Easy way to test the calibration of a DMM (Fluke 45)?

[CalibrationGuy]

(I'm not sure why they'd do that, was it just so that people could have accurate clocks?)

In Ye Olden Days the 60Hz line was used for a whole lot of timing purposes. In cheaper devices, it eliminated an internal clock circuit. (Yes, I am that old.)

TomG.

[mawyatt]

(I'm not sure why they'd do that, was it just so that people could have accurate clocks?)

In Ye Olden Days the 60Hz line was used for a whole lot of timing purposes. In cheaper devices, it eliminated an internal clock circuit. (Yes, I am that old.)

TomG.

Ironically the 60Hz Power Grid was not that accurate short term, over the span of precisely a day it was dead on by design. As mentioned during the day the frequency would drop, this was caused by the on-line loads in the Grid (1/2 the US on each Grid, East and West coast). However during lighter loads late at night and early morning the grid would "Speed Up" to catch up time-wise (think of an old clock which was based upon a synchronous motor).

Exactly 60 cycles/sec * 60 seconds/min * 60 minutes/hr * 24 hours/day or 5,184,000 cycles/day were created by the Grid and "tuned" to such every day. This was why old wall clocks were always in time from day to day, week to week, month to month and only needed setting when power was interrupted. In fact, we used the wall clock to discover if we had a power interruption and how long this was by simply comparing the wall clock to our wristwatch which we synched up every day.

Back in ~1970 we made an instrument to monitor and plot the Grid Frequency. While giving a lecture on such the Nuclear Power Plant Turkey Point in South Florida tripped off-line and we all witnessed a large Dip in the Eastern US Grid Frequency, recall this Dip was something like ~30mHz!!

You could see the Grid "Speed Up" at night/early morning and the "Dip" mid day when viewing the instruments chart recorder output over a 24 hours period.

Best,

[J-R]

Trying quoting for the first time:

It's true that not everyone needs the same level of performance from their test equipment, but these random unknown references really are bad no matter how you look at it.  The technical shortcomings have already been well discussed (and perhaps somewhat ignored), but a different point is that you are frequently contributing to people who would be totally fine scamming you wherever possible. 

That works both ways. There's no innocence on either side, with Westerners openly scamming by unlocking functionality that they have not paid for on Eastern test equipment. I don't choose to automatically mistrust every seller based on the occasional bad experience here or there.

What about all the threads here for hacking the Agilent/Keysight scopes (one over 131 pages long)?  No need for any West/East nonsense... People will hack equipment in their possession for various reasons and it has nothing to do with where they or the equipment came from.  You're moving the goalposts now...
Back on topic: how ELSE would you describe an online seller that prints out duplicate copies of a calibration sheet and applies them to all the "references" they ship out?  That is an extremely solid definition of a scammer.

Instead, send your money to individuals who are not really doing it for the money but for the love of the hobby.  Win/win for everyone.

Not everyone thinks the same. It is possible to "buy local" as well as to help others across the world, who might actually need the money.

I don't know what you are arguing against here.  I never said anything about not buying local.  Support honest businesses.  Local as well.

And again my main point about why these cheap references are "dangerous" is that they have specific voltages printed on them and are sold as "references" so buyers apply additional weight to them.  What ratio of buyers do we think have adjusted their DMM based on the "reference" compared to adjusting the "reference" based on their DMM?

Let's guess at an extreme 90% of buyers adjusting their DMMs based on that board. Even if they did, so what? They would have a DMM that was possibly inaccurate by tens of mV at the most? What are the implications:

(a) For a hobbyist, if it affects their projects, it will be a great learning curve on what to trust. If something is worth doing, it's worth doing wrong.

(b) On the other hand, if you're a professional: a qualified engineer may well rely on prototype or 'uncalibrated' equipment when developing products; there's nothing abnormal in that. But equally, that engineer is in every way liable for trusting a $10 device over a calibration if they ignored a process, or were supposed to be working to a standard, or if a body of other professionals would not have done the same for that specific task.

