R107 MkI

Just arrived for an overhaul is this early example of the R107 receiver.

Note that what you read below describes various problems as they were identified and experiments made to determine their rectification. Sometimes it took quite a long time to arrive at the true causes and some ideas proved wrong before effecting repairs. Some of the parts concerned are now, like myself, just about 80 years old and obviously will not be perfect. Another point to remember is that this example was refurbished by MoD, probably 70 years ago and it's clear that some errors were made in the process.

I first located a lead with an R107 plug, tested this and then checked across the pins on the receiver which proved to be open. The first step was to clean the fuse located behind the mains transformer which had a verdigris coating (this is used for 12 volt DC operation only) then waggle the on/off switch which had duly followed the usual practice of being lazy. Plugging in a mains supply brought up the valve heaters including the rectifier but no sign of life from the set. At the front is a test panel which includes an HT point which slowly rose to 58 volts then dropped back to 56 volts as the valves warmed up. Waggling various controls showed no signs of crackling so time to delve into the innards. I could perhaps take a short cut and connect an external HT supply to the receiver, but before I try this I'll investigate further.

It was at this point I discovered that the HT voltage must be top secret as it wasn't given anywhere in the manual (ZA3220). When you consider that the key to fault finding is the set of voltage readings at the test panel you'd think that the HT value is rather important as it will influence the quoted readings. I eventually found a figure of 250 volts quoted in brackets in a document labelled 3STC Aug '42 VJN. It was added to advise that if all the test readings were bad the HT voltage might be the cause. Many of the power supply components are not exactly accessible but to compensate there are several "tag boards" labelled A, B, C and D which carry important connections, including on Pin 1 the HT output from the HT smoothing condenser. B1, C1 and D1 should therefore have 250 volts present.

You'll note the test point labelling cleverly relates directly to the valves. Measurements are made between HT (+) and the test point.

Circuit	Test Point	Reading volts
V1A	1A	15.0
V2A	2A	11.0 or 5.0
V2A'	2A'	9.5 or 0
V1B	1B	11.5
V1C	1C	16.5
V1D	1D	16.5
V2B	2B	9.5
V2B'	2B'	20
HT	+	250 wrt chassis

The four tag boards left to right A, B, C and D.

A lot of effort has been made to help diagnose faults.

The tags aren't marked but their circuit designations are numbered 1 to 12 top left to bottom right.

As the red lamp wasn't lit I checked and found the bulb was missing. The receiver can be operated from a 12 volt battery so valve heaters are connected in series-parallel. The dial lamps, like valve heaters are 6.3V and are connected in series but the bulb under the red cover is fed via a resistor. The bulb ratings are also top secret as they're excluded from the parts lists. As long as the dial lamps are matched (because two of the three are wired in series) I guess their rating isn't too important.

Having failed to make sense of the connections to the 6X5 rectifier (which was lit OK) I decided to connect a DC supply in place of the set's HT at pin 1 of tag board D and chassis ground.

There wasn't any sign of trouble as I wound up the voltage with 250 volts drawing some 35 to 40mA but nothing from the speaker. I plugged in headphones and was rewarded with a slight hum and, by adjusting various controls and connecting a long wire aerial, turning to Range 3, I was able to tune a strong medium wave broadcast. After fiddling with controls I found the speaker switch was completely unresponsive so I added a jumper wire across the switch and Smooth Radio was suddenly loud and clear. The BFO responded to its switch but was miles off frequency.

So what's the problem? There are a few possibilities including an open circuit HT secondary winding, a bad 6X5 or a bad component buried in the power supply area. My Bernards Valve guide gives 6X5 anodes at pins 3 and 5, heater at pins 2 and 7, and cathode at pin 8 but a quick continuity test failed to prove connections to these to be right.

It didn't take too long to figure out what was wrong and it's something pretty weird. In fact weird is becoming quite a frequent thing with radios visiting my workshop. You can see below what looks to me to be wiring that looks different to other stuff. Those two blue wires are connected to the 6X5 rectifier at pins 4 and 5. The black wires (plus that faded black wire) connect to chassis. Resistance readings across pins 4 and 5 and either one to chassis match the secondary winding (around 320 ohms and two readings of 160 ohms) so it appears the transformer is OK.

I can only explain what I found by accusing a dabbler rewiring this area and mistaking the connections to the valve socket. If pins 3 and 5 rather than pins 4 and 5 had been soldered the HT would have probably been OK. According to the circuit diagram a single condenser viz. C20A should be wired across the secondary not a pair connected to ground. Those two blue wires, plus the orange and grey are very close to metalwork and should perhaps have been routed at the near edge where there's more room.

I should really delve under the power supply chassis and sort out the mess...

Circuit	Test Point	Correct volts	Actual volts
V1A	1A	15.0	7.4
V2A	2A	11.0 or 5.0	5.3
V2A'	2A'	9.5 or 0	0
V1B	1B	11.5	13
V1C	1C	16.5	14.6
V1D	1D	16.5	15
V2B	2B	9.5	10.7
V2B'	2B'	20	24
HT	+	250 wrt chassis	250 wrt chassis

With exactly 250 volts from an external supply I measured the test points (left).

Not perfect but at least Smooth Radio is booming from the loudspeaker and switching from AVC to OFF indicates at least there's partial AVC action.

After rewiring the rectifier and finding the set failed to work, the last owner must have got so fed up he just dumped it....

Rather than stripping the PSU to access the 6X5 pins.. why don't I just insert a shorting wire from pin 4 to pin 5 after checking the latter isn't wired to anything else?

