Silicon Chip Online - MiniCal 5V Meter Calibration Standard

Recently, the need arose to recalibrate an expensive digital multimeter. As the job seemed quite straightforward, I decided to tackle it myself. Like most hobbyists, I don't have access to the high-accuracy voltage standards used in calibration labs. Nevertheless, I came up with a scheme that I thought would be accurate enough for general hobbyist work.

By hooking up five multimeters and two panel meters to a voltage divider across a battery, I figured that the mean reading should serve as a reasonable "standard". However, I was amazed to see that no two meters read the same and the range of values was much greater than I had anticipated.

Although the readings were probably within the specs for each meter, it was a sobering demonstration. In the absence of anything better, I calibrated my upmarket digital meter to the mean value but was determined to find a more accurate method that would give me some confidence.

Main Features

5.000V ±0.02% voltage standard
192.3mV ±0.2% voltage standard (optional ±0.1% or ±0.04%)
Two ±0.1% resistor standards (optional ±0.01%)
Crystal-locked frequency reference
Crystal checker

The MiniCal solution

The Maxim range of IC voltage references proved ideal for this purpose. In particular, the MAX6350 +5V DC reference boasts a very impressive untrimmed accuracy of ±0.02%, with an extremely low temperature coefficient of 0.5ppm/°C.

Generally, voltmeters are calibrated on their lowest DC range (200mV for 3.5-digit meters). The "MiniCal", as this new project is called, divides down the MAX6350's +5V output to generate a 192.3mV reference.

In addition, the board includes a crystal-locked oscillator for checking meters, oscilloscopes and the like. The frequency of the oscillator is determined by crystal selection.

How it works

Fig.1 shows that the circuit consists of two completely separate sections.

With slide switch S1 in the lefthand position, battery power is applied to the oscillator section. Some readers may recognise this circuit and, in fact, it's based on the "Simple Go/No Go Crystal Checker", originally published in the August 1994 edition of SILICON CHIP.

The basic Colpitts oscillator used in the original design proved ideal for the frequency reference section of the MiniCal. Although not strictly necessary, the circuit has been reproduced in its entirety, meaning that it can also be used as a crystal checker if so desired.

Crystal X1, the 150pF capacitor between Q1's base and emitter, and the 100pF capacitor to ground together form the feedback network. The output from Q1's emitter is AC-coupled via a 1nF capacitor to the "FREQ" test pin.

Although we've specified a 10MHz crystal for X1, the circuit should work with values from 1MHz to at least 21MHz without modification.

The remaining circuitry connected to Q1's emitter performs the crystal "go/no go" function. Diodes D1 & D2 and the 100nF capacitor rectify and filter the AC signal from the emitter. The resultant DC voltage is applied to the base of Q2, switching it on and lighting the "OK" LED whenever oscillation is present.

Voltage reference

With switch S1 in the righthand position, the voltage reference section of the circuit is powered. This section is very simple and consists of only a voltage reference IC, three capacitors and two resistors.

The MAX6350 (IC1) can operate with an input range of 8-36V, providing an untrimmed output of 5V ±0.02% (4.999V - 5.001V). Small tantalum capacitors on the input, output and "NR" (Noise Reduction) pins reduce circuit noise to just 3.0μVp/p (typical) in the 0.1Hz to 10Hz spectrum. Battery-powered operation ensures that this is not degraded by external (conducted) noise sources.

Note: the MAX6350 is available in both 8-pin DIP and SO (surface mount) packages. The PC board design accommodates both package styles. We expect that most constructors will opt for the surface mount device, as it is cheaper and easier to obtain (see parts list).

Resistors R1 & R2 divide down the MAX6350's +5V output to obtain the 192.3mV calibration voltage. At a minimum, these resistors need to be 0.1% types (see parts list) to achieve the specified 0.2% voltage tolerance.

As you can see, the use of 0.1% resistors degrades circuit performance somewhat. However, the result is a good compromise between accuracy and cost, and is sufficient for meter checking. If you want to use the MiniCal for calibration, then you will need to upgrade to tighter-tolerance resistors in order to meet the basic accuracy specs of your instrument.

Two alternatives for R1 & R2 are shown in the parts list. The 0.01% resistor pair gives a ±0.04% tolerance on the 192.3mV output but will set you back about $77. Alternatively, you can install the 0.05% 25:1 divider network for a tolerance of about 0.1% and a much lower cost of just $18.

Note: the 25:1 divider network consists of two 0.1% resistors (1kΩ & 25kΩ) with a ratio accuracy of 0.05%. The device is supplied in a 3-pin surface-mount (SOT-23) package.

So why did we choose an odd calibration voltage of 192.3mV instead of a nice round figure? Well, it was simply a convenient choice using available resistor values. Other division ratios could be used but for best results the reference voltage must be close to (but not exceeding) 200mV.

Construction

All parts mount on a single PC board coded 04112031 - see Fig.2. If you have surface-mount devices for IC1 and/or R1 & R2, these should be installed first (see Fig.3). You'll need a temperature-controlled soldering iron with a fine chisel tip and small-gauge solder for the job. A bright light, magnifying glass and 0.76mm desoldering braid ("Soder-Wick" size #00) will also prove useful.

Next, on the top side of the board (see Fig.2), install all components in order of height, starting with the wire link, resistors and diodes (D1 & D2). Obviously, if you've mounted the R1/R2 divider on the bottom side, then you shouldn't install anything in the R1 & R2 positions on this side!

Note that all the tantalum capacitors are polarised devices and must be inserted with their positive leads aligned with the "+" symbol marked on the overlay.

Install the battery holder last of all. It should be fixed to the PC board with No.4 x 6mm self-tapping screws before soldering.

To complete the job, attach small stick-on rubber feet to the underside of the PC board to protect the assembly as well as your desktop.

Operation

Due to the expected intermittent use of the MiniCal, a power switch has not been included. Simply plug in a battery and use the slide switch to select between the oscillator function ("FREQ") or voltage reference function ("VOLTS"). Note that the battery voltage must be at least 8V for correct operation of the reference IC.

When measuring the oscillator frequency, the crystal checker function must be disabled by removing the jumper from JP2. This is necessary because the checker circuit loads the oscillator, reducing the signal on the "FREQ" test pin below the sensitivity level of most multimeters.

Follow the instructions provided with your multimeter regarding calibration. In general, most multimeters should be calibrated on their lowest (basic) range, which is normally 200mV for 3.5 digit models.

As described earlier, accuracy will be about ±0.2% using ±0.1% resistors for R1 & R2. This figure is good enough for many general-purpose instruments, which typically specify an accuracy of ±0.25% at best. Note that calibration instructions usually specify a standard of ±0.1% or better.

Calibration is normally only applicable to the basic range, with all other ranges depending on that calibration. The 5V output and 0.1% resistors should therefore only be used to check the accuracy of your meter, not to calibrate it. Note that, in use, the jumper shunt (on JP1) must be removed before measuring the 0.1% resistor values.

Note also that some meters may require special tools and/or knowledge for successful calibration. When in doubt, read the (service) manual first!

Meter loading effects

A resistive divider was chosen to generate the millivolt source because it's simple and requires no adjustment.

However, the down side to this simplicity is that the meter's input impedance loads the divider network and therefore reduces the reference accuracy.

For example, when a meter with a 10MΩ input impedance is connected, the reference voltage will fall by about 0.02mV. This corresponds to a 0.01% reduction in accuracy.

Assuming you know your meter's input impedance, the loading effect can easily be factored into the calibration where maximum accuracy is required.