Reworking the WA7GIE 10 GHz
propagation beacon


The WA7GIE beacon was installed in 2004 atop one of the higher mountains in the Oquirrh (pronounced "Oh-Kerr") range, west of Salt Lake City and with the locals just starting to build/buy narrowband 10 GHz gear, it was a welcome addition, being useful to both test their gear to see if it was working, but also as a common frequency reference when making contacts to simplify coordination of QSO as in "I'll call you 25 kHz below the beacon."

The beacon originally used the DB6NT beacon transmitter, model "MKU10 BAKE" with an internal crystal oscillator and was set for somewhere in the area of 10368.15-10368.20 MHz range (I'm not sure...) but over the course of the day it would drift 5-10 kHz - depending on ambient temperature at the radio site - and it seemed to be drifting slowly down in frequency over time.  When Dave, WA7GIE, had the opportunity to do so, he'd occasionally nudge the frequency back up, but during 2011, he ran out of tuning capacitor range and the beacon was beginning to encroach on the weak-signal calling frequency of 10368.100 MHz. 

This left Dave with two options:
The second option - the use of an external oscillator - was chosen as a good-quality OCXO (Oven-Controlled Crystal Oscillator) could be used to hold the frequency to much greater precision and stability than had been possible with the oscillator that was built into the DB6NT unit and I took this on as a project.


From the factory, the DB6NT unit is available with either an "external" or "internal" reference - the WA7GIE beacon originally using the latter.  Converting it to use an external input simply involved coupling the external reference via a 680pF capacitor to the appropriate side of the crystal's connection (e.g. the one "closest" to the input of the transistor in the oscillator circuit as viewed on the schematic)  To accommodate this new connection a right-angle female SMA connector was soldered to the outside of the case to which the RF output from the OCXO could be connected.  Because the MKU10 uses a 96-fold multiplication of the crystal to obtain the 10 GHz frequency, a stable signal source in the 108 MHz region was required.  In this case the target frequency was 10328.200 MHz, so the oscillator's output frequency was to be 108.00208333 MHz due to to the multiplication by 96 in the DB6NT beacon transmitter itself.

Using a die-cast box, I modified the DB6NT unit slightly moving the voltage regulator - the main heat-generating component - to the same side of the box as the RF output SMA connector and in the side of the die-cast box a large-ish hole was also drilled to accommodate the SMA connector and its mounting hardware.  A hole was then drilled through the DB6NT box that matched the hole in the regulator chip's heat sink and used both as a mounting point and to transfer heat from the regulator to the die-cast box while another hole was drilled on that same side of the DB6NT box - as far away from the regulator as possible - and used as another mounting point, both of these serving to hold the DB6NT unit inside the die-cast box. 


Figure 1:
An OCXO of the type used in the rebuilt WA7GIE beacon - a Vectron "229" series oscillator.
Click on the image for a slightly larger version.
Vectron 229-series OCXO
Re-working the OCXO:

Having had recently worked on the K7RJ beacon I was already familiar with what was involved in putting together a stable frequency reference and as with the K7RJ beacon, a surplus Vectron "229" series OCXO was chosen - see Figure 1 for a picture of the OCXO used in the K7RJ beacon.  This unit operates from a single +24 VDC source, requiring about 400 milliamps when "cold" and taking 10-15 minutes to warm up to reasonable stability and was designed to operate in the 90-110 MHz region (something in the 101 MHz area was stamped on the case).  Because the target frequency was 10328.200 MHz, the oscillator's output frequency was to be 108.00208333 MHz due to to the multiplication by 96 in the DB6NT beacon transmitter itself.

These units are soldered shut, so the tuning access screw was removed to allow air to escape and to prevent it "spitting" hot solder and the device was wrapped in a rag and firmly clamped upside-down in a vise while the outside edges of the end of the case with the soldered connection were quickly and carefully heated with a propane torch - always keeping the flame moving - while firmly (but gently) tugging on 6-32 nuts spun onto the four mounting studs.  Just as the solder flowed on one side, the flame was moved about while each corner was eased up and with a bit of care, the "bottom" portion will came off while I was being very careful not to yank out any wires!

If you were careful, the circuity and wiring in the "bottom" of the case will remain intact, but I've frequently had to re-work the circuits, usually due to something becoming un-soldered by the heat (usually a ground) or a blob of molten solder bridging across something that it should not.  Once it has cooled, it should be possible to extract the oscillator unit in its foam cocoon from the case and you should note the arrangement of the wires.  It's not unusual to break the bottom foam piece into 2 or 3 pieces, but just as long as you keep it from breaking apart even more and save all of the pieces, you'll be fine.  Once it is apart, use a hot soldering iron or soldering gun to remove "blobs" of solder that might impeded reassembly.

