R107 Overhaul
|
|
Looking forward to my next receiver
overhaul, I picked the best example from the four R107 receivers
I have but found to my dismay that a previous owner had been
hamfisted and broken three important grub screws. One secures
the outer and two the inner tuning knob. The tuning mechanism
wasn't working because the lubricating grease had hardened and
the mechanism needed to be completely removed to remedy this.
I wonder if the fine tuner needed fixing back in the 1960s, the
owner tried to remove it for lubrication, broke the grub screws
and then set the receiver on one side... for ever? Certainly
everything looks very original, unlike most other examples I've
seen. From the lettering on the badge I think this example was
made by either McMichael or McMurdo Radio? |
 |
|
|
 |
|
I tried to drill out one
of the screws but soon found that this was virtually impossible,
but remembering how I detached knobs from ancient domestic sets
I tried an alternative method. Placing a flat screwdriver blade
under opposite sides of the outer knob I was able, little by
little, to gently lever it off. It wasn't too difficult and revealed
a very rusty shaft. I guess the front of the set must have been
damp at one time and this had expanded the screws slightly making
them impossible to remove. In a situation like this it's essential
to use a well fitting screwdriver blade otherwise the slot in
a grub screw will shatter. Once the knob had been removed I was
able to work out the next step. |
 |
 |
|
As the securing screws in the larger
knob were completely unusable I decided to use a hammer on the
end of the spindle. This worked and the knob just fell off. |
|
|
|
The slow motion tuner
comprises several parts. There's a large brass slow motion drive
assembly which connects to a metal bellows which deals with mechanical
alignment inaccuracies, manufacturing tolerances and also provides
some tension in the tuning condenser shaft. The drive assembly
is held in place by a brass ring that's secured to its inner
rim by a couple of screws. To remove the drive assembly for cleaning
the bellows needs to be removed and you'll soon discover that
the most of the parts are an interference fit and must be juggled
around to remove them. This applies to the pointer as well as
other parts. The pointer assembly has a collar which is secured
to the main shaft by a 6BA screw mating with a tapped ring crimped
to the collar. The crimping had failed and it proved very tricky
to remove the collar from the main shaft. It seems the resin
used to hold all the parts together against loosening from vibration
was stronger than the crimping. Detaching the bellows was not
easy. This is held in place by six grub screws treated with resin.
I dabbed an alcohol solution on the resin but this hadn't any
effect so I used a tight-fitting screwdriver which cracked the
joints and allowed the grub screws to be slackened. Once the
bellows had been removed the pointer could be removed and the
brass slow motion assembly carefully manoevered through the front
panel. At first sight it looks impossible but after a bit of
jiggling it came out. In fact the tolerances are such that the
assembly can be drawn straight out if it's exactly square to
the panel. |
|
|
|
The only method I could
think of to remove the inner knob was tapping the end of the
rusty spindle with a hammer. This worked OK and the whole assembly
slid out. You can see the groove in the brass where one of the
two grub screws was tightened. The brass ring came off after
slackening its two screws. As all the screws were locktighted
in place a tight fitting screwdriver is essential. |
 |
|
The old grease is in a
dreadful state and I had to boil the loose metal parts in washing
up liquid to remove it. The bearing surfaces above are part of
the metal dial backplate and are cleaned in-situ. |
|
 |
|
 |
|
Removing the damaged grub screws
from the knob was impossible but I did drill out enough of the
old screws to tap new 4BA holes and fit a couple of new screws.
Because the new holes were not now centrally located I had to
file the heads of new screws to the same diameter as their threads
(electric drill plus file).. ending up with grub screws... |
|
|
Above the brass slow motion
drive which has the code ZA11840 and serial number 8414 marked
on the rim. This example of the R107 dates
from 1943 (marked on its block condenser). |
 |
|
The bellows adaptor which
has six securing screws and the spindle coupler. Both need removing
before the spindle could be knocked out to detach the seized
larger knob. |
|
|
 |
The outer knob which had
to be levered off (awkward because it's sunk into the larger
knob) and the dial pointer. The latter was also very difficult
to remove because its captive nut crimping had failed and the
screw was locktighted and access was poor. I had to use a pair
of angled pliers to grip the captive nut (in fact it's not a
nut but a small tapped circular brass bush not easy to grip). |
|
|
Getting smooth action
in the slow motion drive was difficult. From the feel of the
input shaft I suspect the ball bearings are rusty and splitting
the drive looked problematical. I dripped an alcohol solution
into the six screw holes used for securing the pressure plates
and eventually I was able to turn the output shaft (the one with
the peg for limiting the range)... prior to this the output shaft
had been seized. Then I dripped oil into the six holes and alternately
spun the input and output shafts with an electric drill, turning
them until rotation felt unrestricted. It wasn't perfect but
infinitely better than when I'd removed it from the receiver.
Finally I inserted a lithium grease into the six screw holes
and repeated the freeing up with the electric drill then refitted
the end pressure plates and adjusted these until the input shaft
felt free. Now, when the input shaft is rotated, the output shaft
turns properly.