Quite an incredible stance.  You are OK with people adjusting their previously factory calibrated test equipment with some random reference?  I think you should reconsider your position.

[shabaz]

What about all the threads here for hacking the Agilent/Keysight scopes (one over 131 pages long)?  No need for any West/East nonsense... People will hack equipment in their possession for various reasons and it has nothing to do with where they or the equipment came from.  You're moving the goalposts now...

No I didn't, except perhaps in the most minor of ways. You were the one that stated that buying from AliExpress means frequently contributing to scammers. I merely responded. But, now you mention Keysight, you think it makes it any better that the scammers that I mentioned in return, are perfectly willing to do the same to fellow colleagues in our own countries too?

Quite an incredible stance.  You are OK with people adjusting their previously factory calibrated test equipment with some random reference?  I think you should reconsider your position.

I'm perfectly OK with what I stated. Not my problem if you misinterpret. The text is there forever, and anyone can read it to judge for themselves.

[Finnaaah]

Not sure if it's that relevant or useful since it's only resistance, but I stumbled across this series of resistors when I last ordered from Mouser.
RNCF0805TKW
e.g. 1K would be RNCF0805TKW1K00

I definitely wouldn't use them to adjust calibration, but 2ppm/C and 0.01% for $4.50 AUD lets you pick up enough to check multiple ranges on a budget. I made a 4W measurement PCB for Vishay's hermetically sealed metal foil resistors and left a few SMD footprints on it, which I've attached. Not saying this is (remotely close to) the pinnacle of accuracy, but it's very handy to have a bunch laying around if you're buying second hand equipment or just want a quick n' dirty idea of your meters resistance accuracy   

[Fungus]

Quite an incredible stance.  You are OK with people adjusting their previously factory calibrated test equipment with some random reference?  I think you should reconsider your position.

The sort of person who would do that deserves to do it.

[tggzzz]

What about all the threads here for hacking the Agilent/Keysight scopes (one over 131 pages long)?  No need for any West/East nonsense... People will hack equipment in their possession for various reasons and it has nothing to do with where they or the equipment came from.  You're moving the goalposts now...

No I didn't, except perhaps in the most minor of ways. You were the one that stated that buying from AliExpress means frequently contributing to scammers. I merely responded. But, now you mention Keysight, you think it makes it any better that the scammers that I mentioned in return, are perfectly willing to do the same to fellow colleagues in our own countries too?

Quite an incredible stance.  You are OK with people adjusting their previously factory calibrated test equipment with some random reference?  I think you should reconsider your position.

I'm perfectly OK with what I stated. Not my problem if you misinterpret. The text is there forever, and anyone can read it to judge for themselves.

Yes, the text is there - but you have snipped the context so a new reader has to actively search out the text. Snipping too much of a conversation is a technique frequently used by those that use dodgy debating tactics, e.g. strawman arguments, moving goalposts, ignoring valid counter-arguments, etc. This isn't stackexchange or EDABoard; multi-level quoting is encouraged.

IMHO J-R comes out of this discussion better than you do.

[shapirus]

You could see the Grid "Speed Up" at night/early morning and the "Dip" mid day when viewing the instruments chart recorder output over a 24 hours period.

You have just boosted my motivation to finally implement visualization for the frequency data that I'm already collecting with my homemade mains voltage data logger. I plot voltage (which btw is not even remotely suitable to check the accuracy of any voltmeter), but not frequency, yet.

[tggzzz]

You could see the Grid "Speed Up" at night/early morning and the "Dip" mid day when viewing the instruments chart recorder output over a 24 hours period.

You have just boosted my motivation to finally implement visualization for the frequency data that I'm already collecting with my homemade mains voltage data logger. I plot voltage (which btw is not even remotely suitable to check the accuracy of any voltmeter), but not frequency, yet.

