small logo Publications Department

Detecting Weak Signals at Receiver IF
by Chris Cadogan, G3XWB (chris@cadogan.u-net.com)

A recent article reminded me of some development work I did some 20 years ago as a new graduate RF engineer at Racal in Tewkesbury, England. I worked on a signal detection system, in principle a reliable squelch circuit. What might make it of interest for SETI is that it was good at detecting weak signals in a very wide bandwidth.

The SETI League is using DSP to pick out narrow bandwidth signals from the receiver's audio output. What I did is a variation on a theme, and involves two phase locked loops working together to pick out narrow band coherent signals from within a wideband IF output of a microwave receiver - typically 4 MHz wide at 70 MHz.

This wouldn't be as sensitive as audio DSP filtering, but does have the advantage of 'seeing' the whole of the receiver wideband IF at once. The wider the receiver bandwidth the better it works. The sensitivity comes about by virtue of the PLL bandwidth being much narrower than the receiver IF bandwidth, the PLLs only see the noise in (or close to) their own filter passbands.

Click here to see Block Diagram

The vital bit is the Phase Sensitive Detector that multiplies the two PLL VCO outputs together (the PSDs are straight analogue multipliers, not the digital coupled D types variant). The output is a DC level more or less proportional to signal strength, up to limit of about 4 dB sig/noise, after which it flattens out. I grant you that any signal with a 4 dB sig/noise in 4 MHz is a fairly beefy signal, but I do remember that there was a detectable and reliable change in DC at very low signal to noise ratios. At the time I was concerned with reliability - not false triggering - rather than sensitivity. It was the radio ham in me that was interested in weak signals, the professional just wanted to know if there was a good signal or not.

There are a number of different ways of thinking about how it works. Here's one of them: with no signal present the two PLLs both see the wideband noise output. Their loop bandwidths are much narrower than the receiver bandwidth, so at any instant their 'centre frequencies' lie randomly within the IF passband. The VCO control voltages are noise like, and there is no correlation at all between the two VCO frequencies, giving a zero volts DC output.

If a weak signal appears, then each PLL will wander over the top of it for some of the time. During the time the signal is within the loop bandwidth the phase detector will 'see' the signal and generate a DC term to pull the VCO towards the signal - the PLLs start to capture the signal.

As each PLL starts to see the signal, the two feedback loops make each VCO spend more time near the signal, and less time randomly wandering about. This shows up in the two VCO starting to become correlated, which generates a DC term at the output - you've detected a signal. At low signal to noise ratios the DC output has a straight line relationship to signal strength. It is a good 'S' meter, between S0 and S1.

For the amateur SETI site I'd think of it in terms of an alarm signal - something's there in the passband, go and find it with a narrow band filter switched in. Alternatively, it could be used in a scanning system to decide whether this wideband chunk of spectrum is of any interest. This is what it was developed for in the first place.

The Theory

I'd put quite a lot of effort at the time into analysing how it worked, and frankly didn't get very far. What puzzled me was why each PLL did its own thing and stayed uncorrelated with respect to each other. I'd put this down to the non-linearity of the PLL phase detectors - the output is proportional to the cosine of the phase of the two inputs - and that large phase excursions moved the operating point over the top of the cosine wave and dumped an uncorrelated block of noise into the feedback loop. With twenty years hindsight this looks like chaotic behaviour - a small change in input could make the output flip over the top of the detector curve or not. It then looks reasonable to regard both systems as being chaotic and thus independant.

What is puzzling is that it continues to work - although not as well - with small phase inputs and a loop bandwidth of the same order as the IF bandwidth. As the receiver bandwidth narrows, then the IF noise signal looks more and more coherent, and the DC output goes up. It gets harder to tell the difference between a weak signal and noise. The converse is that the wider the IF bandwidth is the less coherent the IF noise signal becomes, giving a lower DC output on no signal, and hence a larger change when there is a signal.

Enough of the theory - the pragmatic answer is that it works reliably :)

Practicalities

This circuit idea is only of use to you if you can get at the first or second IF - not common on amateur receivers. However, if you're developing your own receiver, then the circuitry is inexpensive. You have to take care with layout and supply decoupling so that the two PLLs don't see each other.

I used a PLL bandwidth of a few kiloHertz, giving recognition times of about ten milliseconds. There's an optimum ratio of IF bandwidth to centre frequency, it works well two or four megs wide at 21.4 MHz : you really want the IF bandwidth to be as wide as possible, until you start to strain the swing of the VCOs. I'd made several variants on the basic idea for differing applications, and there's nothing particularly critical needed in the circuitry or that's difficult to set up. I'd recommend varactor diode tuned LC VCOs - alignment is done simply by tuning the core so that the PLL control voltage is noisy when it sees the receiver IF frequencies, and is centred on zero volts.

I'd imagine you could make it more sensitive by reducing the loop bandwidth, at a cost of making the detect time longer. As I pointed out in the theory bit, the relationship between receiver bandwidth and the phase locked loop bandwidth isn't a simple one. It is a splendid system for detecting receiver spurii :)

Racal took out a patent on this system about twenty years ago. I was one of the co-authors. At the time they were more interested in detecting narrow band HF signals, in which case using two phase locked loops was overly complicated.

I'm pretty sure the patent would have lapsed, because I've not heard of the idea being used commercially. I can't be too sure about this, because I've not had any contacts with the radio industry for quite a while now - these days I work on DSP in the pro-audio industry. I do know that the particular company in Tewkesbury - one of a large group - no longer exists.

In any case, amateurs are free to build whatever they choose.

I'm afraid I no longer have access to the sort of RF test gear needed to develop this sort of thing, but perhaps other SETI League members might wish to pursue it. I have a ham radio licence - G3XWB -, but I've not done anything with it for years - I gave a whole garage full of electronics away to a school a long time ago, when I decided that homemade high voltage transmitters and babies didn't mix.

It occured to me that any real SETI comms using radio - as opposed to perhaps narrow beamwidth lasars - would be optimally coded and hard to distinguish from noise, I wouldn't necessarily expect to pick up a narrow band signal or a carrier. Just a thought.

I hope this will be of interest to SETI League members. At the time I thought it might be really useful, I just wasn't sure what for.


Click to email the Webmaster
email
the
Webmaster
| Home | General | Memb Svcs | Publications | Press | Technical | Internet | Index |
entire website copyright © The SETI League, Inc.
this page last updated 28 December 2002
Click for top of page
Top of Page