PIC-A-STAR Software Transmitter and Receiver

 

The PIC-A-STAR is a home-build SSB/CW transceiver project designed by the extremely talented Peter Rhodes, G3XJP. So called, because housekeeping processes within the transceiver are handled by Microchip Technology Inc. PIC microcontrollers and the generation & detection of signals accomplished in software running on the Analogue Devices AD2181 DSP chip. A series of articles supporting the construction of Star was published in Radcom, the monthly journal of the Radio Society of Great Britain (RSGB) between August 2002 and March 2004. Very detailed current documentation and support for the project is available via the PicaStar Yahoo group A brief summary of Star features follows:

I commenced construction of my Star during December 2005 and on 25th February 2006 made my first qsos on 20m using Star as seen below. Construction has been a fascinating and thoroughly enjoyable experience. That said, the project is not yet finished, it has merely entered a new phase, in which I can make contacts using Star, whilst I continue its integration.

This is my Star as used "on air" from 25th February 2006 . At the front can be seen the Pic'n Mix synthesiser display board with the DDS pcb just behind. Behind that and slightly to the left is the Magic Roundabout H-mode 1st mixer board and to the left of that the integrated DSP/IF assembly. Just to the front of the power meter is my bpf assembly and on top of the meter itself is a G6ALU design 20W p.a. Audio reinforcement is currently provided by an old computer sound system. The PC is monitoring a multitude of operational parameters using a monitor utility developed by Peter, G3XJP and provided, along with all other operating software "free for personal use".

 

In the photograph above are a number of kitchen table/kitchen sink manufactured pcbs for my STAR project. These were made with a transfer process using PnP Blue. TIF files of the PCB layouts are printed directly on to PnP Blue with a laser printer and then ironed onto scrupulously clean PCB material. Some boards are single sided and some double. I had a few early disasters but once I got used to the process it became very easy and repeatable.

 

 

Above is the DSP motherboard with two daughter boards attached, making the complete DSP assembly. I had put off any idea of surface mount homebrew for years, believing it would be impractical. Oh but how wrong I was. I discovered that a whole bunch of folks were having success with these devices and I decided it was time to have a go. I did think it might have been better had I started a few years sooner when my eyesight was better but in practice my poor close-up vision was no significant barrier. Working successfully with SMT devices requires little by way of specialised equipment. The above assembly was put together using my standard Weller type temperature controlled solder station and using standard 60/40 multicore solder

The important thing is good visibility. I purchased an illuminated desk magnifier from Maplin and this, combined with the magnification of my reading glasses, provided everything I needed. I worried that my hand might not be steady enough to deal with soldering these very small devices. I need not have done. It is amazing just how steady one's hand becomes when provided with good optical feedback. Most of my jittering came from not being able to adequately see what I was trying to solder. The combination of specs and desk magnifier solved that problem.

The complete DSP assembly on the bench and under test. All very easy with the first class software tools and support provided by G3XJP and other Star builders. To the rear and just left of the computer screen is the IF assembly into which the DSP unit was about to be integrated.

By 10th March I made well over 100 contacts with Star on 160 - 80 - 40 & 20 metres using both SSB and CW. Reports have been excellent and Star has been a joy to use, despite the Heath-Robinson style lash-up on the desk top. Time to move on to a more permanent and more durable integration scheme with more appropriate screening of various modules.

The DSP assembly is now installed in a four sided box made from double sided FR4 copper clad pcb material. The DSP assembly is back to back with the completed IF assembly which can be seen below.

The foil covered assembly at top centre is the H-mode 2nd mixer, which combines signals from the 1st IF at 8830kHz with local oscillator at 8845kHz to produce 2nd IF signals at 15kHz to be fed to the DSP assembly for processing. Top right is the 8845kHz local oscillator and beneath it, also screened, is the balance of IF circuits at 8830kHz. At the bottom and to the left of that is the screened underside of the 8830kHz crystal roofing filter, which removes the 2nd IF image, displaced only 30kHz from the wanted signal. The miniture coax cable exiting stage left is 8830kHz in/out, to/from the Magic Roundabout 1st mixer and amplifier, which can be seen below.

This is the Magic Roundabout (MR) H-mode 1st mixer and quad J310 pre-amp/post-amp. The MR uses bus switches to effect deployment of the quad J310 amplifier either before the 1st mixer as an rf amplifier or after it, as an IF amplifier, depending upon whether best IP3 (LF bands) or best NF (10 metres) is required. On transmit the quad J310 amplifier is always in its best IP3 location in the IF chain. As can be seen all transformers are hand wound using low-cost cores. The twin blue cores located centre right are a diplexer. The MR is surrounded on four sides by a box made from FR4 and is about to be completely enclosed with a lid which will be soldered in place to complete screening. All inputs and outputs feed through the bottom pcb as can be seen below.

Band pass filtering for my initial on-air tests has been courtesy of the bpf set for the CDG2000 transceiver. Another of my on-going projects. Now that I am moving towards final integration of Star, I am laying out for manufacture a pcb for a 9 band bpf set for permanent inclusion in Star. For this, I plan to use Lodestone Pacific formers in an arrangement similar to that used very successfully in the CDG transceiver project.

I have currently configured Star to produce ~20W output on all 9 bands using the G6ALU design FET p.a. This is a very solid design, which has shown itself to be very stable and very easy to reproduce at low cost. It can be seen below. The design uses a three stage topology with a single ended pre-driver and a pair of RD06HHF1 in push-pull, driving a push-pull pair of RD16HHF1 in the final stage. Overall power gain is >50dB.

The completed G6ALU P.A. with the final bit of screening about to be soldered in place.

Having tested Star on-air using my CDG2K front -end board as the bpf the time has come to make a bpf board for permanent integration into Star. No sense in reinventing the wheel. The CDG2K front-end is a proven high performance unit so I have replicated the filter set for use in Star. The vacant locations at the left are to provide for adding a switched attenuator should it prove desirable to do so. There are 9 filters, one for each band 160 - 10 metres physically switched using high quality signal relays. The two additional formers at the top left of the picture form a notch filter centred on my first IF at 8830kHz. All formers are from Lodestone Pacific. Both wide and narrow sweeps for each filter can be viewed here.

As of 22 April 2006 Star is housed in a case. This was constructed using double-sided FR4 pcb material of 1.6mm thickness. Internal cross members of pcb material make the completed assembly very strong and provide the required rf screening.

Star's user interface (UI) is dramatically different to the 20 knobs and a hundred switches approach seen on commercial tranceivers. Star has only two controls on its front panel, a "What?" and a "How much?". These take the form of a 4 x 3 keypad and a rotary optical encoder. Not yet added, Star has an overlay for the keypad which illustrates the functions accesible with the pad.

The rear panel provides the usual connectivity options and houses the G6ALU P.A. heatsink.

The picture above shows the top-side of Star. Lower left is the screened compartment housing the Pic 'n Mix synthesiser, with the keypad assembly to its right. The keypad dodn't need to be screened from the rest of the Pic 'n MIx assembly but a cross member was added for physical strength. Above the Pic 'n Mix and to the left is a compartment containing the TDA 2004 audio amplifier and the seperately screened H-mode first mixer and quad J310 amp, the Magic Roundabout. The bandpass filter assembly is clearly visible in the central compartment with the DSP/IF assembly to its right. Above the bpf and to its left is the G6ALU P.A. and to the right is the T/R sequencer and the diode T/R switch.

The above photograph shows the "traditionally less beautiful" underside of the transceiver. Lower right the under screen of the Pic 'n Mix with the keypad to its left. The low pass filters can be clearly seen in the central area and are located directly underneath the band pass filter set above. To the left of the lpf is the IF assembly. The parts of the assembly housing the 8830kHz 1st IF are seperately screened, as is the 8845kHz conversion oscillator and the Mix Adapter H-mode 2nd mixer.

The desk, which in February housed Star as an array of boards in rats nest formation, is in May home to a much more complete Star. There is still work to be done but Star is now looking rather more like the fully functional 9 band HF transceiver it is.

Star incorporates an Analog Devices 9850 Direct Digital Synthesis chip as its means of frequency determination. This 9850 is incorporated in a Pic 'n Mix synthesiser, designed by Peter Rhodes prior to development of Star itself. Avoiding the use of a PLL, the resulting synthesiser has very good phase noise characteristics and is very effective in use. However, a minor concern has been the existence of a number of low level but nonetheless audible spurs on 10m. These are most noticeable at the bottom of the sunspot cycle, when for the most part, the band is completely bereft of signals.

The march of technology is as relentless as ever and since development of Pic 'n Mix was completed, Analog Devices has introduced a more powerful 9951 DDS chip. Whereas the 9850 incorporated a 10 bit DAC (digital to analogue converter) the more up-to-the-minute 9951 incorporates a DAC with14 bits of resolution. This development leads to a rather dramatic 25dB reduction in the amplitude of unwanted spurs.

In October 2006 I set out to follow Peter Rhodes example (and detailed instructions) for the replacement of the 9850 with a 9951. At the same time, I adjusted the Butler oscillator to utilise the 7th harmonic of Pic 'n Mix 25 MHz clock crystal, as opposed to the 5th, used with the 9850. I also took the opportunity to oven the Butler, incorporating a very effective proportional control heater, the work of Harry, G3NHR. This improves long term frequency stability and substantially eliminates the undesirable effect of drafts.

 

G3XJP carefully crafted the design of a 9951 assembly, as a direct plug in replacement for the old 9850. The photograph above is of the completed 9951 assembly ready to be installed for testing in Pic 'n MIx. The size of the photograph is misleading. It should be kept in mind that the 9951 chip at the approximated centre of the assembly is 7mm square with 48 legs each 1mm long. The overall dimensions of the 9951 PCB are 35 x 29 mm. Such is the quality and accuracy of Peter's assembly instructions, that assembly was entirely straight forward and on installation, it worked first time.