Comments? email to
Many stock microwave transverters have a problem: DRIFT. Some do it just a little, others a lot more. This is especially troublesome when operating a portable "rover" station. Drift is only annoying on the lower UHF and microwave bands; but on the upper ones it can make the difference in making the contact or not. After all, you have enough to do aiming the antenna, etc. without worrying about being way off of the other station's frequency.
Microwave transverters are usually stable enough at home if left to warm up for a long time in a controlled environment; even then, there are those that begin to drift after a couple of long transmissions when their innards warm up.
The two most common solutions for this problem are:
Phase-lock the local oscillator to a stable source (usually a 10 MHz oven-controlled oscillator) -or-
Retrofit the existing LO with a temperature controller
I like the second solution. There are pros and cons for both methods, but what makes me avoid the phase-lock approach is the higher phase noise and spurs, both of which degrade the receiver's ability to hear the weak ones. These things also cause trouble on the transmit side, especially if you are sharing a mountain-top with another station.
I use the oven controller circuit shown below; it is very effective on many of the commercial transverters on the market today. Development data and theory follow next, but If you want to fast-forward to instructions and examples of how to retrofit the more common units out there, select the appropriate link below.
First of all, I have to admit that I'm no expert on crystal ovens and temperature control; what I've learned and will describe here is the result of about 3 weeks of experimentation and testing, but I did manage to learn enough to figure out how to get more than a 10x stability improvement on all of the stock units I worked on.
One of the before/after tests I made was how much the LO moved when a muffin fan was directed over the transverter case for 10 minutes (I thought this might be a reasonable parallel to windy conditions in the field). My unmodified 10 GHz unit moved 12 KHz at 10 GHz. After retrofit, it moved less than 1 KHz. On my 1296 unit, the movement was reduced to less than 100 Hz.
Here are some key factors to consider when designing temperature control for any crystal oscillator:
- The crystal turn point. This is the temperature at which the crystal produces the least change in frequency for a given temperature change. If you don't know where the turn point is, you can find it experimentally. I usually buy crystals with a turn point between 50 and 60C. This allows a reasonable elevation over ambient temperature to cover most environmental conditions without running the temperature up so high that it stresses or prematurely ages components. With the component values shown above, the temperature will actually be 66C at the sensor (see the controller setup table for details).
The sensor. I've been using an inexpensive SMT thermistor with a negative temperature coefficient. As the heat goes up, it's resistance goes down, allowing the controller to sense the change and compensate.
- Loop gain. Too little amplification of the sensor signal, and the control is poor. Too much, and it's like a dog chasing it's tail...the LO frequency runs up and down in an endless loop. In most installations, I set the loop gain to 200. This gain is set by resistors R4 and R5.
- Placement of the heating element. It's common to see a heater attached directly to the middle of the crystal can, but this is not always the best spot to put it. The crystal is the most sensitive component, but there are other components around the oscillator that are temperature-sensitive and also affect the frequency. Heating all of these components together is usually the best approach. You also want the heater as close to these components as practical to reduce thermal transmission delay, allowing as much loop gain as possible.
- Placement of the sensor. Again, it is common to see the sensor on the crystal can. This gets you close, but placing the sensor on the ground foil of the board closest to the crystal element wires often works better. I'm just speculating , but it might be that because these wires are attached directly to the crystal element inside the can, they have more thermal influence than the surrounding can.
- Black magic and luck. Despite all common sense about placing the sensor, there is usually a bit of luck involved in finding the perfect spot. Locating it as described above will get you very close to optimum. Moving it around a few mm one way or the other on the ground foil (staying in the vicinity of the crystal) can often locate a place that's just a bit better. The ebb and flow of heat is the cause of this anomaly; think of it like waves in water, canceling themselves at certain spots, adding in others, etc.
- Insulation. This often helps. It slows the heat transfer in and out of the controlled area, making the controller's job much easier. I use common household fiberglass pipe insulation for this; it comes in a roll 3" wide by several feet long by 1/2" thick (uncompressed). It forms easily, and is inexpensive.
