[hpsdr] Reference Clock Information

[Chuck Clark]

Clock Oscillator Selection Guide

There are three different types of clock or reference oscillators that 
are useable. All are quartz crystal controlled and may be varied in 
frequency to compensate for temperature or aging drift. They all have 
slightly different features.

VCXO

The VCXO is the least expensive of these references. The features are 
dependent on the crystal but include the ability to discipline the 
frequency. Frequency error none under phase lock. Phase noise, jitter 
and residual FM are all expressions of the short term stability of the 
oscillator. The phase noise would vary from good (loaded Q of the 
oscillator ~100-200k at 10 MHz) to excellent (Ql>10^6) depending on the 
oscillator and crystal.

Problems. No hold over if GPS lock was broken. I think temperature drift 
would be too high.

TCXO

Temperature compensation of a VCXO. This improves the temperature drift. 
Possibly useful in a hold over mode.

/Holdover mode./

/If GPS lock is broken by movement through a canyon (think NYC) and 
satellite view is blocked it is useful to have the control voltage held 
pending satellite reacquisition. A microprocessor monitoring satellite 
lock and tuning voltage can provide the correction loop with the average 
voltage out of the phase detector to hold the oscillator to the last 
known tuning voltage. Holdover is useful for operation when the GPS lock 
is lost. Lock could be lost during heavy rains as well as movement 
through an area with poor sky visibility/.

The TCXO is characterized by the ability to hold the frequency range to 
within x parts per million (ppm) over a specific temperature range. 
Wider temperature ranges usually have wider frequency windows.

Problems. Holdover requires additional knowledge of the temperature to 
keep the frequency extremely accurate.

OCXO

Oven controlled crystal oscillators the ultimate quartz controlled 
oscillators. The approach to the frequency drift is to hold the 
temperature fixed.

Problems. Oven needs to be on all the time. Oven warm up can be up to 30 
days for finial stability. Cost, ovenized oscillators can be ten times 
the cost of a TCXO. The $260 Crystek ovenized oscillator is a low cost 
reference. Some of us would be happy to pay that on Ebay for a used HP…… 
Ovens always require current, the Crystek requires 300 ma during warmup 
and 110 ma for normal operation.

Obliviously I like OCXOs but they may not be better than a VCXO if you 
don’t require holdover. The phase noise doesn’t need to be superior to a 
VCXO or TCXO. It depends of the oscillator and the crystal.

HPSDR Reference Requirements

SDR 1000 200MHz for the DDS.

500 MHz for AD9959. Using the internal clock multipliers increases the 
spurious levels relative to using a high frequency clock This is an area 
which might be used to attack costs and board area. The 9959 output 
should be good through 1.25 meters with a 500 MHz clock.

10 MHz. To lock other downconverters.

HPSDR Solution

Current plan is for a 10 MHz OCXO. This will require a second loop to 
control a 200 MHz reference for the Flex SDR or a HF DDS

A third clock is required to be locked to the 10 MHz for the 500 MHz 
clock for the VHF DDS.

I am attaching a first cut block diagram for thought. There are probably 
better idea’s than I have listed here, please pass them along.


Data, gotta have data.

Engineers must have data. Its part of our genetic makeup. The above 
discussion doesn’t help most of us decide what oscillator is the best 
for the HPSDR. I’ve collected phase noise data on several frequency 
references.

Austron. This is a surplus 11.52 MHz ovenized reference that I bought at 
the Central States VHF Conference a number of years ago. Austron is a 
manufacturer of Loran receivers. Equipment from them should be 
considered excellent. Date coded 1985. The noise floor is the worst 
measured.

TCXO. This TCXO came from a Qualcom synthesizer board. Performance is 
excellent! This cost surplus about $25 at a MUD. This is a 10 MHz unit.

Morion. This is a current production ovenized 13 MHz reference. About 
the best close in noise performance.

Crystek. This is the proposed reference oscillator and data is from 
their data sheet. It certainly looks competitive.

Rohde 10 MHz. This is taken from Microwave and Wireless Synthesizers, 
Theory and Design by Dr. U. L. Rohde, Pages 573 (rf elements) and P258 
(dc biasing). I took the power off the collector as per P573. The 
harmonics are at the same level as the signal, -10 dBm. I followed it 
with a three pole filter.

The crystal was one from my junkbox, where its resided for more than a 
decade! When I measured the crystal the Q was about 100k. This is much 
lower than many good references and that can be seen from the plot. 
Where is just doesn’t measure up. It probably good enough for many 
applications.

We also need some perspective on the accuracy of a phase noise 
measurement. My daily observations of phase noise in oscillators says 
that +/-2 dB is about all I can claim. If one VCO is greater than 2 dB 
better or worse than another its probably true.

I would be happy to collect measured data on a Crystek if someone wants 
to loan me one for the measurement. Likewise for an HP….

The next question is how good does it need to be?