In the UK you could compare that with stats from official sources: http://gridwatch.templar.co.uk/download.php

(Great information, but appalling skeuomorphic GUI: click "white none" then click "white voltage" and "white frequency" and select the dates and times.)

[shabaz]

What about all the threads here for hacking the Agilent/Keysight scopes (one over 131 pages long)?  No need for any West/East nonsense... People will hack equipment in their possession for various reasons and it has nothing to do with where they or the equipment came from.  You're moving the goalposts now...

No I didn't, except perhaps in the most minor of ways. You were the one that stated that buying from AliExpress means frequently contributing to scammers. I merely responded. But, now you mention Keysight, you think it makes it any better that the scammers that I mentioned in return, are perfectly willing to do the same to fellow colleagues in our own countries too?

Quite an incredible stance.  You are OK with people adjusting their previously factory calibrated test equipment with some random reference?  I think you should reconsider your position.

I'm perfectly OK with what I stated. Not my problem if you misinterpret. The text is there forever, and anyone can read it to judge for themselves.

Yes, the text is there - but you have snipped the context so a new reader has to actively search out the text. Snipping too much of a conversation is a technique frequently used by those that use dodgy debating tactics, e.g. strawman arguments, moving goalposts, ignoring valid counter-arguments, etc. This isn't stackexchange or EDABoard; multi-level quoting is encouraged.

IMHO J-R comes out of this discussion better than you do.

I might try multi-level quoting in the future, to see if it is helps. Others (like me) may well prefer not to see repeated walls of text nested. You can't please everyone.

It's up to the reader if they wish to read previous comments or to step in part-way. I've been speedy with the responses so one doesn't need to traverse too far back. Anyone freshly googling how to test their calibration will easily be able to see the discussion around various options. It's not a difficult discussion to follow. If anyone is really stuck, I'm happy to clarify my comments at least.

My contribution started on page 2, with the following entire text:

Very kind, generous offer to unofficially calibrate.

Just to put some sample numbers out there, for the AD584 cheap references (sub-$10 on AliExpress), by chance, a friend and I have both measured these devices (one sample each, purchased separately at different times) using two separate DMM6500 meters.

The label on it, claiming to have been measured with HP 3458A, is, of course, completely untrue (unless it's a completely broken HP 3458A!). They probably used (at best) a 6.5-digit meter, perhaps not even calibrated, and even then, they have typos in the printed values (the label on mine had a noticeably erroneous extra digit for one of the values!).

However, on the plus side, it turned out that the nominal voltage values (2.5V, 5V, 7.5V, 10V, were actually close enough to the real values, that it's very likely that one would be able to tell if that 0.02V discrepancy you're seeing between multimeters is a problem with one meter or the other, or both. In other words, a 20 mV discrepancy should be identifiable if it's present at one of those four voltages (2.5/5/7.5/10V).

Beyond three digits, you can't really rely on it with no measurement with a known calibrated instrument. Nevertheless, for the price (under $10) with no further measurement, it is a crude finger-in-the-air type of check if absolutely nothing better is available.

Just in case this helps, this was my sample result:
Label              My measurement
-------            ---------------
2.49954            2.49963
4.999501           4.99964
7.39875            7.49899
9.99823            9.99854

My last message in its entirety that you were referring to, which was on page 4:

Quote from: J-R on Yesterday at 08:26:43 am
It's true that not everyone needs the same level of performance from their test equipment, but these random unknown references really are bad no matter how you look at it.  The technical shortcomings have already been well discussed (and perhaps somewhat ignored), but a different point is that you are frequently contributing to people who would be totally fine scamming you wherever possible.

That works both ways. There's no innocence on either side, with Westerners openly scamming by unlocking functionality that they have not paid for on Eastern test equipment. I don't choose to automatically mistrust every seller based on the occasional bad experience here or there.

Quote from: J-R on Yesterday at 08:26:43 am
Instead, send your money to individuals who are not really doing it for the money but for the love of the hobby.  Win/win for everyone.

