Sppeding up a voltmeterPieter-Tjerk de Boer, PA3FWM email@example.com
(This is an adapted version of part of an article I wrote for the Dutch amateur radio magazine Electron, September 2021.)
The ICL7106 voltmeter chipEven if you've never heard of the ICL7106, chances are you have several in your home. It's a very common chip for building a 3.5 digit voltmeter with LCD readout (ICL7106) or LED (ICL7107). These chips are at the heart of many digital multimeters, and can also be found in voltmeter modules in e.g. a bench power supply.
This chip is already quite old. It was designed in the late 1970s by Fluke, a manufacturer of professional test equipment also back then, for their type 8020A digital multimeter . Fluke had Intersil manufacture the chip. Intersil expected a bigger market for the chip, removed a small part from the design which was patented by Fluke, and marketed the resulting IC as the ICL7106. Even the Fluke logo was still on the chip! Of course this lead to a lawsuit, which was settled. Since then, the ICL7106/7 (from Intersil, or a clone from another manufacturer) is practically the standard solution for a simple digital voltmeter.
In fact, such a digital voltmeter is an analog-to-digital converter, whose output is not a bunch of binary digits (bits), but decimal digits. In case of the ICL7106, the decimal output is between -1999 and +1999, so some 4000 possibilities. Thus, the chip is comparable to a 12-bit A/D converter (since 212=4096), doing about 2 conversions per second.
By current-day standards, these are very modest specifications for an A/D converter. Compare this to the sound input of a personal computer: already since the 1990s those feature A/D converters with at least 16 bits and at least 44100 conversions per second. And for SDR we use even much faster ADCs. But for a multimeter 12 bits and a few conversions per second suffice.
Accelerating a voltmeterMy bench power supply has two 3-digit meters, for voltage and current, each measuring about twice per second. But that's inconveniently slow. At this rate, one can easily overshoot while setting the output voltage; and if one's circuit draws too much power, one would rather see this as soon as possible. In short, can't we make those meters faster?
The ICL7106 datasheet explains this chip's timing. It has an RC oscillator, whose frequency is first divided by 4. From the resulting signal, 4000 periods are needed for a single measurement. That's beause the ADC uses the "dual slope" principle, see the figure. At first, a capacitor is completely discharged. Next, during 1000 cycles, the capacitor is charged with a current that is directly proportional to the voltage being measured. Finally, the same capacitor is discharged with a precisely known current, and the number of cycles until complete discharge is counted. That number shown on the display. And then this whole process repeats.
The R and C of the RC oscillator are connected as external components to the chip. The datasheet suggests 100 kOhm and 100 pF, resulting in about 40 kHz and thus about 2.5 measurements per second. Equipment manufacturers of course blindly follow this. But according to this same datasheet, higher frequencies are allowed, up to 240 kHz, which would give 16 measurements per second, although it's questionable the human eye could follow that.
I haven't been able to ascertain whether my power supply really has ICL7106 chips, because the type number is hidden from view. But since on pins 38, 39 and 40, which in case of the ICL7106 connect to the oscillator RC network, a squarewave of about 37 kHz was found, it most likely is. I've soldered another 100 kOhm resistor in parallel to the existing one, and now the meter measures twice as fast. That's much handier in daily use!
According to the datasheet, for optimal measurement acccuracy the integration capacitor discussed above should now be made smaller, so that it, despite the shorter times, still gets charged equally far. I didn't do this though, and haven't notice a reduction in accuracy.