Last updated on November 08th 2008.
From my teenager years I was interested into HAM radio, digital modes in particular. I was active in building packet radio network in OK. Over university years my interest in HAM radio ceased and was sparkled again by QRP movement and homebrewing. After I built two single band CW QRP transceivers (DOB80 and SW40), my dream was to build a digital modes able multiband QRP transceiver. I was able to make my dream come true with Steve Weber's ATS-3a. The transceiver was donated by Ron Pfeiffer N1ZSW for my digital modes experiments. By extending firmware the transceiver is now able to work various PSK and MFSK modes. My firmware extensions were accepted by Steve into ATS-3b  .
The original ATS-3a has two weaknesses I wanted to tackle. First its crystal ladder IF and audio filters are very narrow for a lot of digital modes. Second, the transceiver is not able to modulate a clean PSK31 amplitude envelope. An approximation of PSK31 is generated using CW carrier shaping. Excessive side bands are created, which have low amplitude, espetially at QRP levels, but I still believe the amplitude envelope shaping is worth of improvements.
On October 2007 I received a prototype of ATS-3b from Steve Weber KD1JV, which I used as a test bed for my experiments. I pondered about putting it into a bigger box with a battery pack. Using bigger box would give more space for experiments. At the end I have to agree with Mike W1MT, that "the transceiver wants to live in the Altoids tin". Having battery outside the box allows experimenting with various battery pack formats and with new cell chemistries.
On the top side of the ATS-3b PCB there is plenty of ground plane to be used for Manhattan construction. I am cutting islands from headers of computer memory module boards and PCI card headers. I file the islands to about 1/3 of the original thickness. Then I spread legs of SMD ICs flat and solder ICs between two filed header boards. ICs soldered between two headers are superglued on top side of the ATS-3b board and islands are interconnected using magnet wire. The technique is suprisingly effective and enables highly integrated homebrewing, see figures 1-4. One needs to be careful not to scratch magnet wire with sharp tools.
Passive parts are mostly scrounged from waste electronics. Homebrewing with SMD requires quite small junk box. I found out by trial and error that it is counterproductive to desolder SMD parts before they are needed. It is much easier to browse through PCB boards than search for a part in a heap of parts already desoldered. To desolder SMD part I am actually adding solder to all pads first. With a blob of solder added I am able to heat both ends of passive component or three leads of transistor at once. Excessive solder is removed from the part after the part is desoldered.
Original ATS-3b prototype from Steve Weber KD1JV was extended by following modules:
Display is no more connected to the MSP430 controller directly. The display is driven by a shift register 74HC164 instead. From 8 controller pins used to drive 7 segment display + dot, 7 were released for other purposes. The 8th controller pin is used for clocking shift register. Shift register data are multiplexed with DDS data.
Rotary encoder is glued by epoxy to the bottom side of the ATS-3b board next to the filter board header. The part I used is Mouser #858-EN11-VNB1AQ15, 20 pulse per revolution encoder with detent and 11mm shaft. After the detent spring is removed, which is quite easy to do, the software recognizes 80 positions per revolution. My firmware prototype tunes 1kHz per revolution when rotating slowly and maximum 10khz per revolution when rotating fast. The tuning speed changes exponentially based on angular velocity between the two boundary values.
Variable bandwidth IF filter. I copied the design from Elecraft transceivers. Three epiplanar varactors (27 to 500pF) are used as shunt capacitors in IF crystal ladder filter. The varactors are sourced by Microchip MCP4725 D/A converter. To achieve high IF bandwidth span, one needs to utilize full span of battery voltage. Output of D/A converter is amplified from 3.5V span to full battery voltage span by an op amp Microchip MCP601, one general purpose NPN and one general purpose PNP transistor, very much in the way Steve KD1JV designed PA modulation in his Unique PSK31 Transceiver  . In contrast Elecraft KX1 is using analog pot and K1 D/A converter sourced from 6V. My design allows wider filter settings with higher battery voltages.
