2008-03-01
One Transistor Voice Transceiver
The Inspiration
Michael Rainey's recent adventures with minimalist transmitters and receivers has been an active topic of discussion on QRP-L. Michael's One Transistor Direct Conversion RX inspired me to finally attempt a project I've been toying with for years; a one transistor voice transceiver. John Kirk VK4TJ also exchanged some emails with me about single transistor TRXs, in particular John reminded me about the QSK-1, Bob Culter N7FKI's single-2N7000 based CW transceiver.
A one transistor CW TRX isn't that difficult to achieve, but I wanted a phone rig. I was under no illusion that this would be extremely difficult to achieve, openly doubting if it was even possible. I agreed to be happy if I managed to achieve any kind of "toy" walkie-talkie style performance.
The General Concept
I decided to go with minimalist WBFM transmitter and a super-regenerative detector. The topology of both circuits is extremely similar, and I had previously got a single transistor FM broadcast receiver working fairly well. It is pretty easy to turn a self-quenched super-regenerative detector into an oscillator by just altering source/emitter circuit to prevent quenching. This is the approach I took.
I picked the FM broadcast band as a candidate frequency band. It is not excessively high of a frequency, WBFM is already common there, and locally there is an empty gap around 100 MHz where Low Interference Potential Devices are allowed to operate (wireless microphones, toys, etc). My spectrum analyser also works well on 3 metres, and includes WBFM demodulation. There is no reason in principle why the resulting device could not be rebuilt for 10 metres, which is very sparsely populated, or even maybe 70 cm. 2 metres is too busy now days for a free-running LC oscillator to be unleashed, even at mW levels. Wideband emissions on 70 cm are more common, with ATV operations there, but frequency stability would become a big problem. Antennas for HTs are much easier at UHF than HF, which might be considered an advantage of using 70 cm.
A Starting Point
I started with a grounded-gate JFET oscillator. (Using a JFET simplifies the biasing components, while I wasn't trying to build a minimum-component receiver, simplicity is always preferable.) A SPDT switch in the source provided the switching between straight oscillation and quenched super-regeneration. For RX, a crystal ear piece was used as both the audio output device and the quenching capacitance (about 14 nF). The resistor paralleling (4k7) it was chosen to give a quench of about 25 kHz which corresponded to the best sensitivity. (Slightly more AF output was available when the quench was dropped to around 18 kHz, but the passive filtering required to remove the quench tone killed any extra AF output, so I kept the quench supersonic.) For TX just a single resistor was switched in resulting in CW oscillation. A trimmer (not show below) in the drain circuit provided tuning from about 70-180 MHz.
At this point, the RX was working fine. The audio output level is quite low, requiring a quiet room, but the receiver is quite sensitive, picking up even the weaker FM broadcast stations with no antenna at all. (The 8 component RX is in itself quite impressive in its minimalism!)
Coupling to the source for an antenna proved to be practical for both TX and RX but did cause some frequency instability with hand capacitance. Coupling to the cold-side of the tank offered no better stability. The antenna was left off for further experiments, it was just getting in the way, and was something that could be left out for now.
The frequency shift between TX and RX turned out to be only about 50 kHz for my ugly birds-nest of a prototype. This was completely unexpected. Previous experiments had delivered circuits with more than 800 kHz of shift between CW and quenched oscillation. The 50 kHz shift is pretty acceptable with WBFM, considering the selectivity of the RX.
Getting TX Modulation
To FM modulate the TX carrier I injected 100 mV of audio from my signal generator into the junction between the source resistor and the RFC. This resulted in an acceptable deviation, similar in "loudness" to commercial FM stations. However, here I had a problem. 100 mV of audio into a low impedance point would require some kind of audio amplifier, no passive microphone could deliver this.
For the purposes of further experimentation I built a two transistor AF amplifier which could amplify a dynamic or electret mic signal up to the required level and drive the low impedance of the source. Note that this particular circuit is capable of providing RX audio amplification as well, and can drive 32 Ohm headphones to a reasonable volume.
In an attempt to avoid the need for the AF amplifier I lifted the J310's gate lead and took it back to ground with a 1M resistor, paralleled with 1 nF of RF bypassing. This gave me a higher impedance point into which I could inject much smaller AF signals and get the JFET to amplify the signal itself. The result was fairly good deviation with just a dynamic Mic and a DC blocking capacitor connected to the gate - at least as long as you yell into the microphone!
