Precision series resonancy crystal oscillator - The SJOSTROM KLOK


We talked about crystal oscillators to CD players. The best oscillators are those which uses series resonancy crystals. I have got requests about those. Here comes some design ideas.

Tips: Gain = 1.5 -10, lower is better but maybe the oscillator get hard to start.

Supply voltage 5-10 volts, not too much because of higher drive level of the crystal.

Transistors, BC5xx etc for <5 MHz, HF types above

Currents, a few mA's

Well stabilized power! Very important in order to get high stablity. Important also not to load the oscillator. An extra buffer in essential.

Careful: You must choose a crystal for the purpose!!!! You can't tune a parallel resonancy crystal all the way to series mode. You loose lots (almost everyting) of Q-factor and stability!!! Very important!!!!
Time for a good old fashioned arms race!

Try this one out.

It also uses a series resonsant crystal, has low distortion, and comes with a low-noise regulator. Puts out 1.6 V p-p.

This is the first version, so more revisions are probably in the works.

It is easy to measure if a crystal is series or parallel resonant. If you need to order one, and the supplier doesn't know which one it is, find a different supplier.



  • oscillator.gif
    6.8 KB · Views: 2,106


2002-05-25 3:45 pm
Might it be better to:

a) run the LED string directly from the current source (that's the way it already is - just reviewing)

b) connect a 100 Ohm resistor (or maybe even some higher for a lower -3 dB point) from the top of the string to the base of the emitter follower

c) connect the 120 uF and other filtering capacitor to the base of the emitter follower?

The idea is that the LED string is a more or less low impedance (hence the reason for using it as a voltage reference). Bypassing this low impedance directly isn't really as effective as using an RC lowpass filter to get rid of noise. The voltage regulation under dynamic conditions shouldn't be affected.

Good? Bad?


2002-05-25 3:45 pm
Gotcha. For a 400 Ohm source, the resistor would be waste.

Have you ever measured the noise voltage density across an LED string? I'm curious how low it is.

Do you find blacking the LEDs out with goop of some kind, or otherwise shielding them from light, lowers the noise? (LEDs - and semiconductor lasers - also exhibit photoelectric effects, although they aren't real efficient.)


2002-05-25 3:45 pm
Good article, thanks for reminding me.

I was just curious about the LED in specific. Some guys I know have abandoned LEDs in this type of application because they were too noisy.

These same guys also found that you really needed to cascode the current source (hope I didn't just step on sensistive toes) and add a series resistor from the drain of the cascode FET to the rail in order to get the isolation they were looking for.

They were looking mostly in the audio band, and that may not be so important here.

Thanks for the guidance. I've been too lazy to measure these things myself...
FET current sources do not have the high impednce BJT types do. Both work fine, but you have to decide what you are trying to accomplish in order to decide which (if any) you should use.

Used to have a lot of data about current sources using LEDs. Can't find it though. Yes, there is a noise issue there. Wish I could remember what exactly the results were.......Oh well.



Paid Member
2002-03-29 5:19 am
I investigated the oscillator/crystal thing recently. I finally
settled on a canned crystal oscillator sold by a guy named
Guido Tent in the Netherlands (I think). A friend and I ordered
one for each. His is installed and he's very happy with it. He
finds it much better than a discete design (LC Audio?) that he
had previously purchased and used.

My oscillator, like other projects, is on my desk awaiting time.

I supposedly has about 3ps (2 sigma limit) jitter. One reason
it's still on my desk is I've got to round up the stuff to attempt
to measure it.

About $25 ea. with shipping. A bargain.
Re: Time for a good old fashioned arms race!

Jocko Homo said:
Try this one out.

It also uses a series resonsant crystal, has low distortion, and comes with a low-noise regulator. Puts out 1.6 V p-p.

Jocko, your oscillator is a colpitts (you use the big "L" in crystal = parallel resonance)

What is the advantage to take out the signal via a resistor? You lower the Q-factor by that.

It is easy to measure if a crystal is series or parallel resonant. If you need to order one, and the supplier doesn't know which one it is, find a different supplier.

Most 99.99% (or even more) are parallel mode crystals. You may have to order a special cut crystal. Not so expensive! Check Kvartselektronik in Sweden.

BTW: My oscillator is a "Butler"-type did I read in the document below. Didn't know that.

The advantage of a Butler oscillator is that the crystal is very hard connected between the low input impedance of the CB stage and the output of the emitter follower. When I tested the circuit I noticed that it works better with very low output impedance of the emitter follower.

I must add though I'm not an expert so take my advice "with a pinch of salt" as we say in Sweden. Crystal oscillators is a very wide subject if you want to do it totally right.


2001-06-12 10:15 am
Re: LED noise

I've done lots of measurements on regulators, discrete and otherwise.

I recently did some measurements comparing LED's to my favoured LM329 buried zener reference.

The noise levels aren't that different, but obviously temp. stab. is orders of magnitude worse.

Personally I think LED's are great for low voltage supplies, particularly 5V reg's for digital. For higher o/p voltages the ratio of noise to o/p voltage for the LM329 makes it lower noise in these applications, since less gain is required in the regulator circuit.

The latter point is less of an issue with regulator topologies that allow bandwidth tailoring to reduce noise amplification (Jung, Sulzer et al).

Also remember an LED needs a lot more current than a zener to work.



2001-06-12 10:15 am

Are you implying you bias the LED below it's normal operating current?

The shape of the V/I curve in this out-of-spec region of the diode's operating range means greater voltage variation for a given current change, i.e. higher dynamic impedance. If one is relying on it as an accurate voltage reference I'd suspect temperature stability to be worse at lower currents.

This means that the current source temperature stability (if biasing this way) becomes a factor in voltage stability. I've only used LED's in discrete reg's where the LED power is bootstrapped from the reg. o/p, and most of these have necessitated resistor biasing.

Personally I only use the LED within it's 'normal' bounds which means higher operating currents than LM329.

W.r.t. the 329, output noise as a proportion of output voltage is lower for the 329 than an LED. Whilst LED's (or most forward-biased semiconductor junctions) are very quiet they do not compare with the LM329, which offers the lowest self-generated noise of almost any device.