On this site:

Philips PM2433 Digital Voltmeter

PM2433-front.jpg
This is the first nixie DVM I acquired. This digital voltmeter from 1967 must have been an expensive instrument. It uses a lot of discrete logic distributed over a number of circuit boards.

This digital voltmeter dates from 1967. The display consists of 4 ZM1030 nixie tubes, one ZM1031 "+/-" nixie tube and 4 small neon lamps for the decimal points. The nixies have nice large digits. It's logic circuitry is built up from discrete transistors. The DVM has 4 ranges: 0.3999 V, 3.999 V, 39.99 V and 399.9 V. It uses the "potentiometric principle" to measure the input. The display shows the value of a digital counter, the numerical value of this counter is fed through a D/A converter whose analogue output is compared to the input signal. The counter counts until the input value is reached, which is like turning a potentiometer until a desired value is reached, hence the name. It uses some kind of successive approximation instead of a linear increase of the counter to speed up the measurement cycle, which takes a maximum of 111 counts instead of 3999, the manual claims.

PM2433 inside

A lot of circuit boards inside.

I was lucky to receive a complete photocopy of the manual with the device. Without it, it would have been hard to figure out the connections and to make the necessary repairs. Finding your way in a TTL counter is one thing, but analysing discrete logic is something different. As the inside photograph shows, the DVM is crammed full of circuitry, despite its large size of 44x13x30 cm.

The designers have done their best to economise on components. The whole device contains only 160 transistors, including the power supply. Even the choice of the nixie tubes is a design trick to save transistors: the digits in a ZM1030 are divided in two groups, each group having a separate anode. By activating only one of the anodes, one group of digits is selected. There are only 5 cathode connections, each connected to two digits. Needing no more than 9 pins, the nixies fit into a Noval socket.

display board

Display board with one nixie,
one counter and driving circuitry.

The decoder circuitry uses 7 HV transistors to drive each nixie tube on their 2 anodes and 5 cathode pairs. So the nixie actually does part of the decoding, being driven with a bi-quinary code to display a decimal digit. A complete counter/decoder decade has 17 transistors. Very clever.

Although the DVM appeared in good condition, I did get some strange readings when I connected a 9 V battery to the input. When connected to the two input pins, this did not give any reading, but when I connected it to one of the inputs and the signal ground, I did get a reading. On the high ranges I got a reading of 1.7 V, which was wrong, on the 3.999 V range I got an overload condition, which was correct but inconsistent. When I reversed the input voltage, the meter started cycling between a + and a - reading, but displaying "0000". The other input pin did not give a reading.

In the "calibration" position of the range switch, I got a reading of 3.899, which is only one count off. After I opened the case and tried again, the readings were correct. Using a variable input from my power supply to check the result for different inputs and ranges, I only saw 1 or 2 counts difference from my modern pocket DVM with LCD display. Not bad.

So the measuring circuitry did function but the input voltage did not arrive there from the input plug the way it should. This led me to suspect the input polarity relay. This relay consists of two SPDT (single pole double throw) reed relays connected as a cross-over switch to reverse the input signal if necessary.

Range switch assembly

Circuit board with range switch.

It is mounted with the range switch on a sandwich of two circuit boards. I pulled out that subassembly and checked for continuity to discover that one the reed relays was faulty. Its NC (normally closed) contact was open and the switching reed did not move. Interestingly, two PCB tracks leading to the relay were also damaged, one burnt, the other interrupted. One was the "-" input pin, the other the connection from the relay to signal ground. Apparently, somewehere in the past of this meter, something went wrong and the reed relay shorted some powerful input signal, damaging the PCB and one of the reed relays. Thinking of it, connecting a voltage from a source that can deliver some Amps of current between signal ground and the "-" input pin (the one that normally connects to signal ground), may result in a damage just like this.

In the process, I decided to check the input connections, especially the two ground connections. The PM2433 has a complete separation between signal ground and safety ground ("Earth" from the power socket). Checking things out, I discovered that there was no continuity between some of the metal parts of the case. Tightening some screws improved this. The device must have been stored in a less suitable environment, witnessing some rusty screws and a few dead spiders and cobwebs I had to clean out.

reed relay

The reed relay that was stuck.

Initially, I could not yet find a replacement reed relay. Now the reed is quite large (the glass tube is 5x40 mm), because it must be able to switch voltages up to 400 V. I temporarily replaced the relay by a miniature relay from a modem. This works, but I kept looking for a direct replacement. After a few years, I got a mail from a friendly person who had a numer of these relay. He sent me a few. But I still haven't finished the job..

Testing the DVM, I came to understand why I got some strange readings when I first tested it with a battery: it seems the input is shorted in tha overrange condition. This made the battery collapse to recover shortly after the event. Have to look into this.

Copyright © 2001 by Onno's E-page         published 2001-07-22