Some Thoughts on Singly Balanced Mixers with Two Diodes and One Transformer

In 2001, out it in the Azores, I built a 17 meter version of Doug DeMaw's Double Sideband transmitter ("Go QRP with Double Sideband" CQ Magazine, February 1997).  I struggled to understand the balanced modulator -- how it mixed, balanced, and how it produced DSB.  I later presented my understanding of the circuit in my book "SolderSmoke -- Global Adventures in Wireless Electronics" pages 132-137.   In essence, I figured out that you had to think of the balancing and the mixing as two separate operations: The transformer provided the balance that eliminated the carrier (the LO signal) while the diodes presented the two signals (audio from the mic amp and LO from the VFO) with a highly non-linear path.  The LO was successively turning on both diodes then turning off both diodes. The audio signal was being "chopped" at the rate of the LO.  This produced a complex waveform that contained sum and difference frequencies -- the upper and lower sidebands.  The carrier was balanced out by the transformer because the two outputs of the transformer were always of opposite polarity, and they were joined together at the output of the mixer.    

Fast forward to 2013.  I built a 17 meter version of Farhan's famous BITX 20 rig.  Above you can see the balanced modulator stage, which also serves as the product detector. As you can see, it is essentially the same circuit as the one used by Doug DeMaw in his DSB rig. 

In 2018 I built a simple direct conversion receiver for my nephew.  For the mixer I used what I considered to be just a cut-down  version of the circuit used by DeMaw and Farhan.  I got the idea for this from Olivier F5LVG and his RX-20 receiver from SPRAT.    It had the RF signal coming in on L1 and the VFO signal coming in to the wiper of the 1 k pot.  But with this arrangement, the diodes were NOT both being turned off on half the VFO cycle, then both being turned on during the other half.  Instead, as the VFO signal swung positive, D2 would conduct and D1 would shut down.  When the VFO signal swung negative,  D1 would conduct and D2 would shut down.  It worked, but the diodes were being switched in a very different way than they had been in the DeMaw and Farhan circuits.  If you have the strong LO signal going in on L1, BOTH diodes conduct, then BOTH don't conduct.  But if you have the LO going in through the pot, one diode conducts while the other does not conduct. 

After I concluded that the BJT product detector circuit in the HA-600A was causing distorted SSB and CW reception, I tried the old DeMaw/Farhan circuit, this time in product detector mode.  See above. This worked better, but I realized that this configuration was balancing out the BFO signal, and not the IF signal.  My problem with the original product detector had been that IF signal was getting simultaneous envelope detection AND product detection.  So I decided to just switch the inputs and put the IF signal into L1 (where it would be balanced) and the BFO into  R1/R2 (the 100 ohm pot). 

This seemed like it would reduce the envelope detection problem, right?  I mean, L1 is the balanced input, right?  But I wonder if we need to consider how the diodes were being switched in this arrangement.  Instead of having both conducting and then both not conducting, in this arrangement one would be conducting during half the BFO's cycle, while the other was not.  That means that at any given moment, the two output sides of the transformer would be looking into very different loads -- hardly a condition conducive to balance. But I used LTSpice to look at the audio output under the two different port arrangements.  Sherwood advised looking at the output of the product detector with the BFO turned off --there should be no output with the BFO off.  And indeed, putting the IF signal into L1 and the BFO into the R1/R2 pot resulted in less of the distortion causing envelope detection.  The way the diodes were being switched didn't seem to adversely affect the balancing out of the IF signal.  I am not sure why this doesn't seem to cause trouble. 

There was, however, another problem with the use of this circuit in the Lafayette HA-600A:  port isolation.  The BFO signal was getting back into the IF signal input on L1.   I could see it on the S-meter.  This was worrisome not only because of the S-meter, but also because the same circuit was driving the receiver's AGC -- in effect, the BFO was turning the gain down.  Theoretically, this should not have been happening.  Look at the transformer.  the BFO currents going through L2 and L3 should be of opposite polarities and should be cancelling each other out in L1.  But obviously this was not happening.  Perhaps this was the result of the sequential way the diode are switching in this arrangement.   On the bench, if I put the BFO into L1, I saw very little BFO signal at the R1/R2 junction. If I put the BFO signal into the R1/R2 junction, I was a lot of BFO signal at the top of L1.  And that is what I saw on my S-meter when this circuit was used in the HA-600A. 

On the bench,  if I turned off the BFO and put an AM modulated signal into the junction of R1/R2, I can see audio getting through once the input signal reaches 1 volt peak.  I do NOT see that kind of "breakthrough" envelope detection when (with the BFO off) I put a modulated signal into L1.  So the singly balanced circuit is doing that it is supposed to do -- it is balancing out the the signal going into L1. 

So it seemed that with the singly balanced circuit I would have to choose: suffer from the poor port isolation or AM breakthrough.   Clearly it was time to go for a doubly balanced circuit.  And that is what I did. 

Finally, I took a look at another two diode detector, the Polyakov or "subharmonic" detector. This is a really interesting circuit that can teach us a lot about how mixers work.  Here you can run the local oscillator at 1/2 the signal frequency.  With two diodes back to back, the incoming signal is being sampled TWICE during each cycle of the local oscillator.  That is equivalent to having the signal sampled at twice the local oscillator frequency.   This circuit allows you to run the oscillator at a much lower frequency -- this could allow much greater oscillator stability.  In the circuit above, with both diodes connected, a 7 MHz incoming signal would produce a 2 kHz tone. 

Another big plus of this circuit comes if you take D1 out of the circuit (as shown).  In this configuration the circuit becomes a normal diode detector.  Here it will receive a signal at 3.5 MHz, converting that signal into a 1 kHz audio tone.  So you can get a direct conversion receiver for 40 and 80 meters fairly easily.