Kits

QRP Labs
Multiple crystal oscillator PDF Print E-mail
Written by Hans Summers   
Monday, 03 April 2017 19:29

Kostas SV3ORA informed me of his oscillator circuit which has a single J108 Field Effect Transistor and three crystals, oscillating all at the same time on their three different frequencies. His article on this oscillator met with some scepticism. People thought that one crystal should dominate, ignoring the others. You can read the forum discussion that Kostas copy and pasted and has on his web page article.

Kostas SV3ORA's triple crystal oscillator: http://qrp.gr/multiosc/index.htm

Personally I can't think of any practical uses for this right now in any of my current or future projects. But I was interested to try out the circuit and see for myself, what happens... and it shouldn't take too long, there aren't very many components to connect together, are there.

WRONG! Yes it's a simple circuit... but it is deceptive! It's a circuit that I feel I could spend days and days studying, or weeks, even. But I don't have days. So this is a simple write-up of what I did, with some points for further investigation another time (or by another person).

The first problem was that I didn't have a J108 FET. I do have some J310 FETs though. As Kostas pointed out, the J108 has a significant amount of gate-source capacitance. Therefore he didn't need to add any external capacitance, he could rely on the internal capacitance of the J108. He said a much lower capacitance device like the J310 should have maybe 150pF added across the gate to source leads. Actually in my case I found the J310 worked (2 crystals oscillating at the same time) even without the additional gate-source capacitance. My construction is rather messy. Nevertheless I don't think there would he a relatively significant amount of stray capacitance to make up the difference in gate-source capacitance.

Refer to the following circuit diagram:

And here's a picture of my messy construction, with two 'scope probes, a spectrum analyser, and a DVM hanging off it:

Firstly, ignore the buffer circuit on the right hand side. Ignore everything to the right of the arrow marked "TEST POINT". This left half of the circuit is equivalent to what Kostas did. I added the 10uF capacitor at the 12V supply connection only because I felt guilty about the rather long (maybe 1m or more) twin-lead cable to my power supply.

The two crystals I played with first were a 3.579MHz TV colorburst crystal and a 5.9904MHz crystal, both used old junk box items in HC49 cases. I found that I could adjust the 22K trimmer potentiometer and at different settings, either one or the other, or seemingly BOTH crystals would oscillate. It was exactly as Kostas described in his article!

Here are three oscilloscope screenshots of the three states of behaviour if this circuit. Firstly, at lowest trimmer resistor value (highest gain) the 5.9904MHz crystal oscillates (see frequency counter measurement at the bottom left of the screen). Next as the gain is lowered, the 5.9904MHz crystal stops and the 3.579MHz crystal starts up. In both cases, a reasonably large amplitude and spectuclarly clean-looking sinewave. At even lower gain (higher source resistance) the oscilloscope waveform gets all messed up and the oscilloscope synchronisation can no longer lock on to it. One could speculate that this is because both crystals are oscillating, And the waveform we are seeing is the summation of 3.578MHz and 5.9904MHz sinewaves.

But speculation isn't good enough... so now I turned to the spectrum analyser for a more deep investigation.

Now the next problem is that the spectrum analyser has a 50-ohm input impedance and such a low impedance directly connected to the crystals killed all oscillation stone dead, in my case. This is where the right hand part if the circuit comes in. This is a transistor in the common collector configuration. This configuration has a high input impedance but a voltage gain of less than 1. In this circuit I measured that the output voltage is approximately 1/150'th of the input voltage. But it converts a high input impedance to a low output impedance which is what we care about here. The spectrum analyser has plenty of sensitivity and dynamic range to see tiny signals way way down.

When only the oscilloscope probe is connected to those two crystals, the probe itself is acting as some capacitance to ground. When the probe was removed and the crystals connected to the common collector transistor buffer base, I found it necessary to add some capacitance to ground. I used 10pF which worked well.

The results are very clear, it really is oscillating on both crystal frequencies simultaneously.

First at high gain, here's the 5.9904MHz crystal oscillating:

Next at a lower gain setting, here is the 3.579MHz crystal oscillating. Note the 2nd harmonic, approximately 24dB down. After passing through this common collector buffer, which is NOT a low distortion linear amplifier, the spectral purity of the nice clean sinewave is lost. The same applies also to the 5.9904MHz oscillation, but the spectrum analyser sweep is only set to 0 to 10MHz.

Finally here is the result at low gain - both crystals are oscillating. There is a peak at 3.579MHz and another at 5.9904MHz. Note also the numerous other mixing products at lower amplitude, caused by the non-linearity of the common collector buffer.

Now I measure the amplitude of the 5.9904MHz oscillation and the 3.579MHz oscillation at different source resistor values and plotted them on a graph. To make my life relatively easier, rather than having to remove the trimmer resistor for measurement, I found in the junk box a nice dual-gang 100K linear potentiometer (this is the one shown in the photograph above). Using this I could measure the resistance on one of the ganged potentiometer while the other was active in circuit.

The RED line is 5.9904MHz and the BLUE line is 3.579MHz. You can see clearly at the lower resistance values, the 5.9904MHz crystal oscillates. Above about 1.5K, the 3.579MHz crystal takes over. From about 3K to 10K, both crystals are oscillating together at the same time. The levels of the two frequencies were surprisingly similar. Above 10K, only the 3.579MHz crystal oscillates.

The two graphs below show the same information, but one uses a linear vertical scale in milli-volts over the 50-ohm load, the other is in dBm (to 50-ohm impedance load). The range 3K to 10K where the RED and BLUE lines are both present together, is where both crystals are oscillating simultaneously.

What didn't work...

I couldn't get three crystals to oscillate simultaneously the way Kostas did. I could observe at highest gain the highest frequency oscillating, then handing over to the next lower, then the lowest frequency crystal as I reduced the gain. But I could not find a point where all three oscillated at the same time.

Furthermore I could not succeed in getting EVERY pair of random crystals from my junkbox to oscillate simultaneously.

Further investigation...

So clearly this oscillator really does operate on multiple crystal frequencies simultaneously. I feel that a lot more study could be done, to completely understand how it works and why; and what are the necessary conditions for it to work. As I said, I really don't have time now... but here is what I would like to investigate, if I did have time:

1. Effect of adding different amounts of gate-source capacitance to the J310

2. Effect of different amounts of capacitance to ground, or loading, at the other end of the crystal (where I put 10pF)

3. How to make three crystals oscillate together, as Kostas did?

4. Investigate a variable amount of series resistance (some 10's of ohms) in series with each crystal to try to equalise the activity of the crystals - which might make make it easier to get them to oscillate together at the same time

5. What is the effect on the actual oscillation frequency of the crystal? Is it pulled some amount?

6. Use of a more linear buffer, to investigate the spectral purity of the oscillations

Last Updated on Monday, 03 April 2017 19:36
 
© 2009-2017 Hans Summers
Web services by DataState Solutions