OK, let's get to the modification of some of the more well-known transverters. And by the way, these modifications are all completely reversible in case you ever want to restore your transverter to it's original configuration.
Modifying DEMI transverters with MICROLO LO (13 through 3 cm)
These transverters are constructed in an aluminum extrusion clamshell. The LO is in the top half on the right, and the rest of the transverter is in the other half on the left.
Normally, there is only one piece of coax coming from the LO. I temporarily added the second piece you see running over to the BNC connector on the rear panel, which allowed me to measure the LO frequency during testing. Otherwise, you see it here in unmodified form.
The first step in the retrofit is to temporarily remove the
four 4-40 mounting screws securing the MICROLO board to the chassis. There are
usually two foam end blocks at each end of the board, presumably to create some
dead air space around the board. Set these aside also.
Turn
the board over and locate the crystal (middle right in this picture).
There is a heater/sensor (PTC thermistor) soldered to the crystal case, and to a 9v feed on the PC board. The job of this part is to heat the crystal case to a given temperature (50C); it does do that, but has only limited success in maintaining stability.
You'll be disconnecting this part from it's 9v feed, effectively eliminating it from the circuit.
Unsolder the 9v feed lead, and tack-solder it to the ground
foil nearby. Or you can unsolder it from both the crystal can and 9v feed and
remove it completely if you prefer.
Here's
a closer look at the crystal on the unmodified board. Note that there is a wire
soldered to the can and the foil at the upper-left of the crystal.
A lead was apparently removed from one side of the PTC thermistor, and the thermistor then soldered directly to the crystal case.
The remaining lead from this part is soldered to the 9v
pad on the PC board, and the crystal element wires go to their respective pads
on the board.
As
you can see here, I removed the old heater component completely and cleaned up
the excess solder left behind.
Then I soldered one end of the new sensor to the ground foil just between the element wires, and ran a small wire from the other end to the 9v feed.
In one of the following steps, the 9v feed wire will be removed from this pad on the other side of the board, and the pad then converted to a sensor connection point.
By the way, this spot was excellent, and the controller was
able to hold the 10 GHz output to within 1.5 KHz during the 10 minute stress
test. I did find an even better spot, though (see the notes on
sensor placement and black magic and luck).
OK, here's that magic spot I was telling you about. I spot-soldered the back of the crystal can to the foil, and buried one end of the sensor in the joint as shown here.
Why this spot was best makes no obvious sense, but the movement during the stress test narrowed to less than 500 Hz at 10 GHz.
When all is said and done, either spot is fine; but I'm glad I
took the time to find this one.
Cut
a 10" length of 3" x 1/2" fiberglass insulation and place it in the clamshell
channel where the LO board mounts. This insulation will be folded over the board
when it is mounted later on.
Next, place 1/16" thick fiberglass washers at the mounting holes. These washers can be nylon, fiberglass, or any other material that is a poor heat conductor. I used fiberglass washers from www.mcmaster.com (p/n 93493A106).
You are insulating the LO board from the cold,
cruel, heat-exchanging chassis. This will not affect the R.F. performance of the
LO., but will aid the controller board in maintaining a fixed temperature.
Using
the same mounting screws you removed earlier, fasten the LO board back into
place. Don't over-tighten.
Assemble the controller PCB using this setup table as a guide; then, tack-solder the controller PCB (a prototype in this photo) to the location shown, just above the oscillator transistor, and just beside the 9v pad that fed the PTC thermistor. The soldering should take place at the ground end of the load resistors near the frequency netting capacitor and the oscillator transistor. Placing the controller here will position it to deliver the most heat to the proper area, and keep it out of the way of the pipe cap filters when the transverter is put back together.
Disconnect the 9v feed wire from that pad to the left of the controller board, and solder it to the 9v pad on the controller board. Inrush current for the controller is about 500 ma, and drops to 200 ma or so as the temperature stabilizes. The existing 9v regulator that supplies the MICROLO has enough capacity for this, and has a proper heat sink.
Run a wire from the sensor pad on the controller
board to the old 9v feed pad. This pad is now the sensor feed pad.
