On The Art of NDB DXing
by Sheldon Remington
©
1987-2000 All Rights Reserved
|
CHAPTER EIGHT: THE FREQUENCY DOMAIN
Part Two - QRM, Frequency Clusters, and Selectivity In the last chapter, we saw that the more populous countries adhere to 3 or 5 kHz channelization. With thousands of beacons active, each of these channels will have dozens of beacons assigned to them. Therefore, co-channel QRM is very much a factor in NBD DXing, just as with other channelized bands such as CB and the broadcast bands. Furthermore, we saw that the most widely used pitch for A2 modulation is 1020 Hertz, especially in the U.S. Each major channel will then end up with a large group of carriers very close to the nominal channel frequency, and a smaller group of lower-sideband idents about 1020 Hertz below that frequency. Each of these groups is known as a cluster. With these facts in hand, we can diagram a typical segment of the NDB band, showing two major (3 kHz-spaced) channels, each with its retinue of three clusters:
In addition to these busy 347 and 350 kHz channels, recall that there will also be a few NDB's assigned to the 1-kHz channels in between, e.g., 348 and 349 kHz. And each of these will have its three (carrier, USB, and LSB) clusters, too. When we add these to the 347 and 350 clusters, we find each cluster superimposed atop another, leaving mostly empty space between clusters as shown here:
When we start out in NDB DXing, we naturally tend to treat each cluster just as we would treat a busy BCB channel: sitting squarely atop the cluster and praying for propagation to bring up the weaker idents to temporary dominance over the regular pest(s). We might also hope for a silent period from the pest, but that's even less likely on the NDB band than on the NSP-plagued BCB. This method can bring in quite a few loggings if we spend a lot of patient nights on a good cluster over the years. However, the good clusters turn out to be reduced in number because a lot of clusters are completely dominated by strong pests that never seem to take a fade. Sooner of later, we'll have milked all the good clusters dry, and the log total will plateau. Cluster Spread When we look more closely at ident pitches, we find that the 1020 Hertz ones actually may be anywhere from about 1000 to 1050 Hertz, in the great majority of cases. A similar range will be found among the 400-Hertz beacons favored in Canada and in eastern Australia; they generally sit between 370 and 430 Hertz from the nominal carrier frequency. Additionally, we find that any given beacon will typically hold its ident frequency stable within perhaps 5 Hertz over the long term, just as with the carrier frequency. Within the ranges given above, the distribution resembles a bell curve, with more beacons near the middle of the range than at the extremes. To illustrate, here is an expanded diagram of a single busy cluster, centered on 349 kHz, based on actual measurements over the past couple of years (ignoring any 349 carriers):
The total width of this cluster is only 130 Hertz, which is quite narrow by the standards of DXers on other bands. Usual bandwidths in use elsewhere in the RF spectrum range upwards from 2000 Hertz, which on the NDB band won't even be able to separate clusters from each other, let alone idents within a cluster. IF Selectivity By itself, the ear/brain system has a bandwidth of about 50 Hertz (at a pitch of 250 Hertz). This is known as the equivalent rectangular bandwidth of the human auditory filter (see "The Filters In Our Ears," Audio, Sept. 1986). However, a really loud tone can be far outside this bandwidth and still constitute a major distraction, leading swiftly to ear fatigue if we're trying to concentrate on a weak ident inside the bandwidth. This, then, represents the minimum acceptable selectivity for our receiver systems: a width sufficiently narrow to pass only one cluster at a time. For DXing in North America or Down Under, this width is 400 Hertz. Again, this seems narrow compared to the bandwidths normally included in general coverage receivers. However, hams who engage in CW contesting often install a 250 Hertz wide crystal filter in their receivers. A wide variety of these are available commercially for the popular ham receivers and transceivers, both from the equipment manufacturers and from independent sources such as Fox-Tango Corp. (Box 15944, W. Palm Beach, FL 34983). Even if you can't find a 250 Hertz filter designed specifically for your equipment, you can often find a match to your IF frequency. You could also build one yourself (see QST's cover story for July 1987). Some of the surplus receivers have narrow bandwidths. Many of the newer receivers have passband tuning, slope tuning, and the like; this is very worthwhile to get additional selectivity once you've inserted your 250 hertz filter. Remember, every decibel helps in DXing. Audio filters can substitute