Alas, refitting proved to be
a problem. It's essential to fit the parts in the correct sequence
and it took me several attempts before I realised this. The direct
drive from the larger knob relies on smooth action from correct
alignment, proper lubrication and correct tension in the mechanism.
Both inner and outer phosphor bronze surfaces on the steel dial
backing plate must be cleaned and greased and the brass securing
ring postioned just right. It was when tightening the screws
in the ring that I noticed the thread on one screw appeared to
be stripped as the screw kept popping back when it was tightened.
I removed the ring and found the thread was fine but the ring
was cracked right across the tapped hole. As the ring is an interference
fit and an essential part of the mechanism I decided to repair
it rather than make a replacement. |
|
|
|
I have a decent large
soldering iron so after cleaning the ring surface with fine emery
paper I tinned the outer surface and sweated a flat metal strip
bent to fit across the break. I have a box of oddments of phosphor
bronze strips so used a part of one cut to size and with a hole
positioned over the tapped hole. To prevent solder messing the
thread I fitted a temporary screw through the hole. To ensure
the ring kept its correct diameter I completed the soldering
with the ring held in a small vice. Once the ring was repaired
I rubbed its bearing surface to ensure it was perfectly flat
and fitted a pair of new screws. |
|
During reassembly I noticed
the dial plate could move slightly under pressure. If not attended
to this would result in a sort of backlash effect in tuning.
I noticed two newish looking 0BA brass screws and although seeming
to be tight, I found one could be screwed tighter, removing the
dial wobble. |
 |
|
 |
The outer bearing surface needs
to be clean and greased. The opposite, inner surface likewise.
You'll also note the dial locking
screw. When refitting the large dial it needs to be positioned
to match the gap in the locking screw clamp. In fact it may be
easier to detach the knurled nut and plate before you fit the
larger knob.
The reassembled parts is shown
below. The dial pointer needs to be carefully positioned so it
passes cleanly over the dial. The short bar welded to the output
shaft extension is used to restrict movement of the tuning condenser.
The two screw heads can be adjusted to limit clockwise and anti-clockwise
rotation to prevent straining the tuning condenser.
Later, I examined a second R107
with a working tuner and noticed it had a large phosphor bronze
washer fitted between the ring and the rear of the dial plate.
This is missing on the example I'm working on and would improve
the feel of the direct tuner which is a bit sticky because the
tuner drive shaft is not precisely in line with the tuning condenser
shaft. |
|
 |
|
During assembly the tuning
condenser needs to be positioned initially at full capacity for
the pointer to be secured then at minimum capacity for other
parts to be secured. There are 6 adjusting screws for setting
the proper resistance within the slow motion assembly, 3 front
and 3 rear. These need to be adjusted for smoothest running without
slippage under load and without any backlash. You can see two
of these screws close to the brass ring above. The screws are
arranged so that they can be re-adjusted with the main parts
assembled if necessary. As the circular brass plate backing onto
the larger knob had lost some black paint I removed what was
left. The paint can be removed by rubbing with a piece of wood
with a good edge so that the soft brass isn't scratched. Now
that I've proved the tuning works properly I'll repaint the brass
plate later. I'll also need to clean the dial which is dusty
and slightly blemished. Because of the way the R107 is constructed,
access for this is a bit awkward.
The centre knob now tunes the
set extremely smoothly. |
 |
|
I powered up the receiver
but, instead of hearing a comforting hiss from the loudspeaker,
I heard a slight hum.. then nothing and occasionally the slight
hum would reappear. Switching the wavechange knob didn't result
in any crackling, just continued silence. The set, like others
from WW2, is fitted with a set of front panel test points but,
in the case of the R107, these are not intended to be read as
voltages with respect to chassis... instead one probe goes to
the reference point (=HT+) and the other probe to the test point
and thus measures the voltages across 3Kohm resistors which are
located in the valve anode feeds.
The designers, or whoever specified
this receiver, only used two types of valves in the receiver
proper, with a third type used in the power supply. This was
done for perhaps two reasons... firstly to aid field maintenance,
and secondly, to help valve manufacturers and the logistics chain
to minimise overall holdings. Most superhets of this vintage
used the 6K8 or ECH35 as a mixer, but the R107 used the same
type of valves employed in other functions. The following table
lists these. I can't explain why the dashes are used and why
the codes didn't use sequential letters for V2 valves as for
V1 types. I recommend making a simple lever-tool to help extract
the valves because some are located in awkward spots. |
|
Circuit Code |
Function |
Type |
Commercial |
Description |
Monitor Resistor |
Test Point |
|
V1A |
RF Amplifier |
ARP34 |
EF39 |
Pentode |
R6A |
1A |
|
V2A |
Local Oscillator |
AR21 |
EBC33 |
Double Diode Triode |
R6B |
2A |
|
V1B |
Mixer |
ARP34 |
EF39 |
Pentode |
R6D |
1B |
|
V1C |
1st IF Amplifier |
ARP34 |
EF39 |
Pentode |
R6E |
1C |
|
V1D |
2nd IF Amplifier |
ARP34 |
EF39 |
Pentode |
R6F |
1D |
|
V2B |
Detector/AVC/AF Amp |
AR21 |
EBC33 |
Double Diode Triode |
R6G |
2B |
|
V2B' |
Output |
AR21 |
EBC33 |
Double Diode Triode |
R6H |
2B' |
|
V2A' |
Beat Oscillator |
AR21 |
EBC33 |
Double Diode Triode |
R6I |
2A' |
|
V3A |
HT Rectifier |
6X5G |
6X5G |
Full Wave Rectifier |
None |
None |
|
|
Reading the servicing details it's
apparent that the exact HT voltage wasn't seen as too important,
or maybe just varied so much that it would have proved puzzling.
