Tuning fork crystal oscillator circuit needed

Those tuning fork watch crystals need very low drive. This circuit works. Use unbuffered CMOS gates:

CD4049UBN |\\ .----------| >O--------o------ | |/ | | ' | | | | | ___ | o---------|___|--------o | 6M8 | | | | | | | | .-. | | | | | |270k | '-' | _ | | | | | o-------|| ||----------o | |_| | | | --- --- ---10pF ---10pF | | | | === === GND GND (created by AACircuit v1.28.6 beta 04/19/05

"Andrew Holme" schrieb im Newsbeitrag news:cNARk.11995$% snipped-for-privacy@newsfe18.ams...

Hello, The low power drive is really necessary. I once tried to measure such a crystal from a watch in a 50 Ohm system. After I increased the voltage to a very few Volts, the crystal immediately "died" forever. Helmut

On a sunny day (Sun, 09 Nov 2008 04:19:49 -0800) it happened Robert Baer wrote in :

I have such a 32 kHz watch crystal working with an 8 pin PIC, 16F629.

It is not exactly 'discrete' but neither is a CMOS gate. The advantage of the PIC is that you can divide and get other frequencies as well. Start-up is also slow. And the size of that 8 pin PIC is really small, compared to a bunch of 'discrete'. All extra I have uses is 2 caps, info is in the PIC data sheet.

Indeed. The low frequency tunning fork crystals are high impedance and very low drive level. You can use them with the standard circuit with C ~ 10pF R ~ 1M. If you need a bullet proof operation, you'd better divide a conventional crystal.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

That capacitance sounds high, as backed up by other posters.

My understanding is that those teeny little crystals are very high impedance, so even though you would think you should scale the capacitance up as frequency goes down, you shouldn't.

Do any manufacturers publish the crystal parameters, i.e. motational inductance and parallel capacitance? With those in hand you'd be able to model the oscillator is SPICE and _really_ find out how to get it working.

Why not buy an oscillator?

John

Oh, that's no fun.

OK, fun: make it out of an opamp.

John

I knew that they required a low drive, but did not know what to do. Wow; 270K! Thanks.

Not too surprised; i understand that one can crack any crystal with sufficient drive.

Well, like i said, i started with the PCB from a mouse and graduated to trying a CMOS inverter in standard configuration. What would be really cool is to have *one* circuit where i can plug in a tuning fork crystal or a standard (say 2-10Mhz) crystal.

well.

'discrete'.

Well, that datasheet is unavailable. A search in the Microchip site gets: Results from Microchip.com Results 1 - 1 of 1 for PIC16F629 Results per page:

PIC12F629/675 Memory Programming Specification

As usual, a real piece of art.

ECS gives more info than Citizen: R1 (motional/series resisance)

Regarding start-up time, consider that the resonator Q roughly equals the number of cycles in the transient.

Pere

AFAIK that drive level is the power dissipated in the motational resistance; 1uW across 30k-ohms works out to 170mV or so at the crystal, which is consistent with the big series resistors you see in the circuit diagrams for oscillators for these things.

Note the load capacitance -- the thing's designed for 12.5pF total series equivalent capacitance, or a 25pF capacitive load (capacitor and IC) on each end of the crystal -- your 470pF load capacitors are out in the stratosphere compared to that.

I'd find a published circuit, pour it into SPICE with a crystal model using the above parameters, and see how it goes.

pF and

se.

ed

d text -

The Pierce oscillator you already built from logic gates comes very very close:

But with generic C1 and C2 you will not always hit the design parallel resonant frequency.

Characterizing a crystal to find its series and resonant parallel frequencies, Q, series resistance, motional inductance is fairly easy with a signal generator, scope, and a simple test circuit. See, e.g. Fig 3.34 in _Experimental Methods in RF Design_.

Many of the parameters I name above are inter-related and for crystal matching, a simple Colpitts oscillator attributed to G3UUR is quite reliable. Google G3UUR.

Tim.

pF and

se.

ncy

Tony Williams pointed this schematic out to me several years ago. The transistors are from a CD4007

----+---+--Vdd | | / | R1\\ | / | +-------------|---|----------------+ | |-+ | | +----------||->---+ | | |-+ _ | | 15meg | | | | +---/\\/\\------+--------+----| |----+ | | | |_| | | |-+ |/ | +----------||- That capacitance sounds high, as backed up by other posters.

When I was an undergrad the joke of modeling a LC oscillator in SPICE was in the same category as challenging city folks to go cow-tipping :-)

It would've taken weeks of computer time to simulate a crystal oscillator startup back then!

Tim.