Design
This transmitter design is based on the control circuitry of the Universal QRP Transmitter which appeared in the April 2006 edition of QST. The case, band pass filters and synthesiser are the same as those used in my Universal Receiver.

The main features of this design are:

The final amplifier is the QRP-PA 2008 kit by QRP Project producing up to 10 Watts on all bands. Following the PA is a relay switched low pass filter, the LPF-100 from HF Projects, which is another kit item as I struggled to find suitable capacitors and decided to take the easier option of buying a kit of parts.

The current drawn on standby is approx 350 mA from a 13 Volt supply. On transmit at 10 Watts output the current is 2 ~ 2.5 Amps depending on band.

There is no ALC or high SWR protection, the PA can stand a high SWR and is run well under the maximum rating of the PA transistors.

The transmitter control board improves on the original QST design by shortening the delay between key down and RF appearing, which is now approx 10 mS. The image below shows the letter "i" sent at the beginning of a transmission at roughly 30 WPM, using an external keyer. The K1EL internal keyer can compensate by lengthening the initial character.

Spurious and harmonic outputs are better than -48 dB. See spectrum analyser plots below for each band:

The case used is a "Unicase 2" by Metcase (www.metcase.com), part number M5502119, which measures 260 x 90 x 250mm (W, H, Depth). These cases are stocked in the UK by RS Components (769-4908). An aluminium sheet is fitted in the centre of the case, this is not supplied with the case. I bought a sheet 250mm square by 1mm thick, which is a perfect fit that just needs cutting to length (a sheet thicker than 1mm can be used, but would need to be narrower than 250mm to fit the side channels of the case). At 1mm thick the sheet would normally be considered too thin, however the bandpass filter mother board imparts sufficient stiffness into the sheet. The knobs, including a 31.8mm diameter tuning knob, are from eBay suppliers. The "signal" meter is from Maplin, with illumination by a white LED.

The front panel was produced in Corel Draw and printed on photo card, which in turn was glued to the aluminium panel with photo mount spray adhesive. A clear plastic self adhesive "book protector" film was stuck over the card to protect it.

Some test equipment is essential to the building of this transmitter, the following will be needed:

An L/C meter to measure capacitors and inductors (an LC200A from eBay is adequate and costs around £25), Spectrum analyser with tracking generator, or signal generator and spectrum analyser (or possibly a general coverage receiver if an analyser is not available) to align and check the band pass filters, a power meter and 50 Ohm load, a low cost multi-meter and a frequency counter. An oscilloscope is useful to check the timing of the TX control board but is not strictly necessary. A vector network analyser (VNA) could also be used to align/test the band pass filters.

For a view of the underside of the transmitter click here, for an upperside view click here.

Module and circuit details
Below are diagrams in PDF format:

* Updated 31 Dec 2013

Board layouts in PDF format for etching:

A Zip file of the above boards in Sprint-Layout format can be downloaded here. The viewer version of Sprint-Layout software can be used to print board layouts rather than use the PDF printouts above.

The control board is made using double sided fibreglass board, the upper side is a plain copper earth plane with clearance for non grounded connections obtained by countersink drilling, the other three boards are single sided.

The bandpass filter boards are 90 x 25mm single sided, pads are drilled using a 7mm diameter diamond tipped cutter. These are sold for drilling kitchen tiles, members of the GQRP Club can purchase similar pad cutters from the members area of their web site.

I do not have a board layout for the keyer, this is a K1EL board and can be obtained from them, or you can use one of their kits which include a PCB.

Power distribution board

The "Off-Stby-Operate-Spot" switch turns on a power relay on the distribution board. An electrolytic capacitor across the relay coil prevents the relay dropping out while rotating the switch between standby, operate and spot. There are two "polyfuse" fuses on the board, check the current needed to supply your synthesiser and uprate the appropriate polyfuse if necessary. 500 mA polyfuses should suffice for most purposes.

The relay was a surplus one I had in my junk box, if your relay has different pin spacing adjust the print layout to suit before etching a board. Note, the print board layout shows several terminal posts for -ve connections, unfortunately my original board only had one.

Control and mixer board

This board is based on the MKII Universal QRP transmitter by W7ZOI with changes to the keying output to minimise shortening of the initial "dit" at the start of a transmission, C13 controls this timing and is reduced from the original 680nF to 330nF. Keying wave shape is controlled on the leading edge by the combination of C16 and R19, the trailing edge shape is controlled by C16 and R15. With the original values the rise time was a little hard, by increasing R19 from 3K3 to 5K6 the wave shape is much better.

Tr3 switches on the crystal mixing oscillator, holding the supply on for a duration timed by R18 and C12. Tr1 forms a Colpitts oscillator operating at an IF frequency of 8.215 MHz, which is buffered by Tr2 and fed by a low pass filter to a TUF-3 double balanced diode ring mixer. Attenuators formed by RA, RB, RC, RX, RY and RZ reduce the 8.215 MHz and synthesizer levels to feed the TUF-3. The output from the mixer is -16 dBm.