Instead valve anode currents are supplied to the fault finders
(but expressed as voltages across similar resistors = 3Kohm).
This is a really good idea, as is the instruction to monitor
the readings and to swap a valve if there was a change. |
|
Test Point |
Normal Reading Volts |
Measured Test Volts |
Voltage to chassis |
|
1A |
15 |
19 |
165 |
|
2A |
11 or 5 |
12 |
177 |
|
1B |
11.5 |
0 |
188 |
|
1C |
16.5 |
7.8 |
179 |
|
1D |
16.5 |
11.3 |
167 |
|
2B |
9.5 |
12.6 |
170 |
|
2B' |
20 |
19 |
180 |
|
2A' |
0 or 9.5 |
7.2 |
164 |
|
The usefulness of the
table is clearly demonstrated as immediately you can see that
the mixer valve isn't drawing any current and there are a couple
of anomalies w.r.t. the 1st IF amplifier and the output valve.
Another clue to problems are high readings for the RF and LF
amplifiers which to me look like leaky condensers. The low value
readings might be due to resistors gone high or leaky screen
decoupling condensers (or both).
I measured the HT as 203 volts
but yesterday it read only 188 volts. Yesterday the HT reservoir/smoothing
condenser block was running very warm and today slightly warm.
This will be a tightly packed paper condenser (dated 1943) which
may or may not recover from its leak... I should really disconnect
it because the similar block condenser in my R1155 failed catastrophically
all of a sudden... Because of the obvious leak the HT supply
will be reduced.
A second fault diagnosis is
also possible (and this is interesting) because you can see an
HT feed to all the valves (earlier readings). |
|
|
Clearly the silence from
the loudspeaker can be at least partly explained. Firstly there
is no signal into the IF amplifier because the mixer is duff
and secondly there is something amiss with the audio section.
I can also see that the heater voltage is down a little, reading
12.3 volts at the lamp terminals rather than something slightly
in excess of 12.6 volts. |
|
Test Point |
Normal Reading Volts |
Equivalent to Current mA |
Measured Test Volts |
Represents Current mA |
Diagnosis |
Voltage to chassis |
|
1A |
15 |
5.0 |
19 |
6.3 |
Fair |
165 |
|
2A |
11 or 5 |
3.7 or 1.7 |
12 |
4.0 |
OK |
177 |
|
1B |
11.5 |
3.8 |
0 |
0 |
Bad |
188 |
|
1C |
16.5 |
5.5 |
7.8 |
2.6 |
Poor |
179 |
|
1D |
16.5 |
5.5 |
11.3 |
3.8 |
Fair |
167 |
|
2B |
9.5 |
3.2 |
12.6 |
4.2 |
Fair |
170 |
|
2B' |
20 |
6.7 |
19 |
2.4 |
OK |
180 |
|
2A' |
0 or 9.5 |
0 or 3.2 |
7.2 |
6.3 |
Odd |
164 |
|
|
I'd like to see if all
the receiver features are functioning because this will give
me an idea of the length of the overhaul exercise. Once the major
anomalies in the measurement table above are cleared up, the
receiver should work to some extent on all bands using a long
wire aerial. The actual performance will depend primarily on
the previous owners expertise.
The R107 has three wavebands. |
|
Range 1 |
17.5MHz-7MHz |
17m-43m |
|
Range 2 |
7.25MHz-2.9MHz |
41m-103m |
|
Range 3 |
3MHz-1.2MHz |
100m-250m |
|
|
Here's the crcuit diagram
for the R107 split into front end, IF strip/output and power
supply. The aerial connections cater for either a long wire or
a dipole and the power supply can be driven from a range of mains
voltages or a 12 volt battery. The set has three chassis which
can theoretically be removed completely from the main chassis
after unsoldering groups of jumper wires on tag panels at the
rear of the set.
