Crystal oscillator circuit design
 

I need to make a fixed frequency oscillator at 4.194304MHz.
I thought that a crystal-based circuit would be best and have chosen this crystal as it's ubiquitous and cheap :https://uk.farnell.com/qantek-techno...9us/dp/2508442
The datasheet states that the maximum drive level is 1.0mW - can anyone tell me what the best circuit to use with this crystal is, that will not overdrive it and also has the best frequency stability with lowest jitter ?
I'd like to know how to calculate the drive power in general as well.

Cheers,
Andy

Re: Crystal oscillator circuit design
 

The data sheet doesn't even tell you whether its a parallel or series mode crystal (all crystals have both modes it's a matter of which one is on the quoted frequency. I expect it to be series mode so you need to look for series mode oscillators.

Dissipation depends on the crystal's ESR. It's a cruel game giving you a dissipation limit and no spec on esr.

This is a miniature crystal of low-ish performance intended for clocking CPUs, serial links etc.

Have a read at the amateur radio books like 'Solid state design for the radio amateur' or there's my oscillators chapter in the ARRL handbook.

David

Re: Crystal oscillator circuit design
 

I found a link to a datasheet which does seem to be of some use.

Low frequencies are fundamental mode and parallel operation with load capacitance of 18pF for this particular part I think.

Higher frequencies are 3rd overtone and series operation - all quite usual.

Also has ESR values so it's a start.

I was wondering how stable you need that frequency to be as you specified it down to 1Hz? Are you multiplying it to a higher frequency perhaps?

Re: Crystal oscillator circuit design
 

It's for a clock timebase, divided down to 1Hz. Like 4.43361875, 4.194304 is a number you never forget!

Re: Crystal oscillator circuit design
 

Ah - and I multiplied that by all the usual numbers and it came out nowhere useful that I could see!

Just for fun I put a Pye Westminster Colpitts Oscillator - as best I could remember it - into LTspice. I have not played with oscillators there before.

At the moment I am lacking a good model for a quartz crystal but with an Inductor and a Capacitor at least I have something that produces output - after messing about with the feedback capacitors. With my memorised values oscillations would start but die away. With a tweak of values it sustains and it runs.

You might try to do the same since oscillators are something of a dark art. I'm not sure they are very amenable to analysis by written equations in a simple way. LT Spice is free but there is an associated learning curve.

David explains oscillators far better than I - read his writings!

What do you want to use as active device? Transistor or logic gate?

The main benefit of LT Spice is the virtual probe which can tell you Voltage across the crystal and then with ESR value you know the dissipation.

Re: Crystal oscillator circuit design
 

These xtals were generally designed to be hooked-up to one of a range of CMOS-based divider-chips

Have a look here: http://www.npc.co.jp/en/products/xta...r/5027-series/

Re: Crystal oscillator circuit design
 

If you can find a crystal in an HC-6 or HC25 package, the larger quartz disc is a lot less delicate: the dissipation rating is 10x higher and the ESR is lower. win-win-win.

The small crystal would have been for timing and would have used a CMOS gate in a not-quite-all-there Pierce circuit.

David

Re: Crystal oscillator circuit design
 

Thanks for all the replies.
My project is basically a heterodyne audio oscillator : The crystal oscillator circuit I'm asking about here will be used as the main reference frequency and will feed a divider circuit.
I read (somewhere) that oscillators based on a CMOS inverter (Pierce?) are not that good... also that you shouldn't use Schmitt type inverters as they basically have too much gain. I don't really want to have an extra IC (e.g. 74HCU04) just to use one out of the six inverters in it : So my thoughts turned to a simple Colpitts circuit based around a discrete transistor. This would feed a spare Schmitt inverter (74HC14) that I'm using before going to the divider.
I've seen some circuits that advocate using high-value bias resistors so as not to 'load' the crystal and ruin it's Q, I've also seen other circuits where the bias resistors were less than 10k !
There's never any mention of 'drive' levels.
Does this help to explain things ?
Of course, maybe I'm over thinking any perceived problems.
I have not had any success with crystal oscillators in circuit simulators either.
- thanks for that info, makes a lot of sense !

David (Radio Wrangler), I'll try and read up more about the subject in the references you've given.

Andy

Re: Crystal oscillator circuit design
 

Pye used this colpitts circuit for just about everything in the days of the Westminster Radio telephone. It is what I was playing with in LT spice.

Re: Crystal oscillator circuit design
 

I have just entered that circuit in LT Spice with a 2N2222 as the active device and 10V as the power supply.

I used a guesswork crystal equivalent circuit which resonates at 4MHz and has an ESR of 120 Ohms.

It takes about 5mS to start and build to full output (high impedance load assumed) after which there is 0.5V pk-pk across the resistor representing crystal ESR.
That is around 0.26mW of dissipation and it should be safe against the rated 1mW for the crystal.

The 39pF capacitor will probably need adjustment to pull to exact frequency.

Re: Crystal oscillator circuit design
 

Hi Jon,
Thanks for trying that out - that's useful information to have, at least I know now that the crystal I'm considering will be OK.
I think that Colpitts circuit (or a 'tweaked' version of it) is probably the best to use for my needs.

Cheers,
Andy

Re: Crystal oscillator circuit design
 

It was fun to play with it. My first outing with an oscillator in LTspice.

Nice to see the effect of emitter resistor on crystal dissipation and at the same time the parallel capacitor on start up.
If the 100pF is made 47pF instead it has not produced any output by 30mS - probably it never will.