It's been several years since
looked at my Racal receiver.
At the time I checked and replaced some resistors and capacitors.
I have a vague idea there's something not quite right. Tuning
a strong station reveals three peaks, like three stations side
by side. Sensitivity isn't great either and generally it seems
a bit docile. Also I notice that the megacycle dial has to be
set half way between the numerals.
So I reckon it needs re-alignment.
Since I now possess a spectrum analyser with built in tracking
generator I am able to sort out the notorious Racal band pass
There are two complex filters. One centred on 40Mc/s and the
second 37.5 Mc/s.
The former is a problem as it needs to have a pass band of around
plus and minus 650KHz. ideally it should pass frequencies within
the range 39.35 to 40.65 Mc/s.
When I last looked at the RA17 I'd discovered it was very deaf.
This turned out to be a faulty silver mica capacitor in the 40Mc/s
band pass filter.
At the time I just twiddled each of the eight trimmers (having
marked their original settings) and discovered one didn't show
any signs of peaking. The others when turned right or left had
an effect on the signal strength of a station but one didn't.
I fitted a new fixed capacitor of identical marked value and
found the associated trimmer now peaked a signal.
The set then worked quite well. That
was perhaps ten years ago
I was now quite keen to see the shape
of the response of the 40Mc/s filter.
click here to for practical info
After removing V5 and V9 and the 1Mc/s crystal, I connected the
DSA815 output to Test Point 2 and its input to Test Point 3 using
miniature crocodile clips together with a third to a local chassis
I set the centre frequency of the DSA18 to 40Mc/s and the span
to 5 Mc/s, giving 500KHz horizontal divisions.
The top of the curve peaked at around -67dBm and the noise level
was around -90dBm (these are relative figures and dependent on
the high impedance amplifier I'm using).
A word of warning here. As spectrum
analysers are very expensive and modern models often have a relatively
low maximum DC limit at input and output it is advisable to isolate
these from any inadvertent HT from the RA17. Test points 2 and
3 are basically OK in this respect but one can never be too careful.
Note that I used a homebrew high impedance
(greater than 1Mohm) amplifier at the input and a DC blocker
(2 x 0.22uF 500v capacitors in series with back to back 15V zener
diodes to ground at their junction) at the output. This inserts
quite a lot of loss into the measurement process but as the loss
is constant over a very wide range of frequencies this is inconsequential
if the internal amplifier in the DSA815 is used and the tracking
generator is set to a high enough level (0dBm)
Surprisingly the shape wasn't too bad,
however, although the top was more than 20dB higher than the
skirts the centre frequency was way over to the right. In fact
it was nearer a 41Mc/s band pass filter rather than a 40Mc/s
Using the spectrum analyser I experimented
by twiddling the eight trimmers to see the effects of each. Some
shifted a peak to the right and some peaked to the left. Some
made the top of the response wider and some made it narrower,
but nothing clear cut, all very vague and woolly.
After half an hour or so I got the filter centred on exactly
40Mc/s. Initially the top was half a meg wide being nice and
flat from 40.3 to 40.8 Mc/s but not wide enough in respect of
the RA17 spec. A bit more fiddling and the top got wider. Now,
the top was a Meg wide having a peak to the left at 39.5 Mc/s
and a peak on the right of 40.5 Mc/s with a rather saggy centre,
but was now a respectable 3Mc/s wide at the noise floor.
The next day I read a bit about overall alignment and decided
to sort out the misalignment of the Megacycle dial.
This dial is on the shaft of the capacitor that tunes VFO1 which
tunes in the range 40 to 60 Megacycles. The figures on the dial
represent the number of whole megacycles to which the RA17 is
tuned in the relationship where "1" equals a VFO setting
of 41.5Mc/s and "29" equals a setting of 69.5 Mc/s.
The reason for the wide response of this filter now becomes meaningful.
If the VFO drifts due to warm-up, or if it shifts due to mechanical
vibration affecting the oscillator valve or due to an imperfect
setting of the megacycle knob the set should still work.