The official repair manual includes
lots of resistance and voltage measurements to aid fault finding,
however the chief type of fault which will be met some 76 years
after its manufacture will not be at all like those for which
the authors of the repair manual were familiar. Now, the majority
of problems will be due to ageing of resistors and condensers
(re-named capacitors at varying times from 1926 to after 1950) |
 |
 |
 |
|
Here are some slightly
better circuit diagrams
|
|
Below you can see the
first three duff resistors I found under the chassis. Most of
the R107 resistors are of the same style commonly found in receivers
dating back to the mid-1930s. They're carbon types and usually
oxidation at the junction of the wire ends and the carbon resistor
body is the reason for a drift in value. The photos below provide
the evidence...
The first is R8A which is supposed
to be an 80Kohm resistor.. clearly open circuit..hence the mixer
is turned off.
The second is the RF amplifier
screen resistor R18A, 25Kohm reading over 660Kohm
The third is R3A the 300ohm
cathode resistor for the RF amplifier reading 857ohms.
I also found several other resistors
wide of the mark, such as R7A marked 400ohm but reading about
850ohms. |
|
|
|
As there was no output to the loudspeaker,
I plugged in a pair of headphones. These worked, but with a very
loud hum. Clearly the block smoothing condenser is not just very
leaky, it's lost much of its capacity. The next step will be
to remove this and try a pair of new capacitors before stuffing
the original block. I'll also check continuity of the speaker
wiring and its on/off switch because I see the headphone output
is supposed to be derived from the same transformer. I checked
this and found the switch was in fact intermittent but after
waggling it lots of times it was much better. The block condenser
is different from the last one I looked at because it is 8uF+8uF
and has three wires protruding from a central grommet in its
base. Because of overcrowding in the power supply the condenser
is secured by four 6BA screws mating with captive threaded bushes.
Once detached I found the two sections measured 17uF and 17nF
on one meter and open circuit on my ESR meter. Resistance-wise
these measured 470Kohm and 2.2Mohm. Whatever were the readings
neither section behaved as a capacitor in-circuit. Fitting temporary
new 10uF electrolytics increased the HT from 188 to 270 volts
without much hum. The adjacent 4uF block condenser measured 4uF
and 0.12 ohms so seems to be fine. |
|
Test Point |
Normal Volts |
Measured Volts |
Voltage to chassis |
New Readings figures (xx) later |
Comment |
Final Volts |
Comment |
|
1A |
15 |
19 |
165 |
5.9 (52) |
Low |
22.7 |
High |
|
2A |
11 or 5 |
12 |
177 |
15.3 (15) |
High |
6.6 |
|
|
1B |
11.5 |
0 |
188 |
17.4 (17) |
High |
21.3 |
High |
|
1C |
16.5 |
7.8 |
179 |
20.4 (27) |
High |
13.7 |
|
|
1D |
16.5 |
11.3 |
167 |
18.4 (27) |
High |
15.8 |
|
|
2B |
9.5 |
12.6 |
170 |
17.6 (18.4) |
High |
12.5 |
High |
|
2B' |
20 |
19 |
180 |
24.8 (25.6) |
High |
27.6 |
High |
|
2A' |
0 or 9.5 |
7.2 |
164 |
9.7 (11.5) |
High |
0 |
|
|
|
The voltages at the test
points with new resistors and new smoothing condensers are now
all on the high side. An obvious reason for this is that the
increased HT voltage (from 188 to 270 volts) has increased the
leakage through each of the various decoupling condensers (one
of which accompanies the resistors feeding valve anodes, screens
and cathodes if a bias resistor is present) and adds to the anode
current. I measured the test panel voltages again after an hour
or so and some had increased indicating worsened leakges.
The next obvious thing to do
is to see if the set receives anything. I turned on my signal
generator and set it to 465KHz with amplitude modulation and
100mV output and discovered the IF amplifier was tuned to around
459KHz. The IF transformers are all tuned by trimmers, a lot
easier to twiddle than dust cores which can jam and break. The
initial plan (before adjusting overall IF response) was to judiciously
reset the trimmers. After some twiddling (you need to detach
three front panel blanking plates to access some trimmers) I
got a decent response at the correct frequency of 465KHz. One
or two trimmers didn't peak very well (indicating a problem)
but the other six were OK. I managed to get the sensitivity down
to 5mV before checking Range 1. Setting the dial to 10MHz I found
I could hear a test signal of around 90mV with the generator
connected to the leftmost aerial terminal and maybe 80mV when
connected across the dipole terminals. Connecting a long wire
fails to bring in any stations. Clearly the set is very very
deaf and I think the best way forward is now to test the valves.
Because the heater voltage is a bit low, any lack of emission
will be critical. Once the valves are in good order I'll continue
component testing. I found all the EF39s were OK but two EBC33s
had poor emission so I replaced these but without too much effect.
The receiver isn't quite as deaf as it was but still pretty poor.