for IF filters in non-demanding situations; more on these in a moment. AGC and BFO Requirements It is absolutely imperative for successful NDB DXing that you find a way to defeat your receiver's AGC. There are two reasons for this, either of which alone would justify the above statement. One is that every lightning impulse causes the AGC to practically shut down the receiver for as long as the AGC time constant holds, typically 1 to 2 seconds. A lot of DX lurks between static crashes, and you won't even know it's there if your AGC is on. As we all know too well, longwave (and even on up to 80 meters) is dominated by lightning; in fact, sometimes it seems the best non-winter propagation coincides with lots of distant thunderstorm noise. If you try a receiver with AGC switched in and out when there is any amount of T-storm activity, you will hear instantly what I mean about AGC hang-up covering up the DX. The other problem with AGC is that a strong signal anywhere in the IF bandpass will similarly tend to reduce receiver gain. If the QRM is a carrier, the gain will simply stay low; worse, if its a loud ident, the receiver will ride up and down with the Morse Code keying, causing weak signals on the frequency to produce false negative keying (for what negative keying is, see Chapter 9, Beacon Defects). Either way, we'll be unable to make much headway with audio filters when it comes to DXing in the parts of the band close to our worst pests. AGC defeat is included in many fancy receivers; if yours doesn't have it, it shouldn't be a hard modification with any except the smallest portables. Defeating the AGC may also defeat the S-meter, but our receiver's amplitude response is now made linear from antenna to headphones. This means we can use a calibrated audio meter, such as a VU meter, to read signal strength, and it will include the effect of audio filtering (which an S-meter wouldn't). The best overall receiver gain levels, adjusted by preamps, attenuators, and the receiver's RF gain control, then become just a matter of keeping signals at least a few dB below the point where some receiver stage starts to become nonlinear (overload). As mentioned earlier, audio filtering can be used in lieu of, or as an adjunct to, narrow IF filtering. Besides defeating the AGC and watching out for overload, there is one other condition that must be met; the BFO must be correctly set just outside one edge, preferably the lower-frequency edge, of the IF passband. As we tune across a carrier, it should go from high to low pitch and then disappear before the pitch starts rising again. Otherwise, two dial settings will produce our reference pitch on every signal, which can get pretty confusing. Different receivers have different controls affecting BFO adjustment, such as passband tuning. Some receivers may require an internal tweak to get zero-beat positioned on the edge of the passband; others might be best in USB mode rather than in CW. Bandpass Audio Filtering The basic kind of audio filter response we need in NDB DXing is the bandpass. A minimal bandpass response can be achieved with just a low-pass filter, if necessary, because the intentional gentle roll-off built into most receivers will establish a bottom edge to the passband. It is not essential that the passband be readily tunable, once we settle on our chosen standard pitch. Sometimes it is useful to be able to vary the width of the filter, especially if the upper and lower edges can be moved independently. With the latter feature we can chop a cluster in half without making the overall bandwidth so narrow that the ear's filter is excluded from the differentiation process. The filter skirts should be as steep as possible for such close-in work. The steepest filter alignment is the elliptical (Cauer) response, which has a notching action right at the skirt's edge. For implementation, there are three general technologies of audio filter design: passive, op-amp, and digitally-switched. The passive type has the advantages of no internal noise and no power requirement, but its components become enormous at low pitches, and achieving steep tunable skirts is impractical. The op-amp types need many stages to achieve steep tunable skirts, and getting those stages to track when tuning is difficult. So that leaves the digitally-switched-capacitor filter technique, which solves all these problems nicely. Their chief drawback appears to be the presence of low-level spurious RF signals in their output. But all in all, such SCFs are probably the easiest way to achieve a cluster-busting response. A number of SCF ICs are available from National Semiconductor, Reticon, Motorola, AMI, Mostek, and others, at low prices. For homebrewing, the National MF-10 series is simple to work with and well backed by documentation; contact them at 2900 Semiconductor Drive, Santa Clara, CA 95052. At least one commercial audio