I tried shorting pins 3 and 4 and the HT increased to 90 volts so I rechecked continuities.... about zero ohms from the bottom leg of the 250-0-250 winding to pin 4 of the 6X5 and about 1Mohm from Pin 4 to pin 8. About 350 ohms across the full winding and a good ground point.

Powered, the winding supplies about 600 volts AC to my meter.

Either there's a wiring error or possibly small resistors to act as fuses have been added as shown here.

I'll need to dismantle the PSU to find out... not too easy a job as the picture below suggests? Those screws look suspicious??

In fact the 6X5 valveholder is to the right of this metalwork underneath an RFC (just visible under the diagonal strengthening bar) marked as L17A on the PSU circuit diagram. It's difficult to see the wiring but I decided on a simpler approach by soldering a couple of wires to the 6X5 pins and connecting these via tiny 51 ohm resistors to act as fuses to the 250-0-250 winding. This worked and gave me about 170 volts HT. A little lower than expected, maybe due to a bad reservoir or smoothing condenser and listening on headphones indicated there was some hum.

I checked the voltages across the resistors and measured 700mV (14mA) and 100mV (2mA) to pins 3 and 5 respectively. Apparently the 6X5 is not perfect and will help explain the low HT. An external HT supply drew around 40mA at 250 volts.

Now that the set is working I'll see if rewiring the 6X5 base is a practical proposition and look for a replacement valve or even cheat and fit a couple of silicon diodes... in fact this would be a very simple mod as all the connections are on top of the transformer.

I fitted a pair of BY299 diodes (800V x 2A) as these were robust and left my two 51 ohm (fuse) series resistors in place. No provision for surge as I'd previously checked the HT line by pulling it up to over 300 volts without a problem. The HT was around 270 volts at switch-on dropping to 235 volts with the valves warmed up. The voltages across the resistors were equal and measured 750mV RMS (15mA each leg). Plenty of volume for Smooth Radio and Absolute Radio but the set was very deaf on the two shortwave ranges with just a faint squeak of 80m SSB. Next I'll recheck the test panel voltages, start alignment, beginning with the IF amplifier which runs at 465KHz and fix the BFO frequency. Many years ago I heard the R107 BFO was so potent it could be heard miles away by ships at sea. I wonder if this was true.

I'll cut the two blue wires, isolating whatever weird circuit is present under the chassis then tidy up the wiring to allow the 6X4 to be refitted as a dummy then the transformer cover can be replaced.

Circuit	Test Point	Correct volts	Actual volts 1	Actual volts 2
V1A	1A	15.0	7.4	13.8
V2A	2A	11.0 or 5.0	5.3	4.7
V2A'	2A'	9.5 or 0	0	0
V1B	1B	11.5	13	11.5
V1C	1C	16.5	14.6	16.4
V1D	1D	16.5	15	16.5
V2B	2B	9.5	10.7	9.2
V2B'	2B'	20	24	20.3
HT	+	250 wrt chassis	250 wrt chassis	216 wrt chassis

With the PSU using a pair of BY299 diodes instead of the 6X5 the test points are reading slightly differently and with no anomalies. The IF amplifier was pretty close to optimum at 465KHz but Range 3 needed tweaking at its HF end. Ranges 1 and 2 are still deaf probably due to misalignment.

In order to make RF adjustments, using an audio wattmeter across the loudspeaker, I fitted a large smoothing capacitor to eliminate hum.

One design feature to help servicing the R107 is the provision of holes in the front panel. These are covered by removable plates (which you can see in the picture at the top of this page), allowing access to the IF transformer trimmer condensers. Some similar wartime receivers need to be adjusted using a short screwdriver in a cramped space.

A stroke of luck for a change was the BFO whose slow motion drive wasn't fastened to the tuning shaft making tuning rather hit and miss. Once I'd positioned the drive and tightened the loose grub screw the BFO tuned across 465KHz.

Front end alignment is quite easy as the trimmers and coil adjustments are nicely grouped together in a logical fashion. Front to back.. Aerial, bandpass input, bandpass output and oscillator. Top to bottom Range 3, Range 2 and Range 1.

Incidentally, to align the IF amplifier I used my new TinySA. Coupled via a 20dB attenuator and a BNC flying lead with croc clips. It worked perfectly. Initially set to 465KHz AM at 1KHz I reduced the audio to 500Hz as my ears got fed up with the former. It was running at its default of -7dBm and via the attenuator, used to protect it against any stray voltage, I initially clipped it to the top of V1C the first IF amplifier then for checking Range 3 at 3MHz and 1.2MHz to the aerial connector. I then checked the results with my HP8640B and found it was set up perfectly. Later I'll use my DSA815TG because merely tweaking the IF trimmers rarely produces perfect IF response. To check the BFO frequency a fully tunable signal source is far better than a digital type, and when confirming you have the true and not an image or spurious response a properly tuneable signal generator is far better. For example sweeping across a particular shortwave frequency might reveal dozens of spurii if the amplitude of the signal is too strong. Sometime (when I've retired) I'll do an experiment to reveal this...

Below the front end adjustments... you follow the general rule to first check the mechanical tuning setup, then trim condensers at the HF end and twiddle coils at the LF end until the tuning range exactly fills the required band and you're happy with the flatness of the background noise level. Bear in mind that in this day and age the background noise level picked up by an aerial may fluctuate considerably. I'm sorry that many listeners will never have the opportunity afforded, in the 1950s for example, when with headphones you could (metaphorically) hear a pin drop in New Zealand on 40m.

