The Inaccuracy of Digital Meters

mshchem · Apr 10, 2019

AgX said:
Well, both are terms out of metrology, I consider rather puzzling in this context, out of themselves and due to varying terms in different languages.

Precision means the range of a values yielded after repeated measuring of the same state.

Accuracy means the deviation of the median of the range of precision.

See also this graphic:
https://en.wikipedia.org/wiki/Accuracy_and_precision#/media/File:Accuracy_and_precision.svg

Thus accuracy is NOT the greatest error...

Now you likely will understand why I do not consider these terms helpful here.

Every time someone reports a finding without a understanding of the uncertainty. It loses credibility.
When I went into manufacturing in the US after working in a lab. I was stunned. One thing was the nightmare of English measures, BTUS etc versus SI. Secondly no error determination associated with a statement. People would compare 0.124 vs 0.120 and thought they had made some breakthrough . I would say they were both 0.1 It would drive these guys crazy. I think it's directly related to the development of electronic displays. People think the more digits they report the better the result.

Jim Jones · Apr 11, 2019

Gee, can't my slide rule indicate the value of pi to 3.14159. . . . . . . .

RalphLambrecht · Apr 12, 2019

AgX said:
Well, both are terms out of metrology, I consider rather puzzling in this context, out of themselves and due to varying terms in different languages.

Precision means the range of a values yielded after repeated measuring of the same state.

Accuracy means the deviation of the median of the range of precision.

See also this graphic:
https://en.wikipedia.org/wiki/Accuracy_and_precision#/media/File:Accuracy_and_precision.svg

Thus accuracy is NOT the greatest error...

Now you likely will understand why I do not consider these terms helpful here.

makes sense.

Malcolm Stewart · Apr 12, 2019

Higher up in this thread, there's a mention of a health professional...
Several decades ago my 3 year old daughter was diagnosed as having leukaemia, and the treatment then was an injection of a drug as part of the chemotherapy. The treatment started in the evening, and the agency nurse was not able to work out the reciprocals needed for the correct dose. Fortunately, thanks to my darkroom practice, I hadn't forgotten basic arithmetic, and the correct dose was administered - and my daughter is still here. But, I guess it was a near thing.

KenS · Apr 12, 2019

I was offered (FREE) a new 'small' digital scale (for weighing out the chems for Pyrocat Hd in my darkroom) I declined the offer since I believe my 'smallish' 3-beam balance to be much more accurate... and does NOT need batteries

Ken

Deleted member 88956 · May 4, 2019

faberryman said:
Why was he using a 200g scale to measure 105mg? Therein lies the negligence.

Exactly !

Deleted member 88956 · May 4, 2019

BrianShaw said:
Woulda jury concur with that as a good definition for negligence?

It should, but that is not how juries end up working out a lot of times.

Deleted member 88956 · May 4, 2019

AgX said:
Again and again I read at Apug quite a nonsense relating to digital readouts of scales and thermometers.
This dramatic case hopefully makes you aware of the potential errors:

These days in Germany a healing practitioner is at criminal court for killing 3 of his patients by applying by negligence an overdose of a drug.

His aim was to apply by infusion 105mg of that substance and he had to weigh the substance before.

He employed a digital scale from a laboratory scale manufacturer (Kern PCB 200 at about 250€)

of

200g max load

and

-,-- g display

and

d= 0,01 designation

-) He did not realize that with any digital readout the last digit (10mg in this case) is insecure and can be +/- 1digit

-) He did not realize that, more important, a metering device by design may incorporate an error by far greater than that last digit, especially at the verge of its range.

A expert witness showed that the scale in question at 1160mg can have an error of 340mg. Thus 30%.

I looked up at the data sheet:

linearity: 20mg

minimum load: 20g

(Linearity here means the max. absolute error over the complete load range. In this case it is lower than the error shown by the expert. But that data sheet error only applies at the allowed load range, here thus above 20g.)

I hope this makes things more clear...

A pun saying amongst german electronic technicians is "Wer mißt, mißt Mist".
Which can be translated as "Every metering results in bullshit". This is applied on electronics metering where the effect of the circuit on the metering complicates things, but doubt on what one is doing should be applied on any metering.

This brings up good points to those who want to be a "doctor" (darkroom or operating room), but do not fully understand the equipment limitations or even effect of range of measurement vs. expected measured values.

The case is quite a drastic departure from majority (and nearly all) of darkroom requirements, but does highlight some few not so apparent inherent problems with digital displays. But, and in light of darkroom use specifically, how does that compare to any analog equipment's expected/required/actual accuracy? Majority of such require at least a consistent way of looking to get comparable/repeatable readouts as most are affected by parallax. A needle on a dial thermometer can be easily read to worse than 0.5 a degree accuracy, depending how one looks at it. Same goes for an analog lab scale where a pointer must be judged of its location vs. set reference marker. Then using a 100 degree range thermometer for a 20 degree measurement will have similar error effect as this "doctor" using scale way beyond weight applied.

