A cautionary tale for DVM users, and some stuff on instruments.

One of the first things I learned about moving-coil multimeters was to treat them gently, avoid 'pinning the needle' by starting on a high setting and working down, assuming the circuit to have more of whatever than you expected.

You get used to the things a moving-coil meter does when mis-used, like accidentally looking at the mains on a DC volt range and seeing the needle madly quivering around zero. And they hum or buzz to highlight the fact.

Then came the Digital Volt Meter and New Age technicians, a kind of silicon version of the valve jockey who could fix any TV as long as it was a failed valve heater. They don't need books or instruction but will figure it out themselves leaving a trail of wreckage in their wake. When working on other people's equipment the first priority is a personal learning experience, not an overriding duty of care to the owner.

The modern version discovers TTL, asks a question about something you've been working with for twenty years, then explains where you got it wrong.

Enter James, who doesn't deserve anonymity.

Early in the computer revolution appeared a thing called a 'Superboard'. About a foot square this had a QWERTY keyboard, a 6502 processor and talked BASIC, using a TV or video monitor for display. At the time it was a very exciting development and a number of people jumped on these despite their considerable expense.

It was still early days and the only way you could get a Superboard was buy the bare kit and build it, but it was mostly just a board stuffing excercise with lots of IC's. So James got his kit and set about furiously stuffing it for two days in the office. (Not everything that goes on in volunteer community organisations is strictly 'on-topic', but as a volunteer it is still your time.)

So James gets it finished and bounces into my office searching for 5 and 12 volts at a few amps. This wasn't unreasonable as I was running mailing lists on an Alpha-16 minicomputer at the time and just happened to have a couple of spare power supplies sitting in the corner.

These were a 2U rack-mounting OEM supply intended to go closely integrate with the Alpha and supply all its various needs. It was even more closely coupled than a modern ATX supply and a classic series (non switching) supply - you could only lift one at a time. All connections except mains-in appeared on a big multi-pin Molex connector.

“I've just been working on those and have some clues if you want.”

“Nah, I'll work it out” and rushed off highly impatient for a result. I looked at the remaining spare and tried to be philosophical.

To rewind a bit; a few weeks before I suddenly started getting frequent and major crashes. It was stupidly simple but took a long time to figure out.

The Alpha's had a brace of cooling fans, and being of US origin these were 115 volt. This was supplied from across half of the 115/230 volt tranny primary. Correctly connected the two lines that went in among the logic were mains neutral, grounded at our front door, and mains active via the mains transformer primaries as an autotransformer, the 'top' half of which incidentally served the function of a mains filter.

One day, unknown to me, somebody swapped a long lead I didn't really need for a shorter one, and my troubles started. 'Upside-down' the filter was in the neutral and the fans beween 230 and 115. Yes it was a simple active-neutral crossover putting unfiltered active into the logic frame, but the chase led me through the power supplies in detail, hence my offer.

So James fires up the power supply on his bench and starts probing the molex for the connections which I had just offered and he refused, with his trusty DVM. No Old Hat analogue for this Digital infatuate, thanks.

Some time passes then there is a brilliant blue-white flash and the circuit-breaker leaps off the switchboard.

Naturally having nothing better to do now the computer has been de-powered in the middle of a run, I stroll into James' domain to find him still trying to absorb what has happend. The flash was the Superboard 3AG fuse going off like a Xenon flash, and James is trying to convince himself this isn't a bad omen.

Still in shock he tells me the awful truth. He has grounded his DVM to the supply case and probed all of the 50-odd molex connections, but he can't find any 5 or 12 volts. In fact he can't find any voltage because the supply is alive but hasn't been told to wake up yet, ATX-style.

“But wait, what's this?” One pin with a small voltage that drifts up and down a few volts around zero. What he does next explains why James will never amount to anything in electronics (and perhaps engineering generally).

“Since that is the only place I can find anything” reasons James “that must be where I connect the +5 volt rail” (of my brand-new $350 and about 20 hours work, computer). So he did, and switched on. The next few milliseconds as above.