C15 and R24 give a hold time sufficient for semi break-in keying, a value of 10uF for C15 is suitable for keying speeds around 22~25 WPM, rising to 22uF for slower hand keying. The actual value is a matter of personal preference.

It is important to use a high speed relay otherwise there is a risk of hot switching the PA, with the values shown the time between the collector of Tr9 going low and RF appearing is 8 ~ 10mS. The Panasonic relay specified has a change over time of 4mS. One pair of relay contacts change over the antenna and the other pair mute the receiver by open circuiting a ground connection, this method of receiver muting is used by my Univeral Receiver and many boat anchor receivers.

Sharp eyed readers will notice the board above is a mirror image of the layout for etching...

Synthesiser
I used a standard kit from Cumbria Designs, however these may have been discontinued. There are several similar synthesiser kits around which are suitable, the requirement being an output level of 7 ~ 10 dBm and a frequency range of 10 ~ 40 MHz. SDR Kits produce a suitable unit which also includes automatic band switching for filters. The Cumbria synthesiser is based on an Si570 device outputting +10 dBm into 50 Ohm and is used with a Bourns 64 pulse per revolution optical shaft encoder in place of the supplied mechanical encoder. The encoder push switch is used as a "tune" switch, wired across the key input, although a separate button could be added to the front panel. The"F/S" (fast/slow/variable tuning speed) button is wired in place of the one originally used with the encoder switch.

Once left on a frequency for around one minute, the frequency is internally stored in the synthesiser allowing the transmitter to be powered up on the last used frequency. This feature may not be available with other synthesiser kits.

Frequency stability is far better than the specification of the Si570 suggests. Frequency drift after a couple of minutes warm up was 30 Hz in the first half hour, followed by virtually no drift afterwards, or at least only 1 Hz or so. This was measured in a centrally heated shack at the final transmitted output frequency on the 10 metre band, using a Racal 9916 counter which was locked on frequency with a GPS derived standard. I didn't try the effect of cooling or heating.

Band Pass Filters

There are 10 band pass filter boards selected by a 1 pole 12 way front panel switch. Each board is constructed on single sided PCB to avoid possible shorts between adjacent boards. Each filter was individually tested/aligned before being added to the main board. The filter design is the same as used in my Universal Receiver and was done using the freeware Windows program "Elsie", using a topology of "Mesh Capacitor-coupled band pass" and "Chebyshev" Family of 3rd order (uses 3 inductors and 5 capacitors per filter, actually produces "5th order filter"). The design was altered to use the nearest 5% capacitor value with inductors wound for the required inductance using a low cost L/C meter. For those not familiar with this program, beware of the default Q settings of capacitors and inductors being excessive (a value of 200 ~ 500 is more realistic) and also set the transmission to "Absolute", both these settings are found under the "Analysis" tab.

The image below is a spectrum analyser plot of the response of the 14 MHz band pass filter, measured between the in/out buss bars. The loss is 3dB.

While it's not strictly necessary to use such elaborate filters in a transmitter, the virtually total elimination of image and other spurious frequencies is worth the small additional cost and effort. Purists may wish to add an attenuator to the output of some filters in order to match the loss across all the filters. Typically the loss will vary from around 1 ~ 3 dB for each filter when measured through the relays and buss bar connections. Up to 4dB loss can be accommodated by the gain of the low power amplifier.

Capacitor and inductor values are critical, being off by a few percent can make a huge difference to performance. Unless you can source close tolerance capacitors, you will generally have to make up the required values by connecting 2 or 3 in parallel and measuring the value with an LC meter. The toroids should be wound for an extra turn and then fine adjusted by removing a turn or two and/or moving the wires on the core until you obtain the required value, estimating the value based on the number of turns will not produce repeatable results. For the higher bands, trimmers were used for the series capacitors (7 MHz and above). T50-2 cores were used from 1.8 to 10 MHz with T50-6 on the higher bands.

The switching relays are generic double pole change over signal relays with 12 Volt coils, a 1N4148 diode is wired across the coil to stop back EMF when switching and each DC switching line is de-coupled to the ground plane of the band pass board.

Each band pass filter is "tack" soldered to a mother board, with the relay switched input/output soldered to a common bus bar. Note the low pass filters I used have polarised relays which require a permanent +ve supply, which is grounded through the relay, this differs from the switching arrangement used in the Universal Receiver.

