CMOS Crystal Oscillator Adjustment

I have built an oscillator using an HC4060 chip and a 4MHz crystal. The circuit is that in the manufacturers data sheet. The crystal is designed for an 18pF load, so I used 39 pF capacitors to ground from each end of the crystal. The oscillator starts up instantly and operates with a supply down to less than 3 volts. My problem is that it is slightly high in frequency (about 25 ppm). This is just inside the crystal manufacturers spec, but not accurate enough for my application. I believe that the oscillator frequency can be pulled by increasing the values of the load capacitors, and my question is how much pulling is possible? Is 25 ppm plausible, and if so how do I calculate the capacitor values required? I tried increasing the capacitors to 100pF to see the effect, but the oscillator became reluctant to start up. Any help appreciated.

TIA

Mike

Why not try it?

Hello,

Have a look at the attached application note.

Bertus

Thanks very much Bertus, that looks as though it has the info I need.

Mike

Well, I digested the very useful document linked by Bertus, but my problem remains. The theory suggests that I should easily be able to pull the oscillator frequency by the 20 ppm I need, but changing capacitor values has no effect. The circuit is the standard Pearce arrangement, with an input capacitor c1 on one side of the xtal, and an output capacitor c2 on the other (isolated from the cmos gate output by a 2k resistor). The xtal load capacitance stated by the manufacturer is 18pF. I started with both capacitors at 39pF, so the series combination was close to the 18pF. The oscillator started up fine and ran very reliably, but was 20 ppm too high. I increased c1 to 82pF and c2 to 100pF. The oscillator was a bit slow to start, but was otherwise fine. However, it was still 20 ppm high! Out of curiosity I changed both capacitors to 27 pF. Still no change. The oscillator is built on strip board, with the shortest possible track lengths, and all unused nearby tracks removed. The cmos chip is well decoupled. One observation which seems a little strange is that the cmos gate output connected to c2 via 2kohm will not tolerate the load of a x10 scope probe. When the probe is touched on the ic pin the oscillations quickly die and the oscillator stops.

Ideas welcome

Mike

You can't scope the oscillator! You need to use a sniffer that does not connect to the oscillator, or scope a buffered output after the oscillator.

How are you measuring the frequency? Maybe your measurement process is affecting the frequency?

Changing the capacitance on the out pin of the xtal (the xtal pin that connects to the IC input) should pull the frequency significantly.

Thanks for the comments THE_RB.

Your comment re. scoping the oscillator is confirmed by my experiment, but I'm still surprised that the loading of a x10 probe (10M and 8pF) kills it, and I'd be interested to understand the reason. The point I'm probing is the output of a CMOS gate, which I expected to be a relatively low impedance point. Equally strange, to me at least, is that probing the input to the gate, which is connected directly to the xtal, does not kill the oscillator. I see a clean sine wave.

The oscillator is the basis of a precision clock (the kind with a dial). I am essentially comparing the output of this with that of a radio-controlled (MSF) clock, and looking at the drift over 1 day. This is around 2 seconds (around 23 ppm). I have two radio clocks which track each other very precisely, so I am fairly confident of their accuracy.

Yes, I would have expected the frequency to be pulled by the changes I tried, but the results were as reported. I am now wondering if the circuit layout on stripboard is the problem, with stray capacitances dominating the circuit.

Your comment re. scoping the oscillator is confirmed by my experiment, but I'm still surprised that the loading of a x10 probe (10M and 8pF) kills it, and I'd be interested to understand the reason. The point I'm probing is the output of a CMOS gate, which I expected to be a relatively low impedance point. Equally strange, to me at least, is that probing the input to the gate, which is connected directly to the xtal, does not kill the oscillator. I see a clean sine wave.

Click to expand...

This is odd. I wonder if it is the length of the scope's ground lead. It should be the shortest one available for your scope probe. Where you are connecting the ground also matters. The scope ground should be connected near the circuit output to keep it away from the input circuit.

The scope connection can cause a ground loop if there is more than one probe and ground wire, is used. More than one piece of test equipment connected to the circuit can also cause ground loops .

Yes, I would have expected the frequency to be pulled by the changes I tried, but the results were as reported. I am now wondering if the circuit layout on stripboard is the problem, with stray capacitances dominating the circuit.

Click to expand...

Yes, this is a big problem. Many years ago I measured the capacitance between a pair of 5-contact columns and got 13pF. Keep in mind that you usually have a contact on both sides of your signal so the capacitance seen is more like 26pF. This also assumes that there is no wiring capacitance! -- seldom the case.

Well, the fog has finally cleared!

It's now clear that I had two problems.

The first was my own lack of clear thinking when I measured the clock error. It was actually running SLOW by 25ppm, not FAST.

As soon as I realised this I looked harder at the stray capacitances around the circuit, and changed the layout to minimise them as much as possible. Immediately there was an improvement in accuracy. Reducing the two capacitors to 39pF then got me as close as I need to be to the right frequency.

Thanks to all those who chipped in with advice, which undoubtedly helped.

Mike

Congrats!