Given the unmitigated joy I've had from building and operating my Das DereLicht transmitter, it seemed only natural to extend the project with a matching receiver. Doubtless, the 80m direct-conversion receiver described on this page is only one of many possible solutions to the puzzle.
As of January 2, 2009, I've made 24 QSOs using the new, Das Derelicht transceiver. I'd like to thank you all for your kind support and shared enthusiasm for this homebrew QRP "puzzle."
AF Amplifier
I noticed the transistors that I salvaged from my second (otherwise, identical) CFL were labeled as BUL128A, instead of the BUL128B devices that I found inside my first CFL. Reading the datasheet, I discovered these transistors are graded by the manufacturer according to their measured DC current gain. "Group B" devices have an hFE in the range of 25 to 40, while "Group A" devices have an hFE in range of 14 to 28. In contrast, the hFE for the ubiquitous 2N3904 transistor ranges from 70 to 300, depending on the collector current. In fact, a low DC current gain is generally part of the trade-off required to produce high-voltage transistors, such as these BUL128s. Given these devices were never intended to be used as small-signal amplifiers, is it impossible to employ a pair of them in order to build a useful audio amplifier for our receiver?
Not at all! In fact, plenty of workable audio amplifiers existed in the early days of transistors; built from devices having specifications quite inferior to our BUL128A's. The 2N107, for example, had an hFE of ~30, whilst the well-known CK722 transistor could only manage a value of 12! The two transistor amplifier in this receiver has an overall voltage gain (Q1 base to headphones) of 58dB at 800Hz. The impedance looking into the base of Q1 was both calculated and measured at just over 1000 Ohms. These are respectable specifications for a two-transistor audio amplifier.
Ah, but just look at L2 in the schematic; seven hundred turns of #39 wire on an FT37-72 core..."Is this guy nuts?" I'm afraid so. In fact, I don't expect anyone to repeat this bit of insanity! It is perfectly acceptable to replace L2, C9 and C10 with a (wideband) audio frequency transformer having roughly a 22:1 impedance step-down ratio. In fact, I used a small AC power transformer (11H primary) from my junk box in just this way throughout most of my first week on the air with this receiver.
Only, this transformer had a way of getting in my face :o) And so I began thinking about the the little FT37-72 core that came with this CFL "kit." It turns out that I'd need a primary winding of approximately 3500 turns on this core in order to produce a usable wideband output transformer; a seemingly impossible task! However, if I were only interested in a narrow bandwidth impedance-match (which I am), the number of turns may be reduced proportionally. I settled on a loaded network Q of 5. I calculated that it ought to be possible to lay-on the required 700 turns if I used a "hair-like" #39 enameled wire that I'd previously salvaged from a small electromagnetic relay coil.
I fashioned a tiny bobbin from a bit of the tough, clear plastic material that so many thing come packaged in these days. Then I donned my Optivisor, put on a Telemann trumpet concerto and set to work. Some hours later (seriously!) I measured the inductance of my winding at 453mH (XL ~2.2kOhms @ 800Hz, Rp ~ 5*2.2kOhms = 11kOhms). A capacitor "divider" resonates this inductance at 800Hz whilst providing the required impedance step-down ratio. The audio selectivity gained is a nice bonus. Again, it's crazy, although I must admit that I'm happy to have been able to push the puzzle that much further. :o)
Product Detector
This was a serendipitous discovery. I had been using a pair of (necessarily, unmatched) 1N4937 diodes taken from my CFL, in a standard, single-balanced mixer; or, rather, a quasi-single-balanced mixer. It suffered from severe SWBCI and hum until I included a small balancing potentiometer at the common node of the two diodes. While the null produced was impressive, it did not entirely remove the offending signals.
In the course of my bedtime reading one evening, I happened to open an analog circuit design textbook to the chapter on "mixers." There, I noticed a pair of diode commutating mixers that I'd not seen before. The caption beneath them explained they were single-balanced mixers; each built of four diodes. One was configured as a transmission gate and the other as a shunt switch. As I traced the operation of these simple mixers it occurred to me they ought to be fairly forgiving of mismatched diodes.
The next day's work on the bench convinced me this really is the case. Luckily, my CFL ballasts were chock-full of diodes, and so it was no problem to replace my two-diode single-balanced mixer with a four-diode version. I chose the shunt-switch configuration for no particular reason. One down-side of this circuit is that only half of the incoming RF signal is utilized. I feel this is a small price to pay in my simple receiver circuit. The hum has all but vanished, and in over a week's use of this mixer on the air I've yet to hear as much as a peep of SWBCI.
This mixer is dead simple. When the BFO signal at the left-hand-side of the T3 secondary goes positive, all four of the diodes turn-on; shorting both the BFO and RF input signals to ground. Half of a BFO cycle later all of the diodes turn-off; isolating the BFO from the remainder of the circuit and allowing the input signal to pass unmolested to the LPF at L1-C5.