Typically, the OCXO itself is in a small, copper case wrapped with the oven's heating element.  On one end (usually that with the access hole for the frequency adjustment capacitor) there are blobs of silicone adhesive holding two protruding bumps of the circuit board in place and scraping these clean and then pushing on them will allow the copper case to come apart.

Inside, one will be able to see a TO-5 crystal case (e.g. one that looks like an old metal-cased transistor) and attached to it - or floating freely within the box - will be a small, round, paper tag on which is written a number - the operating temperature of the original crystal in Centigrade:  Keep that number as you will need it when you specify the temperature at which the new crystal will be cut!

At this point, you'll be ready to order a new crystal and expect to pay about $50 plus shipping from International Crystal.  The folks there actually have the "formula" of this particular OCXO in their records, but to save cost, you can order a crystal cut with a lower precision:  Since you'll be tweaking it with the tuning capacitor, its "natural" frequency is less importance.  When ordering the crystal, also specify that it be cut so that it operates at the same temperature of the original device using the information written on the small, round paper tag that you should have saved!

The actual frequency of the crystal is 1/2 of the operating frequency - or, in the case of a 10 GHz beacon with a crystal operating near 108 MHz, the crystal will be "cut" for 54 MHz or so and the doubled in the OCXO to the final 108 MHz frequency.  When the crystal arrived, I carefully removed the old one noting the lead length and its orientation, re-using the spacing hardware that was  been used on the original and cutting the crystal leads only after they have been soldered to avoid damaging it from the mechanical shock!

Using a service monitor with a spectrum analyzer I applied +24 volts to the OCXO and observed that it was outputting a signal on the appropriate frequency - in this case, close to 108.0020833 MHz.  While observing the output power, stretch and squeezed the two air-core inductors on the board next to the one with the crystal until maximum power was obtained at the 108 MHz frequency which was around +10dBm.  Now that the output tuning network had tuned, I was ready to put it back together!

Reassembling the two halves of the copper case back together, I put new "blobs" of RTV adhesive (the "non-vinegar" type!) on the knubs from which they were originally removed, allowed the silicone to cure overnight, and then put it back in the foam, taking care to orient it so that one can access the tuning capacitor through the hole in the case.  After this, I carefully put the pieces of foam (it's nearly impossible to disassemble these things without breaking some foam!) and firmly pressed the bottom of the OCXO back into the case.  I then powered up the OCXO again and waited 20 minutes before determining if the mechnical adjustment easily tuned through the desired operating frequency.  Once satisfied that it was operating properly, with solder at the ready I used a torch to quickly re-seal it with minimum heat.

Once it had cooled, I re-tested the oscillator and found that everything was still working so kept it powered up and set it aside for several days until I got time to continue work on the project.  In that time, the oscillator drifted down 10-15 kHz at the 10 GHz frequency - but this was entirely expected:  This oscillator had been powered off for years and it likely took some time for the oven to stabilize in temperature.  In addition to this, it was likely that the parts inside were "baking out" and losing absorbed moisture, mechanically settling in, and most of all, the brand new "green" crystal was aging.  In the several weeks that I had the completed beacon before handing it over to WA7GIE, I kept it powered up, testing it on my roof, and it was heard throughout the Salt Lake valley and in this time, it drifted down a few more kHz indicating that the effects of aging and "bake-out" had already slowed.


If you don't want to re-crystal the OCXO yourself, you may be able to find someone on a microwave-related internet group that may be willing to do it for you for a reasonable cost.  It's possible that International Crystal may be willing to re-work the oscillator - also for an additional cost!
Figure 2:
Diagrams of the power supply and oscillator
keying (top) and beacon keyer (bottom)
Click on an image for a larger version.
Power supply and RF drive of the beacon
Beacon keyer


Power supply and oven section:

Figure 2 shows the diagrams of the external portions of the beacon.  While the DB6NT beacon module is not shown schematically, it is just a matter of feeding the output of the oscillator into the position formerly occupied by the crystal, using a blocking capacitor, as described above.  The output of the oven oscillator is at the 108 MHz frequency of the original crystal and because of this, no further modification is needed to the DB6NT unit.  If it was originally ordered with the "external" option, the output of the crystal oven oscillator would be connected there.

This Vectron ovenized oscillator has an external frequency control and as shown in the diagram, this is used to provide "fine tuning" of the crystal frequency using a multi-turn potentiometer (R202) but as we can see, an external FSCW (Frequency Shift CW) signal is applied via R201.  The other components - C201 and C202 help to stabilize the keying, both reducing "chirp" from keying as well as a low-frequency warble that emanates from noise of the Zener diode regulator internal to the oscillator and present on the "+REF" pin.