Moral to me is: know the limits of technology used and take no specification for granted without understanding what it actually means.

Theo Sulphate · May 4, 2019

BMbikerider said:
...
In testing them, I used the white vertical blinds in my living room back-lit by the natural daylight outside. It is a nice even light ...
.

In the Puget Sound region of Washington state, "sunlight" is often a theoretical concept and the luminosity in a room is unlikely to be consistent from one hour to the next.

What I use for reference, out of total convenience, is a "bathroom" which has six clear medium wattage bulbs on a bar across a mirror; the light from them, reflected on the dull white countertop, has been remarkably consistent over the years. Checking with my Sekonic L308S, the proper exposure off this surface is always f/5.6 at 1/30s, ISO 200. Whenever I test a camera's meter, I first perform a reference check with the Sekonic.

AgX · May 4, 2019

Witold said:
... how does that compare to any analog equipment's expected/required/actual accuracy? Majority of such require at least a consistent way of looking to get comparable/repeatable readouts as most are affected by parallax. A needle on a dial thermometer can be easily read to worse than 0.5 a degree accuracy, depending how one looks at it. Same goes for an analog lab scale where a pointer must be judged of its location vs. set reference marker. Then using a 100 degree range thermometer for a 20 degree measurement will have similar error effect as this "doctor" using scale way beyond weight applied.

The point is that digital readout provoke a feeling of security in reading.

With analog readouts, some issues are more or less obvious, as having to find the magnitude of the figures on the scales, to interpolate between values, having to cope with mirror scales etc. Some of us may even remember clicking on a barometer before taking a reading....
All this means at least some activity, and hopefully, questioning oneselves.

With digital reasout one gets a seemingly absolute true value. One even may lead that via USB to a computer and having those machines sorting it out...

Theo Sulphate · May 4, 2019

Students have so much confidence and familiarity with results being displayed in digital format, that often they don't think beforehand what the expected valid range of values should be. This problem is called innumeracy, "mathematical incompetence":

https://en.wikipedia.org/wiki/Innumeracy_(book)

Deleted member 88956 · May 4, 2019

AgX said:
The point is that digital readout provoke a feeling of security in reading.

With analog readouts, some issues are more or less obvious, as having to find the magnitude of the figures on the scales, to interpolate between values, having to cope with mirror scales etc. Some of us may even remember clicking on a barometer before taking a reading....
All this means at least some activity, and hopefully, questioning oneselves.

With digital reasout one gets a seemingly absolute true value. One even may lead that via USB to a computer and having those machines sorting it out...

I fully agree, so long as the analog person is aware of potential pitfalls of looking at his display in a wrong or inconsistent way.In the end it is still about awareness of tools and processes, although false sense of data accuracy from a digital display adds another layer to this.

Sirius Glass · May 4, 2019

I bought a Pentax Digital Spotmeter in great condition, but it was off so I sent it to Quality Light Metric and had it calibrated.

7095 Hollywood Blvd
Ste 550
Hollywood, CA 90028
Phone number (323) 467-2265

AgX · May 4, 2019

A meter being off can happen to both, analog as digital meters. Such has nothing to do with the inherit way the read-out works or is looked upon.

Sirius Glass · May 4, 2019

Of course but I was staying on topic. If I had mentioned analog meters too I would have been off topic.

Chan Tran · May 4, 2019

RalphLambrecht said:
most interesting thread but, can somebody knowledgable now summarize how to use a digital scale properlyplease?

Simple he should read the specs and use the correct scale needed for the required accuracy.
Most digital readout instrument would specify the number of bobbing digits which means the number of the last digits that you can't count on. Another specs is the percent accuracy for the range. Most instruments with digital readout have higher resolution than their accuracy.
While the experts said it can be off by 340 mg at 1130 mg. Except for the specs requires at least a 20g so what he weighted was less than the minimum required so I think all bets are off. With weight higher I would expect the accuracy of +/- 30mg.

Alan Edward Klein · May 4, 2019

I bracket my medium format shots. It's cheap insurance not so much from meter inaccuracies but from my own errors where I point it.

Deleted member 88956 · May 5, 2019

Chan Tran said:
Simple he should read the specs and use the correct scale needed for the required accuracy.
Most digital readout instrument would specify the number of bobbing digits which means the number of the last digits that you can't count on. Another specs is the percent accuracy for the range. Most instruments with digital readout have higher resolution than their accuracy.
While the experts said it can be off by 340 mg at 1130 mg. Except for the specs requires at least a 20g so what he weighted was less than the minimum required so I think all bets are off. With weight higher I would expect the accuracy of +/- 30mg.

