Winner of the G2NJ Trophy 2022 !!!
The double-balanced cross-coupled product detector had a brief stint of popularity in the 70s and 80s. It’s popularity quickly faded once integrated product detectors like the Plessey SL640, Motorola MC1496/1596 and the CA3028A came on the market. Offering ease of use and further integration. My attention was first drawn to the cross-coupled product, when casually browsing some Technical Topics columns from Pat Hawker G3VA from the 80s. In his June/July 1980 article briefly mentions the product detector. Out of curiosity, I decided to build it and do some tests with it.
The nice thing about this whole circuit it can be build without any transformers! The only thing needed is a 1mH choke that be bought of the shelf for a couple of cents.
The circuit is copied directly from the Technical Topics. The only thing I’ve changed is the use of 2n2222a transistors for Q1 and Q2. The article mentions the use of BC108, BC182 and other old-school transistors that I fondly remember.
Operation
Details on the actual operation of the circuit are sparse. The only hint to it’s functioning was found in the Technical Topics column of June/July 1980 [1]. Here it was stated that “The bfo injection should provide a signal about 10 times that of the i.f. signal“. Based on this, the following operation can be deduced: if the BFO input goes positive, the voltage on the base of transistor Q2 will rise, as well as the voltage on the emitter of Q1. Thus transistor Q2 will conduct more. However, because the base of transistor Q1 is held a at a fixed voltage and the emitter of Q1 rises, Q1 will conduct less.
The RF signal is offered, at the base of Q1 and the emitter of Q2. For the RF signal, Q2 is configured as a common base amplifier and thus the input and output signal will be in phase. On the combined collectors of Q1 and Q2, the RF signal will show up in phase.
If the BFO signal goes negative, the opposite of what has just been described happens: now Q1 will conduct more and Q2 will conduct less. For the RF signal Q1 is configured as common emitter amplifier and thus the input and output signal will be reversed in phase. On the combined collectors the RF signal will show up in reversed polarity.
This whole scheme will only work if the BFO signal is dominant, e.q. the BFO signal determines if either Q1 or Q2 will increase of decrease conduction. Hence the requirement for the signal to be 10 times larger.
Build and test
The whole circuit is quickly build on a piece of copper clad board. You can easily see the metal can TO-18 2n2222a at the bottom. The 10nF caps are the four yellow blobs at the bottom. The 1mH RFC choke is the fat mint-green resistor like device in the middle. The transistors were not matched, just came from the same batch. For the biasing 1% metal film resistor were used, however.
After firing up the PCB, the static DC voltages were measured. It all pretty much agreed with the calculated values. Initially started with 9V DC but I could not get the circuit to work (later on, it was no problem, so I must have make a mistake somewhere). Then I decided to up the ante a little bit and go for 24V. This worked beautifully! You’ll find the results in the table below. The photo at the beginning show the scope output.
| BFO [ mVpp ] | AF [ mVpp ] |
| 1000 | 92 |
| 900 | 260 |
| 800 | 260 |
| 700 | 260 |
| 600 | 250 |
| 500 | 250 |
| 400 | 210 |
| 300 | 90 |
| 200 | 76 |
All test were done with fRF = 1 Mhz, a fBFO = 1.001 Mhz and VRF = 20mVpp. The optimum value for the BFO input voltage is thus somewhere between 700-900 mVpp, featuring a voltage gain of 13x = 22 dB. Any lower and it will not sufficiently put Q1 and Q2 in and out of conduction. That at even higher voltages the gain would decrease was a surprise. However, this is probably because Q1 and Q2 are being pushed into saturation and thus takes a longer time to recover and thus makes the conversion operation less efficient.
The bandwidth of the output filter (C4, L1, C3) was measured by increasing the fBFO in 100 Hz increments and was found to be f-3dB = 2.8 kHz.
Gain improvement
The gain of the mixer is set by resistor R4. The value of this component cannot be increased unlimited, however. The standing current through the resistor creates a voltage drop of R4. If the voltage drop becomes too large, the voltage at the collector will become too low for proper operation. Another option would be to change the resistor to a current source. For this purpose R7, Q3, R8 and C7 were added. Even if the AF output changes momentarily, the voltage over C7 will keep the Vbe voltage and the voltage over R7 constant and hence the current. So, this section effectively works as a dynamic constant current source. Dynamic, in the sense that it automatically adjust to the standing current through Q1 and Q2. Constant, in the sense that it does not change because of changes in the output.
The tests above were repeated, with the following results:
| Vcc = 9V | Vcc = 24V | |
| BFO [ mVpp ] | AF [ mVpp ] | AF [ mVpp ] |
| 1000 | 500 | 460 |
| 900 | 480 | 480 |
| 800 | 480 | 440 |
| 700 | 480 | 440 |
| 600 | 470 | 440 |
| 500 | 470 | 400 |
| 400 | 460 | 320 |
| 300 | 440 | 140 |
| 200 | 430 | 64 |
| 100 | 320 | 24 |
All test were done with fRF = 1 Mhz, a fBFO = 1.001 Mhz and VRF = 20mVpp. So, the conversion gain has almost doubled. The parallel resistance to the current source is R3 (10k3), so that why is is now only a 2x increase. The odd value is because the original 5k6 was augmented with a 4k7 resistor.
In his seminal book “Ontvangers” F.A.S. Sterrenburg, briefly mentions a product detector with a JFET [2]. In general, those devices have a large spread. So the inter-device spread needs to be compensated. In the referenced design, this is done by making the source resistance of one of the FETs adjustable. I fail to see that using jFETs over bipolar transistors seem to offer any significant advantages in this application, so it is not further pursued here.
Although ordinary BJT transistors, do a have a large spread as well, this is easily remedied by fixing the bias conditions. In this case this is done with R1,R2, R3, R5, R6 and R7.
Literature
[1] RadComm, Technical Topics, June/July 1980, page 639
[2] Ontvangers, F.A.S. Sterrenburg, page 135
[3] Electron 1980, Reflecties, May 1980, page 204
[4] RadComm, Technical Topics, May 1982, page 411
[5] RadComm, An sic transceiver of ssb and am, June 1971, page 378ff
Recent Comments