filter is available based on the MF-10, the Palomar Engineers unit, which has independently tuned passband edges. Several SCF bandpass designs have appeared in the pages of the amateur and hobbyist magazines. One excellent project was the superSCAF in QST April 1986, Page 13, which has exactly the response we need. It is based on the Reticon SCF's, and partial or complete part kits are available; the Junior kit at $50 is great value for the money. One ready-built filter that at least provides a flat-topped bandpass response and is inexpensive is the popular Bencher XZ-2 which was reviewed by W.R. McIntosh in the Nov. 1982 Lowdown. Unfortunately, many of the so-called bandpass filters on the market do not have flat-topped response. Instead, they have a peaked response, often variable in roundness by a Q control. At first glance this might seem ideal, but experience indicates that this goes too far in bypassing the powerful frequency-differentiating function of the ear. The result, as with ultra-narrow flattop filters, is that it becomes difficult to separate the signal from the noise and QRM, since everything comes out with the same pitch. Some filters suffering from this problem are the Autek, the MFJ's, and the Datong FL-1. Notch Filtering Since we've used bandpass filtering to restrict our selectivity to just one cluster, or portion of a cluster, there will often still be one or two fairly pesky signals riding on top of the DX. Since a properly-designed NDB notch removes only a small percentage of our 50-250 Hertz wide passband, it doesn't impair the ear's ability to discriminate. Many clusters have one really loud carrier or ident; it can be like a breath of fresh air to be able to completely wipe it out, opening up the channel for DX without removing the DX too. In fact, the notch is the real secret for cluster-busting, using the bandpass filter just to exclude adjacent clusters. The drawback of sharp notching is that key-clicks will be produced when a very strong ident is being notched. Since a notch needs fewer parts than steep bandpass filters, the op-amp technique is perhaps better suited than the SCF technique. Some commercial filters are equipped with a notching function, but probably would need a bit of modification to make the notch narrow enough for cluster-busting. Also, with an extremely narrow notch, it becomes difficult to locate the notch with your ears until it's squarely atop a signal. So, it's helpful to get extra bandspread capability in the notch tuning control by restricting its overall range. Some receivers have a built-in notch function, but, again, these will probably need minor modification for the above reasons. If manufacturers provided NDB DX narrowness in their notches, the casual user would think it wasn't operative since they couldn't find the notch! Some receivers have the notch circuit in the IF stage rather than the audio, which, in theory is superior. But for NDB DXing with the AGC off, either type is equally effective if they have the same width and depth. Shown below is a circuit for a 1 Hertz wide notch. Originally, this was the notch in the Kenwood TS-430, but John Seamons and this writer felt it was too wide, so we tinkered and redesigned until it was optimized for cluster-busting. Construction is non-critical, and any standard op-amp can be used. It has unity gain and should be inserted into the early stages of a receiver. Some DX situations benefit from two or more notches; this circuit can be built in multiples as desired.
The depth of the notch can be trimmed for total rejection of any stable signal, probably exceeding 70 dB. A linear pot at R7 gives a lot of bandspread in the low end of the tuning range and squeezes the frequencies above 1000 Hertz into about 20% of the rotation. For use primarily with voice-bandwidth reception, try an audio taper pot at R7. If extreme narrowness is not needed, the notch can be widened by reducing the value of R10 and proportionally reducing the R8 and R9 values. Tuning range covers about 145 - 2300 Hertz with a constant bandwidth. These limits can be changed easily. To change both ends, change the value of R7. For the high end, change R5. Here are the formulae to calculate the frequency range:
To sum up, the minimum filter would consist of a low-pass with a cutoff tuned to slightly above our standard pitch. The most versatile filter would consist of several notches combined with a bandpass filter having steep, independently-tunable edges. The bandpass will also be handy for isolating the between-cluster split idents from foreign countries, which then effectively become clear-channel beacons. Below is a diagram of the response of such an ideal filter. Compare it with the diagram of the 349 kHz cluster (back a few illustrations) and the possibilities will be self evident.
Chapter Nine will discuss ways and means of measuring exact ident
frequencies.
|