Exactly. It was negligence indeed, the guy gives bad name to all "healers" out there who just stick to cannabis which is administered as one pleases and at some point any measurement errors become moot.

Chan Tran · May 5, 2019

AgX said:
Again and again I read at Apug quite a nonsense relating to digital readouts of scales and thermometers.
This dramatic case hopefully makes you aware of the potential errors:

These days in Germany a healing practitioner is at criminal court for killing 3 of his patients by applying by negligence an overdose of a drug.

His aim was to apply by infusion 105mg of that substance and he had to weigh the substance before.

He employed a digital scale from a laboratory scale manufacturer (Kern PCB 200 at about 250€)

of

200g max load

and

-,-- g display

and

d= 0,01 designation

-) He did not realize that with any digital readout the last digit (10mg in this case) is insecure and can be +/- 1digit

-) He did not realize that, more important, a metering device by design may incorporate an error by far greater than that last digit, especially at the verge of its range.

A expert witness showed that the scale in question at 1160mg can have an error of 340mg. Thus 30%.

I looked up at the data sheet:

linearity: 20mg

minimum load: 20g

(Linearity here means the max. absolute error over the complete load range. In this case it is lower than the error shown by the expert. But that data sheet error only applies at the allowed load range, here thus above 20g.)

I hope this makes things more clear...

A pun saying amongst german electronic technicians is "Wer mißt, mißt Mist".
Which can be translated as "Every metering results in bullshit". This is applied on electronics metering where the effect of the circuit on the metering complicates things, but doubt on what one is doing should be applied on any metering.

In this case I think everyone agrees that it was the guy fault. However this is not because the digital readout. The readout would clearly show that it has the resolution of 10mg (that is what it can display and not speaking of accuracy) and if he want to measure 105mg with that is his fault. It's not the digital display that gives the false sense of accuracy. It's more likely that such a guy would have tried to guess between the mark when he used a scale with analog pointer.

MattKing · May 5, 2019

It occurs to me that a better title for this thread might be something like: "How digital readouts can mislead you".

Chan Tran · May 5, 2019

Perhaps that is what the OP intention but in this case it's not. The digital readout doesn't mislead. When you put a 105mg weight on the scale and it read 0.10g or 0.11g how do you decide that it is 105mg?

AgX · May 5, 2019

MattKing said:
It occurs to me that a better title for this thread might be something like: "How digital readouts can mislead you".

In my comments I emphasized that misleading effect.

However both analog and digital meters have inherit shortcomings. They though differ.
See that last-digit issue. Or the minimum load issue. I got an amateur beam balance, with knife-on-glass bearing which has much lesser minimum load than those laboratory digital balances.

Deleted member 88956 · May 6, 2019

Chan Tran said:
Perhaps that is what the OP intention but in this case it's not. The digital readout doesn't mislead. When you put a 105mg weight on the scale and it read 0.10g or 0.11g how do you decide that it is 105mg?

This only further supports the moral of the story ... know your equipment and use it within its limitations and never try to outsmart it.

KN4SMF · May 6, 2019

Digital displays are nothing but a cheap way to manufacture something in many cases. Accuracy is incidental. But is satisfies its targeted market long enough to retain the profits without being overwhelmed by returns and refunds. Fortunately in many cases, the product actually was highly accurate. I recall a cheapo chinese frequency counter I recently bought. It was about $5.00 or so and it took a month to get here. But after using it and testing it against different crystal oscillators, broadcast stations, and WWV, it turns out the thing was dead on the money down to the thousanth of a megacycle or kilocycycle. I was astounded. That frequency counter allowed me to use my 50 year old low-end tube signal generator to align Hammarlunds and Hallicrafters to be dead on the money for sideband and CW. I'm talking about 1950's vacuum tube technology that I was able to calibrate to 2019 standards expectations with a $5.00 frequency counter.
Digital readout in a set of scales is an entirely different matter. I wouldn't trust it one iota to compete with my Ohaus triple beams. I KNOW they're accurate down to the fraction of a gram. I guess if there was a moral here, is to know your equipment. Just because a digital readout says something and you haven't tested it, then that's just irresponsible.

Theo Sulphate · May 6, 2019

KN4SMF said:
... I recall a cheapo chinese frequency counter I recently bought. It was about $5.00 or so ... turns out the thing was dead on the money down to the thousanth of a megacycle or kilocycycle.
...

If you're young enough, for an extra $1 they'll include a conversion chart fron Kc, Mc to KHz, MHz.

The Inaccuracy of Digital Meters

Deleted member 88956

Deleted member 88956

Deleted member 88956

Deleted member 88956

Deleted member 88956

Moderator

Deleted member 88956

Privacy & Transparency

Privacy & Transparency