SP-9: A Compact Five band SSB/CW QRP Transceiver

‘Summit Prowler 9’ is a homebrew five band SSB/CW 5 watt transceiver designed for and tested on the summits near Melbourne Australia. This project further developed my interest and ideas on the right mix of features and design choices in a moderately compact case that any keen radio builder could reproduce in the home workshop with modest equipment. The transceiver project was completed over an 18 month period to April 2021.

Schematics

Page 1 is the receiver:

SP9-p1Download

Page 2 is the transmitter:

SP9-p2Download

Page 3 is the microcontroller, PLL and associated control pieces:

SP9-p3Download

Listening to G4 station on 20m.

VFO/BFO/Controller

The VFO, BFO, CW keyer and all control functions are provided by my usual Arduino Nano and si5351 combination. This module is built sandwich style. The front double sided board supports front panel encoder, encoder switch and three pushbuttons, all soldered directly to wide pads. The reverse side hosts a carrier oscillator buffer for CW and its DC supply switch, the T/R relay driver FET, and a PCF8574 decoder and five filter relay driver FETs. It also mates with the second board via 0.1″ headers.

This board hosts the Arduino Nano, si5351 breakout, a VFO buffer (MMBT3904), R/C sidetone filtering and a 7805 voltage regulator. The module is conveniently self contained and can be built and tested standalone.

Front view of the VFO, BFO and Controller module showing the SSD1306 OLED, LED-illuminated pushbuttons and mechanical encoder.

Rear view, self explanatory. Hidden in the middle board are a 5V regulator, MMBT3904 VFO buffer, carrier oscillator DC switch and MMBT3904 buffer ( for software-generated CW).

Receiver front end

The front end is almost identical to that developed for an earlier transceiver project, SP7. It consists of a switchable, AGC controlled dual gate MOSFET RF amplifier, a Minicircuits JMS-1 double balanced mixer,  a balanced tee diplexor and a post-mixer Class A amplifier stage (2N2219A). Fifty ohm pi attenuator pads are used for impedance stabilisation and level-setting to the L7 mixer and the IF amplifier that follows. 

The switchable RF amplifier is a conventional broad-band NTE332 dual gate MOSFET stage, but departs from that used in SP7 in that the switching is done with SA630D BiCMOS RF switch from NXP (2014). In an earlier rig (SP7), this amplifier stage was switched using a front panel toggle. This time, I decided to use an Arduino digital output to switch this stage for the 20m band only.

MOSFET RF amplifier test, demonstrating times 6 voltage gain or 15dB into a 50 ohm resistor. 

Diplexor test rig.

Diplexor bandwidth.

Diplexor sweep, showing -1.37dB at 9MHz, and a 2MHz -3dB bandwidth.

The mixer (a JMS1) is a double balanced mixer; with the LO and IF used, these appear to exhibit around 5 dB conversion loss. The post mixer amplifier stage has a fairly flat gain of about 15dB. The band pass filters (see below) show a loss of about 2dB. So overall gain is as follows: -2 (BPF) -5 (mixer) -2 (diplexor) +15 (post mixer amp) = 6dB overall gain. Add about 10dB with the RF preamp switched in.

IF filter and amplifier

Unlike previous transceiver projects, I had no crystal filter in mind for this rig. Peter DK7IH has been using these tiny 8-pole 9MHz SSB filters from Germany. They look ideal, but the delay involved in landing one in Melbourne was unknown. So I layed out the IF board with space for a 9MXF24D, but make up a simple 4-pole experimental filter using on-hand 9MHz matched crystals from Minikits. The board is encased in brass sheet sides and top — as the crystal filter sits at the front of an 80dB gain stage, good shielding is essential.

Simple four pole 9MHz crystal filter, sweep shows it centred on 9001kHz, -40dB bandwidth about 2.2kHz, roughly 5dB of passband ripple; not great but good enough to develop and test this transceiver.

