The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator

[andrea_mori]

I start this thread to investigate the opportunity for a diyer to build a real low jitter oscillator.

The great issue for a diyer is that typically he doesn't own the suitable gear to test the performance of an oscillator, so the results are uncertain.
The equipement to test the phase noise of an oscillator starts from some tens thousands dollars, and a diyer can't afford them.

The true guru in this matter, that you probably know as Jocko Homo, has demonstrated that one could build a low phase noise oscillator taking a crystal from the shelf with a simple pico gate. He did reach -122 dBC@10Hz from the carrier for a 11.2896 MHz oscillator, that's an impressive result.
But he did a strong selection of the crystal using a phase noise measurement system that costs 30,000 USD or more.
Moreover the crystals he had used, around 0.5 USD each from Mouser or Digi-Key, are no longer manufactured, and the supplier don't have anymore in stock. They were the cylinder crystal CSA309 from Citizen.
He also tested the HC49 type, but seems it's not perform like the above type.
You have to consider also that after the selection, only a few crystals can reach similar performance when placed in the oscillator, usually not more than 5%.
The other crystals were throw out, so the cost increases.

That's the reason I believe the goal is to start from a very good crystal, with high standard and repeatable features.

I will experiment with 3 kind of oscillator circuit to implement the above crystal: the Clapp crystal oscillator, the Butler two emitters, and the Driscoll.
The first oscillator is almost ready, with its own PCB, then I'll investigate the other two.
In the second and third circuits, the crystal see a very low impedance, and this usually guarantees a very high loaded Q.

When all the circuits will be ready, I can access a university lab to test them with an Agilent phase noise measurement system.


TWTMC project. For Schematics, BOM, PCB and Assembly guide see post #1506
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator

Group buy thread for this project is located at
The Well Tempered Master Clock - Group buy - diyAudio

New oscillators, frequency doublers and sine to square converters from post #3008
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator

User manuals
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator

[andrea_mori]

Since a very good crystal is the goal of this project, I got a custom device from the canadian Laptech Inc.

The most important features for a very good crystal are: cold welded package, very low ESR and very high Q, but overall its surfaces have to be strongly polished.
If the surfaces are not polished with the maximum care, the crystal shows some kind of perturbance that decreases its performance in phase noise close to the carrier.

The selected crystal is an HC/49U cold welded polished AT crystal.
These are the typical characteristics for a 11.2896 MHz crystal:
Mode: fundamental
ESR: less than 10 ohm
Motional capacitance: less than 12 fF
Q: better than 150 k

But features aside, the most important feature of the above crystal is that its features are standardized and repeatable.
The crystal is individually polished and tested from the manufacturer, that throw out the devices that not performs as specified.
When I got the first batch of 5 crystals, they were accompanied with a list where each crystal is tested individually. This list contains 12 tested crystals, but 5 only of them passed the selection.

[dmills]

You can make PN measurements without needting tens of thousands of dollars of kit you know.

Look up the quadrature mixing method of phase noise measurement for example, Martein Baker PA3AKE goes into some details on using a relatively low cost spectrum analyser to measure close in PN by this method, personally I would build two identical clocks and just subtract 3dB from the measured output rather then trying to source a state of the art reference source.

Wenzel associates have some stuff in their library.
Low-Cost Phase Noise Measurement | Wenzel Associates, Inc..

How are you planning to do the limiting? Keeping the gain stage operating in its linear region is important to avoiding the conversion of AM noise to PM.

Have a look at some of the stuff folks like Driscoll have been doing for sources intended to be multiplied up into the GHz bands, there is some interesting work there, for all that series mode seems to be preferred in those (overtone) designs.

Regards, Dan.

[andrea_mori]

Originally Posted by dmills

Have a look at some of the stuff folks like Driscoll have been doing for sources intended to be multiplied up into the GHz bands, there is some interesting work there, for all that series mode seems to be preferred in those (overtone) designs.

Indeed I'm planning to use a 3rd overtone crystal for the Driscoll oscillator.

I got also some pieces of 3rd overtone crystals from Laptech.
These are the specs:
Frequency: 33.8688 MHz
Mode: 3rd overtone
ESR: less than 15 ohm
Motional capacitance: 1.35 fF
Q: better than 260 k

Andrea

[andrea_mori]

The first circuit I have built is the Clapp oscillator, a variant of the Colpitts.
The circuit was designed by a smart dutch guy, using only a few components.
The oscillator include one jfet as the active device, and a few capacitors and resistors.
Then a simple '04 inverter was used to square the sine waveform coming from the oscillator.
The circuit is intended for use with fundamental mode crystal, from 5 MHz to 25 MHz.

This oscillator was measured with a R&S FSUP system.
As you can see in the attached plot the performance is impressive:
-132 dBc at 10 Hz from the carrier and -101 dBc at 1 Hz
RMS jitter below 1 ps (0.4060)
One of the best AT crystal oscillator I have ever seen, not far from an state of the art OCXO.

Note that the oscillator performs much better using a 74HC04 for the squarer, rather than using a comparator such as the LT1016.

The crystal used comes from the german QT Quarztechnik GmbH.
It's a HC-49U resistance welded package, 15 ohm ESR, laser engraving strongly polished crystal. Very similar to the Laptech, that on the paper should be a little better, since it show lower ESR and comes in cold welded package.

The dutch guy measured also the oscillator with Laptech crystal (I sent him 1 piece) with his DC receiver. The DC receiver cannot measure the phase noise very close to the carrier. The results was:
at 200 Hz the noise is at the bottom of the DC-receiver: -157 dBc/Hz@200 Hz
at 50 Hz the noise is -155dBc/Hz@50Hz.
He said about the Laptech crystal "...one of the best Xtals I measured before! Moreover its microphonics is very low which makes it very suitable for a transport in the same room as the loudspeakers"

Next step: the circuit diagram.

[jcx]

how about just buying a cheap cell tower surplused ovenized osc

[andrea_mori]

Originally Posted by jcx

how about just buying a cheap cell tower surplused ovenized osc

Could be an option, maybe is a little difficult to find it at the right frequency.

[esgigt]

Nice thread

[andrea_mori]

The schematic of the Clapp crystal oscillator is enough simple, one J310 jfet as the active device and a few other components.

The capacitors C1 and C2 depend on the oscillator frequency:
- 100 pF for 11.2896 MHz and and 12.288 MHz
- 200 pf for 5.6448 MHz
- 50 pf for 22.5792 MHz and 24.576 MHz

The capacitor C3 is required only to fine tune the oscillator for measurement; in the real operation it has to be replaced with a 0 ohm resistor or a link.

The resistor R2 has to be choose to get 18-20 mA flowing through the jfet when the crystal is removed from the circuit.

The resistors R3, R4 and R5 are required to get 50% duty cycle.
Finally an inverter is used to square the sine waveform coming from the oscillator.

The circuit requires two low noise power supply: +15V for the oscillator and +5V for the squarer.

To get 3V3 at the output, the 74HC04 should be replaced with a 74LVC04, and the +5V power supply should be replaced with 3V3 voltage.

[randytsuch]

Cool thread.

I'll be watching.