Improving Crystal Oscillator Signal Purity by G8ACE
Many of us Microwavers purchase low cost crystals from providers such as QuartSlab, Klove and Eisch. At 24 Ghz and above it sometimes happens that the received signal using a new low cost crystal can sound full of LF jitter or perturbations . This makes copying of weaker SSB signals quite difficult. The blame for this effect can be laid directly at the door of the crystal itself. I have been pursuing a cure for this problem for some time, it being quite difficult to get suppliers to discuss the problem and therefore track down a solution. The crystal blank quality on low cost crystals appears to drop to very poor quality at times. Possibly quality that is adequate only for computer crystals. Both G4BRK and myself purchased 100.2 MHz crystals from Klove recently and both these showed large amounts of jitter. Other crystals purchased at the same time from Klove are perfectly good but because the frequency was different they where almost certainly made from different quartz blanks of better quality. So the first unknown is the crystal blank quality. It seems likely that once a bad quality crystal frequency appears its likely to persist until all that stock of blanks is exhausted. Additionally there appears to be a likely hood that variable stresses are contained in the bond wires the crystal element within the can adding to this jitter problem. QuartSlab suggested some time ago that holding a crystal vertically and using a soldering iron to heat the lead out wire for ten seconds to allow the heat to travel up inside the unit would reduce the jitter. This was tried and an improvement in the amplitude of the jitter of perhaps 2:1 was obtained. I also understood that professional users often store new crystals at around 80º C until they are required. I had kept some crystals under the jacket of my hot water cylinder for several years but on finally using these the jitter was no different to a new crystal. This warm storage may help with ageing though I have no measurements to support this aspect. Another crystal supplier then suggested cycling the troublesome crystals between around +80º C and -10º C for one hour at each temperature for 48 hours.
A simple arrangement to do this cycling was to use my OCXO¹ unit itself to do the crystal heating. A resistor of 4K7 was added from TP1 on the circuit diagram to the +10.5v regulated supply. This allows the upper oven temperature range to be extended. Additionally if the heating is done using the complete OCXO it is important that high temperature Epoxy was used to bond the heater plate to the PCB board. For the low temperature simply placing the bare OCXO module in the freezer was the simplest solution. Only the crystal itself needs to be cycled so if you have a means then it can be done outside the OCXO.
|
|
An evaluation method needed to be established to verify any improvements that might be made. The system diagram shows the arrangement employed. Two Adret units are employed to establish a reasonable quality reference. The Adrets themselves will not be perfect but the results show them to be considerably better than a ‘bad’ crystal.
The IF output from the transverter is connected to the FT290 Rx and the resulting CW audio note connected to the sound card in the PC and the DL4YHF² Spectrum Lab software.
Figure 1 shows the frequency perturbations of the test set-up. Vertical scale is 10Hz/div. Horizontal divisions are 10 seconds. The amplitude of the jitter is around 10-15Hz pp. This amount sounds perfectly satisfactory on CW
|
Figure 1 |
Substituting the test crystal in its OCXO for Adret No. 2 a measurement of that can be made. Figure 2 shows the new 100.2 MHz crystal as considerably worse than the test setup. Vertical scale is now 25 Hz/div. The jitter amplitude being as high as 300 Hz pp. This is very audible as a nasty wobbly note.
|
Figure 2 |
The method of heating and cooling the crystal is rather arduous and was done as time permitted. The crystal being re-measured at intervals for signs of improvement. Temperature cycling during the daytime over a period of around five days gave the result in Figure 3. Here the vertical scale is 20 Hz/div. So its possible to see around 30 Hz pp. jitter and a result about 2:1 worse than the Adrets but an overall improvement of about 10:1 for this particular crystal . Another crystal for which a graph is not available improved to match the Adret result so its possible that the Adret jitter would be the limiting measurement factor for this case.
|
Figure 3 |
The improvement is not necessarily as good for all crystals. If you have the jitter problem then only by doing the temperature stressing will you find how good the crystal is capable of becoming. The sharp frequency transistions shown in Fig. 2 are assumed to be stresses in the bonding wires which presumably are annealed in the temperature cycling process along with some stresses within the crystal resonator. Luis Cupido, CT1DMK, on a recent visit suggested increasing the crystal drive level. This I assume will help accelerate the shedding of any rubbish on the crystal itself.
The design drive level in my OCXO is nominally 750uW to the
crystal. By shorting R2, the effect of
the amplitude limiter diode D1 is removed and by shorting R8 the crystal drive
level is raised to 10mW whilst maintaining around +4dbm output level. A really high power oscillator is possibly
needed but there is some evidence that running the OCXO for several days at
this higher drive level has some improvement effect. The tests have been done at 24 GHz because the writers portable
transverter design allows easy testing.
Testing at 10 GHz would be possible but the amplitude of the
measurements would be around 2.3 times smaller. It is assumed from the crystals evaluated so far that this jitter
only manifests itself as an operating problem at 10 GHz and higher. Regrettably the degree of improvements that
can be obtained can only be found with the extended cycling and
re-testing. So far all crystals have
improved to a greater or lesser extent.
It has been reported and observed that the problem does improve on its
own over a very long period but I assume this is due to the slower action of a
crystal being heated and cooled in its application at a much lower cycling
rate. If you have a problem crystal
then testing and cycling should reward you with an improved LO note purity.
References.
1. http://www.microwaves.dsl.pipex.com/