Various AGC-controlled IF amplifiers were considered; the design by Eamon EI9GQ was chosen on the basis of overall gain, dynamic range and size. Three pairs of BF246A RF JFETs arranged in a cascode fashion each behave like dual gate MOSFETs with signal on the lower FETs gate and AGC on the upper FETs gate. The stages increase gain with increasing AGC voltage, just like a dual gate MOSFET.

As built, the IF strip had way too much gain and oscillated on tuned circuit peaks ( at or around 9MHz). The first stage’s tuned circuit was damped down with a parallel 5k6 resistor. The board measures 95 x 38mm.

IF amp test rig.

30dB gain.

IF amplifier.

Cascode BF246As.

IF board. In test, 100mV at output of crystal filter yields 3V-pp or 30dB gain.

When the tiny 9MHz 9MXF24D filter from our friends at FunkAmateur arrived it was dropped in and resulted in a noticeable improvement in the passband.

Product detector, audio and AGC

These receiver stages occupy an irregular 95 x 50mm board. The product detector is an SA612, which is followed by a low pass RC filter, a discrete audio preamp and a NEC uPC2002 audio amplifier.  A two-transistor AGC circuit occupies a small vertical double sided board that doubles to screen the audio power amplifier.   The transmit-receive relay occupies the right side of the board for antenna and DC switching. 

Receiver back end board.

AGC on vertical divider.

Clockwise from top: SA612 product detector, T/R relay, 7812 regulator, AGC (vertical), upc2002 audio power amplifier.

Five-band Band Pass Filter module

This board contains five individual Band Pass Filters, relay switching for each, and transmit- receive switching to allow the selected filter to be used in both receive and transmit modes. 

Band Pass Filter module, from left, 20, 30, 40, 60 and 80m. Relays at the top select one of the five filters. Relays at the bottom do T/R switching.

The filters are by Eamon EI9GQ (Radcom homebrew columnist).  Each filter uses four adjustable parallel  LC tuned circuits with coupling. This module is obviously critical for the spectral performance of the transceiver, and is probably the most design and labour-intensive.

EI9GQ BPF sweeps.

The filters are fairly consistent, 400 to 500kHz wide at the 3dB points and with insertion loss between 3 to 5dB. This figure is higher than EI9GQ reported (1.5 to 2dB). I used regular 1206 X7R 50v ceramic caps. One DPDT relay was used to switch each filter.

The 80m filter initially came up with a good shape but insertion loss of 8dB. I breadboarded it and found that using the regular shiny blue leaded 50v ceramic caps in the four resonant tuned circuits brought the insertion loss down to < 2dB. From subsequent discussions with Nick N1UBZ and David VK3KR, I learned that X7R surface mount capacitors are not great for RF. Minikits stocks caps with a better dialectic, I should use these in future. That said, the difference is 3dB which may be accommodated in the receiver’s overall gain distribution.

Receiver tune-up

Connecting these boards together was straight forward, and yielded band noise and signals. As a result of independent module testing and getting the IF gain about right, no major changes were necessary to get it working well on all bands. The receiver has plenty of gain, and on 80m rides on the,AGC line to keep tamed.

Transmitter

To get this bunch of boards to transmit required four more modules: a microphone amplifier and balanced modulator,  a transmit mixer, the driver and PA chain, and a five-band software-selectable Low Pass Filter.

Exciter board, before the addition of brass shielding.

Mic amp and balanced modulator

The three transmitter stages were laid out on a single board. These modules reproduced those used in SP7 and are copied from SSDRA p202/203. The mic amp uses a FET for a high impedance microphone preamp and an op amp for gain. The LM1496 in balanced modulator configuration has a tuned output at 9MHz.

Transmit mixer

Another LM1496 configured as a transmit mixer with broadband output, and followed by a broadband gain stage (MPSH10).

Drive

The broadband driver uses a BFU590GX. This is a fairly hot little RF amp transistor that is good to microwaves. It is my attempt to employ a modern replacement for the 50 year old 2N5179. The first one blew up on the test bench, not sure why, I may have let a clip lead go astray. The stage delivers 250mW into a 50 ohm load flat to 30MHz, depending on drive.