One potential difficulty with the Vectron oscillator - and many similar oscillators - is that they require a +24 volt supply, an inconvenience since lower-voltage supplies (nominally 11-15 volts) are typically already available.  To accommodate this, an LM2577-12 switching up-converter is used to supply the +24 volts for the oscillator, using R102 and R103 to set the voltage to 24 volts:  If the "adjustable" version (e.g. LM2577-ADJ) had been available, I would have used it, along with the appropriate-value resistors to set the voltage.  As can be seen in the diagram, while this supply is quite simple to build, suitable voltage up-conversion modules are easily found on Ebay and similar sites, many of them having a voltage adjust capability.  If one uses a pre-built unit, be sure to test it and adjust the voltage - preferably under at least 100 mA load - before connecting it to the oven!

As can be seen in the diagram there is fairly extensive input and output filtering on the switching supply.  On the input there is L101/C101 located at the power supply input to the enclosure containing the DB6NT unit and the oscillator as well as C102/L102/C103/C104 on the switching power supply board itself along with C106/C107/L102 and C107 on the output of the switcher.  This additional filtering is absolutely necessary to keep the switching supply's noise out of the oscillator and the DB6NT unit!

If you happen to buy a pre-built switching converter, it is strongly recommended that one adds additional capacitance and inductance as shown in the diagram, particularly when one considers the fact that most of these inexpensive, pre-built converters tend to use "cheap" capacitors that are probably not well-suited, by themselves, for the task of removing the switching energy!  Adding additional capacitance to the input and output - directly on the converter - will likely go a long way to prolonging its operational lifetime.  Also, these inexpensive converters do tend to be a bit "dirty" when it comes to the amount of switching energy that they impart on the input and output leads and this additional filtering is arguably more important.

For an example of a project that uses a pre-built voltage up-converter and extra filtering, see the page:  A Portable 10 MHz Rubidium Frequency Reference using the FE-5680A.

Keying section:

For generating the keying signal a simple keyer based on an 8-pin PIC is used.  When I wrote this code I wanted it to be as flexible as possible and have produced several different versions, but they all have one common trait:  In FSCW mode they all output differential keying.  This is a fancy way of saying that when one of the keying pins goes high, the other goes low in voltage and vice-versa.

Connected across the keying line is a 10k potentiometer and with this differential keying, the net result - if the potentiometer is set exactly in the middle is a voltage that is one-half of the supply voltage (2.5 volts in this case) no matter what the keying state may be.  The advantage to this is that simply by adjusting this potentiometer, one may select both the magnitude (amount of shift for the FSCW) and the sign (whether a "key-down" is a higher or lower frequency) with one simple adjustment.

Our preference has been to set the magnitude (amount of shift) of the keying to something on the order of 1.5-2.0 kHz while the sign of the keying is such that a "key-down" condition causes a shift upwards in frequency.  The reason for doing it this way are three-fold:

In addition to keying, the differential lines also connect to a 2-leaded dual-color LED and key-up/key-down is indicated by a red or green LED as desired.  Of course, one could use just a single LED or two separate LEDs for this indication!  The advantage of the dual-color LED it is always on no matter what the keying state so that one could easily tell if it was powered up - plus it required only one hole to be drilled!

The PIC used also has an A/D converter and it is used for two purposes here.  The voltage from the wiper of potentiometer R201 is read and the speed of the Morse keying is adjustable in this way, while the voltage appearing across U203, an LM335, is read and converted into a temperature that is included on the transmitted Morse message as telemetry.  On the WA7GIE beacon it was decided to place the LM335 inside the enclosure and as such, it typically reads between 110 and 125 F (43-52C) since it is located next to the crystal oven and the DB6NT module.

The results:

The re-configure beacon has been on the air since late 2010 and during the first year, it drifted down 12-15 kHz from its original frequency of 10250.200, this largely occurring due to aging of the crystal and due to some "settling in" of the oscillator components.  After about 18 month of operation, the opportunity arose and the beacon frequency was re-set to something near its intended and since then, it has continued to drift downwards, but at a much slower rate.

With this rebuild, the beacon frequency is much more stable and appears to vary less than 1.5 kHz in frequency over the course of a day - far better than the 10-15 kHz of the original configuration!  Since the beacon is easily heard over a wide portion of Northern Utah, it is still used as a frequency reference when setting up QSO's, as in "10 kHz below the beacon" and as a "sanity check" to see if our transverters are anywhere near where they are supposed to be!

If there is interest in obtaining a PIC for your beacon project - or, if you have other questions about this beacon, please contact me using the email link below.

Go to the KA7OEI microwave page, or go to the KA7OEI main page.

If you have additional questions, you may send email using the link at this page.

This page updated on 20130107

Since 1/2013: