Precision series resonancy crystal oscillator - The SJOSTROM KLOK

[peranders]

I haven't made any measurements but I image that of forward biased PN-junction is much quiter than a reversed.

A LED has charateristics similar for a diode of pure silicone. The only difference is a factor in the EXP() function. If you want low leakage at a higher level, LED's works fine if they have a black hat! Light will create photo currents.

The disadvantage of diodes or LED's are rather high temperature coeffcient, 0,3 %/degree and similar for LED's (don't remember the exact value). They are also dependent of process techology, especially LED's! 1.5 - 1.9 V (with same current) are normal distributions (my experience). "Bad" LED's have usually higher voltage and it comes from the "resistive" part, higher losses.

When we talk dynamic impedance diodes are rather poor compared to a simple LM431 or similar. We talk of a factor of 20-50! LM431 has 0,5-1 ohms and a diode has 25 ohms at 1 mA.

[Jocko Homo]

Herr Per:

Looks like a Colpitts, works more like a Butler. The signal passes
through the crystal, to remove harmonics and noise sidebands. So it is operating close to the series resosant point. Possible it is just above that point a bit. You can also tune a Butler to operate close (but not exactly) to the series point with a parallel mode crystal.

You can verify series mode by building one with a LC tank, and seeing which one works: series or parallel. It has to be close to series, otherwise the signal at the resistor where the signal is picked off will be too small to use.

Bulters work fine too. This one has the advantage of one less active device. Big deal, eh?

Jocko

[peranders]

Originally posted by Jocko Homo
Herr Per:

Looks like a Colpitts, works more like a Butler. The signal passes
through the crystal, to remove harmonics and noise sidebands.

This is true but the main thing (for digital purposes) is jitter and phase noise. Harmonics are ADDED if you convert the sinus wave into square wave.

So it is operating close to the series resosant point. Possible it is just above that point a bit. You can also tune a Butler to operate close (but not exactly) to the series point with a parallel mode crystal.

You still use the "L" in the crystal and as I said before you get much lower Q when you tune and low Q is bad if you talk phase noise. You must have a crystal for series mode if you want good results.

You can verify series mode by building one with a LC tank, and seeing which one works: series or parallel. It has to be close to series, otherwise the signal at the resistor where the signal is picked off will be too small to use.

The resistor serves as a shunt. You measure the current through the crystal, still the big "L" instead of the low "R".

This one has the advantage of one less active device. Big deal, eh?

You still need at least three active devices to be able to use the signal.

I wonder how these high precision oscillators are designed. Anyone who has a schematic over these metal cans (100 USD and above)?

[Jocko Homo]

Jitter and phase noise are caused by noise sidebands.

As I recall.........someone complained that making an quality oscillator was not that easy and one of the problems was radiation from a board that is a few inches above the main PCB. This one has no harmonics. The circuit that converts it (read the filter chip in most CD players) is on the board. You can drive them with a sine wave of sufficient level.

The resistor is not a shunt. Ever measure the series R of a crystal??

Jocko

[peranders]

This topic was very narrow, a dialogue between me and Jocko Homo only. Maybe this is a signal of the nature of crystal oscillators?

My contribution was to lift forward the Butler oscillator as an alternative the common Colpitts. A Buttler oscillator is not an alternative for most people since it requires special crystal (series resonance is rather unusual these days).

Still, I'm very interested in how a precision oscillator is made (those in a rather large metal can).

[Jocko Homo]

How large of a can? And what frequency? Soundslike you might be refering to the large ovenized types, usually in the 5-10 MHz range.

Jocko

[peranders]

I was thinking of the size from a matchbox and up, TXCO, OCXO etc.

[Jocko Homo]

If you are refering to the large ovenized ones, you might find some info at www.bliley.com .Those types use a special crystal cut, usually in the 5-10 MHz range, as that is where the "sweet spot" for stability is.

Jocko

[peranders]

You see Jock Homo, I'm not interested in the oscillator products. I'm interested in the design, inside the can.

[HarryHaller]

http://www.minicircuits.com/application.html

You could always go buy some or get some samples and cut them open and reverse engineer them.

http://www.google.com/search?hl=en&i...=Google+Search

Show some inititive and do a little research. You are delving into a topic that requires much more than a few paragraphs for an answer. Your questions could be more specific. "I am interested in ......" doesn't help a reader to figure out what you really to know.

H.H.