Saturday, January 13, 2018

A simple crystal oven/heater that uses no power resistors

There are times where one needs to elevate a component to a consistent temperature to better-maintain its characteristics, the most common being the need to heat a quartz crystal to attain better temperature stability.

I had such a need when I was constructing my 24 GHz transverter - link - and needed to have the 99 MHz crystal oscillator (multiplied by 240 to yield 23.76 GHz for a 432 MHz IF at 24.192 GHz) that was locked to a high-stability 10 MHz reference.  This 99 MHz oscillator uses an overtone crystal oscillator, but these are notoriously difficult to electronically tune over much of a range so I needed to keep the crystal and oscillator at a constant temperature to keep it "close enough" to frequency to be within its narrow tuning capability to allow it to be locked precisely at frequency.

To do this I needed to construct a "crystal oven" - a circuit/device that holds the critical components at a (fairly) constant temperature to accomplish this.

How it works:

Typically, one uses several power resistors to heat a crystal but there are ways that this can be done using no power resistors at all:  Consider the circuit in Figure 1, below.

Figure 1:
Crystal oven using a power MOSFET as the heating element.
R320/Q306 form a constant-current limiter circuit so that the FET can never be fully "on", keeping its resistance higher so that it can function as a heater.  The FET itself, along with thermistor R316 are "thermally coupled" to the crystal being heated.  The (nominal) 0.6 volt turn-on threshold of Q306 and the value of R320 limit the maximum current to about 600mA, but this could be adjusted by selecting appropriate values for R320.
Click on the image for a larger version.
In this circuit Q305, a generic N-channel MOSFET power transistor, is used as the sole heating element:  Its tab could be soldered or bolted directly to the device that needed to be heated such as the crystal or soldered/bolted to a substrate to which the crystal and other components are mounted, etc.  Because the tab of  typical power FET is the drain lead, make sure that it has DC isolation from ground.

This heart of this circuit is the Q305/Q306/R320 combination.  Assuming that voltage has been applied to the gate of the FET via R319, when the current through R320 exceeds that which is required to effect an (approximately) 0.6 volt drop, Q306 turns on, pinching off the gate drive and regulating the current through R305 to that level.  With the components shown, the maximum amount of current that will flow is that which causes 0.6 volts to appear across R320, which, in this case is (0.6 volts * 1 ohm = 0.6 amps) 600 milliamps.  By limiting the current to a reasonable level the FET's "on" resistance is regulated, allowing it to act as a heater.  Without this current limiting U303 would simply turn FET Q305 "on" and its resistance would be a fraction of an ohm, effectively shorting out the power supply - but more likely it would dissipate a lot of power and blow up if there was sufficient power supply current.

Assuming that there are 9 volts available across Q305 (e.g. the 10 volt supply shown, the drop across R320 and a few other, miscellaneous losses) Q305 will produce about (9 volts * 0.6 amps) 5.4 watts of heat, conducted from its metal tab into whatever it is that is being heated.

To regulate the temperature a simple controller is built around an op amp - in this case, a generic 741 (U303) and a thermistor (R316):  Because of the typical 2-4 volt threshold voltage of typical power FETs, the fact that the output voltage of a common op amp like the 741 doesn't get very close to its negative rail isn't a problem - but this should be kept in mind if you happen to use a "logic threshold" FET that starts to turn on at about a volt.  R316 is "thermally coupled" to the device(s) being heated - but not coupled too closely to the heat-generating component, Q305 that there is constant over/undershoot when the heater is active.

The actual temperature at which the oven will stabilize is determined by matching the value of fixed resistor R315 that of thermistor R316 at the desired temperature - a resistance that can be determined from the thermistor's 'spec sheet or by experimentation.  The actual value of the thermistor at the operating temperature is not particularly important but it is recommended that it be in the range of 5k-100k for practical reasons.  The thermistor that I happened to use had a nominal resistance of 30k at 25C, decreasing to about 11k at 50C, the target temperature, so I used a value of 11k for R315.

When the oven is cold the resistance of thermistor R316 is going to be higher than that of fixed resistance R315 which causes the non-inverting (+) terminal of the op-amp to be higher than that of the inverting (-) terminal which is biased at mid-supply by two equal resistors.  When this happens the output of the op amp goes high, providing gate voltage to Q305 via resistor R319, allowing it to heat up.