A few things to do are.. remove and clean the loudspeaker switch, detach the HT connection from the rear of the muting plug. A better idea might be to remove the muting plug and fit a new IEC mains plug in its place, using the wires currently going to the two pin mains plug. The reason being that no safety earth is provided with the existing plug. The loudspeaker switch one detached very reluctantly woke up after applying switch cleaner and jiggling it a few hundred times (due to a rusty return spring). I suppose I should also check valve top caps for any positive DC voltage...

Next I tackled the slow motion drive as this felt rough and slipped in places. I decided to remove the whole thing and take it apart. Not too difficult but tricky reassembling because you need to know precisely which parts to fit first. The method is to assemble the bearings before sliding into the outer sleeve, taking care to get the outer parts the right way round. The one with the locating screw slides in first with the notch last. The outer three balls are fitted first then the inner three balls slip into the gap at the end with the lever and note that they only fit in one sequence. The inner three 6BA screws need to be quite tight and the outer ones adjusted for drive smoothness.

The rough feel was because grease had migrated out of the bearing surfaces leaving metal-metal contact which had worn the bearing surfaces. This was much worse at the inner bearing.

Left, the slow motion drive removed from the receiver and below the component parts. The two bearing shell pairs are each connected by a set of three 6BA screws and linked together within the outer cylinder to which the outer shells are held in place. One by a tiny screw and the other by a pin engaging with a slot in the cylinder edge as you can see on the left. The inner bearing is driven by a shaft coupled to the outer bearing and just enough tension is applied to the sets of screws to overcome friction and rotate the tuning condenser. In this design the cylinder carries the direct drive knob which remains stationary as the inner slow motion knob is turned. The dial pointer is clamped to the thicker output shaft which also carries a pin to limit travel to 180 degrees.

I'd noticed that there was a lot of play in the pointer both before and after I'd put in new grease. The cause is the copper disk under the three screws tends to move backwards and forwards as drive is applied. I'm unsure how to fix this other than perhaps supergluing the disk in place. Maybe star washers between the copper disk and the bearing would stop movement? The root cause is turning force from the inner bearing pushes the 6BA screws across the hole width in the outer bearing because the holes in the bearing are larger than 6BA. Even though the movement is quite tiny it's amplified at the dial pointer causing backlash. The solution was simple. The three screws at the rear of the drive were too tight. Slackening these resulted in well nigh perfect tuning and I was able to listen to SSB on 40m, however, turning on the set this afternoon resulted in the pair of HT fuse resistors burning up... the last time this sort of problem occurred it was a large grey condenser. These are made in the same way as older waxed tubular versions wound with paper and aluminium foil. I guess a better solution is to fit a proper fuse in the centre tap ground connection and maybe a pair NTC thermistors in place of the fuse resistors?

When I investigated the reason for the burnt fuse resistors I found the main circuit was OK but one of the diodes had failed short-circuit. I then realised the full wave rectifier was driven by an HT winding that supplies 250 volts DC when loaded with say 50mA but when unloaded I measured 300 volts RMS whose peak is 424 volts and across the whole winding double this at about 848 volts. Bearing in mind these figures are probably even a little low because of mains fluctuations and noise, the use of an 800 volt diode is definitely a bad idea. I swapped the pair of BY299 to BY255 which are rated at 1200 volts and also fitted a pair of 120 ohm eighth watt fuse resistors in the diode cathode legs which now have 1.7 volts DC across each. This represents only about 15mA or 30mA HT drain. The HT voltage at the smoothing condenser measured at 210 volts with test points below. These (Actual volts 3) are now very close to optimum.

Circuit	Test Point	Correct volts	Actual volts 1	Actual volts 2	Actual volts 3	Final Test
V1A	1A	15.0	7.4	13.8	12.9	6.2
V2A	2A	11.0 or 5.0	5.3	4.7	11.0 and 5	5.7
V2A'	2A'	9.5 or 0	0	0	8.2 and 0	0
V1B	1B	11.5	13	11.5	11.3	13.1
V1C	1C	16.5	14.6	16.4	15.6	14.4
V1D	1D	16.5	15	16.5	15.8	14.3
V2B	2B	9.5	10.7	9.2	8.9	11.5
V2B'	2B'	20	24	20.3	19.75	26.3
HT	+	250 wrt chassis	250 wrt chassis	216 wrt chassis	210 wrt chassis	263 wrt chassis

As you can see in the picture above, all the components are readily accessible apart from those under the diagonal re-enforcing bar but as HT current is only relatively small and the various test points now show normal conditions I don't believe wholesale swapping of parts is necessary. I measured most of the resistors and found all were high in value. Typically 20% high. During initial testing I'd noticed swinging the HT from 250 volts down to less than 100 volts had no real effect on the audio output and that being so resistor values being universally high doesn't matter. Keeping the colourful originals therefore is a sound idea. Alas, there's an exception as for some odd reason the panel lamp doesn't come on because the 100 ohm series resistor in the manual is actually 150 ohms and this measured 180 ohms. The latter requires a 6.3 volt lamp rated at say 50mA. Rather than fit a standard 200mA lamp and change the resistor (R1B) to 30 ohms 1W I fitted a red LED soldered to an old lamp base. This works OK and runs with 4.6 volts RMS across it and, being a diode, requires no external rectifier. My guess is the panel lamp was too bright and a mod was issued to raise the series resistor by 50% to dim it.

Replacing a valve rectifier with silicon diodes in an old receiver is not as straightforward as it first seems. A valve rectifier warms up much the same as the receiver's valves so a potential voltage surge isn't that important. Diodes on the other hand will immediately place their rectified voltage at their cathodes and this voltage can be very much higher than the receiver is used to accommodating.