In fact if one slowly tunes the megacycle knob whilst tuned to
a strong station the response of the 37.5 megacycle filter can
In my case there are two peaks (39.5 Mc/s and 40.5 Mc/s) and
a dip of around 2 or 3 dB in the centre at 40 Mc/s and this shape
can clearly be seen as a weakish station is slowly tuned using
the megacycle knob.
The position of the knob is best set in the zone in between the
two peaks. If any drift occurs the signal will increase slightly
First though let me explain what I found when investigating the
megacycle dial offset.
Station peaking resulted in the megacycle dial ending up half
way between the numbers.
This could be due to the dial being fixed in the wrong place
on the capacitor shaft or the VFO not tracking properly.
I decided to investigate the latter.
This is where I had a brainwave. I have a communications receiver
that tunes the VHF range and can be quickly set by keying in
a number. I connected the aerial input of the VHF receiver to
a length of cable terminated with crocodile clips and clipped
these to the Racal. The ground clip to the chassis and the live
clip to a wire passing close to the VFO valve, V5.
I then tuned the VHF receiver to 69.5 Mc/s AM and adjusted the
Racal megacycle dial to "29". By slowly tuning the
latter backwards and forwards I was able to hear the 69.5 Mc/s
Racal VFO. Setting the latter to "1" allowed me to
hear the 41.5 Mc/s signal also, however finding the exact signal
was not too easy as a mixing process took place and resulted
in a whole set of medium wave signals emerging from the VHF receiver
loudspeaker. In fact using dial setting "2" turned
out to be easier ("2" is equivalent to 42.5 Mc/s).
After some experimentation it was apparent that the Racal VFO
was out of sync with its dial.
Basically the VFO frequency range is shifted with respect to
the dial so that instead of 41.5 Mc/s at "1" the frequency
is about 41 Mc/s or 41.5 Mc/s ends up half way between "1"
Clearly, if the frequency needs to be aligned, I need to adjust
and track the VFO so it lines up with the dial. Where is the
VFO trimmer and coil tuning slug?
The answer is; underneath the metal box carrying V13 and V15.
This box carries the calibrator and is connected via a coax cable
and an octal socket. A securing screw needs to be slackened off
and the calibrator pulled away and lifted clear to access the
Once this is done the trimmer capacitor and coil tuning slug
adjuster can be seen.
The latter is locked with a nut and I could see from distortion
of the screw slot that it's been tackled before.
Using the VHF receiver I was able to fiddle with the trimmer
and the tuning slug to bring the VFO into line with the dial
The trimmer adjustment is extremely critical as we're dealing
with frequency changes of around 10Kc/s, but I was able to set
this correctly by persevering.
For reference I checked the VFO frequencies at either side of
the black "1" marker.
The leftmost edge was 41.47 Mc/s and the rightmost edge was 41.55
Mc/s, a range of 80Kc/s.
41.5 Mc/s ended up almost in the centre of the black "1"
At the other end of the Mc/s dial the leftmost "29"
was 69.41 Mc/s and the rightmost was 69.57 Mc/s (a range of 160Kc/s).
If I'd carried on fiddling the settings could have been more
accurate, but the final settings were good enough.
Once I'd finished the VFO settings I plugged the calibrator back
in and tried the RA17 on an aerial.
Performance was now an order of magnitude better and the megacycle
dial was within the black markings for all settings.
I confirmed the BFO was aligned correctly by switching the receiver
to the BFO calibration setting and confirming that zero-beat
was at the zero position of the BFO adjustment knob. Tuning to
5, 10, 15 and 20 Mc/s gave me very weak reception of standard
frequency stations and these corresponded exactly to the zero
position of the Kc/s dial setting (adjusting where appropriate
the sliding plastic dial marker to conform to the relevant receiver
I checked the "drift-ability" by moving the Mc/s knob
whilst tuned to a strong medium wave station. This is essentially
moving the VFO frequency downwards and upwards from the centre
frequency and observing signal strength. The station was lost
at settings of 41.31 and 41.83 Mc/s. The frequency difference
of about half a meg, to a large extent represents the pass band
of the 40 Mc/s band pass filter.
