Racal 40MHz Filter Alignment

 The Racal 40MHz filter has assumed some notoriety as being very difficult to align and one can see why when doing the job.

The first step is to disable the receiver and attach a couple of probes, one for signal injection and the second a monitor device. A signal of 40MHz is input at Test Point 2 which is located close to V7 and a monitor probe connected to Test Point 3 near to V9. See below..



 You'll notice that V5 and V9 have been removed, also the 1MHz crystal inboard from the mains transformer.

I used my spectrum analyser which has a lowish voltage limit on both its input and tracking generator. Under no circumstances must you connect a spectrum analyser to a valve radio before checking its specification. I used a DC blocker with a voltage protection device to buffer the tracking generator plus an active probe with a 1 megohm input impedance. Because of these items the actual voltage/power readings are meaningless in themselves but, comparatively speaking, measurements of filter response are fine.

First, the response of the filter before realignment. Markers are set at 40MHz and plus/minus 650KHz. Vertical divisions are 10dB and horizontal divisions are 600KHz.

As the numbers below refer to power the 3dB points are important. Bandwidth of plus/minus 650KHz (39.35MHz to 40.65MHz) at 6dB given in the official spec will read 3dB in the picture below and is something like -3dB at 39.1MHz to 40.8MHz or a bandwidth of 1700KHz (instead of 1300KHz). Also, although the leading edge isn't too bad the trailing edge is poor. The official spec says -20dB (= -10dBm) at 38.6MHz to 41.25MHz or a bandwidth of 2.65MHz. Below this is about 2MHz which is OK.

Attenuation outside the main curve at 38.75MHz should be -20dB or -10dBm reads -72dBm or an attenuation of -72+44= -28dB which is about right.

 Below I've pictured the final response. The picture shows markers 1 to 4 which indicate the power level at specific points of the curve.

Marker 1 is the 40MHz centre frequency at -43dBm, 2 and 3 show the 650KHz points at -43.8dBm/-43.5dBm and 4 shows the skirt response at 1.5MHz from the centre at -62dBm/-61dBm. Vertical divisions are 3dB and horizontal divisions are 500KHz. The aim of the official guide was to produce -6dB at plus/minus 650KHz and a trough at plus/minus 1.5MHz of something less than -20dB wrt 0dB. The troughs on my curve are -69dBM and -65dBM wrt -43dBm.

The final setting has squeezed the bandwidth narrower. The 3dB points are 39.3MHz and 41.8MHz, a bandwidth of 1500KHz. Ripple is about 1dB. Ideally the 3dB bandwidth should be 1300KHz, so I'll have another go at it later.

 How does the curve above affect operation? Tuning the Megacycle knob should be easy. Tuning into the specific 1MHz band should be a fairly sharp process, tuning up the filter front edge, then there should be a plateau of contant noise until further tuning drops down the filter tailing edge. Tuning should place the noise output roughly in the centre of the plateau where any oscillator drifting might push one way or another, but without dropping off the filter front or back. When the filter has a peak either side of a central peak the three peaks will be visible on the S-meter. Tuning to the central peak is best. The official filter shape has one peak either side of a central dip. If the plateau is too narrow tuning will be too sharp. If the response is too broad or the skirts are too shallow then strong stations in adjacent 1MHz bands may break through. I think the ideal shape should have a slight droop at 40MHz of say 1dB to allow tuning to be dead centre, or as above a slight 1dB peak.

During alignment, I started with the cover removed but found the response changed when the cover was refitted, so it's best to finally align with the cover in place. Initially (or once completed) it may be an idea to remove the cover and confirm that each trimmer is not either fully meshed or fully open as a common falure are the silver mica capacitors shunting the trimmers.

This also applies to the 37.5MHz filter.

back to realignment