G4ALG's QRP Radio Pages

[ Previously GW4ALG (QRT in February 2007) ]

Home ] Antennas ] Equipment ] Introduction ] Links ] QRP Operating ] Station Summary ]

Up ] Audio Filter ] FL-50B TX ] FL-50B VFO ] FR-50B RX ] IC-R71E receiver ] IC756PRO ] 'Multi-Five' VXO TX ] Experimental Valve Transmitter ] Morse Sidetone Oscillator ] SP220 Power Meter ] [ All-Band Valve Transmitter ] QRPP VXO TX for 5 MHz ] 80 m Transceiver ]

All-Band Valve Transmitter

This page describes a project that is nearing completion, and follows my decision to abandon an attempt to make a broadband valve transmitter using untuned stages.  As Colin, G3VTT had correctly predicted at the start of that project, I was indeed unable to get enough voltage gain using untuned ferrite cored transformers.  But I had fun trying, and I re-learned a lot about valves along the way.

Happily, many of the parts from my now defunct Experimental Valve Transmitter project are being used for this project.  It's fun to be using valves that were made about the time I was born!  Old valves being used by an old timer.

This project was started following a discussion with Jan, PA3GSV who has significant experience of building well-made valve receivers and transmitters.  I have received several suggestions from Jan regarding the use of xtals in valve oscillators.  

While soldering together the first version of the xtal oscillator, I happened to hear Martin, G4ZXN calling CQ on 60 m CW.  Knowing that Martin has done much work with small crystals in his Paraset and other valve rigs, I naturally told him what I was building.  Martin boosted my confidence by saying that I shouldn't worry about breaking modern crystals in a valve circuit.  Then he added, " . . although, they might squeal a bit".  

The short term objective was to make a 30 m (10 MHz) valve QRP CW transmitter while keeping in mind my eventual desire to produce a multi-band transmitter.   I have now taken that next step to make this a 10-band transmitter, covering all bands from 160 to 10 m.  A variable frequency crystal oscillator (VXO) operating at signal frequency is used to control the transmit frequency.  This requires the use of modern, often tiny, quartz crystals in the EF80 valve oscillator.  The key requirement was to use a VXO configuration that keeps the crystal current low to avoid damaging the crystal, and to prevent frequency instability.


Variable Crystal Oscillator (VXO)

Here's a photo of the early prototype crystal oscillator, including the prototype power supply:


Following tests with several different oscillator circuits using crystals stocked by the G-QRP Club, I took the best aspects of each circuit to produce the oscillator design shown below.  From the outset, I sought to use plug-in crystal units because this would  provide some of the nostalgia associated with the early 'crystal controlled' valve transmitters that used the large plug-in crystals.

I figured that the large housing (or shroud) available for the connector would permit installation of several small components such as multiple crystals, a miniature switch, and a small inductor that, together, could be used to determine the VXO range within a single plug-in unit (Pins 12 and 24).

As the design developed, I also determined that some of the oscillator feedback components could, if required, be placed within the plug-in unit to provide the correct level of feedback for the selected band (Pins 10, 22, and 24).  But it didn't end there.  I also decided to make the plug-in unit 'band aware' by using spare pins in the connector (Pins 1-5, and 14-18).   I also made the connector 'state aware' by extending the transmit/receive status to Pin 19 which is grounded whenever the transmitter is in the 'Spot' and 'Operate' conditions. 

It should now be possible, at some later date, to control the transmit frequency and to undertake band switching from an external DDS VFO, using the transmit/receive status to disable the DDS VFO output when on receive.

This electron coupled Colpitts oscillator draws about 3 mA from the HT supply, and generates about 10 V p-p on 10 MHz at the output, dropping to 5 V p-p on 28 MHz.  It is possible to short the output to ground with no discernable change in frequency.  This is the beauty of the electron coupled oscillator!

So far, I have not observed any indications of stressed crystals when using this circuit.  28 MHz crystals are a bit slow to start, but I have no intention of keying the oscillator.


Plug-In Crystal Units

Below, I have shown some example plug-in modules that provide a useful VXO range.  Some modules will rely on the internal capacitance of the EF80 plus the in-circuit 22 pF fixed capacitors to provide a suitable level of feedback to start and maintain oscillation.  Other band modules may require additional components within the module to achieve reliable oscillator start-up (pins 10, 23 and 24) and the required VXO range (pins 10 and 12).  

In particular, note the possible use of a series coil to bring the centre frequency lower in frequency, and the trick of using parallel crystals to raise the centre frequency.  There is considerable scope for experimentation with different configurations, including the use of parallel crystals having different resonant frequencies up to, say, 20 kHz apart.