After a couple of hours only Range 1 was working, pulling in
several strong signals, and with the BFO on I could hear some
40m CW. Next, I'll need to replace the decoupling condensers
and continue checking resistors and figure out why Ranges 2 and
3 are dead. |
|
This table shows the condition
of six R107 paper condensers chosen at random from around 20-30
used in the receiver. Resistance measurements showed up as over
1Mohm each but in a test set their condition was very poor. Using
a test voltage of 200 volts applied across the selected condenser
in series with a 100Kohm resistor a high impedance voltmeter
gave the following readings across the resistor. What this means
is when one of these is used to decouple for example a valve
screen fed by a 100Kohm resistor only about half the theoretical
screen voltage will be present at the valve pin. In addition
the decoupling quality will be reduced also. |
100nF = 0.1uF
|
Sample |
Marked Value |
Measured Value |
Condenser voltage |
Resistor voltage |
Leakage |
|
1 |
0.05uF |
9nF |
108v |
92 |
0.92mA |
|
2 |
0.05uF |
8nF |
138v |
62 |
0.62mA |
|
3 |
0.1uF |
163nF |
121v |
79 |
0.79mA |
|
4 |
0.1uF |
296nF |
27v |
173 |
1.73mA |
|
5 |
0.1uF |
250nF |
35v |
165 |
1.65mA |
|
6 |
0.1uF |
155nF |
120v |
80 |
0.8mA |
|
|
|
I fitted a sample quantity
of new capacitors and, once receiver performance had improved
to the point I could hear background hiss and several strongish
broadcast stations on the 7-17.5MHz band, I was surprised to
still find nothing on the two lower bands. Looking at the front-end
circuit diagram shows that wired across the wavechange switch
near to the oscillator valve are two resistors marked 25Kohm
and 80Kohm (being R4B and R8B). These measured as 32Kohm and
open circuit so I fitted new parts. Now all three wavebands are
OK. Presumably the local oscillator output voltage was deemed
by the designer to be excessive for frequencies below 7MHz when
the EBC33 anode voltage was optimum for the highest frequency
range (The local oscillator anode feed for Ranges 2 and 3 has
R4B in series with R8B but R8A is switched out for Range 1).
This explains why there are two voltages given for the front
panel test points. Presumably the performance of the EBC33 drops
off at frequencies over 7MHz so needs a higher anode voltage
to produce an optimum RF level to feed the mixer valve (suppressor
grid). In fact, if you study the oscillator circuit you'll see
that it's pretty odd. There's no self bias resistor in the cathode
and its heater is connected to its cathode. To establish an RF
voltage at the cathode for feeding the mixer there's an RF choke
shown in the heater connection to pin 2. The description of the
oscillator in the technical write-up is a bit strange describing
the choke as blocking RF from reaching other parts of the circuit.
All things considered... I suspect the final circuitry may have
been arrived at by trial and error?
It seems the reliability of
the 80Kohm carbon resistor in particular is pretty bad as both
R8A and R8B were open circuit. |
|
I decided to just fit
100nF x 500 volt chip capacitors in place of the waxed paper
types. Here's one example replacing a 1st IF amplifier decoupling
condenser. The lead is solder braid which is flexible to avoid
stressing the capacitor body and easy to solder.
Later I used a thin insulated
connecting wire taking care to use the same grounding points
for the new capacitors. |
 |
|
|
There are still about a dozen
or more old decoupling condensers to replace before the receiver
specification can be reliably checked. Also, whilst looking at
the oscillator circuitry I noticed that a few screws had been
drilled out at the bases of the coil cans. This is probably because
the chassis uses a crimped threaded bush rather than just tapped
metalwork and the crimp had failed. Maybe this was a standard
practice within the manufacturer's factory and the R107 mechanical
designer just continued the policy? Alas, the crimping is not
too reliable and the bushes can just spin round preventing screws
from being removed. The only recourse is to then drill out the
screw. I did note that there are official modifications relating
to these coils so maybe the problem was encountered when these
were being incorporated rather than the last owner drilling the
screws, maybe to track down the reason for dead wavebands?
Each time I've carried out a
few component changes I tested the receiver. The audio volume
is improving gradually but I've noticed that the IF selectivity
seems too sharp. As the IF amplifier relies heavily on decoupling
condensers I guess that once I've replaced them all the IF alignment
will need re-doing. The IF amplifiers are merely tweaked for
maximum output rather than stagger tuning (read the Information
Sheet by clicking below). I've also noticed the receiver is on
the verge of oscillation. Again, this problem will have to wait
to be sorted out until I've changed all the waxed paper condensers
and replaced all the resistors that are too wide of their marked
values.
I finished changing the paper
condensers. The receiver is a lot more lively but the IF alignment
has moved and I notice there's instability as the RF gain is
turned to maximum. I'll next do a final check on resistor values
then make sure all the screening is in place before realigning
the IF.
The next step was to measure
all the resistors. These were mostly high by 30% but this is
not too critical in valved receivers so I left all in place.
During the improvements to overall sensitivity I'd noticed instability
and once all the old condensers had been replaced the instability
was too much to ignore. I found it originated in three separate
areas, all associated with EF39 valves. Two (the mixer and the
1st IF amplifier) had lost most of their metallizing so I wrapped
baco-foil around these and wound bare wire around this connecting
it to pin 1 of the valve. This worked fine, comletely stopping
the instability around the front end. The third EF39 (2nd IF
amlifier) looked OK but the glass was slightly loose and a continuity
check between Pin 1 and the metallization showed the wire had
broken. The instability here was microphony... tapping the valve
produced a howl. Again, I wrapped baco-foil around the valve
and connected this to pin1. All the instability had now cleared
up.