Not everyone thinks the same. It is possible to "buy local" as well as to help others across the world, who might actually need the money.

Quote from: J-R on Yesterday at 08:26:43 am
And again my main point about why these cheap references are "dangerous" is that they have specific voltages printed on them and are sold as "references" so buyers apply additional weight to them.  What ratio of buyers do we think have adjusted their DMM based on the "reference" compared to adjusting the "reference" based on their DMM?

Let's guess at an extreme 90% of buyers adjusting their DMMs based on that board. Even if they did, so what? They would have a DMM that was possibly inaccurate by tens of mV at the most? What are the implications:

(a) For a hobbyist, if it affects their projects, it will be a great learning curve on what to trust. If something is worth doing, it's worth doing wrong.

(b) On the other hand, if you're a professional: a qualified engineer may well rely on prototype or 'uncalibrated' equipment when developing products; there's nothing abnormal in that. But equally, that engineer is in every way liable for trusting a $10 device over a calibration if they ignored a process, or were supposed to be working to a standard, or if a body of other professionals would not have done the same for that specific task.

Not one person in the discussion suggested that it was a good idea to calibrate a test tool with a $10 device. In fact, everyone went out of their way to mention that their suggestions were a spot-check to determine any significant glaring error at best.

[tggzzz]

I might try multi-level quoting in the future, to see if it is helps. Others (like me) may well prefer not to see repeated walls of text nested. You can't please everyone.

True, but good taste goes a long way.

The same is true in creating electronic hardware and software.

It's up to the reader if they wish to read previous comments or to step in part-way.

Not entirely. Your position assumes that readers are devoting attention to your sub-thread, and are prepared to spend more their time than is necessary. Both represent bad taste.

It is polite and considerate to consider the readers. That's the opposite of most yootoob videos, which mainly consider how easy it is for the creator. Waste of my remaining life.

[shabaz]

(No quoting because I'm responding to the immediate message above):

I can't argue with that. I love learning from elegant designs, and hate seeing poor-quality circuits (or badly drawn schematics) or code, although sometimes one can learn a little from that too.

I am as considerate to readers as the next man and always try to speak in plain language, annotate, take photos, provide diagrams, or do whatever it takes to help the average reader. I am always genuinely happy to help clarify things because some comment style choices will please some readers and not others. But no one likes it when people try to put words in their mouths that they didn't state, e.g. strawman arguments etc. Sometimes, that's accidental, of course.

Same with YouTube videos; you can't please everyone, although there some additional compromises are needed from both creators and viewers because of the extraordinary length of time, learning curve, and skill set it could take for individuals. I believe I have just as negative an opinion about most YouTube content as you might, because I don't like the fact that the viewers are just a means to an end, advert delivery, have to sit through a lot for little information, and so on, for the majority of the content out there. Sometimes people want to see how things are done as if they were peering at a circuit or screen as if they were sitting next to a colleague, but it stinks for the remainder 95% of the time.

[mawyatt]

You could see the Grid "Speed Up" at night/early morning and the "Dip" mid day when viewing the instruments chart recorder output over a 24 hours period.

You have just boosted my motivation to finally implement visualization for the frequency data that I'm already collecting with my homemade mains voltage data logger. I plot voltage (which btw is not even remotely suitable to check the accuracy of any voltmeter), but not frequency, yet.

The technique we used way back in ~1970 was for measuring and plotting Frequency Deviation from ideal Mains (60Hz).

This involved starting with an isolation step down transformer, followed by passive and active BPF to remove noise, then zero crossing. This squared up signal was used in a low noise PLL that multiplied the Reference (60Hz) by 1000, so the output was now 60KHz, or 1000*Fmains.

An NBS traceable OCXO was divided down to 60KHz and the two signals were phase compared. The phase pulses were integrated and sampled on each cycle at 60KHz. This integrated & sampled phase signal was then filtered by two cascaded high order Elliptic LPFs with the 1st null located at 60KHz to remove the residual 60KHz. The signal then was precisely differentiated to yield the frequency deviation, as deltaF = K*dP/dt, where K is the Differential Scaling and Amplification/Attenuation as applied for range settings for display and chart plotting.

