The Direct Conversion Receiver (DCRX) has many faults. If you want the full story so far, you can do no better than to read Chapter 8 of "Experimental Methods in RF Design" (ARRL 2003). Something can be done about most of these faults, but Direct Demodulation, also known as AM demodulation, is probably the worst. That's because, once the Local Oscillator (LO) leakage has been brought under control, there is little that can be done about DD. Or is there?

DD is known to radio amateurs as "the Radio Moscow effect". It doesn't matter what frequency you tune your 7MHz receiver to, if you are listening from Europe and it's night and you have a good dipole antenna and you aren't using an input attenuator, Radio Moscow is what you'll hear. It's simply the strongest signal on the band. This is not to say that the problem is confined to HF. I wouldn't recommend taking a 144MHz DCRX to a hill top during a contest. What checks out as a perfectly functional RX at the bottom may turn into a Banshee at the hill top. (Banshee: an Irish/Scots fairy who shrieks & wails outside houses as a portent of death. A bit melodramatic perhaps, but you get the picture.)

The commercial boys would also like a railway carriage full of passengers to be able to talk simultaneously on their DCRX mobile 'phones. And it also makes sense to pack the frequencies in as tightly as the commuters. This too is a recipe for DD problems.

Figure 1 shows a representation of a DCRX, consisting of the LO, AF amp and, within the dotted area, the mixer. For the purposes of this discussion, the mixer is presumed to be a ring mixer of the SBL-1 class.

The contents of the mixer are diagrammatic. The LO to RF leakage which is so important to the DCRX is ignored; all that is shown are three signal paths. These are:

1. RF envelope detection, represented by the upper diode symbol in Fig. 1. This is the signal path responsible for the DD. The incoming "Radio Moscow" signal appears as a dc offset proportional to the AM carrier amplitude at the mixer output, together with the detected modulation.
2. The wanted signal, which is the RF input frequency shifted by the LO to audio frequencies.
3. LO envelope detection. This appears as a dc offset proportional to the LO drive. The LO's am sidebands are also demodulated to audio, so a noisy LO is, as ever, not good! Also, given some phase delay in LO leakage, LO phase noise sidebands may be demodulated as well.

A ring mixer relies on its balance to reduce RF DD, LO DD and signal leakage from the LO to the RF input. All of these can be influenced by applying a small dc bias to the mixer's diodes via the audio port. However, the minima of RF DD, LO DD and LO leakage do not occur at exactly the same bias current, and none of these minima are infinitely deep. However, providing the mixer with a good clean input signal will reduce LO DD so that its effects are insignificant. Good screening and use of pre-amplifiers with a high S₁₂ will also render LO leakage insignificant. So it seems sensible to use the dc bias of the mixer to reduce RF DD.

My first experiment was to pass the LO through a very stringent low pass filter to remove all the 2^nd harmonic. Then detecting the amplitude of the LO 2^nd harmonic leakage from the mixer at the dc coupled port, and using this to adjust the dc bias for minimum harmonic signal. This worked OK at 7MHz, and gave a useful 15dB odd of additional RF DD suppression. But then, so did connecting a low offset op-amp to the mixer input with a low pass feedback loop in order to force the mixer's dc coupled port to exactly 0V dc offset. Furthermore, neither scheme gave useful RF DD suppression at 14MHz. The problem is that the ring mixer's DD performance is not very linear. So what happens when the RF DD is nulled in this way, is that the audio fundamental component of the DD is eliminated, but the harmonic content of the RF DD remains.

Fig 2 is intended to show the harmonic contributors to the dc offset of the mixer. These were omitted from, and an overlay to Fig 1. This all has serious consequences: a 7MHz DCRX has a third harmonic response at 21MHz, and in theory, it is only 13dB less sensitive here than at the fundamental. And of course, the way of getting a better spurious free dynamic range is by driving the mixer harder, or with a square-wave, so there is plenty of 3^rd harmonic floating around: of course it is 13dB less than the fundamental. So the problems caused by LO leakage at the fundamental frequency are repeated at the harmonic frequencies. Now I have not attempted a 10GHz DCRX yet, but keeping 30GHz and up signals under control is probably a bit of a challenge. It's lucky you automatically get a measure of low pass filtering by using microwave parts at their maximum working frequency.

I'll post a further report when I'm not so fed up preparing this, my 1^st web page.

See also "Carrier Suppression in the Ring Mixer", J.v.Parpart, VHF Communications 3/1993 pp130-138

Incidentally, my personal preference is for Weaver receivers. See "SSB: third method, fourth explanation" Electronics World + Wireless World" April 1993 pp278-284. There are some problems with this article:

1. I miss-connected the LO divider in Fig 5: the D inputs both connect to the opposite Q bar outputs, clearly one should go to the Q output to make it work.
2. The graph in the box on Signal Cancellation is wrong. This is partly my fault, and partly the magazine's.
3. The input to the audio amplifier chain is a bit noisy. This generally doesn't matter on HF unless you are on 28MHz, and the harder you search for the noise floor, the more af microphony you find

Now then… Anyone tried using their Phasing/Weaver receivers dc coupled to oscilloscope X & Y inputs with a Smith Chart graticule with the LO connected to the RF input via a directional coupler to make a Vector Network Analyser? You'll have to be careful of the harmonics.

©Nic Hamilton 2003