IF realignment was now straightforward
because the instability had cleared up and the receiver sensitivity
had increased enormously, receiving broadcasts even with the
RF gain turned down to zero. The IF bandwidth switch seems to
do what it's designed to do whereas previously not much happened
when the switch was turned (probably bad condensers).
Any remaining problems? There
are lots of relatively minor problems. I noticed the outer knob
of the tuner doesn't rotate when the fine tuner is turned but
this didn't really matter (in fact this makes the slow motion
tuning extra smooth). AVC doesn't seem to work properly and turning
AVC on or off doesn't really change the volume of broadcasts
so that needs checking. My guess is there isn't enough voltage
being produced by the AVC diode. Maybe there's a bad condenser
or an open circuit resistor that I missed in circuit, or even
a bad diode in the EBC33?
I added a set of test panel
readings entitled "Final Volts" in the table above
(note that because all the old resistors are high in value these
monitor voltages will also be on the high side. For example..
take V1A, if the monitor resistor is 30% high it will be 3,900
ohms so a 15 volt reading across 3,000 ohms representing 5mA
anode current would now result in 19.5 volts)
During testing I found that
receiver gain varied by a significant amount if the centre IF
transformer was pushed slightly to one side so I temporarily
put an elastic band across this and the adjacent tansformer.
I spotted this because when gain was down the lower trimmer had
no effect on IF tuning. Just before I finished for the day both
the lower headphone jack sockets from where I pick up the output
for an external speaker failed mechanically (oddly, both 4BA
fixing screws securing the socket to the inner front panel seemed
to have sheared). Ordinarily, fixing this this wouldn't be much
bother but the weight of the R107 makes manipulating the chassis
very awkward.
Below is a view showing the
power supply module partly detached from the main chassis. The
R107 has a set of three modules designed to be removed from the
main chassis.. however, complete removal requires lots of wiring
to be unsoldered. By removing four 0BA brass screws, and the
knob from the headphone volume control, the power supply can
be moved back from the front panel sufficiently to examine the
securing screws for the headphone jack socket. It's held in place
by two long countersunk 2BA bolts secured by a nut and a locknut.
I found something very strange. Instead of proper countersunk
bolts two standard bolts had been fitted with their heads filed
down to allow the module to fit properly. Unfortunately, too
much metal had been removed and the modified heads had squeezed
through the panel allowing the double jack socket to have be
pushed backwards by a tight-fitting jack plug. The socket is
a bit special because it needs to be fitted in place behind a
front bezel and the metal panel in order for headphone jack to
be located correctly. |
 |
 |
These are the two bolts. Both
were very difficult to remove because what remained of the heads
had pulled into the jack socket body and jammed solid laterally
but turned freely preventing the nuts from being unscrewed.
Below the double jack socket back in
place at the upper right of the picture. I needed a cramp to
hold the heavy power supply in place while I refitted the four
securing screws (the heads are visible through holes in the chassis). |
|
 |
|
If the audio filter is selected
by its switch the receiver goes completely silent because there's
no filter fitted. At first I thought that whoever last had this
R107 removed the audio filter can, but reading one of the 1946
technical papers it seems that because these audio filters were
problematic, a number of R107 sets had been manufactured without
the filter fitted in order to expedite delivery. I recalled that
a similar type of filter in my R206 hadn't worked because the
paper condensers were in an absolutely awful state. There was
a general manufacturing problem with filters of this type because
they needed very high tolerance condensers which were not readily
available. Because the whole circuit is missing I'll need to
find a couple of suitable coils to make a new filter. Looking
at what remains of this area of circuitry, I see C10C couples
the anode of the AF amplifier to the grid of the output valve
and the reading across R6H (3000 ohms) is a trifle high (27.6
volts instead of 20 volts) indicating a possible leak in C10C
which is placing a couple of volts of forward bias on the valve
raising its anode current from 6.6mA to 9.2mA. I tested C10C
but this had absolutely no leakage so the valve itself may not
be within spec. Then the penny dropped... a drifted monitor resistor
was probably giving the high reading (ie. 38% increase in the
3000 ohm resistor would give the 27.6 volt reading from the correct
the anode current).
When I'd first started the overhaul
the audio level was very poor so I added a 100uF decoupling capacitor
across the cathode resistor of the audio output stage. This dramatically
increased the audio level but of course affected the negative
feedback so I'll need to step back and see if I should remove
the extra capacitor now that there's plenty of receiver gain.
Interestingly the output valve is a mere EBC33 whose anode is
designed to run at around 6mA and 250 volts which is about 1.2
watts. In class A the output would be around 0.5W. In the R206
there's an EL32 running at 3.5 watts output for a decent loudspeaker
volume. In practice I guess the use of the EBC33 may account
for some of the audio distortion I can hear on strong broadcasts?