Low Power Amplifier

This amplifier is based on a design in Experimental Methods in RF Design, by the ARRL. The original design had a far from flat frequency response curve and was a poor match to the 50 Ohm band pass filters. By careful tweaking of the components of the first stage, the amplifier can give a response which is flat to within 1 dB from 1 MHz to 50 MHz. The gain ranges from 23 to 31 dB depending on the setting of the power control, that includes the loss through the 3dB output attenuator. The 1N4007 behaves as a PIN diode in this circuit allowing for gain adjustment between typically 2 and 10 Watts from the PA (the lower power level depends on the band pass filter loss). In the above image the 3 dB output attenuator resistors are mounted under the board. A later version is shown in the layout diagram, and includes track changes for the above.

A metal can 2N2222, or MPSH10, could be used in place of a BSX20. GQRP Club stock MPSH10 transistors.

Power Amplifier
This is an easy to build kit from Germany which gives a flat response from 1.8 to 30 MHz, dropping to 3dB down at 50 MHz. The output is 10 Watts and can withstand a high SWR without needing ALC protection. Following investigation for lower than expected output power on some bands, a sweep of the combined low power amplifier and the PA showed an unexpected peak in the 32 MHz region, adding a 3 dB attenuator to the output of the low power amplifier cured this instability.

The transistors are bolted to an aluminium heatsink approximately 111mm wide x 60mm high and 25mm deep.

Low Pass Filter
Having struggled to source capacitors, I took the easy route and bought a kit from HF Projects in the USA. There is nothing special about the low pass filter requirement for this transmitter and any similar design could be used. I consider the attenuation of the 2nd harmonic on 1.8 MHz (3.6 MHz) and also the 2nd harmonic of 5.3 (10.6 Mhz) to be marginal at roughly 48 ~ 52dB below peak output, however, the FCC requirement for harmonic attenuation is 40 dB, or better, for 10 Watt transmitters and this filter meets that. The Cauer filter used on 40m/60m is very effective when used on 40m but the notches are not effective against the harmonics on 60m. Better filters can be designed using Elsie, and if you opt for an extra low pass filter for the 60m band, the overall performance with regard to harmonics can be improved. However, obtaining the required capacitor values can be difficult and the overall harmonic output is lower than some commercial kit transceivers that run higher power.

Keyer
The keyer board uses a K14 PIC by K1EL and provides an internal keyer with memories and side tone. I bought the IC and made a copy of the keyer board in order to keep the cost under the UK import limit for charging VAT and handling, however a complete kit, including PCB is available from K1EL. The side tone is fed via a level pot to a rear socket in order to provide a direct feed to my Universal Receiver. Beware the sidetone output has DC on it, add a 0.1 uF capacitor in series with the output. The external hand key input doesn't have sidetone, but adding side tone is fairly easily done.

I added an extra 7mS to the initial character in order to avoid character shortening in this transmitter, there is a parameter in the K1EL keyer to accommodate this (extended command "E").

SWR and Power Board

The output of an SWR and Power measuring bridge is amplified with a CA3130 operational amplifier. The meter output is fed via D5 and D6 to improve the linearity. My meter is a basic "signal" meter marked 1 to 5, which corresponds to half the measured output power. The meter reads the peak RF level, holding the reading for several milliseconds by the charge held by C19. C12 is adjusted for minimum reflected power when terminated with a 50 Ohm dummy load. R28 can be reduced slightly if you cannot obtain full scale with 10 Watts output. R23 and R24 set the forward and reflected levels.

Alignment
Each band pass filter board was tested/aligned before adding them to the mother board. Those with trimmer capacitors were adjusted for maximum signal towards the middle part of the CW portion of each band, there was no need to stagger tune them. A final adjustment for maximum power can be done with the boards soldered in place. A spectrum analyser was used for alignment.

The low pass filters were checked for correct frequency response prior to installing the board in the case, this was done with a spectrum analyser.

The PA bias was set as per the kit instructions and the output checked with a signal generator providing the +10 dBm drive. Beware you need a heatsink on the PA transistors, even when setting the bias.

The 8.215 MHz crystal trimmer was set using a frequency counter connected to the output of the low pass filter, the synthesiser was similarly adjusted. A final check was conducted at 29 MHz to ensure the dial correctly read the frequency.

Conclusion
The transmitter works remarkably well and receives excellent reports. The keying and sidetone work really smoothly, sharp eyed readers will have noticed I cheated and used the sidetone from the K1EL board which means there is no sidetone from a manual key input. As I use paddles and the internal keyer, it's not an issue for me. However it's simple enough to make a two transistor oscillator to produce sidetone from any key input.

The circuit diagrams and board layouts were drawn using sPlan 7 and Sprint-Layout 6 from Abacom. Click here for a link to their site.
Elsie can be downloaded from here.

HF Projects (source of the low pass filter) are here
K1EL, the keyer board, is here
And finally, the source of the PA board is here

Please do not build this design without checking the values for yourself, in particular the band pass filter components. The information is presented here to assist experienced constructors who may wish to build something similar, it is not suitable for novice constructors.