On one hand, it probably doesn’t make too much sense to try to refine the MMM (the construction of which was described in an earlier post): it’s more an oscillator demo than a building block of any more complex radio, but there are a couple variables that I thought would be fun to explore: choice of transistor, supply voltage and emitter resistor value. Tables and pretty x-y graphs follow.
The design calls for pretty much any off the shelf NPN transistor, and there are reports online of folks succeeding with common parts such as 2n2222 and 2n3904; however, other builders have used “beefier” transistors with higher fT to try to get more output. I haven’t seen a formal comparison, so I went to the junque box and tried what I had on hand.
The circuit was built as previously described including a LPF on the output, with a fixed capacitance of 151 pF instead of a variable capacitor and an emitter resistor of 27 ohm. My power supply was set to 13.8V, but voltage at the transistor collector would be lower by the about 0.7V of the protection diode and whatever losses I have between the supply and the board. Let’s say about 13V then at the collector.
Choice of Transistor
To allow swapping in of transistors, I used a SIP socket (well, a DIP socket cut in half). The leads on some of the transistors did not line up in the right order, so I also made an adapter SIP to swap two pins around. In any event, lead lengths were under a centimeter for both adapters and on the transistors. Some of the transistors were bought, some scrounged. I measured three replicas for each type of transistor, and where possible, picked the from different sources (e.g., bought vs scrounge) or lots of bought. Results are expressed as Vrms measured at a 50 ohm dummy load plus/minus standard error of the mean. For convenience, Vrms is also converted to mW.
| Transistor | Vrms | SEM | mW |
|---|---|---|---|
| 2n2222a | 3.06 | 0.04 | 188 |
| 2n4401ta | 3.10 | 0.01 | 193 |
| 2n3904 | 3.48 | 0.05 | 243 |
| s8050 | 3.06 | 0.02 | 187 |
| s9014 | 2.74 | 0.02 | 151 |
| 2n2219a | 3.06 | 0.01 | 188 |
| 2n3866 | 3.71 | 0.03 | 276 |
| h945 | 2.19 | 0.03 | 169 |
| c945 | 3.00 | 0.01 | 180 |
| mpsh10 | 2.35 | 0.01 | 111 |
| 2sc5706 | 2.80 | 0.01 | 157 |
| 2n5109 | 3.94 | 0.02 | 310 |
| nte 108-1 | – | – | – |
(the NTE108-1 did not oscillate for me in this set-up).
Supply Voltage
What about supply voltage? One would expect that more juice would yield more RF, right? So, with the same circuit as above, just varying the bench supply voltage, here are the Vrms measurements.
| Supply Voltage | 2n3904 | 2n3866 |
|---|---|---|
| 10 | 2.55 | 2.66 |
| 10.5 | 2.69 | 2.81 |
| 11 | 2.82 | 2.96 |
| 11.5 | 2.95 | 3.12 |
| 12 | 3.1 | 3.28 |
| 12.5 | 3.21 | 3.42 |
| 13 | 3.34 | 3.57 |
| 13.5 | 3.49 | 3.71 |
| 14 | 3.63 | 3.86 |
| 14.5 | 3.76 | 4.00 |
| 15 | 3.89 | 4.14 |
With short key downs (up to 5 sec for measurements), I did not notice any heating of these transistors.
Emitter Resistor
It also seemed likely that at a given supply voltage, I could get more output by lowering the value of the emitter resistor, with some expectation of increased instability and/or toasting of the transistor. This time, I set the voltage to 13.75 at the input to the board, and measured 13.05V after the dropping diode. The rest of the circuit was as above. Vrms was measured for each transistor into a 50 ohm dummy load.
| R (ohms) | 2n3904 | 2n3866 | 2n5109 |
|---|---|---|---|
| 33 | 3.48 | 3.57 | 3.74 |
| 27 | 3.67 | 3.79 | 3.97 |
| 22 | 3.82 | 3.96 | 4.17 |
| 18 | 3.96 | 4.11 | 4.36 |
| 15 | 4.07 | 4.24 | 4.51 |
| 10 | 4.26 | 4.33 | 4.76 |
| 7.5 | 4.34 | 4.40 | 4.89 |
| 6 | 4.41 | 4.46 | 4.98 |
| 5 | 4.45 | 4.54 | 5.04 |
| 2.7 | 4.55 | – | 5.15 |
| 2 | 4.63 | 4.58 | 5.23 |
| 1 | 4.63 | 4.59 | 5.26 |
| 0.33 | 4.63 | – | 5.31 |
The output from the 2n3866 tested remained sinusoidal, but the amplitude began to vary a bit below 5 ohms. I tested a couple 2N5109, and one had no problem going down to 0.33 ohms, but another from the same bag began to distort at 15 ohms and yielded noise below that value.
With 15 seconds of keydown, the 2n3904 increased temperature 3 celsius degrees at 33 ohms and 8 at 0.33 ohms. For similar keydown, 2n3866 increased 5C, and 2n5109 increased 2C.
Conclusions
The methodology wasn’t all that rigorous, but I’d hazard a few conclusions:
1. The joy of oscillation can be experienced with most of the NPN transistors in the junque box using the circuit shown in the schematic.
2. With that circuit and powered at 13.8V, the power output didn’t vary much between transistors of a given type, even when they came from different sources.
3. With that circuit, there was a linear relationship between supply voltage and power output, within the range tested, for these two transistors.
4. The power output can also be upped by decreasing the value of the resistor between the emitter and ground. Doing so in some cases results in instability. Power dissipation in the transistor is increased (as reflected in increased case temperature); more so for the 2n3904 and 2n3866 than the 2n5109. Looking closely at the graph, at some point lower resistor values yield diminishing returns.
Given equipment on hand, it looked to me like lowering the value of the resistor significantly did not distort the output wave form (or at least, what I could see beyond the LPF).
5. Some of the power output values reported on the web for this design should probably be taken with a grain of salt. Often this is presented as a half-watt transmitter. That may be true of the oscillator before the LPF, but in terms of on-frequency output, most of the dime-a-dozen transistors in the junque box will yield an optimistic quarter watt with the basic design and typical 13.8V power supply.
Between picking an efficient transistor, increasing supply voltage and decreasing the emitter resistor value, the power output can be increased usefully. With the 2n5109, 15 V bench supply, and using a 5 ohm emitter resistor, I measured almost 700 mW power output. On the other hand, we’re still a long way from QRO.
There’s a limit to how far this can be taken. When I got above 16V on the supply, I couldn’t get oscillation to kick off again. I noticed that increased voltage pushed the resonant frequency up a bit, so perhaps the circuit could go a bit further were the tuning tweaked. If I raise the emitter resistance, I can crank the supply voltage higher, but it is a trade off in terms of affecting output.
What passes for an 80m antenna at my house is not even close to resonant and far from efficient. Given that and the lack of any reverse beacon station on 80m within thousands of miles of me, I think that’s the end of the road for the MMM, at least for now.