Original Prototype
Current Design
My original One-Tube Transmitter was built using a 6M11 tube which I just happened to have on hand. Although it works very well as is, I decided that the design could be improved and simplified, by using a tube having characteristics specifically suited for the various functions. The first change was to eliminate the separate oscillator section. With frequency being controlled by a crystal or a ceramic resonator, a modulated oscillator should have more than adequate frequency stability and freedom from FM distortion. With this decision made, the possible candidate tubes increased considerably. Dual section tubes are much more common than triple section tubes.
I also decided that this new transmitter would use carrier control, as did the previous design. This is because of an issue which crops up when using simple transmitters. That is the need for some sort of compressor/limiter audio pre-processing to prevent over-modulation and under-modulation. If the audio source is from audio files stored on a computer, then it's possible to use audio software to perform the compression and limiting function. But, if the audio source is some other device, then one is left with the choice of either:
Option a. costs nothing, but tends to be unsatisfactory if the program sources vary widely in dynamic range, as is usually the case. Option b. involves additional circuitry which offsets the simplicity of the original transmitter design.
The carrier control option avoids the problem by adjusting carrier level automatically as the program source level changes, maintaining a high modulation level at all times, and thereby compensating for changing levels. As was mentioned in the original transmitter article, the carrier control is implemented by biasing the input triode grid to ground potential, and using the resulting grid clamping effect. A bit of research revealed that this works best with a high transconductance triode. On that basis, I chose a 6CQ8 dual section triode/tetrode for the transmitter. The triode transconductance is a suitably high 8000 µMhos. Another requirement for carrier control is the direct coupling from the triode plate to the modulator tube screen. Hence, screen modulation is used rather than plate or suppressor modulation. Tetrodes exhibit better linearity than pentodes when screen modulated. Therefore, the second section being a tetrode made the 6CQ8 an ideal choice.
Here is the schematic:
The original prototype was constructed on a sub-chassis salvaged from an old fax machine. The underside of the chassis is shown below:
The tetrode section functions as a modulated crystal oscillator (Colpitts with cathode feedback), using the screen grid for modulation. Both a quartz crystal and a ceramic resonator have been used with good results. The value of C1 is 22pF when a quartz crystal is used, and 150pF when a ceramic resonator is used. Also, when using a ceramic resonator, it may be necessary to replace 100pF capacitor C3 with a 10-100pF trimmer to fine tune the frequency.
Coupling from the tetrode plate to the antenna, in the original prototype, was by means of a pi-network which was adjusted for maximum power transfer. Pi-network component values depend on the transmitter frequency, and antenna length and configuration. The inductor had taps ranging from about 300 to 900 micro-henries in six steps, which covered the medium wave band quite well when used with a 3 meter long antenna. When I created the original prototype, I was still very inexperienced with the science of antenna matching, and the highly adjustable pi-network allowed for a lot of adjustment. However, it was not the most efficient for transferring power to the antenna. When I designed the printed circuit board, I replaced the pi-network with a parallel LC tank in the tube plate circuit, this is simpler and more efficient. The parallel LC matching is shown in the current schematics. For more information on both the pi-network and the parallel LC matching, see the Antenna Matching Page.
The triode section of the 6CQ8 functions as the audio pre-amp and carrier control. It uses a fixed grid bias near zero volts (with no audio input signal). There is no cathode bias. The significant point is that the combination of the grid resistor, the input DC blocking capacitor, and the diode action between grid and cathode, function as a diode clamp which clamps the positive peaks of the audio input signal to approximately ground potential. To be precise, due to grid-cathode contact potential, quiescent grid bias is about -0.5 volts, positive grid current starts to flow at about this same voltage, and so this is where the positive input peaks are clamped. As a result, the grid bias is always equal to minus half the peak to peak audio input level minus a half volt. It's important to point out, that with this circuit, the term "quiescent" does not mean "average DC level", as it often does in AC coupled circuits, where the meanings are often interchangeable; it refers to the state with no audio input signal. With varying input signal levels, the average DC levels will shift. Consequently, the signal at the plate will have its negative peaks near the quiescent DC plate voltage, and the positive peaks will vary with the audio signal. The average plate DC level will be proportional to the average audio level. This average DC level on the plate is the desired carrier control. By direct coupling the triode plate to the screen grid of the tetrode, the combined audio and DC carrier control will modulate the tetrode oscillator section of the tube. The predecessor to this circuit, using the 6M11, required a DC level shift which was accomplished quite simply by placing an NE-2 neon lamp between the triode plate and pentode screen. However, no DC level shift is required in the present design, further simplifying the circuit.