WW2 and 1950s/1960s equipment might have used an electro-mechanical time delay device to prevent damage from a voltage surge but is there a simpler method using modern components?

The above sketches show the basic R107 HT power supply, the silicon diode version I'm using and a theoretical circuit using thermistors.

My BY255 diode solution is very compact and uses a very high value smoothing capacitor to remove hum and as a side effect delays the HT from establishing across the valve circuit. This delay is due to the fact that the capacitor is initially seen as a short-circuit fed through resistance made up from a number of elements. The HT transformer winding has a resistance of about 330 ohms, the RFC about 10 ohms, the LFC about 100 ohms and a pair of small resistors for fusing open in the event of HT failure each of 120 ohms. If the circuit resistance is say 600 ohms the initial surge, given an open circuit HT of 350 volts will be around 0.6 amp reducing rapidly as the capacitor charges and the valves warm up. Because the target voltage at the R107 circuit is say 250 volts (and it may not even reach this) we're looking at a time constant of only say 20% of RC so our 600 ohms and 470uF is effectively a couple of seconds and by this time the valves will be warming up and drawing current thus limiting the voltage surge.

A more elegant solution might be shown in the third sketch above. T1 and T2 are thermistors having a cold resistance of say 100 ohms dropping to say 1 ohm when hot. T3 is a PTC thermistor having a low value when cold rising to a very high value when hot. The circuit looks like a good solution but there's a technical difficulty with T3. Thermistors are designed to limit current surge not to limit voltage surge so are primarily low resistance devices. There are some that have a quite high resistance but have only a small current handling capability. They also have a limited voltage capability which means faced with say 350 volts of HT we need to use lots of these devices in series. Current handling being fairly low means we may need to use two or more of these series-connected chains in parallel.

Another technical difficulty is to arrange the time delay for resistance changes to take place. Thermistors are quite small and need to react swiftly in order to minimise heat dissipation, whereas our valve circuit would like maybe 10 seconds of delay. So there's a conflict which may lead to the thermistors getting too hot.

Another option is this which uses T1, T2 and T3 as they were intended, to handle current surge. It also helps to arrange a calculable time delay for the establishment of HT. A downside is the power handling capacity of T3 as well as its current and voltage handling capacity. In fact such a device as required for T3 may not exist.

Can T3 be constructed from available devices? In the previous circuit T3 is a PTC type but here it's an NTC type of thermistor. Perhaps an NTCLE100E3102JB0 would work? Will it just go POP at switch on or maybe BANG or would a suitable choice for T1 and T2 help? I'll invite readers to work this out for themselves.

Having been diverted I got back to R107 alignment. The IF response curve was OK but slightly shifted HF. I corrected this initially in narrow then checked it on wide with the following results. I injected the signal at the top cap of V1B and measured the response at the detector diode in V2F.

I'm using my high voltage probe (to safeguard my DSA815TG) which has an inderterminate gain so ignore the power levels in the traces. The relative values are fine but the actual numbers are irrelevant.

Narrow response.

Wide response.

All the IF trimmers peaked nicely except one.. this was the lower trimmer in T1C which suggests something wrong. Overall gain is fair with Range 3 pretty good but the shortwave ranges are a bit deaf so I'm guessing there's a bad component in the connections between V1C and V2F. Oops.. yet another senior moment... I asked the WS19 Forum if anyone had any advice concerning T1C and was informed I'd published the very reason for the problem and here it is below **. The IF cans rely solely on earthing on a couple of metal bars being in contact with the inside surface of the IF can and in the earlier example these bars were not only badly tarnished but painted. It makes me wonder if a large number of R107s have exactly the same fault as a result perhaps of the same person doing the assembly. My guess is, if that were so who suspects the loss of gain is due to this? Maybe umpteen examples are in use with a deaf IF stage with owners blaming bad band conditions? Rather than strip out the can, a rather tedious exercise, maybe a small hole drilled through the centre of the base and a pair of self-tapping screws used to renew contact with the can? Click either picture to see the original refurbishing article.

** In this case Sod's Law ruled that the fault should be due to something completely different

A practical solution is to fit a couple of screws with solder tags and provide a new grounding point for the IF can.

Before I tackled this I decided to check everything that was supposed to be grounded and everything tested fine so that bad earth on the previous R107 displayed the same fault but this time the problem lies elsewhere.. there's no option but to remove the IF can. This is very tricky as some screws are hidden and most are loctited but after a bit of ingenuity I managed. My original thought was a bad micamould condenser and two are connected to the lower untuning coil.

The circuit diagram on the R107 lid and in the manual have an error. The same error, which is a missing connection, associated with T1C.

Is it Sod's law that dictates this very schematic error points to the fault I'm experiencing? Two condensers joined in series are designed to tune the lower coil in conjunction with the trimmer. I looked at them.. they're bolted together to the base of the IF can with two 6BA screws, loctited and each with a locknut. I removed the screws and cut off the condensers from the trimmer. One marked 0.001uF measured 989pF and the second marked 400pF measuring 388pF. Pretty good BUT they fell apart when I measured them because they should have been.. but weren't soldered together! Was it the very same bloke that made a mistake wiring that 6X5 rectifier that forgot to solder the condensers?

I think I'll assume this...

Above, T1C with the two micamoulds removed (right). They're held together because one leg is just bent around the other but not soldered! The lower one (400pF) in series with the upper (0.001uF) were in parallel with the trimmer with the junction soldered to the coil and the brown covered wire at bottom left.

