how to detune a crystal oscillator ?

[feedbackkiller]

Hi everybody !

I need an oscillator which runs at 32,83 MHz with +7dBm output power.

I want to keep it as simple as possible, so I want to use a crystal
oscillator and an op-amp.
As this design will be unique (no series production), I searched
some catalogues and found some off-the-shelf crystal oscillators
with 32 MHz and 33,33 MHz and one with 32,768 MHz which might be
suitable.
I know that it is possible to (de-)tune a crystal oscillator with
a series or parallel capacitor (depending on the type of XO).

At the moment I have that problem that I can´t figure out how big the
capacitor needs to be to reach the desired frequency of 32,83 MHz or
if I will ever reach the desired frequency with the oscillators I can get.

Can any one help me ?

P.S.: I know that I also could use a XO with a higher frequency together
with a prescaler/PLL, but that seems too sophisticated to me at the moment.

thanks,

feedbackkiller

<font size=-1>[ This Message was edited by: feedbackkiller on 2002-02-20 15:04 ]</font>

 

[Pim]

I hardly think it is possible to detune an "of-the shelf" oscillator since they use to be fully encapsulated (you won't be able to get at the inner electronic circuits). Furthermore, it is usually not possible to detune a crystal-based oscillator by that very much (I would guess about +/- 300 ppm very roughly).

You can tune them by either connect a series inductance and/or parallell capacitance (to ground). Using a combination of this gives you the possibility to tune the frequency both above and below the crystal's nominal frequency of oscillation.

I would also assume that it is much more convinient to base the design on a single or dual transistor circuit rather than trying to acquire an OP-AMP that can cope with such high frequencies (they usually are not very cheap). You can find many useful simple designs based on transistors (bipolar, FET, MOSFET) in the ARRL's handbook.

If wanting such a specific frequency (which I believe is not any standard frequency), why not try to locate any standard crystal with a frequency that is 1/3, 1/5, 1/7... of the desired frequency and tune the oscillator (by the help of a LC-tank) to this particular harmonic that constitutes the frequency that you have in mind?

If you want this oscillator made in just a very low number of units, I do not think that you want the cost of having a special crystal being manufactured for you with the expences that follows.

Should you choose to base the design on a PLL (which is pretty feasible), then you also will get spurs of the reference frequency into the spectrum. Whether or not that is of any consideration to you depends on how clean the signal must be. The output power of +7dbm would be no problem to achieve in any case.

/Pim


<font size=-1>[ This Message was edited by: Pim on 2002-02-20 19:04 ]</font>

 

[unkarc]

Hi Feedbackkiller,

Please see my answer on crystal detuning at **broken link removed** where just a low frequency (32,768kHz wristwatch) crystal was with the same problem.
That rule of thumb is valid maybe up to 20-30MHz in respect of the 20-30kHz detunability: if you try to detune such a high frequency crystal more than 20-30kHz then stability suffers more and more and finally the oscillation stops. Using lumped high-Q inductance may help a bit more detuning as Pim wrote correctly but I do NOT think you can reach the needed 62kHz detuning which is the difference between 32,83MHz and 32,768MHz.

What solution remains?
First: you could ask a crystal manufacturer in your country to learn how much the cheapest offer would cost for you; maybe you would accept it...
Second: Look for ready-made PLL integrated circuits (a rather old one up to 50MHZ is NE564 from Philips/Signetics or a more modern from Motorola MC145161...171 series and up and so on there are others) and calculate the division ratio for the reference crystal to that PLL. This solution needs a dedicated IC + a crystal which still can be or cannot be off-the-shelf!!

If your specific 32,83MHz frequency is needed for a receiver's 2nd mixer for instance, then it is not very good to solve it with PLL due to the spurs/digital harmonics etc. as Pim wrote. In this case the best to use the ordered single crystal run in a one-or-two transistor oscillator.
The +7dBm level can easily be achieved with a one-transistor amplifier.

Good luck, unkarc

<font size=-1>[ This Message was edited by: unkarc on 2002-02-20 17:36 ]</font>

 

[feedbackkiller]

OK, thank you so far.
I think i made a little mistake on my first post: When i was talking about a crystal oscillator, i thought about the plain crystal, no TTL crystal oscillator. I know that you can´t detune these.
My textbooks are not of too much help here, as they only handle

 

[rfsystem]

There is a simple method to detune.

The crystal is represented by a series tank circuit with an additional parasitic caps across the series tank. The series tank has a small series resistor in the range 25-100Ohm and a motional capacitance in the femto-farad range. The pierce oscillator built with the crystal resonate with two external caps. The effective tank circuit is then the series combination of the two working caps and the motional cap. The effective working frequency is therefore a litle bit higher than the series tank resonant frequency. That is the way to detune the crystal.

 

[feedbackkiller]

Hi rfsystem,

I´ve received the notification about your reply today, 17th of march.
I did a little calculus and will try it out ASAP,
because it seems that this might work.

Thank you.

 

[Bagster]

Firstly i would contact as many crystal suppliers and ask if they have a crystal at that frequency. There are literally only a handful of standard frequencies, but someone else may be using a frequency that is very close, if not the actual one. Ask, you may be surprised, i was when i wanted a custom one.

Bagster