A down side was this change to the circuit caused the RX/TX frequency shift to increase to almost 500 kHz. I had always planned to include a switched trimmer to correct the RX/TX shift, so while annoying, this wasn't a fatal flaw. The RX frequency was higher, so a DPDT switch could be used to switch in extra capacitance on RX. This gives you a RIT feature too which is very helpful in minimising RX audio distortion.
The low deviation was a problem though, it was no where near as good as with the source injection with the AF amplifier, and combined with the fairly poor AF output of the RX this configuration is just barely usable (unit-to-unit anyway)...
After much fiddling around I decided to cheat and use a electret microphone instead of a dynamic one. A electret mic has an internal FET and can't be classified as "passive" (but more importantly for the circuit, its output is significantly greater). Along with adding a AF bypass capacitor at the source resistor, this provided acceptable deviation with no additional stage of AF gain and only a moderately raised voice.
A carbon microphone might be a way to avoid the extra "active" device. I also contemplated directly pulling the tank frequency with either the capacitance of a microphone diaphragm in close proximity, or a piece of ferrite glued on to a diaphragm which might be rubber or acetate sheet stretched across a PETE plastic bottle top as an acoustic "horn". The mechanics of this were thought too fiddly for the proof of concept unit, the purists might like to explore this route however.
Success
So at this point, I have a working toy VHF transceiver using only a single J310 transistor and just a handful of other components. The miniature switch and earphone socket dwarf most of the other components, and the current drain is next to nothing.
I've contemplated building a pair of these circuits into Eclipse mint tins.
I do really need to build at least two, so I can have a proper QSO unit-to-unit.
Improvements
The proof-of-concept circuit has a lot of limitations, the biggest is no RIT or fixed TX/RX frequency shift compensation. When I tightened-up the initial messy layout, the circuit actually ended up having a much smaller T/R frequency offset. Almost acceptable considering the selectivity of the RX, but wrong enough to limit its performance significantly.
At its most primitive using a DPDT T/R switch and a small compensating trimmer would allow correcting this. I am tempted however to use varicap tuning for the RIT. Some TX shift would be handy for "netting" three or more units, so perhaps a pair of pots; tune + RIT.
The microphone isn't cut off on RX, you can hear it in the RX audio. Removing the microphone supply or shorting out the transistor gate would fix this. More T/R switch poles or some diode switching associated with the RIT implementation sounds the most feasible.
The TX frequency takes a moment to settle while the AF bypass capacitor on the source charges. This isn't a big deal, but you can watch the oscillator sweep up the band on the SA as it charges when you first key the transmitter. Reducing the value of the capacitor helps, but reduces the low-frequency deviation. Some adjustment of TX frequency is available by varying the source resistor, but this also changes the output power and can only be done within certain limits (before it starts super-regenerating at a fraction of a Hertz or the transistor saturates). It is possible to tune out the T/R frequency difference by this method.
Obviously coupling the tank to an antenna needs to happen in the final unit too.
More Devices?
There is a obvious trade-off here in performance vrs device count. For one active device you get a quiet RX with moderate TX deviation, and an output power in the milliwatt region: Usable, but only really to say you've done it!
Adding another transistor gives you either better RX AF gain, or more TX deviation. Two extra transistors lets you have quite good TX and RX performance, but you are still tied to the earphone for RX. Extra switching lets you drive a small speaker which will also act as the microphone, but a 4PDT TR switch is required and circuit complexity is rising fast. An dual Op-Amp IC gives you similar options for more current draw.
If you don't mind extra transistors you can just build yourself separate Microphone and Speaker amplifiers. Three or four extra transistors gives you a quite capable little rig. You can even just build the TX and RX separately and switch the power and antenna, at that point though, you have a transceiver not unlike the Fredbox.
The AM signal of the Fredbox is more compatible with the super-regenerative detector too, in the sense that tuning is not as critical for undistorted output. It would be possible to AM the oscillator, indeed there is some AM already along with the FM in the current circuit, but a modulated buffer would be a better idea to minimise FM. Crystal locking the TX frequency would be an option for an AM rig.
RF buffering on RX would reduce LO radiation. On TX you could develop more power with a stage or two of amplification and both would help prevent the antenna loading pulling the frequency around while in use.
So even if limited to say, 6 transistors total you could make a pretty good transceiver with a tiny footprint and power consumption. Of course, it would have none of the wow factor of using a transceiver with only one transistor!
8 comments.