Look closely at these two drawings. The one on the left is the same as the circuit diagram marked on the inside of the R107 front lid and you can see an error..the right hand diagram is consistent with the assembly inside the IF can.

I noticed a fault when first checking the RF trimmers at the can carrying the anode coils for the RF amplifier V1A. The gain for Range 2 infrequently increased dramatically (reaching normal level) when touching the can making me suspect a bad solder joint or maybe a bad earthing point. It didn't really happen with Range 3. The proper gain disappeared as suddenly as it had appeared and I couldn't repeat the gain change by tapping the can.

After detaching the cans from the RF stage coils (eg L4A and L4B) and the paxolin board holding the three trimmers and coils L4A/L5A/L6A I found the centre coil and its trimmer very sensitive to touch ***. Inputting -90dBm of modulated AM at 3MHz into the aerial on Range 3, I could comfortably hear the tone. I then tuned to Range 2 and tuned to 3MHz. The input had to be raised to -40dBm to hear the same level tone and by jiggling L5A the tone would increase by 30dB. I found a pair of plastic covered wires stretched tightly across the edge of the RF chassis, one of which might have been shorting but I'm not 100% convinced as the coil still seemed intermittently sensitive to touch. I might detach the coil and see if there's a poorly soldered joint. Because of the cramped layout it's not possible to see connections to the coils and trimmers. In the meantime I tried my homebrew noise source to try and evaluate what's going on. This is really easy to use compared with a signal generator and I was able to roughly equalise the RF response over Ranges 1 and 3.

At this point I decided to deal with the red loctite used on the coil adjusters. I found the one way to get rid of this was to hold a scalpel to the thread and rotate the screw. I also used a 2BA die. First screw the nut downwards against the top of the coil, withdrawing the screw as far as possible then carefully position the die and by hand rotate it down the brass thread.

*** This turned out to be a faulty wavechange switch. The wire from the Range 2 coil wasn't being selected at the switch.

This is the output from the noise source scanned from zero to 30MHz.

A view of the RF amplifier and Mixer grid coils which form a bandpass filter for each of the three wavebands.

The trimmers and coil adjusters are very well placed and logically laid out for alignment. The covers for both coil sets have been detached.

I found Range 2 to be very odd. I can peak the noise at the LF end and at the HF end but tuning across the range shows a huge dead spot about a third down from the HF end. At first I thought this is due to inadvertent switching from true to image reception eg. the LF end and the HF end are tuned separately to the true and image signals. I have come across this problem before which I think was due to a bad coil or too much or too little stray capacitance. I confirmed this wasn't the case by using my spectrum analyser which method I'll describe below.

The R107 coils are unusual, according to the handbook, using eddy currents to modify their inductance. It's said that each coil has a copper disk rather than a dust-iron core and the recommendation is not to insert the disk too far into the coil for fear of losing gain. This effect is demonstrable using the noise source.

In a similar way to checking IF response you can check a receiver's RF alignment. To do this connect the tracking generator to the aerial socket whist monitoring the grid of the frequency changer. See here which is a test of the RF25 unit. Instead of a complete scan as in the case of a fixed frequency IF amplifier you see two signals for a correctly aligned receiver. These are the response at the frequency to which the dial is set plus the local oscillator (LO). The R107 has the LO higher by 465KHz than the tuned frequency so the narrow spike seen on the right is the LO. Two useful checks can be undertaken. First, by tuning end to end, you can readily see any amplitude change in the LO. Ideally the amplitude of the spike shouldn't vary too much. The shape of the RF response is very interesting as it should smoothly rise in amplitude before plunging down with much the same slope. In the R107 there are three sets of tuned circuits which help keep the curve shape relatively narrow. The narrower the curve and the higher is the top of curve the better is the attenuation of the LO and the image. Set the frequency to the high frequency end of the dial and use the trimmers to set the shape of the curve, then retune to the lowest frequency and adjust the coils for best curve shape again. Then repeat a few times (as with standard alignment). Quite tiny trimmer changes can make significant curve changes.

With care you can also check dial settings by monitoring then setting the LO frequency to match the dial (knocking off 465KHz).

What did I see? Well I was able to perfectly align Range 1 and 3. Notable was the effect of altering the coils. The normal method of using a fixed frequency and an audio wattmeter has coil adjustment pretty vague and woolly compared with the SA method where even small coil changes were quite apparent. So Ranges 1 and 3 checked out well; however Range 2 gave the same result as shown above with a pronounced dip somewhere between 5 and 6MHz. I've shown 100% as the norm but Range 2 develops a lot less response to the noise generator than Ranges 1 and 3. I can now confirm though that the notch is nothing to do with a crossover from true to image signals, but instead a resonance effect. Somewhere in the (faulty) front end is a circuit rejecting something like 5MHz. Because the low gain can correct itself intermittently it can't be a wiring error but must be an open or short circuit or faulty component which is introducing a series connected coil and condenser across the signal path or a parallel tuned circuit in series with the signal path.. ie. a rejector circuit.

Springing to mind is something like an IF rejector tuned circuit designed to deafen a receiver to a signal close to the IF. The aim of that is to place a narrow band short circuit across the aerial input circuit.

