EDN visits a calibration lab

Article By : Martin Rowe

At Transcat's Rochester, N.Y. lab, we gave an HP 34401A DMM a checkup. The meter's readings were disappointing, but it was a worthwhile reality check.

As explained in Calibration: Driven by regulations, test & measurement equipment needs calibration to give you confidence that measurements are within required limits. Otherwise, you can get false positive or false negatives. Neither is desirable.

After getting a tour of Transcat‘s Rochester calibration lab from service manager Fred “Dusty” Tank, the time had come to get a checkup for an HP34401A. The most recent calibration sticker was dated 2008. I don’t recommend going that long between calibrations. Here’s why.

For purposes of the video below, technician Steve Trotta (Figure 1) performed a manual calibration, so we could see the steps involved. Most often, calibrations are performed automatically using Fluke METCAL software. METCAL not only controls the calibrator and UUT, it records the measurements and can create calibration reports.

Transcat Steve TrottaFigure 1. Technician Steve Trotta prepares for a calibration.

Trotta started by checking the meter’s offset and noise. Figure 2 shows a Keithley 8620 shorting plug (datasheet PDF) inserted into the HP 34401A’s input jacks. With the meter stable, he connected it to a Fluke 5700A multifunction calibrator.

Keithley 8620 sorting blockFigure 2. A Keithley 8620 shorting plug is used to check an instrument’s noise and offset.

Trotta started the calibration with 10 VDC and 100 mVDC. At 100 mV, the difference in readings between the calibrator and meter was 3 µV. That’s within manufacturer’s specification shown in Figure 3. That’s impressive for 10 years.

textFigure 3. The HP 34401A is accurate to 3 µV with a 100 mV input.

Trotta also checked the meter at 10 VDC. Again, it was well within manufacturer’s specifications.

AC voltage didn’t go so well. Both the 100 mV and 10V (1 kHz) readings were low by nearly 12%. That’s well out of spec (Figure 4).

textFigure 4. HP 34401A AC voltage specifications

At that point, we stopped the calibration. Had this been a complete calibration, we would have checked all the 34401A’s functions and ranges. When you bring your equipment to a calibration lab, you’ll get a complete report on all measurements, comparing them to specifications, indicating which pass and which fail. You’ll also know which measurements were made using a test uncertainty ratio (TUR) that’s less than 4:1. Typically, calibrations are performed using equipment with a measurement uncertainty that is four times smaller than the equipment being calibrated.

What if a calibration revelas a measurement that’s out of spec? You can request that the lab adjust your instrument to bring readings within limits, but you should have a guard band. There’s no point in adjusting an instrument to just barely within published specifications. Should the instrument drift at all, you could end up with false positive or false negatives.

In the metrology profession, “calibration” simply means comparison, not adjustment. See Calibration means different things in different professions for an explanation by Transcat’s Chris Grachanen.

If, however, you never use a DMM to measure resistance and such measurements are out of specification, you don’t care, so there’s no point in adjusting it. In some instances, you might need a measurement uncertainty tighter than the manufacturer’s published specification. Be sure to let the lab know of any such requirements.

As for the out-of-range ACV measurements on the 34401A, I have the service manual. It may be worth attempting an adjustment using some other meter as a standard even if the TUR is unknown. I won’t have a NIST-traceable calibration, but it’s better than nothing and good enough for my needs.

Lesson learned
Give your equipment some kind of check before bringing it to the lab by comparing it to another piece of equipment (or two). Had I checked the HP 34401A ACV readings against one or two other meters, I would have at least known if the error was caused by transport. Alas, I’ll never know.

Martin Rowe covers test and measurement for EE Times and EDN. Contact him at martin.rowe@AspenCore.com

Leave a comment