard to get my DC Voltages, I just use the "Reversing" function of the 5440B to change the direction of the input voltage to the 540B.  I was going to build an external reversing switch in an external box but that got too complicated with the need to keep Force and Sense all the way through the external box.  So I just took out the Reversal switch and wired everything directly.  Works like a charm.
Bill

[essele]

Doh!  Of course -- sometimes the simplest answer is the best!

I've got a freshly calibrated Datron 4000A for my DC calibrator, and I've just checked the cal cert and the positive and negative 1V output was identical to the uV so that's a good test.

0.1% is three and a half small division on the medium sensitivity, and I'm seeing two small divisions difference, so that confirms that I am actually seeing roughly a 0.06% error ... sigh. That said, it's all still way better than I can actually measure so I will live with it for now.

I did notice that the percent input scale goes off scale negative when you apply a reverse polarity input ... I assume you just live with that?

[Bill158]

I did notice that the percent input scale goes off scale negative when you apply a reverse polarity input ... I assume you just live with that?

Yes I just ignore the percent input going negative.  It doesn't seem to hurt anything.  I done this way for years.
Also I just thought about another problem.  That is you can't get the mercury batteries anymore, at least in the USA, for the Reference Power Supply that "bucks" the output of the Thermal Converter and gives you something to null against when applying the AC unknown.  Attached is the FLUKE solution to this problem.  I have installed this modification and it works great.
Bill

[essele]

Thanks Bill - my 540B already includes this, but I am also tinkering with a complete Lipo replacement for the whole battery pack. 9 Lipo’s with charging, and then four low noise regulators for +12.5v, +/-6.25, and 1.35v, and hopefully the ability to still use the existing battery check circuits. I’m just waiting for the second version of the board to arrive (schoolboy error with the wrong regulator footprint on the first one) ... I’ll post details if I get it working ok.

Lee.

[TheDefpom]

Hi,
You might find this slightly useful, my replacement of the mercury cell with a lithium.

[1audio]

Essele;  Any progress on the power supply upgrade? I'm on the verge of getting new nicads for mine but would like to see what you have. Mine is a very late example with the AA battery upgrade for the Mercury battery from Fluke. Factory installed (the cal stickers were still intact). It also has the rotary switch for polarity reversal.

It also needed virtually every contact cleaned after 10+ years of sitting unused.

One really vexing issue is that the DC side resistance is .5 Ohm different from the AC side. It seems there is a shunt resistor switched in on the DC side only. You must use remote sensing to get accuracy. Mine also has an annoying slow drift negative. on high resolution. Its the thermocouple or the slldeback. Not sure yet.

[essele]

Hi 1audio,

I had the power supplies (I did two in the end) working nicely, used them for quite a while, but a few months ago I decided to see the units as they ultimately just started gathering dust ... power supplies were still working nicely.

The only thing I think I'd do differently would be to put a physical power switch in there somewhere ... in trying to avoid too many mods to the 540B it meant that one of the lines was on all the time (can't remember which now) and that drained the battery over a period of a couple of months. Wasn't a major issue, but not right.

I've attached a photo that I used when I sold them ... doesn't really do it justice, but should give you an idea.

Lee.

[1audio]

It is a question for me- I don't use it often (every year or two?) and I'm trying to resolve a key question-What do I trust?
I have the 540B
2 931 AC differential voltmeters and 6 thermal converters.
I had an 8506A until its conversion chip died a few weeks ago. Replacement chip is listed as $700 at Fluke when discontinued in 2006. Extremely frustrating.
The 941 response doesn't seem to match the 540B which doesn't match the 10V fluke converter. Its all small except above 100 KHz. I think I'll use the 10V converter for frequency response but the difference between it and the 540B annoys me. Its all small differences and these are microscopes on those differences.
Given how little I use the 540 maybe fitting it with disposable batteries and regulators makes sense. I can pull the batteries to use elsewhere when I'm done. Or use the LI disposable batteries which have a long (20 yr) shelf life.
Do you still have any current measurements for the different supplies?

[Roman oh]

Batteries frustrate me, as they're always flat when I need them. So I'm taking a slightly different tack with my 540B. I am in the throes of assembling a small DC/DC converter which will sit inside the unit in place of the battery pack, deliberately designed with good isolation, to provide the required voltages from an external 8-14V source. I'm hoping to be able to run it off a 12V wall brick, but can substitute an (external) 12V battery if the mains connection still proves problematic. I'm not even sure how I'm going to establish "adequacy"; but playing around with this sort of stuff is what I enjoy doing...I would appreciate any feedback/suggestions.   
 

[View[+]Finder]

Hi,
You might find this slightly useful, my replacement of the mercury cell with a lithium.

Excellent video! Looking for an easier solution to the "no mercury, no more" issue, I have installed a "D" cell NiMH charged up to about 1.3V. The 540B manual refers to a minimum for the reference battery at 1VDC and the nulling and adjusting procedure to do a transfer to measure AC voltage makes full use of the three pots to null the meter. Based on the schematics, that should make the Hg battery optional. The alkaline 1.5V does not work as the pots deal with under-voltage, not over.

The transfer procedure to obtain a RMS DC equivalent for AC voltage provided an entertaining answer to the question: Why did engineers design 'true RMS' meters?

Question for the 'experts': -- how close should the measurements by a 540B of a 500Hz 500mV signal be to the results from an HP3400 or a modern DMM?

[1audio]

How accurate is your reference voltage? At 500 Hz there should be no AC effects that would compromise the accuracy. You can read to .001% or 10 ppm. However once you throw in all the accuracy/uncertainty stuff .01% is probably best case. And once anything like the power supplies have been changed you can't know immediately if its more or less accurate. One confusor I encountered- the AC load impedance is slightly different from the DC load. When chasing PPM that is not insignificant. Its the overload sensing circuit that seems to only be on the AC side? In any case you should be using 4 wire connections where possible to drive it with its low internal impedance.