I studied the circuit diagram and noticed the wavechange switch not only selected the desired range coils but also shorted out coils not being used. That being so if I looked at the trimmer stators (which are easy to get at and the coils not so) to identify what I confidently exected to be either an undesired short or open circuit within the set of RF anode or mixer grid coils. I checked the resistance of each stator to ground. The mixer grid stators were a solid 6Kohm to ground (in fact the rotors are wired together to ground for the first and third coil sets and decoupling condenser C5C for the second set) whatever position was the wavechange switch but the anode coil stators were very odd. These were either 50Kohm or 500Kohm to ground and these resistances seemed to be completely random at any setting of the wavechange switch. In fact turning the switch sometimes gave 50K and a moment later in the same setting 500K. I donned my magnifying goggles, illuminated the switch wafer with a torch and immediately spotted a switch tab lying on top of the pair of contacts instead of between them. The adjacent contacts were skewed at 45 degrees and all the switch tabs were the same. None were feeding into their contact sets, and mostly either resting on the upper contact or with the tiniest space away. The whole rotating part of the wafer was very slightly out of place, just far enough for its tabs to miss threading between the contact sets. It was a simple enough task to gently pry the contact sets into line with the tabs.

What was happening then is shown in the sketches below. RF energy was being coupled to the mixer Range 2 coil from a combination of wiring plus unshorted coils for Ranges 1 and 3. The unselected coil for Range 2 was absorbing RF energy at the natural frequency of coil and trimmer (without the tuning condenser). Range 3 was working OK with its switch contacts in good order and Range 1 not too bad but in the Range 2 position of the wavechange switch the combination of faults matched those shown below. Once corrected the Range 2 coil was properly selected and overall gain dramatically increased. At this point I hadn't replaced the two screening cans.

The sets of coils are housed in metal cans and the wavechange switch is normally protected with a heavy metal cover.

I initially thought the R107 had been placed on something that had fouled the wavechange switch, stressing the wafer, but because the cover was in place the switch was either poorly manufactured or damaged by a previous owner. When I removed the metal cover I did notice the screws were a mixed bunch and not original.

Checking alignment showed that hardly anything now lined up other than the local oscillator and a fresh re-alignment is required. I also need to identify why the audio output is relatively low (apart from the LF end of Range 3 which covers some local medium wave broadcasts) although I still have an audio wattmeter, which accounts for some loss, connected across the loudspeaker..

Realignment went ahead with the screening cans back in place but the notch had re-appeared around 5MHz in Range 2. Testing is not easy because of the way the receiver is put together but I removed the two screening cans and the paxolin top from the anode coils again. There are a few possible reasons for the notch and I decided to see if I could rule out a bad coil. I connected my multitester to each coil in turn and measured its inductance starting with RF amplifier anode coils. These measured 50uH and 10uH plus "Low inductance" (probably 1.5uH?). Each was checked with the wavechange switch in its correct position and should have been shorted out in the other two settings, however in Range 3 the 10uH coil failed to be shorted so there's still a remaing problem with the wavechange switch. I used a different method to test the mixer coils (as I'd left its paxolin panel in place) by measuring the inductance at the tuning condenser. The coils measured exactly as before so the RF amp anode and mixer coils are OK. The table below shows measurements across the RF amplifier anode and mixer coils. ** missing short.

Tuning Range	Range 1(anode)	Range 2(anode)	Range 3(anode)	Range 1(mixer)	Range 2(mixer)	Range 3(mixer)
Range 1	290mohm	140mohm	140mohm	290mohm	250mohm	250mohm
Range 2	600mohm	10uH	10uH **	250mohm	10uH	250mohm
Range 3	290mohm	290mohm	50uH	250mohm	250mohm	50uH

This second table takes the possible tuning range of the variable condenser plus strays plus trimmer in pF then calculates the natural frequency of an unshorted coil plus trimmer and you can readily see the Range 3 coil left unshorted would absorb 5MHz from Range 2 (exactly the problem). The components list gives C14A as 300pF x 4 which lines up perfectly with the table below given a trimmer of say 30pF plus strays including the tuning condenser minimum value. Coil values were those measured plus a guess at that for Range 1. If this analysis is correct shorting the Range 3 coil with Range 2 selected would remove the notch.

	COIL	CAP	FREQ	COIL	CAP	FREQ	CAP	FREQ
		Tuning	Low end		Tuning	High end	Trimmer	Resonance
Range 3	50uH	350pF	1.2MHz	50uH	50pF	3.2MHz	20pF	5MHz
Range 2	10uH	350pF	2.7MHz	10uH	50pF	7.2MHz	20pF	11.3MHz
Range 1	1.5uH	350pF	6.9MHz	1.5uH	50pF	18.4MHz	20pF	29MHz

I played around with the three wave ranges for ages with the screening cans on and then removed. One puzzle was why I'd discovered that the notch in Range 2 had seemingly gone, when later it had re-appeared. Problem solved.. at least now to my satisfaction. Using the homebrew noise source I noticed the deafness in Range 2 had to a large extent gone away. In fact the noise level in Range 2 was now equal to the noise level in Range 3. Rembering that Range 3 was pretty good means that now Range 2 is also pretty good.. but only with both the bandpass filter screening cans removed... put one anywhere near the coils and Range 2 noise drops and with even one in place Range two is completely deaf. My guess is the coils shift in inductance as the screening can is fitted so they are no longer able to be tracked across the range. In fact even with the cans removed correct tracking is tricky. Oddly Range 1 appears deafer with the screening can detached than in place and is very difficult to track. I now need to carry out some further experiments to resolve the issues.

Range 1 was indeed pretty deaf with the screening can off but with the can in place pretty good, with the noise source producing much the same results as with Range 3. What if I fooled the Range 1 coils? I offerered up a small piece of tinplate between Range 1 anode and mixer coils and up went the trace on the DSA815 by more than 10dB and up went the noise level to the original level. Very odd, but I'm putting it down to "eddy current tuning" (the method used in the R107 handbook).