Next I need to confirm the 37.5 Mc/s
band pass filter is acceptable, the 100 Kc/s IF amplifier is
OK and overall sensitivity is within spec.
During my experiments it became pretty
obvious that the modern listening environment was extremely poor.
When I used to listen to radio in the 50s the background level
between stations was really quiet and you could virtually hear
a pin drop. In those days the short waves were full of fascinating
tunes. Every broadcast station had it's own signature tune, Radio
Moscow, Radio Peking, Radio Sofia and the Voice of America, all
had their own very recognisable tune.
As the years rolled on, in order to comfortably continue listening
one had to switch off fluorescent lights and rely on ordinary
lighting. That's not true any more as modern electric lamps use
integral switch-mode power supplies, even worse than tubes.
Add to this a host of other extraneous noises and short wave
listening can be impossible.
Telephone lines carry broadband signals, dozens of electronic
items on standby (including computers), hosts of plug-in switch-mode
power supplies radiating interference, and mains wiring circulating
all sorts of weird noises.
The answer now, I suspect is to erect a vertical aerial as far
away as possible from telephone wires and mains wiring, and to
feed this via good quality coax.
Another solution may be to employ a
tuneable frame aerial or a ferrite rod mounted high up and away
from buildings. Both can be rotated to reduce interference. To
make best use of these techniques would need a method of tuning.
Maybe a varactor diode connected to a potentiometer located adjacent
to the receiver?
The next step was to re-check the 37.5Mc/s
and the 40Mc/s filters as I didn't care for the slight double
hump response when tuning the Mc/s knob. A second exercise squared
off the top at the expense of a slight narrowing of the pass
band of the filter. It now has a width of 1.0Mc/s at the -6dB
points instead of the recommended 1.3Mc/s. At 20dB down the width
is now 1.8Mc/s and at -40dB it's now 2.4Mc/s.
I might have another go at this if it fails to work out in practice.
The 37.5 Mc/s filter has more adjustments
but can be narrower than that for 40Mc/s (recommended at 300kc/s).
I tried adjusting it using the spectrum analyser but I wasn't
happy with the loss of gain for a relatively small increase in
bandwidth, so I just peaked all the trimmers using a signal generator
set to 5Mc/s connected to the aerial socket.
I also tweaked the 100Kc/s slugs in the final IF amplifier, but
these were relatively flat so just peaked them up.
During proceedings I checked overall
sensitivity. First at the low end of a band, choosing 29.000Mc/s.
I found the set could easily hear 0.5 microvolts modulated at
I then tuned to 30.000Mc/s and discovered the receiver was completely
deaf at this setting.
The sensitivity was OK up to about 29.6Mc/s then the test signal
A look at the circuit diagram indicated the likely problem was
VFO2 which tunes 2.1 to 3.1Mc/s.
I connected the DSA815 to a convenient point which was a socket
close to V11 and set the RA17 back to 29.000Mc/s and adjusted
the analyser centre frequency to 2.5Mc/s with a 2Mc/s scan width.
The display showed a good solid signal at 3.1Mc/s at a level
of -36dBm (this figure is relative) but as I tuned the RA17 up
in frequency the signal started to drop in amplitude as its frequency
decreased until, at 2.5Mc/s it just disappeared into the noise.
Just before it vanished it stood at -48dBm with the RA17 tuned
I hunted around and found a used EF91 and plugged this in place
of the original valve.
After a minute or so the signal reappeared but this time it vanished
at around 2.1Mc/s when it stood at -44dBm. A second hunt through
my boxes of valves revealed a brand new EF91. This did the trick.
The signal tuned right across its range starting at -32dBm and
dropping to -38dBm.
Another test and I found that at 1.5Mc/s
I could comfortably hear a quarter microvolt modulated signal
and with the BFO switched on I could just hear a signal of only
The set now tunes stations nicely with no spurious humps and
the Mc/s dial smoothly increases and decreases signals with each
Mc/s setting within the width of the associated black mark.
However, I'd like to play around with
the set before putting it all back together and maybe making