Earphone volume control. 10k digital pot Microchip MCP4011 is connected in feedback of earphone power amplifier. Earphone amplifier is no more wired as a Sallen-Key audio band pass filter, but as a straight amplifier.
CW/SSB audio filter using Microchip MSP602 op amp as 2.5kHz low pass and 600Hz Sallen-Key band pass filter. 600Hz CW filter may be baypassed by audio switch 74LVC1G3157 if listening to SSB. Polystyrene film capacitors were used in CW filter for pitch precision and temperature stability.
Electronic power on/off using one NDT2955, one 2n7000 and two general purpose silicon diodes. All circuits but power amplifier is switched by the
power FET. I did not want to increase voltage drop on power amplifier by chaining the power on/off FET with the CW shaping FET. The power FET is switched on by first button (was tune up before) and by controller through 2n7000. If one keeps first button pressed for 2 seconds, controller powers down itself by switching off the power FET. Auto power off after 30 minutes of inactivity and on low battery will be implemented in firmware. If off, the transceiver uses about 0.1mA current, which I account mainly for the parasitic resistance of electrolytic capacitor in PA. Replacing it with tantalus one will probably decrease idle current by an order of magnitude.
BFO was pulled below IF frequency by replacing cap trimmer with 22uH inductor. Originally the BFO is pulled above IF frequency, so the RX receives LSB naturally. There is a trick to flip sideband. One needs to tune VFO by IF frequency above RX frequency. Unfortunately this trick is not applicable at higher bands where the DDS is alredy pushed to its frequency limits. The inductor was wound with 20 turns on FT37-61 toroid. Firmware needs to be adapted accordingly to work with the opposite RX BFO. This trick was learned from Elecraft KX1.
Through hole digital input/output jack socket was fitted next to the earphone socket.
Digital input slicer is wired using Microchip MCP602 op amp, see figure 5. Its output is connected to a free controller pin and digital interface is active all the time.
Volume controlled output to PDA microphone. Sound before CW audio filter is buffered by MCP601 op amp and regulated by another MCP4011 digital 10k pot in the feedback of the buffer op amp. This is probably overkill, but it is convenient to be able to adjust volume when one changes computers or PDAs. There is a need for some kind of volume adjustment, because PDAs usually do not have microphone volume control integrated.
Receiver attenuator. Simple 330 Ohm pot shall do. Hopefully it will fit into the Altoids tin next to the rotary encoder.
PSK31 AM modulator. I plan to duplicate very much the same circuit I used for IF bandwidth control (Microchip MCP4725 D/A converter + two transistors). There is still enough ground plane left to fit the circuit.
RF peak detector to estimate transmit power. Exact power will only be detected if the antenna has 50 Ohm real only impedance. Ideally one would fit an SWR bridge, but there will most probably not be enough space in the tin. The power would be announciated during antenna Tune mode.
Firmware changes are needed to support the newly added hardware. I implemented only rudimentary support to test the new hardware, but it needs a lot of effort yet to be really smooth.
Place PA FETs on top side of the board, flat side up. Add two more stand offs across the PA FETs and fit a sheet of aluminium over the top side of the board as a heat sink.
In the ATS-3, 3a and 3b devices one half of the earphone PA stereo chip is used as audio pre-amp just after detector. The difference between the input amplitude of audio pre-amp and earphone PA may be as high as 60dB, depending on the AGC action. There seems to be some feedback between the op amp stages, which influences frequency response of pre-amp based on earphone amp load.
Please forgive me terrible quality of published photographs. My digital camera cannot make good close ups and the pictures were actually taken with a video camera.
Figure 1. ATS-3b board, top side
Figure 2. ATS-3b board, top side, annotated
Figure 3. ATS-3b board, bottom side
Figure 4. ATS-3b board, bottom side, annotated
Figure 5. Digital interface schematic