0
\$\begingroup\$

I have an old XT computer with a crystal oscillator that I would like to test.

My multimeter has a Mhz setting that as far as I know should allow me to test it. (Providing the crystal is powered)

The problem is that most of the time during the measurement the computer crashes. About 1 out of 10 times am I able to see a constant measurement (the actual frequency also displayed on the oscillator), but only for a couple of seconds. Most of the time the pc crashes as soon as I touch 1 of the 2 leads.

I would like to ensure that it is operating correctly. Is this normal and as such not the proper way to test ? Is there another way to determine the crystal is in good working order ?

\$\endgroup\$
5
  • \$\begingroup\$ What is your question? \$\endgroup\$
    – Eugene Sh.
    Commented Feb 7, 2019 at 19:44
  • 2
    \$\begingroup\$ There is no question here but anyway: Crystal oscillators are easy disturbed. Your multi-meter puts an extra load on the pins and stops it oscillating. What do you want measure, after all you have seen the frequency already? \$\endgroup\$
    – Oldfart
    Commented Feb 7, 2019 at 19:45
  • \$\begingroup\$ Sorry if it was not clear ... edited the question \$\endgroup\$
    – ddewaele
    Commented Feb 7, 2019 at 19:55
  • \$\begingroup\$ The computer was running very slowly and unstable. Someone suggested looking at the oscillator. \$\endgroup\$
    – ddewaele
    Commented Feb 7, 2019 at 19:57
  • \$\begingroup\$ Is it a really-o truly-o IBM XT, or a close enough clone that it has a crystal and a chip? Or does it have a little oscillator in a square or rectangular can? If you put a multimeter lead on a crystal pin it'll load the oscillator circuit and kill the oscillations. If you put a multimeter lead on the output of a crystal oscillator it should be OK. \$\endgroup\$
    – TimWescott
    Commented Feb 7, 2019 at 20:26

2 Answers 2

5
\$\begingroup\$

Most of the time the pc crashes as soon as I touch 1 of the 2 leads. [...] Is this normal and as such not the proper way to test ?

Yes, that's expected behaviour. As already mentioned in comments, the actual crystal drive is very low power and so it's sensitive to any additional loading.

Your multimeter is likely designed to measure the frequency of (relatively) low impedance e.g. logic level signals. Even so, its additional loading might still cause problems, depending on the multimeter.

Is there another way to determine the crystal is in good working order ?

On an IBM PC or PC/XT, the crystal is connected to an Intel 8284A clock generator IC, to perform the necessary division of the crystal frequency for the CPU and other clocks, and generate the appropriate logic level signals.

There are a number of tests you could do. I'd start with measuring the OSC output (pin 12 of the 8284A) which is a logic-level buffered version of the crystal frequency, so you should measure 14.31818 MHz there. (Be careful of possibly shorting other pins on that IC when making measurements, especially if you have "normal" and not very fine ("needle") multimeter probes.)


The computer was running very slowly and unstable. Someone suggested looking at the oscillator.

I'll just add that there are other possible causes of an old computer being unstable. The main crystal wouldn't be one of my first suspects. Connections (e.g. socketed ICs) are one area of concern. One of the main components which deteriorates with age, are electrolytic capacitors. Therefore I would include checking the quality of the power rail voltages with an oscilloscope, as the first part of any investigation.

A multimeter can measure "gross" problems on power rails, but is not able to show relatively "subtle" problems e.g. excessive ripple. In other words: If the multimeter shows there's a problem, then there is a problem; if the multimeter doesn't show a problem, there might still be a problem. Only an oscilloscope can show some types of problem.

You might want to visit the Retrocomputing Stack for specialist advice.

\$\endgroup\$
1
  • \$\begingroup\$ Thx a lot for the detailed response. This computer doesn't use a 8284A but rather has a Faraday 2010 chip that implements most of IBM PC/XT LSIs (Intel 8xxx ICs) and glue logic (according to github.com/skiselev/micro_8088). I was able to measure the frequency now by following the OSC pin on the 2010 chip so I could easily measure it. Good point also on the oscilloscope. Thanks again (and also for the retrocomputing stack tip). \$\endgroup\$
    – ddewaele
    Commented Feb 9, 2019 at 12:17
0
\$\begingroup\$

This answer assumed wrongly that the probes were oscilloscope probes, in which the situation is easier. I am leaving this for reference, though, unless it is voted down.

The crystal should have loading capacitors on both end that adjust the impedance introduced to the chip. They should be in the order of pF, so it should be close to your probe's capacitance. I think you can remove one cap and probe there. You should use the cap of the probe that has a small lead as gnd to minimize the loop. I think this measurement should start with boot up, not introduced midway.

Also, you can create a short circuit on the probe's leads and get it close to the crystal. Magnetic coupling would reveal the oscillation. I use it on higher currents, but it might just work.

Other than that, the uC might have registers that could be warning flags.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.