frequency stable oscillator

[mostafah67]

hi
i'm working on design of a non-crystal (and not mems) oscillator, in general i must use an electrical resonator(like LC etc.) , my frequency is 10Mhz and the oscillator must be very stable in frequency(1 ppm !). what is your suggestion?

 

[aProgrammer]

I don't think this is possible except if you use a PLL to control the oscillator frequency,
(but the PLL need a reference frequency provided by a crystal oscillator).

 

[ROB2222]

You could use a GPS receiver module to receive a 1pps signal which is sent out by the obiting satelites referenced to caesium and rubidium atomic clocks. The 1pps out varies from Module to module manufacturer, but there are modules explicitly designed for this purpose (rather than location).

With this highly accurate clock as a reference you could use an off the shelf clock PLL that is designed to use this as its reference from companies like IDT or Micrel.

Clearly the GPS needs an antenna which in turn limits where the unit can be placed otherwise no reception and no 1pps.

 

[mostafah67]

Thank's friends,but i shoudn`t use crystal i want to design circuit only with electrical resonstor (in fact i have to design in this way i work on an academic isue)

 

[FvM]

Consider that a crystal oscillator with 1 ppm stability will at least utilize temperature compensation (TCXO) or possibly oven control (OCXO). A LC oscillator with similar specifcation will very surely need oven control with a much tighter temperature stability than a crystal oscillator. So besides academic interest to show that it "can be made", what's the purpose of this project?

 

[mostafah67]

(At First excuse me for my poor english!)
This is part of my final thesis and i`m not allowed to use crystal because we want to use a method that using circuit solution -like negative feedback -for this isue . Because that nowadays using of crystals are common i have a hard work to find refference or paper. I think finally the circuit will be used in gps systems

 

[FvM]

I don't see the exact point. A cyrstal oscillator will also use negative feedback. May be a misunderstanding?

 

[mikersia]

To design 1ppm@10MHz non crystal oscillator - unrealistic task from my viewpoint. Even with sophisticated digital calibration. You have to provide a stable (<0.2ppm) electrical reference to allow feedback compensate drift.

 

[mostafah67]

Oh i'm sorry,i know crystal oscillator can also use negative feedback but i mean we want to use a method in which a mechanism cancel other that cause a frequency instability, and specially we use an electrical resonator and want to determine which part in- for example LC - oscillator circuit cause more instability, i know first we must use experimental method but what is your idea? Which part have must proportion in producing instability?
Transistors? Resonator? Feedback network? Or the others?

 

[mikersia]

To cancel one mechanizm of instability by other you have to provide compensation region through whole PVT range, including statistical mismatches. Typically each instability factor is uncorrelating with others.
So many criterial task is very very complex. It have to include a lot of investigations or access to such investigations, a lot of simulations and practical checking. I'm in doubt regarding possibility to resolve described task.

 

[E-design]

One of the best non-xtal stability designs is the Vackar oscillator. http://www.qsl.net/va3diw/vackar.html
Some measurements were done and achieved under 1ppm stability. You will have to design this with very good quality components to get this performance as stated in the article.

**broken link removed**

 

[mikersia]

Vackar oscillator seems to be type of low noise LC oscillator. In monolithic implementation it will lose stability along with componenets drift.

 

[mostafah67]

thank you very much my dear friends, your guidelines are very helpfull

- - - Updated - - -

for simulating this circuit specially LC tank which simulating program is better?
in fach i want to see the frequency drift across temperature variations, so the program should be able to simulate correct behavior of the LC tank

 

[FvM]

Any SPICE based circuit simulator will be able to simulate the oscillator, e.g. LTSpice. But your result can't more accurate than the models you put in. There's no standard model that includes exact temperature drift of L and C components so you have to add it to the model parameters.

 

[udhay_cit]

By using a microcontroller you can generate accurate square wave signal(with internal oscillator), but sine wave :shock: i don't know...

 

[mostafah67]

good, so if i have my own model can i embed it in the simulation ?