I was curious about how well distributed the heat was, so I placed sensors in various locations on the board to find out.
The numbers shown in red were measured with the transverter assembled, and the main sensor at the crystal stabilized at 66C.
Even with insulation, it's clear that the temperatures at the extremities, particularly near the mounting screws, is much cooler; lots of heat loss there.
This is a good illustration of the importance of
placing the heating elements (load resistors and driving transistor) in the
right spot; in this case, directly over the crystal on the other side of the
board.
Finally,
fold the insulation over the LO assembly, put the two halves of the
transverter back together, and you're done.
Modifying older model DEMI transverters
The
unit shown here is an older model 900 MHz transverter. In these units, the
filters used are the PC board hairpin style, and the oscillator is not
enclosed within a metal cover, as it is on the current production units. The
LO is in the upper right corner in the picture on the right. This same
arrangement was used in some of their older transverters for other bands as
well.
I don't think the original oscillators used the PTC thermistor seen soldered to
the crystal can in this close-up; it was probably added later on to help
stabilize things. Anyway, you can remove it...it won't be used.
Notice the 9v regulator IC next to the hairpin trace? The
controller board will need about 500 ma of inrush current to do the
initial heating, so the next thing to do is provide better heat sinking for this
part.
To
improve the heat sinking, I cut a 1.75" x .25" copper strip out of .015 copper
sheet and drilled 1/8" holes 1/8" from the ends.
I formed it as shown, holding one end under a mounting nut on the board, and secured the other end to the regulator chip with a 4-40 screw and nut.
You can substitute an aluminum strip here instead, but it should be a bit thicker, perhaps .030. Don't use brass, it's a relatively poor heat conductor.
Assemble the controller PCB using this setup table as a guide; then, tack-solder the controller board (a prototype in this photo) to the ground foil at the edge of the board as shown. Solder where the load resistors are located along the bottom edge. You'll have to angle and tilt the board a bit to clear the input attenuator at the upper left. Not all units have this attenuator (an option), but this one did.
Run a wire from the 9v pad on the controller to the 9v trace feeding the oscillator.
Solder a wire to the crystal can and connect it to the ground
foil at the right corner as shown. Solder one end of the thermistor into this
solder pool on the ground foil. Run a wire from the other end of the thermistor
to the sensor pad on the controller board.
That's all there is to this one; the sensor was set for 66C, and the controller was able to hold temperature well enough to keep the LO within 150 Hz when subjected to the muffin fan 10-minute stress test. You can probably get even better stability by using some insulation around the LO, but even without it, the stability is much better.
Modifying later model DEMI transverters (with a cover over the LO)
These newer models can be recognized by the cover that encloses the LO section on the top of the main circuit board. The one shown here is a 900 MHz unit.
I made some
interesting measurements on this style transverter prior to retrofit, and it
helped me understand why they are not 100% stable in their stock form.
Remove
the cover over the LO section (held on with two 4-40 screws) and put it aside
for now. In a subsequent step, the controller board will be affixed into the
inside of this cover.
The 9v regulator IC inside does not have enough heat sinking to handle the extra 500 ma of inrush current that the controller will draw from it, so the first step is to bond it to the ground foil on the PC board.
This must be done in such a way as to allow the cover to be replaced without interference, and stay out of the way of the controller board, which will be affixed to the inside of the cover's roof.
Bend the IC over as shown so that the top of it's tab is no higher that 5/8 above the ground foil. It's going to get crowded inside that cover.
I used a small piece of .015 copper strip (1/2" x 3/4" with a 1/8" hole drilled 1/8" in from one end. Form the strip as shown above, and before you secure it to the IC with a 4-40 screw and nut, solder the edge of the copper strip to the ground foil as close to the edge as you can get it without shorting out the components to the left of it. You may have to form it a bit better than I did; perhaps straight up for a few mm, then over and up again. After securing it with the screw and nut, check the fit of the top cover.
Note
that there is a PTC thermistor soldered to the crystal can, and connected to 9v;
the other lead from the crystal can goes to a ground pad. Disconnect the
thermistor lead from 9v and solder it to the ground pad to keep it out of the
way.
