How to test an oscillator that's not working

[Radio Wrangler]

So, you've got a set in front of you with an oscillator which won't run. It could be a repair job, in which case you have the comfort of knowing that it once worked, or it may be something freshly built, or worst of all a new design and all bets are off.

You can measure bias voltages and currents, replace the valve or transistor and if nothing wrong is found, you're stumped.

Or are you?

I had a newly laid-out PCB prototype on the bench. It had a rather extreme oscillator using a SAW delay line in a loop near 1000 MHz. The system had been used before so it was known that it could work, but was fiercely sensitive to layout effects. And this one needed a different layout to fit in the available space...

Well, an oscillator without enough gain to get going is an amplifier. An amplifier with regeneration, an amplifier with positive feedback. As negative feedback decreases the gain of an amp, positive feedback increases it.

A running oscillator can be thought of as an amplifier with infinite gain.

A non-running oscillator can still have rather high gain and be hyper-sensitive to anything nearby and near the right frequency. It's a super-regen receiver just waiting to be exploited.

This led to the idea of inserting a signal, very lightly coupled, and seeing what the response was.

I was lucky and had a spectrum analyser with a tracking generator available (posh wobbulator in other words).

You can use ordinary scope probes on spectrum analysers. Yes, the scope probes expect a megohm load and spectrum analysers are 50 Ohms. You get a lot of loss, but the analysers are usually very sensitive if set for low reference levels. The scope probe is very unflat. Essentially it's working as just a series capacitor - the probe tip resistor compensation capacitor, so at HF and above, the loss is less than you'd expect. The unknown loss and bad flatness are not a problem when looking at narrow band things - like oscillators.

So snag the scope probe onto the oscillator output and have the analyser scanning around the expected frequency.

Use a second scope probe on the tracking generator output and set it for a volt or two output.

Now wave the generator scope probe around in the air near the oscillator circuitry. If there's any attempt at life in the circuit, you'll see a bump in the analyser trace. Move the probe nearer, it should get bigger. Is the bump on the right frequency? If not, you're on the trail.

If you have somewhere to rest the generator probe or something insulated to grab it onto, you can get a reference picture. Try changing things in the oscillator and seeing if you make the response bigger or smaller. This tells you whether yoou're getting hotter or colder and you can feel your way around.

You don't get a flashing neon sign forming in free space pointing at the problem, but you have an awful lot more clues than just the DC condition checks.

If you move the probe closer you can often find the oscillator starting up. It's likely to stop when you remove the stimulus, but you're on the trail.

All well and good, but few people have trackers and analysers. So use your oscilloscope to view the output, and use a plain old sig gen to drive the stimulus probe. Rock the sig gen frequency and watch the scope for any response. You can still get the clues and it's not that hard to do. You build a picture of response versus frequency in your mind. Tune for peak response, note the trace amplitude and the generator frequency and you have the clues.

Cheating?

Oh yes, but in a circumstance where cheating is perfectly OK.

David

[trh01uk]

Very clever, David. The only thing I would add is to short circuit the scope probes, by connecting the earth lead to the probe tip. This eliminates annoying things like hum pickup, but allows inductive coupling in the created loop at the probe to couple to what is going on in the oscillator.

Richard

[murphyv310]

Interesting post David and some good tips.
I have been able in the past to get oscillators to run with a signal generator loosely coupled to the Oscillator, it has helped in some cases but not all which has resulted in a bleeding head due to excess scratching.

[lesmw0sec]

Ah, the old law - amplifiers oscillate, while oscillators won't!

[G6Tanuki]

That's cunning!

Another trick I have heard of is to fire one of the old piezo-electric gaslighters near to a recalcitrant oscillator and see if the radiated burst of energy will shock the oscillator into life (even briefly).

Wouldn't recommend doing this around FETs or CMOS circuitry though...

[Oldcodger]

RK- in a establishment to far from you, many years ago, we tuned radio receivers and high powered transmitters. One day I tuned a set and my mate decided to try a mod. Instead of using the "Approved" scope probe, he aligned a pair with a wire lead. Instantly better response and alignment. We suspected loading of the circuits. Possibility?

[Radio Wrangler]

Oh, yes, that's quite possible. Getting the coupling right makes a big difference.

Another interesting trick relating to quartz crystals whether for oscillators or filters was used at crystal manufacturers.

The crystal blanks are ground in machines that look like giant Spirographs - a planetary gear arrangement with holes in the planet gears to take the quartz discs. As the machine runs, the blanks get moved ina spirograph pattern between two flat plates, 'lubricated' with an abrasive slurry of an appropriate grit size.

As grinding progresses, the quartz experiences mechanical random noise. Being piezo electric and resonant, there is a radio emission. So the manufacturers have a shortwave set to listen for the peak noise emission so they can monitor the frequency of crystals even while they are being ground. The operator tuning the receiver upwards to follow progress. Experience tells them when to switch to the next finer abrasive and when to switch to manual finishing.

David

[trh01uk]

I witnessed that process in progress at Cathodeon Crystals in 1977 - there were banks of Eddystone receivers in use. That was part of the graduate induction process at Pye Telecom. But it seems now that custom crystals have virtually disappeared from the market place, having been wiped out by frequency synthesisers.

Richard