It's all very puzzling as presumably the 1939 design team knew what they were doing. My guess is that ageing is involved. Maybe metal oxidation, bad earthing, poor condensers or even the refurbishing carried out in the 1950s provides the explanation? As one possible reason for problems with the front end is bad condensers I decided to remove the decouplers. All were identical 0.05uF x 350V (500V peak) and metal clad. Results are given below. I replaced all seven with new polyester 68nF x 250VAC. Of course as you'd expect with this R107.. absolutely no change.

Measuring the condenser leakages revealed they're a lot better than wax coated types. C5M was the worst being 95Kohm with the others ranging from 194Kohm to 1Mohm. Interestingly as they were connected to a circuit carrying 400 volts in series with 360Kohm about half were slowly improving whilst the remainder slowly worsening.

CONDENSER	C5A	C5B	C5C	C5D	C5E	C5F	C5M
MARKED	0.05uF	0.05uF	0.05uF	0.05uF	0.05uF	0.05uF	0.05uF
CAPACITANCE	300nF	109nF	56nF	204nF	174nF	99nF	589nF
% LOSS	12.0	4.9	1.3	8.6	8.0	4.7	15.0
ESR OHMS	110	-	-	130	-	-	110

I made a small tin bracket which I placed between the two Range 1 bandpass coils and this improved the receiver's response across the band by 10dB. During this process I noticed that pressing on the Range 2 coil increased the audio output by 20dB. I tried poking at the wiring etc but only pressing on the coil had any effect until the imrovement just went away and couldn't be repeated. The investigation would appear to need the coil removing. I also noticed that the same coil refused to track, meaning that the trimmer needed twiddling to improve a -70dBm signal from nothing to audible. Maybe the coupling coil is dry jointed or the coil copper slug has detached? I found waving ferrite or brass near the anode coil altered its tuning so the 10uH coil itself must be intact as are connections to its trimmer. I suppose I could wind a new coupling coil in place between the anode and mixer coils...

After puzzling over the bandpass filter tuning problem I decided to discover what "eddy current tuning" meant and soon realised the reason for my difficulties. It all hinges on the use of ferrous or non-ferrous tuning slugs. Ordinarily mutual coupling between iron cored coils increases inductance so each time an alignment pass is made the coil adjustment reinforces alignment but the reverse happens with "copper" tuning (called "eddy current" tuning). For starters, looking at the drawing opposite, all six tuning coils interact to some extent with each other. Enclose a pair of screening cans over the input (A) and output coils (B) and the major interaction is reduced to four coils, and of course the non-ferrous can has a dramatic effect on the inductance of every coil. I also noted that the R107 uses a special screen over the tuning condenser, presumably to minimise coupling between the different tuned circuits. Of course I'd detached this in order to work on the receiver. Things had not been improved by the bad wavechange switch and an intermittent in L5A. Clearly every part has to be in place when aligning the receiver.

Attempting to resonate L4A and L4B to transfer the optimum RF signal across Range 1 is difficult because as L4A is brought to resonance it raises the resonant frequency of L4B and vice-versa. Also, because unused coils are shorted by the wavechange switch their resonant frequency is dependent on their inductance plus trimmer. If this resonance occurs within the tuning range of the selected waveband it will alter the inductance of the tuned coil in such away as to detune it at one specific point. This results in a notch which is quite obvious using a noise source connected to the aerial connector. Presumably this notch can be minimised by using a specific alignment technique.

Now.. can the R107 be fixed? I decided to check the RF amplifier tuned circuits to see if I could precisely determine their characteristics. What I found does explain the set's deafness because all the coils have too much inductance. They all resonate to a frequency lower than the dial setting. To make things simpler I've quoted the average of the coil/trimmer settings eg. for Range 2 at 7MHz the coil/trimmer range covered 4.8 to 5.2MHz so a false resonance would occur when the setting was 5.2MHz.

It's not too easy to check the bandpass filter resonances so it remains to be seen if these are as bad as those in the RF amplifier.

During checking I discovered the grid leak was a 22K resistor measuring 26K instead of R2A quoted at 250Kohm.

	COIL	LOW DIAL	TUNED CIRCUIT	HIGH DIAL	TUNED CIRCUIT
RANGE 1	L1A	7.0MHz	6.0MHz	17.0MHz	14.6MHz
RANGE 2	L2A	3.0MHz	2.50MHz	7.0MHz	5.0MHz
RANGE 3	L3A	1.2MHz	1.15MHz	3.0MHz	2.25MHz

Is it possible perhaps to use the same technique as used on a local oscillator circuit namely to insert a padding capacitor in series with each of the three RF coils? Access is not to difficult at the switch contacts. S1A/A or even a common padder at C13C/C4A to the switch wiper... No.. I tried this and it wasn't successful. The following table attempts to calculate the actual inductances of the three aerial coils assuming the tuning condenser is 300pF, a trimmer setting of 15pF and varying strays and coil capacitance 7pF. The next step is to roughly work out the number of turns required and then to count them. It's then possible to work out how many turns are required to produce the correct coil value. The latter will be calculated in another table using dial markings. Plans have now changed because I measured the aerial coil inductances and they seem to be exactly as they should be or even less! (see below).... back to the drawing board.. Is the tuning condenser the right one at 300pF? Not easy to measure it in-situ so I measured one that had the same size rotor plates and it measured 470pF. It has 15 vanes compared with 10 which makes about 313pF for the R107 example. Next I checked the grid coupling condenser C13A and it measured correctly as 200pF.