Very effective instrument at the time.

Anyway, hope this helps understand what we did way back then in measuring Mains Frequency Deviation.

Best,

[bdunham7]

The Fluke 45 arrived today and I had a bit of time between tasks to set it up atop the wreckage on my bench and take some initial readings.  It didn't get a full warmup, but that's probably not an issue.  Here's the numbers, read 'em and weep!  As posted elsewhere, my 5220A current amp has gone up in flames for the moment.  When it arises from the ashes, or at least I manage to get it off my bench, I'll get back to this meter. 

[J-R]

I think grid voltage is typically going to be highly variable these days since there is so much interconnectivity and buying/selling.  My grid-tied inverter claims to use frequency to control import/export and I wonder if that is used on the grid as well.  I don't have a way currently to track frequency, but here is a voltage graph for the last year:

[J-R]

The Fluke 45 arrived today and I had a bit of time between tasks to set it up atop the wreckage on my bench and take some initial readings.  It didn't get a full warmup, but that's probably not an issue.  Here's the numbers, read 'em and weep!  As posted elsewhere, my 5220A current amp has gone up in flames for the moment.  When it arises from the ashes, or at least I manage to get it off my bench, I'll get back to this meter. 

I think that this explains the 5.1V discrepency:
"In the medium and fast measurement rates, the a/d converter uses one of two ranges: ±300 mV and ±3 V full scale.
In the slow rate, the a/d converter uses one of two additional ranges (±100 mV and ±1000 mV full scale), for a total of four ranges."

[bdunham7]

I think that this explains the 5.1V discrepency:
"In the medium and fast measurement rates, the a/d converter uses one of two ranges: ±300 mV and ±3 V full scale.
In the slow rate, the a/d converter uses one of two additional ranges (±100 mV and ±1000 mV full scale), for a total of four ranges."

Possibly, but since it measures 10V using a scaled (divided) range (10V or 30V, not 3 as I wrote) it could be either the scaling constant for that range or the ADC calibration.

If I measure 1V using both the 1V and 3V ranges, I get 0.99460 and 0.9969 respectively.  0.99460 x 5.1 = 5.07246 and 0.9969 x 5.1 = 5.084.  So according to that, the range scaling factor is still way off on 10V but close on 30V.  And this is where it might get interesting as the DCV calibration adjustment process sets all 4 ADC ranges, but only sets range constants on the 30/300/1000V ranges not the 10/100. 

100.000 mV on the 100mV and 300mV ADC ranges gives me 96.535mV and 98.84mV respectively. 

So what now?  Does this patient need calibration (adjustment) or repair? 

[shapirus]

Anyway, hope this helps understand what we did way back then in measuring Mains Frequency Deviation.

Sounds like one hell of a complicated system, but I guess it was the way to go back then.

Mine is much more simple (by today's standards, of course): input divider -> MCP3304 -> single-board computer. Frequency measurement is as (in)accurate as the SBC's clock generator is, plus any inaccuracy added by the finite sampling rate, which is 100 kSa/s. I guess that's good enough for my home/hobby use. At least it agrees with my Brymen BM869s to within ~2 mHz. The latter's accuracy spec, however, is +/- 0.014 Hz at 50 Hz line frequency, so not an ideal comparison, but what are the odds of them both being off by the same amount? But I think I can verify this: I have a GPSDO, and I can build a frequency divider circuit to divide its 10 MHz to get a 50 Hz output, and measure that with the BM869s to see what it's showing.

p.s. whenever I measured AC line frequency, I don't remember seeing exactly 50 Hz. It was 49.98-ish every time. Right now its 3 a.m., and it's fluctuating around 49.97-49.99. I don't think I would want to sync my wall clock to that :).

p.p.s. I also log the value of the crest factor (\$\frac{|V_{peak}|}{V_{RMS}}\$) of the actual AC line waveform divided by the crest factor of the ideal sine wave to be used as a representation of how much the sine wave in the wall outlet is distorted. I'm usually seeing values around 0.97-0.98, which means that the wave's peaks are a bit flattened (and this can actually be observed with an oscilloscope). I guess this can be expected. I completely forgot that I had this value logged too -- will of course plot it along with frequency. Time to find (or rather remember) a suitable time-series data acquisition and storage backend for Grafana. There was one that I thought was a perfect fit for the job, but forgot it, maybe someone can remind me.

