VK1SV LF/MF, flexible, efficient, non-linear, transmit converter



Dimitrios Tsifakis, VK1SV


Introduction

The purpose of this page is to describe my flexible, efficient, non-linear, transmit converter for the two newest bands allocated to amateurs: 136 kHz and 472 kHz. The simplicity of the design will hopefully inspire others to join the fun on the longer wavelengths.

Some explanation of the title is in order. First, I call this flexible because it can be used for either 136 kHz or 472 kHz. The only difference is at the PA section. In fact, one could produce both an 136 kHz and a 472 kHz PA and switch between them for a dual band system. Second, this is an efficient design. Unlike the typical linear amplifiers that rarely exhibit efficiency higher than 50%, this design uses a Class-E PA which, with a careful design, can yield an efficiency or better than 90%! The third attribute of this design is that it is a "non-linear" one. This implies that only non-linear modes can be used with it. The transmitter will not have the ability to produce signals of variable amplitude AND frequency, it will only produce signals of variable frequency. This provides some limitations on the modes that can be used which contradicts the "flexible" title, given earlier. Modes such as SSB will not work with this design. However, the value of SSB on LF and MF is very limited, for reasons of bandwidth and efficiency. The most likely to be used digital modes are non-linear and they include CW (QRSS), WSPR, Opera, WSJTX, good old' RTTY, CMSK and so on. Finally, this design is a "transmit converter". Think of it as half a transverter, it will convert the transmitted signal but not the received signal. This approach is taken in order to simplify things, and it is not a huge handicap in my opinion. Most general purpose transceivers will receive on LF and MF. Some will be better than others. My Yaesu FT-817 seems to be useable on 472 kHz, relatively deaf on 136 kHz. The operator has a number of choices regarding receiving. First, this transmit converter can be used as part of a MEPT one way transmissions anyway. Second option is to use "split frequency" on the transceiver: transmit on 10.136 MHz which will be converted to 136 kHz by the transmit converter and receive directly on 136 kHz. The same applies to 472 kHz. Some antenna switching may be required for this operation, in fact I commonly use a different receive (PA0RDT Mini Whip) and transmit (Inverted-L) antenna. Finally, the operator can use a totally different receiver for receiving on LF and MF. For example, the Icom R75 has a great reputation for its performance on the lower bands.

This document is not meant to be very prescriptive of a tight design. Instead, the intention is to inspire the newcomer to use whatever components are available to design a system that suits their needs in terms of IF frequency and power output.

Design description

The transmit converter can be separated into two major sub-systems: the driver circuit and the PA circuit. Parts of the circuit have been based on Roger G3XBM's design. I recommend the reader to study Roger's design before attempting to create their own, based on the instructions on this web site.

The driver circuit

The driver circuit is responsible for accepting a 1/2 W signal from a general purpose HF transceiver such as the FT-817. The frequency of this signal in my case is 10.136 MHz (or 10.472 MHz) because I use a 10 MHz local oscillator. The 10 MHz signal is attenuated and mixed with the 10 MHz LO which results in a 136/472 kHz signal. This signal is converted to a pulse which is suitable to drive a beefy FET driver, the TC4422. The FET driver produces a 136/475 kHz square wave signal which is suitable for driving the next stage. The following is a photo of the prototype driver board. I used a double-sided copper-clad PCB and my Dremel drill. I find the Dremel drill an excellent tool for doing a prototype, rough PCB, however if you have access to proper PCB producing facilities, the results will be far more aesthetically pleasing.

The following comments apply to the parts of the circuit identified by the numbers in the photo:

1. The incoming 0.5 W signal is attenuated and fed to the RF port of the mixer. The mixer is a Mini-Circuits ADE-1. Although this mixer is specified to work on 0.5 MHz and higher, it seems that it works well enough down to 136 kHz. Remember, it doesn't really matter if the losses are higher as long as the output of the mixer is strong enough to drive the next state - which it is. For component values, consult G3XBM's design.

2. The mixer is a MCL ADE-1. There are many available on eBay at a cost of a few dollars each. The IF (output) port of the mixer follows the component selection of G3XBM. The purpose of the 390 pF capacitor to ground is to eliminate all high frequency products of the mixer but leave enough of the LF or MF signal to drive the next stage. This capacitor can be adjusted to change the timing of the FET driver turning on and off (duty cycle). A duty cycle of approx. 50% is suitable for driving a MOSFET operating in Class-E.

3. The local oscillator consists of a 10 MHz, ovenised oscillator module, also purchased on eBay. A square wave output is fine in this case, as the filtering at the output of the mixer will eliminate the high harmonics leaving only the frequency of interest. The small pot next to the oscillator is a frequency adjustment pot - not all OCXOs have that option. The user does not need an OCXO, but having a stable oscillator may help with the slower QRSS modes and their increased requirement for stability. However, the whole transmitter will be as stable as its least stable component. The two components that will influence the final stability are the HF radio and this oscillator. The oscillator is capacitively coupled to the mixer via a 150 pF capacitor - adjust to suit your oscillator frequency and output.