I chose to remove the thermistor completely as shown here, but
you can leave it in place if you prefer; just be sure to disconnect it from the
9v feed.
Assemble the controller PCB using this setup table as a guide; then, mount the controller board into the inside roof of the LO cover. You'll need a hot iron to do this...I used a Weller soldering station with an 800f tip. Prep the two areas of the can you need to solder with some flux, and tin them up a bit before attempting to solder the board into place. Then solder the ground foil to the can in the two places shown. I used a small piece of solder braid to help bridge the small gap where it is soldered at the lower right.
The board must be positioned so that the hole near the "to R2' marking aligns with the hole in the cover. You may have to tilt the board slightly, or file it's sides a bit to get it in place. As long as you can get to the LO frequency netting screw, you'll be fine.
Check to make certain that when you put the cover in place,
the load resistors on the controller board do not touch the tab on the 9v
regulator. I used a small ruler to check this out.
The hard part is done. Next, solder one end of the sensor to the ground foil near the 9v regulator as shown. Connect a short piece of wire to the other end to run to the controller.
Note that the sensor is not connected to the crystal can here;
it was not found to be the best spot. For some reason, this particular spot was
best. See the notes on black magic and sensor
placement.
Connect
a wire to the 9v pad near the regulator, and run it to the +V pad on the
controller; and connect the sensor wire to the sensor pad ("to R2") on the
controller as shown.
Reinstall the cover, taking care not to pinch the wires. I had to remove the screws holding the main board in place to get the nut onto one of the cover screws, but it will lift up just enough to allow this easily. Put the bottom cover of the transverter back on, and it's all done.
After warm-up, the muffin fan torture test moved the LO less
than 30 Hz (at 902 MHz). I didn't use any insulation on this one; it didn't seem
to be necessary.
Modifying Down East VHF Transverters
The
only VHF transverter I have is a 222 MHz unit, shown unmodified in the photo on
the right.
Although it did warm up and stabilize in about 3 minutes, and
control was much better than with the standard PTC thermistor, I didn't get the
orders of magnitude improvement like I did on the 33cm and above transverters. The main reasons for this are
the location of the heat-draining mounting screw right next to the crystal, and
the lack of an effective way to mount the controller board to efficiently heat
all of the oscillator components together.
I think the 144 and 432 units are better suited for retrofit; from the pictures
I've seen, it looks like there are no nearby mounting screws draining
heat, and there appears to be room to mount the controller board properly, as in
the 902 example.
I didn't experiment with this one much because the original PTC heater already did a
fair job; at VHF, there is no multiplying of the crystal frequency, so drift is
not normally a problem. If I wanted to do a proper job, I would probably remove
the mounting screw near the crystal, and perhaps the one near the filter, and
relocate the controller and sensor. But I really didn't think it was worth the
trouble. However, if you'd like to go ahead, here's what I did:
The 9v regulator on this board needs some extra heat sinking to handle the additional 500 ma of inrush current drawn by the controller.
I cut a 3/8" wide piece of .015 copper sheet and formed it as
shown on the right.
Tin
the non-drilled end of the copper strip, and place it into position as shown.
You can pass the screw through the regulator tab to hold it in place, but wait
on the nut until you have a chance to solder the other end to the foil.
Assemble the controller PCB using this
setup table as a guide.
OK, the next thing to do is solder two
small solder lugs onto the
back of the controller board. Bend the one near the load resistors toward the
front of the board, and the one on the other end toward the back.
Unsolder the PTC thermistor lead from the 9v feed and tack it to the ground foil to keep it out of the way. I removed mine completely.
Place the controller as shown and solder the lugs to the foil. The board needs be up above the foil enough to clear the component leads underneath.
Run a wire from the +v pad on the controller to the 9v feed.
Unsolder the crystal can from the transistor can, and spot-solder the crystal can to the back foil on the controller board as shown. Solder one end of the sensor into this joint, and run a wire from the other end to the sensor pad on the controller board.
That's it.