Figure 2:
The oven and its controller.  The tab of Q305 is soldered directly to
a large, electrically-isolated island of circuit board material.  As can be
seen from the picture, the board to which the heater is mounted is actually
smaller than the surrounding enclosure, mechanically "floating" in the center
via four pieces of small-gauge wire that provide both a DC return and RF
ground connections as well as allowing a gap that is filled with an air
space and insulating foam.  The circuit is wired "dead bug" with the op
amp being "leads-up", just to the right of Q305.  At the bottom of the
picture is a 3-terminal 10 volt regulator (a 7810 - not shown in figure 1) that
provides a stable 10 volt source for both the oven and the crystal oscillator.
Click on the image for a larger version.
When the oven comes up to the design temperature (e.g. that which the resistance of thermistor R316 is the same as fixed resistor R315) the voltages at the inverting and non-inverting terminal of the op amp are equal and the voltage being output by the op amp drops, removing the bias voltage from the gate of Q305 and preventing the oven from heating further.  In reality, there isn't an "on/off" action by the oven, but a more gradual "power up/power down" caused by the inclusion of R317 between the output of the op amp and the noninverting (-) input of the op amp.

It's worth noting that using the turn-on voltage of a transistor as a current reference means that the actual current will vary depending on that transistor's temperature (e.g. that of Q306), but because Q306 is located within the oven chamber it, too, will be heated and the maximum "oven" current (e.g. that through Q305) will be quite stable.  What can affect the dynamics of this oven system is a variation in the voltage applied to Q305:  The higher the voltage, the more power (in watts) will be produced in heat.  While the oven controller will help to maintain temperature, if the power supply voltage is quite variable - as could happen when this oven is run from a "12 volt" battery (anywhere from 11.5-13.5 volts, depending on state-of-charge) the thermal input power can change and cause a slight instability in the closed-loop temperature control due to this change in available thermal power.  In this case - with the crystal being externally locked to frequency, anyway, this wasn't too important.

There are a few other component sprinkled about in the diagram as well:
  • C318 and R321 are used to prevent the Q305/Q306 circuit from oscillating.
  • R322 is an optional test point to measure the oven current.  FT304 is an (optional) feedthrough capacitor used to prevent RF ingress/egress along this monitor point.
  • C316 and C317 are power supply bypass capacitors - always a good practice to include.
  • R318 and LED D303 provide an optional "oven on" indication.  The cycling of this LED between full brightness and dim/off after being powered up indicates that the oven is heating/stabilizing.  If it cycles on and off continually this can indicate that there is too much thermal resistance between the heater (Q305) and the thermistor, causing the circuit to overshoot.
  • The values of R313 and R314 are not critical - but they should be equal.
  • The value of the thermistor is not critical, but it should probably be between 1k and 100k at the desired operating temperature.  Select R315 to have the same resistance as the thermistor at the desired oven operating temperature.
  • To provide a temperature adjustment, R315 may be made variable with a good-quality multi-turn potentiometer.  Alternatively, resistors R313/R314 can be replaced with a single 20k-50k multi-turn potentiometer.
What to use this for?
Figure 3:
The "oscillator side" of the circuit shown in Figure 2.  The oscillator is a 5th-
overtone "Butler" type build "dead bug" on a piece of double-sided
copper-clad epoxy board.  The crystal is located directly opposite the
location of Q305, the heater (e.g. the board is rotated 1/4 turn counter-
clockwise from Figure 2).
Click on the image for a larger version. 

I devised this circuit when constructing a homebrew 24 GHz amateur radio transverter (transmit/receive frequency converter) and needed a "fairly stable" source of a 99 MHz signal to be multiplied upwards and to be locked to an outboard, stable 10 MHz reference (e.g. high-stability crystal or rubidium source.)

Because this crystal oscillator was to be externally locked, it didn't need to be ultra-stable - just stable enough to keep its temperature close enough to the rather limited frequency-pulling range afforded by high-frequency overtone crystal oscillators.  With this relaxed requirement, the crystal could actually vary a few degrees about the set point with no ill effects whatsoever.


I didn't need to optimize this circuit for ultimate frequency stability as the 99 MHz oscillator is locked to an external 10 MHz reference:  All that is necessary is that the frequency be "close enough" - which is to say, within the rather narrow frequency tuning range afforded via the VCXO (Voltage Controlled Crystal Oscillator) circuit.

If this circuit is to be used for a "stand-alone" oscillator where the frequency is directly affected by the temperature, additional care will be required to appropriately thermally couple the thermistor and heater (Q305) - and possibly tweak the value of resistor R317 - to prevent the temperature from oscillating about its set point.

Finally, even though the temperature controller is entirely ratiometric - that is, power supply voltage variations will not affect the temperature set point to a significant degree -  remember that at higher voltages the power going into the heating element (e.g. the power FET Q305) will also increase.  This change in thermal input can cause the dynamics of the oven to change somewhat and slightly change the rate-of-change and potentially alter the stability of the feedback loop.  Because this oven was intended to keep an oscillator "close enough to" rather than "dead on" frequency this factor wasn't important.

Additional resources:
  • W6PQL Crystal Oven Controller - link - This discusses a more conventional "heater-resistor" circuit for maintain constant crystal/component temperature.
  • The OCXO/Si5351A synthesizer - link - QRP Labs sells a version of their Si5351A synthesizer board with a built-in crystal oven controller providing a stability of 1ppm or better.  The assembly manual for this kit (linked on the referred page) uses low-power FETs as heaters and a bipolar transistor as the temperature sensor.  This manual discusses the operation of the circuit and is an interesting read.