The AVC switch was a bit intermittent
but after operating it several times it improved and I could
then see the AVC line correctly switching to ground with AVC
off. However the audio with AVC on was poor and got much worse
the stronger the signals. Measuring the AVC diode output at R2E
(250Kohm) showed minus 2.2 volts with the strongest broadcast
signal tuned in. I then noticed I'd missed a decoupling condenser
(C11K) which is located hidden away against the inner surface
of the front panel. I snipped the positive lead to C11K and the
AVC jumped to minus 5 volts. I then fitted a new 100nF chip capacitor
in its place and checked the AVC line. With the IF response set
to wide the AVC was now reading over minus 7 volts and the audio
was now clean and undistorted. Turning the AVC switch to off
resulted in receiver overload (as expected) with the RF gain
pot now controlling the audio level.
I checked some SSB signals on
40 metres and by adjusting the RF gain I got very good results
but I then noticed the slow motion drive for the BFO wasn't working
because of dried grease.
Alignment of the front end was
next on the agenda and the table below lists the RF coil details.
From the inductance values you can see that the oscillator is
always meant to be on the high side of the incoming signal. As
with all superheterodyne receivers it's sometimes difficult to
identify whether you're tuned to the "true" signal
or its image. A simple test is to use a local communications
receiver to listen to the R107 local oscillator. That way you
can confirm you're not aligning the set either on its image,
or as I disovered recently with my R208 the true frequency at
one end of the tuning range and the image at the other! |
|
Waveband (Range) |
RF Amp Grid |
RF Amp Anode |
Mixer Grid |
RF Oscillator |
Oscillator Tunes |
Image |
|
17.5MHz-7MHz (1) |
L1A=1.6uH |
L4A=1.6uH |
L4B=1.6uH |
L7A=1.5uH |
17.965MHz-7.465MHz |
18.43MHz-7.93MHz |
|
7.25MHz-2.9MHz (2) |
L2A=10.2uH |
L5A=10.2uH |
L5B=10.2uH |
L8A=8.4uH |
7.715MHz-3.365MHz |
8.18MHz-3.83MHz |
|
3MHz-1.2MHz (3) |
L3A=60.4uH |
L6A=60.4uH |
L6B=60.4uH |
L9A=40.5uH |
3.465MHz-1.645MHz |
3.93MHz-2.11MHz |
|
|
The coils are mounted
in four metal cans, oscillator (nearest the rear of the chassis),
mixer, RF amp and Antenna with the highest frequency coils nearest
the chassis to reduce wiring lengths. Coils are adjusted by 2BA
brass screws (held in place by nuts) with cores (these are said
to be brass rather than iron dust) and the trimmers are air-spaced
ceramic condensers. Adjustment of the coils is carried out by
tuning the set to the lowest marked frequency on each of the
three wavebands, then tuning to the highest marked frequency
and adjusting the trimmer for maximum output. By repeating this
exercise a few times the dial markings should correspond to received
signals. Years ago I aligned my R206 and found one particular
coil was too high in inductance so I had to change the iron slug
to a brass slug. Inserting a brass slug will reduce the coils
inductance and thus it was possible to track the receiver correctly.
I found this same problem with the R107. Several RF coils refused
to peak at the low end of the three bands. Before going further
I need to confirm I'm not adjusting the set to tune the image
(or possibly the coil cores have come adfrift from the adjusting
screws, which I've heard is a common R107 problem)
Before going further I need to resolve an intermittent
issue. I'd noticed that sometimes the receiver would suddenly
lose loads of gain. I found two points of note... if the third
IF transformer was tapped the gain would improve (or reduce)
and on one occasion whilst pondering the fact that the lower
trimmer had no effect I tapped the screwdriver in contact with
the trimmer and this had dramtically increased the audio output.
Clearly something is amiss. I removed the transformer and detached
the lid. |
 |
|
Two coils resonated at 465KHz by
their trimmers. Note... no metallic contact by any components
to the can. |
|
 |
 |
|
Above... four 6BA bolts
screwed into the two plates and countersunk into bakelite |
Above (top)... the two
metal plates used for grounding the IF can |
|
|
Looking at the circuit diagram
the IF circuit above is shown as T1C and in one (official...
see Part 2 (IF + Audio) link below
the circuit diagrams) schematic there's an error with a connection
missing ... which I corrected. The circuit diagram can be seen
correctly in the second example (shown above). The circuit has
a pair of tuned 465KHz coils inductively coupled but with a bandwidth
switch joining their earthy ends. No matter whether the bandwidth
setting is normal or narrow, both coils should tune using their
respective trimmers. Oddly, both technical documents I have exclude
C26 and C24 but as these are in place and both coils do tune
correctly some of the time, the intermittent fault must lie elsewhere.