4. This part of the circuit contains a simple 12 V power supply based on a 7812 regulator. The LO needs 12 V however, if you have an oscillator that needs a different voltage (commonly 5 V) you will need to provide the appropriate regulator here.

5. This is another voltage regulator with provides the Vcc for the FET driver. I have selected a 9 V regulator (7809) because I want a 9 Vp-p square wave to drive my FET. Most FETs will be fully ON with a 9 V signal at the gate. These includes the IRF510, IRF540, IRF640, IRF840, IRFP450 and so on.

6. The final part of the driver circuit is the FET driver. I used a high peak current, single, non-inverting MOSFET driver such as the TC4420 (6 A peak current) and TC4422 (9 A peak current). The bigger the MOSFET, the higher the input capacitance (Ciss in the datasheet) and the more difficult it will be to drive it. The signal from the LO goes to the gate of a 2n3904 which produces a squareish waveform suitable to drive a small FET or a FET driver. G3XBM has a resistive coupling to the gate of the FET, however I have selected to capacitively couple (2n2) to the FET driver input. This is done in order to avoid a FET 100% ON when if the 2N3904 fails. A 47k resistor from the FET driver input to the ground ensures that the input will not float to the "ON" voltage. Finally, a 150 pF capacitor from the FET driver input to the ground eliminates any unwanted high frequency oscillations.

A DRAFT version of the schematic is shown below.



The final product of this driver should be a square signal suitable to feed the gate of the PA FET. The following oscillograms are representative of the expected waveform on the FET gate:


Driver board output with no load and 10.136 MHz input


Driver board output with no load and 10.475 MHz input


Driver board output with 3300 pF load and 10.475 MHz input

The last oscillogram is representative of the waveform at the gate of a relatively big FET, with a Ciss of 3300 pF. In this case, a bit of ringing is observable on the rising and falling edges of the pulse. At this amplitude, the ringing is unlikely to cause any problems to the FET, however if it is larger than this it needs to be eliminated. Common techniques for the elimination of ringing on the gate include the use of a low value series resistor and the addition of a ferrite bead on the FET gate leg.

The PA circuit

The PA is a simple, Class-E design using a single FET. The description of how to design a Class-E PA and how to select appropriate components can be found here. The following schematic, taken from Nathan Sokal's article, shows the design of a class-e amplifier and is the exact design of this PA.

This PA has been designed to deliver 100 W from 100 V. This combination of wattage/voltage is selected because it is convenient; the output impedance, R, is exactly 50 ohm, removing the need for some potentially lossy impedance transformation. As the power output scales with the square of the input voltage, the same amplifier will produce about 25 W from 50 V or 400 W from 200 V, assuming of course that the components have been selected to survive the higher voltages and currents.

An IRFP450 has been selected to be used in the heart of the PA. This is a 16 A, 500 V device. Two 200 V zener diodes are clamping the drain to a maximum of 400 V. This automatically limits the maximum operating voltage to a bit over 100 V and also protects the MOSFET from damage from accidentally exceeding the 500 V drain-source limit. C1 (from FET drain to source) consists of one 1000 pF/1000V silver mica capacitor plus three 1000 pF/500 V silver mica capacitors in series. This results in a combined capacitance of 1333 pF which will operate up to 1000 V. Because of the zener clamp, this capacitor will not see more than 400 V but one has to cater for not exceeding the current limits of the capacitors used. The remainder capacitance required for C1 comes from the FET's Coss, which is connected in parallel with C1. C2 consists of the following capacitor chains in parallel: three times 2x1000pF in series plus 1000pF/500pF/500pF in series. The combined capacitance is 1700 pF and is rated to 1000 V and more than 2 A of current. C2 will see a voltage close to 1000 V peak to peak when the Vcc is a bit over 100 V. Finally, L1 is a commercial 270 uH/9 A unit salvaged from a power supply and L2 is a T200-2 with 83 turns of 1 mm diameter ECW. The voltage across L2 is also approaching 1 kV peak to peak, so care has to be taken for the two ends of the winding not to be too close.

The following photo is the resulting system being bench tested. I selected a supply voltage of 70 volts or so, which will produce an output of around 50 W. This is a good compromise for my simple backyard inverted-L antenna. However, this system has been designed to safely operate at 100 W with a 100 V power supply.

The power supply consists of a 300 VA toroid with two 50 V windings. Only one winding is used with a full bridge rectifier and 4450 uF of filtering capacitance. This produces a voltage of 74 V under load. The current under load is less than 1 A (roughly measured around 800 mA). This corresponds to an input power in the vicinity of 60 W.

The oscillogram below shows the drain waveform.

The last oscillogram shows the output voltage on a 50 ohm load (after a LPF) and the gate voltage. The calculated power output is close to 60 W which implies a very high efficiency, as expected from a Class-E amplifier. A more accurate input current/voltage measurement will be taken to derive an exact efficiency number. However, after an extended key-down operation on a dummy load, none of the components on the PA board was even mildly warm.

The over-engineering behind the design of this transmit converter will hopefully ensure a long life and and successful operation on 630 m.

Dimitris Tsifakis, VK1SV