You may have now guessed what he was actually looking at was the 115VAC fan power line. It may have been limited by the power tranny primary in series but he must have hit it on the wrong half cycle and forward biassed the bulk diode on every chip on the board. I've seem limited situations where one chip fails and 'crowbar' protects the rest by shorting the lines, but even a MOV would have had a hard time in this case.

So one moral of this story is to know your instruments, particularly their limitations, which may trap you. What James didn't know is that most DVM systems sample in a way that tries to deliberately cancel out mains frequencies when measuring DC. The low drifting DC value James saw on the 115VAC was mainly the imperfections of this cancelling process.

My bench is populated by a few cheap DVM's, a mid-range moving-coil multimeter, a VTVM (Vacuume Tube VoltMeter), a CRO (Cathode Ray Oscilloscope) and audio millivoltmeter. Others less used such as the Megger and Digital thermometer are on the shelf.

Connected to a single point, say the anode of an early preamp stage, each has its own story to tell. Like any group of witnesses their accounts differ somewhat and it is your job as detective to sort the overall truth out from these individual truths.

There are many cunning ways to use instruments, CRO's in particular, to isolate the information you are looking for. But along with these advanced techniques and tricks you need to have a good understanding of the limitations of each of your instruments or you will waste a lot of time chasing shadows and your own tail.

One of the most common shortcomings is bandwidth. If you are working in the range 20Hz to 20KHz you really need surrounding instruments that are flat a decade either side, 2Hz to 200KHz, and quite a bit more in your CRO if you want to catch HF instability.

Many research labs these days operate using traceable standards, so called sub-standards, and blind result comparisons. Most electronics workshops tend to take their instrument specs on trust, which is fine until the first time the apprentice blows it up and it is repaired but not recalibrated. And not just the apprentice. ;)

Try this; connect your CRO, DVM, m/c meter, and whatever else you've got, to the output of your signal generator. Now vary the frequency and watch how they all respond. Perfect agreement from 20Hz to 20KHz? Or does the DVM have its own ideas as the frequency goes up? Or is your sig gen rolling off? Which can you trust?

Test arrangement are often dodgy with clipleads and so on, and if you're deep in the middle of something and start getting odd results it often pays to quickly revert to straight-through and check the outside environment is still the same and the ground hasn't dropped off the sig gen or something similar.

The three main standards needed in a workshop are resistance, voltage and frequency. Current and time can be derived from these. If you work with RF you really need standard capacitors and inductors as well, but they can be avoided for most audio work.

These days resistance is the simplest with high tolerance resistors almost becoming standard and 1% freely available and cheap.

Similarly voltage has become easier with first three-pin regulators, and later band-gap references of very high accuracy and stability, also very cheap.

For many years now the American standards stations WWV and WWVH have been transmitting atomic standard derived 1 second ticks, 440 and 1000 tones, all on similar carriers on 5MHz, 10MHz and some other frequencies. By tuning in WWV(H) on a shortwave receiver the master oscillator in the bench frequency counter can be set to zero-beat with the carrier, and (for an instant at least) your tuning will be atomically accurate.

Now you can tell cheap guitar tuners the frequency is and adjust it using the internal trim pot. And you thought all guitar tuners were crystal-locked. ;)

If you want to go to extremes, a New Zealand Ham built a highly stable high frequency oscillator, a “Vacar” circuit I think, sealed in a large thermos (Dewar) flask against atmospheric pressure changes, and buried under his shack in a car fridge filled with styrofoam for thermal stability. His excellent article in Amateur Radio magazine illustrates how difficult it is to get really high degrees of accuracy and stability in electronics, but that with care it can be done.

How often should you calibrate? How long do you want to be wrong?

Valid HTML 4.01 Transitional <<<OzValveAmps
http://www.ozvalveamps.org/superboard.htm | Last update: 21:57 24/09/05