The first step is to check the values of the trimmers C26L and
C26F and the three fixed condensers C24E, C23C and C24F. These
do seem OK. Also the replacements for C5K, C5L and C11F. There's
also C12A which is a tiny 2.2pF coupling condenser, external
to the cans, carrying the IF signal from T1B (which at 465KHz
has a very high impedance of something like 156Kohm).
Looking at the construction of the transformer
there's a puzzling fact. Grounding of the outer can is done through
two metal plates held in place at the base of the can by four
countersunk 6BA screws. Four further 6BA screws are used to connect
these plates to the receiver chassis. My first concern is a build
up of oxide between the plates and the can plus poor electrical
connection between the eight screws and the metalwork. Sure enough....
see below.
The use of a 2.2pF coupling condenser
is, in itself, probably a sound design technique as long as there
are no problems. However, if the outer metal can became isolated
from ground its action will change. One effect is its change
in capacitance to ground and it's likely to have a change in
value greater than 2.2pF so will shunt a varying proportion of
incoming RF to ground. What about the other three IF cans? Well,
these do not have a tiny 2.2pF condenser associated with them,
so an unearthed can will not have the same effect... maybe just
the odd crackle. |
|
|
|
Above.. inner surfaces
of the two metal plates that were supposed to provide earth bonding
between the end of the IF can and the chassis.
Not only were these badly oxidised
but the red paint used in the factory to confirm the various
parts are correctly installed has leaked between the metal surfaces
making conductivity worse. The plates are secured by small 6BA
countersunk screws whose heads are sunk into the bakelite sheet
holding the six solder tags, not the metal of the can. Ordinarily
screws with star washers would be used and these would make good
electrical contact between their threads and the metal can. As
it is, once oxidation began, electrical continuity would suffer
resulting in T1C can being electrically isolated (to varying
degrees) from the chassis. Earthing of the can was intermittent
and could be influenced by wobbling the can. I added an elastic
band but, although that helped intially, failed to eliminate
intermittent operation later. When the can was wobbled and grounded,
signals were around 20dB stronger than when the can was wobbled
and floating. Perhaps the change in gain was due to a change
in parasitic capacity between the ends of the 2.2pF condenser
(=136Kohm @ 465KHz) and chassis? This would shunt a different
amount of the RF signal from T1B to chassis. For the theorists...
if the parasitic capacitance varied due to the IF can not being
grounded by say 100pF (=3.4Kohm @ 465KHz) then the attenuation
of the incoming IF (voltage) signal would be 34dB. Not only that
but the parasitic capacity might prevent the lower coil from
being resonated (which, apart from a distinct loss in loudspeaker
volume, is exactly the symptom).
Cleaning the plates with a fine
emery paper showed that the metal wasn't even flat. When the
plates were cut (back in 1944) the metal deformed resulting in
very little metallic contact between the plates and the inside
surface of the can especially if a plate with a concave surface
faced the can. In fact one plate was in this position with the
other having its convex surface facing the can. I also cleaned
away oxide from the inner surface of the can with a small rotary
wire brush before refitting the refurbished plates. This cleared
up the annoying intermittent at T1C.
Below a picture showing the
overhauled R107 complete with a full complement of chip capacitors
in place of the original 30 or so wax covered 0.1uF (C11)and
0.05uF (C5) condensers. |
 |
|
I reset the IF amplifier
to precisely 465KHz using my spectrum analyser. At first, tuning
the first transformers appeared to have no effect on the observed
465KHz signal until I realised the RF gain setting set to maximum
was causing the final stage to oscillate at exactly 465KHz. Backing
off the RF gain sorted things out and I was able to adjust the
IF response approximately equally balanced about 465KHz. I found
that either increasing or deceasing the RF gain setting made
the peak IF response move higher or lower in frequency by a KHz
or two. I also found the amplitude of the local oscillator varied
considerably across the two shortwave bands. This was probably
because I was measuring the amplitude at the control grid of
the mixer, which is affected by its tuning condenser, so I'll
repeat the measurements at the suppressor grid of the mixer where
the output of the local oscillator is injected. Sure enough,
when I monitored the local oscillator directly it remained ostensibly
constant across each waveband.
Later, I swapped the final IF
amplifier valve with a ropey metallising for that serving the
first RF amplifier with perfect metallising. The instability
that occurred with maximum RF gain cleared up.... then not long
afterwards the audio at the loudspeaker dropped considerably.
I put my finger on the grid of the output valve with very little
effect so I fitted an EBC33 in place of the old AR21. The audio
reappeared and normal service was resumed. As a final check I
measred the HT voltage. This had risen considerably from when
I'd first tackled the receiver and was now reading 300 volts
primarily due to getting rid of 30 leaky condensers. When I first
turned on the R107 the HT was 203 volts, then dropping to 188
volts when the smoothing condensers failed... then after some
decoupling condensers had been swapped it had risen to 270 volts,
then after some more, 288 volts and finally 300 volts after all
wax condensers had been swapped.
I feel inclined to now close
the overhaul exercise and move onto something new even though
there are still a few things to clear up. |
|
|
|
|
