Strange variations of off the shelf LPM Panel Meters....

[lasersbee]

While working on the latest testing and developments of our
soon to be released DL-Module™ we discovered that not all
Digital Panel Meters are calibrated or created equal....

These are the Digital Panel Meters that are used in the older
Kenometers... nospin LPM... the Alpha and even our own Limited
Edition 5Watt LaserBee with Data Logging...

We hooked a few (brand new out of the box) in parallel... Then
used 2 Regulated variable power supplies and set one to 5volts
to power the Displays... and set the other regulated power supply
to an arbitrary output value..

We also hooked our Fluke DMM in parallel to the DPM inputs for reference...

There seems to be quite a variation from true voltage on some
of the DPMs (and only a small sampling) that makes me wonder
if that would falsify the LPM readings...

Granted some of the variations are minor... but we are trying to make
the DL-Module™ as accurate as possible to the voltage being applied
at the Digital Panel Meter terminals...


Jerry

 

[Trevor]

The worst of those is off by a little over 1.7% (assuming your DMM is 100% accurate). Even high end LPM's state in specs that they are +/- a few percent...

If the voltage signal to the panel meter were what's being calibrated, I could see this being a big problem. If this were the case, accurate signal would go to the panel meter and be falsified.

However, the output of the panel meter is what's being calibrated against some sort of trusted source (at least in LaserBees and Radiant Alphas, I'm not sure what nospin does?). If the signal coming out of the sensor is tuned so that the output of the panel matches a panel meter, then the panel should be fairly accurate (obviously not 100%).

As long as your calibration tool allows a user to calibrate across a full range, accounting for non-linearity, I don't see any problems.

-Trevor

 

[lasersbee]

In the case of the Nospin and our 5Watt that uses a calibrated (by OPHIR)
Thermopile head at 1mV=1mW... the variations of the DPM will effect the
actual calibrated reading...

Like I stated above.... Granted some of the variations are minor...

Thought I would share my findings....
Maybe I'm just a perfectionist....


Jerry

 

[Trevor]

Fortunately, your observations about these panel meters (except for the one over 1%) show less variation than the advertised uncertainty of a Coherent LabMax LPM.

I'm sure the sensors add far more uncertainty into the mix.

-Trevor

 

[lasersbee]

Yeah... and if you keep adding to those uncertainties by introducing
more errors... at the end if the line it will be difficult to see what
the actual reading really is...

I'm not talking about High End new professional Lab Quality LPMs...
I'm talking about the LPMs that were sold here on LPF that use those
Digital panel meters... Who knows how far off the older LPMs were...
and these DPMs don't ADD to the accuracy...

That's all I'm saying....


Jerry

 

[Bluefan]

It's an error, not a total uncertainty. The calibration will compensate for most of this (unless the drift this way is completely unpredictable). But I'm a bit amazed, I usually get 3.5 digit panel meters that are within 1 count correct. I'm not sure about the 4.5 digit panel meters, I haven't used much of them.

 

[lasersbee]

If you look at the 1st post pic... it is the 3-1/2 digit (orange)
DPM that seem to be off the most...
I just grabbed the DPMs I had Brand New in the shop to test
them...
Before doing the tests I was expecting 0-3 counts of error...
but when I saw those numbers... I thought I would share my
findings...

If you have a "new" Calibrated by OPHIR head that is calibrated
to output 1mV/1mW.... I would not go into the OPHIR head's
internal electronics to compensate the readings of a cheap
Voltmeter DPM...

But maybe that's just me...

We only sell one Limited Edition LPM that uses those 4-1/2 DPMs
and the "new" OPHIR heads... but we also include a Zeroing Control
to compensate for any drift...


Jerry

 

[Benm]

Its fairly common for these panel meters to have high tolerances.. often they are also mentioned in the datasheet. For a cheap model, they would be 1% accurate, plus being off by a couple (or even 5) of least significant digits.

In such a case, 1.000 volts could read somewhere between 0.985 and 1.015 within tolerances.

 

[lasersbee]

I understand that 1% is not much... but we are working on a
project that uses the input voltages to those Digital Panel
Meters to record data and the sensor we use now is much more
accurate..
We first thought it was our design or sensor... Then we did the
DPM tests.

It turns that out our sensor solution is spot on and it is the DPM
that is skewed...

Jerry

 

[Trevor]

I understand that 1% is not much... but we are working on a
project that uses the input voltages to those Digital Panel
Meters to record data and the sensor we use now is much more
accurate..
We first thought it was our design or sensor... Then we did the
DPM tests.

It turns that out our sensor solution is spot on and it is the DPM
that is skewed...

Jerry

When you tested the panels, did you find that the variation was consistent across the full range, or did it change depending on the voltage applied?

-Trevor

 

[HIMNL9]

Well, all those panel meters are produced in a big lot of pieces, and calibration is usually made from machines ..... or from "quick reading" process ..... and all gives a 2% tolerance in the specs, so the manufacturer is ok, about this (i mean, he don't sell them as high precision 0.01% instruments, he sell them as generic use instruments, so we cannot complain about that)

And also the vibrations from transport can displace a little bit the trimmers, also the 10-turns ones, so one cannot be sure at 100% to receive a calibrated product, from commercial selling / shipping lines.

They can, anyway, be corrected with the usually present trimmers (as i do for all the applications where i need some real precision), but this require extra time and to have a surely calibrated instrument for the comparison, ofcourse.

 

[lasersbee]

Well, all those panel meters are produced in a big lot of pieces, and calibration is usually made from machines ..... or from "quick reading" process ..... and all gives a 2% tolerance in the specs, so the manufacturer is ok, about this (i mean, he don't sell them as high precision 0.01% instruments, he sell them as generic use instruments, so we cannot complain about that)

And also the vibrations from transport can displace a little bit the trimmers, also the 10-turns ones, so one cannot be sure at 100% to receive a calibrated product, from commercial selling / shipping lines.

They can, anyway, be corrected with the usually present trimmers (as i do for all the applications where i need some real precision), but this require extra time and to have a surely calibrated instrument for the comparison, ofcourse.

Yes I understand that those meters are probably still in tolerance
of the stated specifications for each different meter..

I'm not complaining....
I'm just saying that I found the variations more than what I originally
expected/thought between meters without doing the math or checking
the specs...

Good points on shipping vibrations... and yes.. I know we can re-trim
them with the on board Trim pot to make them a bit more precise...:beer:


Jerrry

 

[Benm]

If there is any on board adjustment available, just use that, somewhere in the middle of the expected range. Linearity might also be an issue, but if you get the center value correct, i doubt that would ever be more than 1% off, except where the tolerance in the least significant digit becomes a problem.

The meters use ADCs to make their readout, and it can be done on a 0-1999 scale with an 8 bit dac and some noise injection and oversampling. Considering the price of these panel meters you can hardly expect more.

The one that reads 1.763 is probably a case of a shifted potmeter, the other are within what i would expect of them.

 

[Bluefan]

Most panel meters I use are 20V FS and based on the ICL7107 chip, usually they're correct within one count, I just tested one a 5V.
It's has a +/- 2000 counts intergrating ADC, definately better than an oversampled 8bit ADC.
The 4.5 digit panel meters haven't got a better precision, mine was 10 counts of and 5.000V. The linearity may just be good enough and the calibration incorrect, I haven't checked that. But the dynamic range alone of the 4.5 digit panel meter is usefull and a 0.2% error isn't that bad after all.