Here are some photos to illustrate how I mounted a miniature toggle switch within a 25-pin connector housing.   Firstly, a piece of plain matrix board is cut along the perforations to produce a small piece of board of 4 clear holes by 8 clear holes (12  x  22 mm).  A 6 mm hole is then drilled in the centre of the board, and the switch mounted in the hole.  This sub-assembly then drops in the plastic shell.  The board is a little undersized, but you can decide how precisely you wish to cut the matrix board, or similar material, to fit the space.  Then it's just a case of wiring in the crystal array and screwing together the two halves of the shell.




And here's a closer look at my 12 m plug-in module:


Note that, in this case, a coil is used in series with the 24.906 crystal to pull it low, providing a large VXO range.  The switch is used to provide a second VXO range by shorting out the coil.   The result is that this single quartz crystal covers nearly the entire CW segment of 12 m, with a useful overlap between the two switched ranges.


And here's how I squeezed 5 crystals, a switch, and a 10 pF capacitor into one of my 40m plug-in modules:


Grid Tuning Module

To get enough drive for the PA, two driver stages are required, with a tuned circuit at the grid of the pre-driver (EF80), the driver (5763), and the power amplifier (5763).  Relay-switched grid tuning modules are used to select the required parallel tuned circuit for each of the three stages.  Each grid tuning module uses ten Omron G5LE-1 relays (one per band).  The relays are powered from a 5 volt 3-pin regulator, and the appropriate relays are operated during all three Spot/Standby/Operate states via the Band Select pin in the selected plug-in crystal unit. 

The resonant frequency calculator at:

and the toroid inductance calculator at:

were very helpful when calculating the number of turns required on the toroid cores.


This photo shows one of the three grid tuning modules.






Power Amplifier

This PA is intended to provide 5 watts RF output.  In practice, the power output falls off above 18 MHz to about 4.5 watts on 21 MHz, and 4 watts on 28 MHz.

Pi Network

Because this transmitter has been designed to work into a load of about 50 ohms, a large value variable capacitor for adjusting the loading over a wide range of output impedances was not required.  Instead, fixed-value loading capacitors are selected using switches SW1 and SW2.  The PA LOAD switch is used to select one of two loading capacitors appropriate to the 10 to 80 m bands.  A significant amount of iterative tests were carried out to find values of inductance and loading capacitance that provided close to 5 watts output on all bands.  (By using optimised values, a lot more power output was possible, but I selected values that provided the desired power output into 50 ohms.)  

Spare positions on the BAND switch are used to provide three loading options on 160 m.  Overall, the PA LOAD switch and associated capacitors save a lot of space normally occupied by a very large variable capacitor, thereby permitting a smaller chassis.

An excellent Pi network calculator can be found at:

When using the above calculator, I guessed the main variables as being: Rin = 4000; Rout as 50; and the loaded Q (working Q) as 13.  The calculator was used only to provide a starting point for winding the Pi Network coils.  The final number of turns on the toroid cores was determined through iterative adjustment of each coil, starting with 10 m.  


Keying and Grid Control Circuit

This keying circuit provides the interface from an electronic keyer (or straight key) for keying the Pre-Driver, Driver, and PA stages using grid block keying.  The Keying Out line is used to key an external sidetone oscillator.    The circuit also provides a Power control for varying the transmitter power from 0 to 5 watts RF output.

If a fault occurs on the grid block keying line, the diode at the base of the 2N6520 prevents a negative voltage reaching the keyer; and the diode at the collector of the 2N6520 clamps a positive voltage to a safe level.

The keying circuit ensures click-free CW by controlling the waveform rise and fall times to all three keyed stages.   The result is a pleasant CW note having rise and fall times of 3 to 4 ms.




Transmit-Receive Control Circuit


Production Version

These photos were taken during the design of the mechanical layout.


And below, you can see the main aluminium parts: all wire-brushed; cleaned in Brasso; rinsed in soapy water, and drying in the sunshine.


Here are the 10 prototype plug-in crystal units used to align the transmitter.  More work is required to optimise the VXO range and centre-frequency of each plug-in unit.


On 19th March 2023, I made my first QSO with the partially completed TX.  I had a nice chat with Fred F8FXA on 10.116 MHz.  I was running 5 watts of RF to a doublet, and Fred was running 4 watts from his K2, also using a doublet antenna. 

On 20th March 2023, more 'firsts' were made, as follows:
20 m: John, SM2FIJ - 2 x QRP
40 m: Cliff, GI4CZW - Running a Mk. 123 transmitter (1950s British 'spy set') at 14 watts
60 m: David, G4HMC - 2 x QRP
80 m: Chris, G3XIZ - 2 x QRP


More photos of the transmitter are shown here.  Click on the thumbnail to enlarge.



Power Supply

The circuit of the prototype power supply is shown below.  Because I no longer have any HT transformers, I used two 30 volt transformers back to back to obtain the HT voltages.  

1) There is poor short circuit protection at the outputs.  Even some fuses here would help.