[time-nuts] Designing and building an OCXO and GPSDO

[Bruce Griffiths]

Philip Pemberton wrote:
> Hi folks,
>    I've been following the mailing list for a few weeks using Pipermail (the 
> web-based archive) and I figured now was a good time to jump in (so to speak).
>
>    I'm working on a GPS-disciplined oscillator, based on a Trimble SVeeSix GPS 
> receiver, and a homebrew OCXO. I've got a pair of 10MHz 50-degree-C oven 
> crystals, and have a pretty good idea how to handle the temperature regulation.
>
>    What I'm planning to do is mount the crystal on a copper plate with two 
> power transistors, using heatsink compound between the copper and 
> transistors/crystal case, and fit a temperature sensor to the top side of the 
> crystal case. I'm planning to use a copper bracket to hold the sensor onto the 
> crystal, and in turn mount the crystal to the copper base.
>
>    As far as temperature regulation goes, I'm going to use a PIC 
> microcontroller (one of the 8-pin chips with an A/D converter) to monitor the 
> temperature of the crystal, and use a PID loop to control the two power 
> transistors to maintain a temperature of 50C +/- 2 Celsius (the accuracy spec 
> of the temperature sensor). I also have other higher-accuracy sensors (Dallas 
> DS18S20 and DS18B20) that I can calibrate with; these are accurate to around 
> half a degree Celsius with a resolution of 0.5C.
>
>    The whole thing is going to be mounted in a metal box lined with 1/2in 
> thick polystyrene, with all external connections made via Molex KK connectors 
> and standard hookup wire. If there's any advantage to doing so, I might use 
> RG174 cable for the oscillator output, but otherwise I'll stick to the KKs and 
> maybe twist the OUT/GND wires together.
>
>    What I'm stuck on is the oscillator itself. The crystals are standard 
> parallel-resonant parts, with a load capacitance of 30 picofarads. I've got a 
> few varicap diodes (varactors) that I'm planning to use to allow external 
> trimming of the frequency, on top of what the ~20pf "coarse" preset will 
> allow. So on one side of the crystal I'll have a 33pf capacitor, and on the 
> other a 20pf load capacitor, the varicap and a low-value DC-blocking capacitor 
> for said varicap.
>
>    The standard oscillator circuit for TTL seems to be a pair of 74HC04 
> inverters and a few passives, or a transistor version that outputs a 
> sine-wave. Are there any particular types of oscillator that are more suitable 
> for high-accuracy timing?
>
>    What I'd like to do is use this oscillator to calibrate frequency counters 
> and check the calibration on oscilloscopes and similar. Being able to lock 
> function generators (a mix of custom DDS sine generators based on Analog 
> Devices DDS chips and FPGA-based complex-signal DDSes) against the oscillator 
> would be very useful as well. Should I be going for a 1V sine output and then 
> convert this to TTL in the generators (which are easy to retrofit with adapter 
> boards) or output TTL from the reference and leave it at that?
>
>    What design parameters should I be optimising for, and how?
>
>    Given that a standard crystal is good to roughly 100ppm, and most 
> commercial OCXOs are specified to be within 1x10^-9 or better, I'm aiming for 
> around 1ppm to start with. Is even this realistic for a homebrew device?
>
>    There seems to be quite a bit of difference between just building a 4MHz 
> oscillator to run a PIC MCU to building an accurate frequency reference source...
>
>    As far as parts are concerned, I'm planning to use either a BB153 or BB148 
> varicap, a Microchip TC1047AVNBTR temperature sensor, a National Semiconductor 
> LM4040CIM3-4.1 voltage reference for the PIC's A/D, two BD139 power 
> transistors and a PIC12F683 microcontroller.
>
> Thanks,
>   
Philip

If you are serious forget the fancy digital or semiconductor temperature 
sensors they aren't good enough.
However with your crude oven structure using higher performance sensors 
may perhaps be unwarranted.
For the best performance, unless you use a bridge oscillator circuit of 
some type, you will need to control the temperature of all the 
oscillator components as well.
Its best to bond the sensor into a well drilled in the oven using non 
electrically conductive thermal epoxy.

An analog bridge using an RTD or an NTC thermistor can have much better 
stability.
If you use an appropriate high resolution sigma delta ADC it can reverse 
the bridge excitation polarity as part of the measurement sequence and 
give you most of the benefits of an AC bridge with fewer devices and 
lower cost.

The next step up from the gate oscillator for fundamental crystals is 
perhaps Wenzel's circuit:
http://www.wenzel.com/pdffiles1/pdfs/xtalosc.pdf

This circuit needs a little optimisation to improve its performance.
A higher base collector voltage on the oscillator transistor is desirable.
This can be done using a pnp transistor to sense the oscillator 
transistor dc collector current and regulate it by adjusting the dc base 
current.

Replacing the Source follower buffer with a common base transistor 
(allow the crystal current to flow into the emitter rather than Wenzel's 
shunt C) will provide higher reverse isolation.
Cascade a few transformer coupled CB stages to provide gain and 
increased isolation.
With a reverse terminated transformer coupled load in the collector of 
the output CB stage any load from open to short circuit can be driven 
without adverse effects.
The inductor shown in Wenzel's circuit wont be required with your 
crystal either with or without a CB output stage.

The limiting action occurs by cutting off the oscillator transistor 
during part of the cycle.
The dc collector current of the oscillator transistor sets the crystal 
current.

I would build a room temperature version first for debugging.

To minimise the phase noise contributed by the varicap the EFC range 
should be as small as is practical.

A very low noise power supply is also required for good performance.
A modified version (uses 2 transistors and larger capacitors) of Wenzels 
active supply filter can be used to reduce the power supply noise by 
30-40dB for frequencies above 1Hz or so.
http://www.wenzel.com/documents/finesse.html

I can provide circuit schematics if you are interested.

Bruce