[J-R]

Repair or adjust...  I admit I'm on the fence.  It does seem a little odd that so many readings are low.  There is a decent amount of information in the service manual, so checking a few things out seems reasonable.

[bdunham7]

Before I take it apart and start poking at its guts, I'm going to test it for stability and linearity.  I wouldn't anticipate problems with either, but I may as well check.  I'll be logging a 10 reference overnight, although the amount it is off by is pretty ridiculous--it reads 9.7470, or 2.53% of nominal.  If that is stable, even being way off, that would be a good sign.

Looking at this meter closely, I'm seeing it is really a 30,000 count meter with a 100,000 count "hi-res" mode, similar to the Fluke 87 and relatives with their 6,000 counts and 20,000 counts in hi-res.  It is also the only meter I know of that has modern closed-case calibration but still uses a plain 6.3V zener (presumably a selected temperature compensated model) as a reference. 

[bdunham7]

And for one final interesting thing before I leave this project for a bit, the calibration constants can be retrieved over the remote connection (RS232 in this case) and so I did.  Here they are.  You can see that the first six are off by amounts similar to the errors I'm seeing.  IDK which way they work or whether they are the cause of the errors or if they are attempting to correct the errors. 

[Fried Chicken]

I am following this with interest!

This is thorough work!

It is also the only meter I know of that has modern closed-case calibration but still uses a plain 6.3V zener (presumably a selected temperature compensated model) as a reference. 

What do meters normally use?  It's my understanding these DMMs are stable in their calibration once calibrated

[Fungus]

It is also the only meter I know of that has modern closed-case calibration but still uses a plain 6.3V zener (presumably a selected temperature compensated model) as a reference. 

It doesn't really matter what the exact voltage is for digital calibration. All you need is something stable.

There's no need for laser trimming to an exact value or anything like that.

It does seem a little odd that so many readings are low.

They're probably all related in some way. At the end of the day everything comes down to measuring voltage (Ohms, amps, etc. are all done by measuring voltage and using Ohms law).

[CalibrationGuy]

It doesn't really matter what the exact voltage is for digital calibration. All you need is something stable.

There's no need for laser trimming to an exact value or anything like that.

It all comes down to your requirements. As an example, if you are using a 12bit DAC then a voltage reference trimmed to 4.096 Volts would be very convenient as each step would be 1mV, simplifying your math and eliminating scaling equations and their associated errors.

TomG.

[bdunham7]

It is also the only meter I know of that has modern closed-case calibration but still uses a plain 6.3V zener (presumably a selected temperature compensated model) as a reference. 

It doesn't really matter what the exact voltage is for digital calibration. All you need is something stable.

There's no need for laser trimming to an exact value or anything like that.

Sure, I'm not concerned about the value.  However, either buried zeners with temperature compensation or on-die heating are used in most decent bench meters these days or else a bandgap reference for those on the lower end.  For example, the 8808A which replaced the Fluke 45 uses an LM399 while the much older 8800A used the SZA263.  The Fluke 45 is a 1990's product, not 1960's.  The only digital meter I have with a plain temp-compensated zener is the 1968 era Fairchild 7000A which has surprisingly good specs--about the same as the 45--but needs a long warmup to acheive them.  In any case, I've run the Fluke 45 overnight and it seems completely stable with no signficant tempco over a 3-4C range.  Apparently the selected temp-comp zener does the job.