The quiescent point of the triode plate and tetrode screen is chosen to be a relatively low voltage which still allows the tetrode oscillator section to generate a modest carrier. As with the earlier circuit, a variable resistance was initially provided in the plate circuit to set this level. However after a considerable amount of testing it was found that these could be replaced with a fixed 820 kΩ resistor. All deviations from the quiescent point will be in the positive direction, and accordingly, these deviations will increase the RF signal output from the tetrode.
A printed circuit board has now been designed for this transmitter. See the bottom of this page for the PCB design files.
I had the component silkscreen applied to both sides of the board to make it easier to mount parts on either side. I did this in order to allow for the tube to protrude through the top of the enclosure without the other parts getting in the way.
In the photo you can see that I used a small transistor radio IF or LO coil for the plate inductor. It happened to have the correct inductance adjustment range needed to match my antenna. Because of the pin spacing, I had to make a small daughter board using a small chunk of veroboard. Although these coils seem as if they’re too small to handle the RF voltage that may appear across them, I haven’t experienced any failures yet.
One must naturally wonder what a controlled carrier signal sounds like at the receiver end. Without any AGC circuit in the receiver, the effect would be much like an audio expander circuit, which is often used in audio systems to provide a "presence" effect. However, the AGC circuit in the receiver compensates by re-compressing the audio, reducing (but not eliminating) this effect. To eliminate the effect completely, the AGC loop gain would have to be infinite, which of course, it is not. Therefore, there remains a net expansion effect, which in my listening tests, I found to be a pleasant departure from the usual heavily compressed audio of commercial broadcast stations. I have to say that this is all subjective, of course.
It is important though, that the time constant of the carrier control, be compatible with the time constant of the receiver AGC. The carrier control time constant is set by the triode grid resistor and the input capacitor. In the present circuit the time constant is 100 milliseconds, which I found to be optimum in my tests, and compatible with the AGC time constant in all receivers with which I tested the transmitter. In fact, there is very little variation in the AGC time constant between different receivers, because the optimum value is a trade off between two constraints. It must be fast enough to follow signal fading, and slow enough not to be affected by the audio modulation. These constraints tend to put receiver AGC time constants in the 100-150 millisecond range.
Another point to consider is the possibility of introducing some distortion due to the triode clamp circuit. It would be expected that positive peaks of the audio signal would cause a considerable increase in grid current and therefore cause a drastic decrease in input impedance at these peaks, leading to a flattening of the top of the input signal waveform. This has proven not to be a problem. There is no visible flattening of the input signal in the scope traces. Very fast transients would be the most probable cause of distortion. However, no distortion was audible to me during listening tests. An audio expert, with better ears than mine, might argue otherwise, but I'm quite happy with the audio quality.
One additional characteristic that this transmitter exhibits, is inherent soft limiting at both the 0% and 100% ends of the modulation envelope, which makes it difficult to overdrive it to the point of audibly unpleasant distortion. This soft limiting is typical of screen modulated transmitters, but it seems to work better with some tubes than others. It works well in conjunction with the carrier control to make set-up virtually foolproof.
Shown here is a trapezoidal trace of the transmitter modulation.
Appended to the right side is the amplitude of the unmodulated
carrier which is about 2.2 divisions on the scope. The maximum
amplitude is about 7 divisions, and the minimum amplitude is 0.2
divisions. From these numbers it's possible to calculate the upward
and downward modulation as follows:
MU=(7.0 - 2.2)/2.2 x 100 = 218%
MD= (2.2 - 0.2)/2.2 x 100 = 91%
The average of these is 154%
Following is a video showing trapezoidal scope traces for a short musical excerpt*. The audio is the actual audio received on a nearby receiver. The receiver has an IF bandwidth of 8 kHz which shows off the fidelity of the transmitted signal.
*The musical excerpt:
Album: African Guitar Summit; CBC Records, 2004.
Artist: Alpha Yaya Diallo
Title: Cette Vie
If you liked the excerpt, then please buy the CD.
Please refer to this page for updated information about:
Matching a
Part 15 transmitter to a short antenna br
It discusses the antenna pi matching network used in the original
prototype, as well as the simpler and more efficient antenna matching
network used in the final version.
Gerber
Files
Component
Layout and Construction Notes
Note that the Gerber files are for a double board, i.e., two transmitters on a single board, because the overall size falls within most fabricators’ 100mm x 100mm minimum charge size. As time permits, I’ll add the Gerbers for a single board.