The pre-driver is a BF961 dual gate MOSFET. This device was chosen for a reason. In other rigs, drive drops off on the highest band, in this case, 20m. I arranged two trimpots, diodes and a switching transistor into an OR gate so that each trimpot sets the gate 2 bias. Trimpot 1 sets the pre-driver gain on the lower bands. Applying 5 volts to the transistor base brings trimpot 2 into the circuit which can be set to lift the bias for the higher bands. A digital line from the Arduino provides the 5v.

PA and 5-band LPF module

The 5 watt power amplifier stage uses a single Mitsubishi RD16HHF1 RF FET. The circuit was copied from one developed by Glenn VK3PE, and is commonly used. It includes a p-channel DC switch to control the 12 volt rail and the gate bias, allowing the entire PA to be enabled or disabled via a 5 volt control line, such as an Arduino digital line. It is important to drop the gate bias when receiving, as this can be set as high as 250mA.

Five watt PA and LPF module.

Three W3NQN LPFs were made to cover the five bands — 80 and 60m, 40 and 30m, and 20m. Each filter is switched via a pair of Telecom relays. Diodes on the five-band select bus ensure the dual band LPFs are shared appropriately.

LPF sweeps.

Testing

RF power output was measured on a 13.7 volt DC supply as: 80m 6.2 watts, 40m 7.8 watts, 30m 7.6 watts, 20m 6.3 watts.

Output waveforms, DC supply 13.7V, 50 ohm load, on 80, 40, 30, 20m.

A spectrum plot was done using the SDRPlay RSP1A running the supplied spectrum analyser software. The rig was connected to a dummy load, with 23dB of attenuation in series with the analyser, delivering around -20dBm to the RSP1A. The test revealed 50 to 55dB suppression of the second harmonic on 30 and 20m, and 34 to 35dB suppression on 80 and 40m. This directly equates to how three LPFs were cut to cover the five bands. Space permitting, a dedicated LPF on each band would achieve better than 50dB harmonic suppression.

Transmitter spectrum on 80 and 30m, second harmonic -35dB and -57dB respectively. Other results: 40m -35dB, 20m -50dB.

Case

The case is made from stock aluminum, hand cut and worked, with pop rivest and M2 and M2.5 bolts, barrel-head and countersunk. The case has two angle pieces running along either side as bearers for top and bottom PCBs. This scheme created receiver (top) and transmitter (bottom) compartments, and allowed open access to the PCB components and tracks for testing. The Arduino controller, si5351 breakout and control components are mounted on a pair of boards that sit parallel and behind the front panel. The front panel is made from 3mm angle, painted black, with Decadry white lettering and several coats of protective clear enamel. The labels on the right side panel are Decadry (black) applied direct onto the cleaned aluminium surface, with clear enamel top coats.

Aluminium case details, 235mm long, 100mm wide, 40mm high.

Wrap-up

This project was an exercise in scratch building a compact 5 band SSB and CW QRP transceiver using approaches and circuit blocks I’d used before. As with anything, doing it for the 2nd or 3rd time is always easier and the results better.

I addressed a number of omissions or weaknesses that bugged me from earlier rigs, including a fully integrated loudspeaker, convenient location and spacing of controls (for me at least), a ‘true’ RF PA transistor (RD16HHF1) to ensure 5 watts on the higher bands, a smooth and functional AGC, loud audio, a rigid microphone connector, and a case that offered top and bottom zones for receiver and transmitter. SP9 met my expectations and has given me an ideal SSB and CW rig for a wide range of backpack or portable outings.

Why include 60m in VK? In the planning, I included 60m in good faith, as there was much anticipation amongst VKs at the time. This project was well advanced before the Australian Communications and Media Authority announced the refusal of the amateur radio community’s request for access to the band. At some stage in the future I may replace the 60m filters to get 17 or 15m. That task has been left for a rainy month.

If you got this far, thanks for reading this homebrew radio story. Please feel free to discuss any aspect of this project, by leaving a comment below. 73 from Paul VK3HN.