This page stolen from

Thursday, January 4, 2018

A quick fix for a Yaesu FT-757GXII blank display

A couple weeks ago I was contacted by an old friend of mine, having obtained his amateur license in mid-2017, who has a Yaesu FT-775GXII - a synthesized, all-mode HF transceiver from the the mid-late 1980s which had been working well until, one day, there was no display.  Clearly the main processor was fine as the front panel buttons would work, he could transmit and receive and he could even "see" and control the radio's frequency and mode on his computer via the serial CAT interface.
Figure 1:
The front panel of an FT-757GXII with a working display!

One clue was that when power cycled, the display would occasionally flash very briefly, a possible indication that something was almost doing what it was supposed to.  Via email (he lives across the country from me) I had him do some preliminary troubleshooting such as the checking of voltages - but based on the rather sparse information available in the service manual and the difficulty in accessing some of the test points:  Even a couple key capacitors in circuits that often cause problems with displays in some radios - namely the switching supplies that provide the odd voltages for the vacuum fluorescent display - were swapped out, but the display remained blank.

I offered to look at it, so he packaged it up and sent it to me.  When I put it on my workbench I started probing the various lines on the display processor with an oscilloscope:  I could see many of the signals that I was expecting - namely the 500 kHz signal from the display processor's clock, the data coming from the radio's main processor that changed as I pushed buttons and turned the main tuning knob and another signal that appeared to be an acknowledgment pulse from the display processor to the main processor.  What I seemed to be missing were half of the multiplexing signals that drove the display:  It appeared that I was seeing the "common" signal lines for the display, but the signals on the pins that appeared to carry information as to which display segment was to be illuminated were missing as if the display was supposed to be blank.  Without both sets of signals activated appropriately, a multiplexed display will remain forever dark.

I'd already consulted the internet and determined, based on postings in various forums, that at least for its predecessor, the 757GX, the failure of the display processor wasn't terribly uncommon - but not surprisingly this part was long gone from the spare parts inventories of Yaesu and other means of repair/replacement such as getting displays from scrapped radios or even the construction of an "alternate" display unit using a different processor and driver transistors was discussed.  What was interesting was that the "important" signals - namely those for data, acknowledgement, scanning and synchronization - seemed to be present, so the display processor clearly wasn't completely dead.
Figure 2:
Annotated picture showing the two buttons that, when both are set to their
"in" position will disconnect the radio's internal memory back-up battery.
If both buttons are in when the power is removed the processor will be
reset to its "factory" state.
Click on the image for a larger version.

At about that point the old adage drilled into me from the early days of computers and Windows came back to me - although it probably should have been one of the first steps to be taken when the display went blank:  "When in doubt, reboot!"  Perusing the user's manual I determined that a complete "memory reset" was done on the FT-757GXII by setting both the "Linear" and "Marker" switches on the back panel (see Figure 2) to the "in" position at the same time and turning off the radio for 30 seconds - and then turning it back on and restoring the two rear switches to their normal position:  It would appear that these two switches have a second, "non-intuitive" function that when used together, disconnects the internal battery.

The result?  The display came back to life!

What had apparently happened was that somehow, the data stream between the display and main processor wasn't what it should be and the main processor was apparently sending some sort of garbage that the display processor didn't understand - probably due to something in the main processor's static RAM.  It would appear that in the absence of sensible data, the display processor remains blank, relying on the main processor to send the various bits and bytes that display frequency, mode, etc. rather than reverting to some sort of static display.  Clearing the battery-backed RAM of the main processor and resetting it apparently cleared whatever junk had gotten into the memory that had caused it to work improperly.

I checked the back-up battery - an innocuous-looking 2-cell NiCd pack that was near the rear of the main synthesizer board - and it read 2.8 volts with the radio having been disconnected from power for over 24 hours indicating about 1.4 volts/cell, which was appropriate for a properly-charged NiCd.  Visually, this small battery pack looked OK in that there were no signs of corrosion, so it is probably OK, despite its age - longevity being one of the virtues of a properly cared-for, high-quality NiCd cell.

How did the main processor's memory get scrambled?  Who knows - it could have been an entirely random event, due to static from a finger touching the front panel, the back-up battery's voltage having sagged below the point of memory retention while the radio was turned off or the results of some sort of spike - perhaps lightning - intercepted by the antenna that found its way into other circuits.  This sort of "display failure" - apparently caused by the processor's memory being scrambled - doesn't seem to be too common, so my friend considers himself very lucky!

After restoring the radio's operation I did a few tests and found that everything seemed to be working as it should, so it will be packed up and returned to its (very fortunate!) owner very soon.


This page stolen from