COIL

LOW DIAL

C in pF

L calc

L actual

TEST

HIGH DIAL

C in pF

L calc

L actual

TEST

L test

RANGE 1

L1A

7.0MHz

322

1.6uH

1.1uH

6.5MHz

17.5MHz

1.6uH

1.1uH

14.3MHz

2.29uH

RANGE 2

L2A

3.0MHz

322

9.3uH

8.8uH

2.9MHz

7.0MHz

9.3uH

8.8uH

7.25MHz

19.5uH

RANGE 3

L3A

1.2MHz

322

54uH

53.8uH

1.2MHz

3.0MHz

54uH

53.8uH

2.2MHz

96uH

Using the experimental results above (results were in the column marked "TEST"), L1A is 2.29uH (43% high), L2A is 19.5uH (110% high) and L3A is 96uH (77% high).

Previously I'd measured the bandpass coils with an LCR meter as L3B=50uH and L2B=10uH which is promising although I'm starting to think that either the coil, due to changes in the wax covering or the former on which the coils are wound has gained extra self capacity so although their inductance might be OK their additional capacitance reduces their resonant frequency. On the other hand their adjusting cores are said to be brass and introducing this material into the coil should decrease their inductance but I'm fairly sure the opposite is happening. Finally I found the true design values of the coils marked in red below from James Cookson G4XWD. If anything these are even more puzzling unless the brass cores can swing their values considerably lower.

Note that unlike a normal receiver that uses dust iron tuning slugs the R107 uses brass cores. Ordinarily a coil will be would to provide the minimum inductance required to match the HF end of the tuning range, but in the case of the R107 the opposite applies. Coils are wound to match the LF end of the tuning range, bearing in mind stray capacitance and trimmers will enable their resonant frequency to be lowered. This is not the case with the R107 as with the core and trimmers set for the LF end the dial must correspond with the HF end. In fact the whole alignment process hinges on the brass cores swinging the coil inductances high enough.. and that's not happening here.

In an effort to align the front end of the receiver I'd noticed R2A had been reduced to 22K from 250K and wondered if this had been a modification. Sure enough whilst atempting to align the bandpass filter the receiver had burst into oscillation which was tamed by resoldering the old R2A resistor back. I've had some experience of this kind of problem so looked at grounding points. There are some very odd examples. Firstly the aerial coils are commoned to a small solder tag held under a post on the left of the screening can. These posts are tarnished and not very tight so I fixed this. Next I looked at the tuning condenser and tightened its securing screws. The grounding of the bandpass filter is also poor and I found that adding additional decoupling capacitors grounded to the edges of the cans considerably improved RF stability. These connect to the live sides of C5C and C5D. I was then able to revert to the 240K grid leak and overall front end gain increased even though tracking was still some way from ideal.

It was easier to add this simple audio amplifier than replace several of the RF coils. The audio from the detector is typically raised from 600mV to 14 volts although the aim is not to produce lots of audio power but to amplify weak signals.

The TIC49 is nothing special except it has a CE rating of 350 volts. I'm running it at around 5mA collector current which places the collector at about 50 volts. The base resistors act to give auto-bias.

While testing the amplifier I found the reason for the low HT voltage... this being an open circuit reservoir condenser. Adding 4.7uF brought the HT up to 310volts.

Above is the simple audio booster located adjacent to the base of V2B' mounted on a 5-way tagstrip. To the right is the new HT smoothing capacitor and I fitted a small HT reservoir capacitor with the pair of silicon diodes inside the transformer cover. The amplifier provides a gain of about 27dB. Of course Sod's Law came into play when, just after checking the amplifier, I carried out a final experiment on the RF coils. During earlier testing I'd noticed Range 2, being very deaf would come to life for a brief moment then die again, but no amount of fiddling with the circuit would repeat the increase in gain. I decided to use a hot air gun to remove the wax coating and maybe warm up the coil in case the wax had somehow added extra intrinsic capacitance which might explain reluctance to resonate. The coil winding emerged and I initially wondered if there was an intermittent short between the coupling and tuned windings. With the receiver switched on I found slight movement of the trimmer produced a slight crackle. The stator was firmly soldered as was the rotor whose black covered wire connected to the coil. I checked the connection from chassis to the rotor and noticed it varied from 1.3Mohm to something higher but very intermittently so, but the Range 3 trimmer rotor showed a solid 1.3Mohm. I pulled the black wire and much to my suprise the bare end slipped out from under the coil.. it hadn't been soldered but before melting away the wax had been fairly secure. I tinned the bare end and soldered it to the outside of the solder tag on L5A.

Before testing further I removed the wax from Range 1 bandpass coils L6A and B whilst checking their resonance. Refitting the two bandpass filter screening cans showed more gain even before fresh alignment and was now quite reasonable for Range 1 and 2 but after proper realignment all three amateur bands 80m, 40m and 20m had decent CW and SSB signals. Using my homebrew noise source I can still see a resonance effect from Range 3 coil in Range 2 and maybe Range 2 in Range 1 but overall performance is now quite acceptable. Some care is now required to get the best results because of overloading and a little instability from the extra audio gain, improvements I made to RF grounding, new RF decoupling capacitors and the substitution of the higher value grid leak for the RF stage.

The final change I made was to fit a 1uF x 600V reservoir capacitor which resulted in an HT of 268 volts rather than over 300 volts using 4.7uF.

I now need to tighten the trimmers and put back detached metalwork. Handling the set shows the amount of audio gain to be good for SSB/CW but needs to be reduced together with RF gain for reception of strong broadcasts, but an essential job is to make the set relatively safe. It uses a special 2-pin mains plug without any provision for a safety ground and additionally carries a connector (used in conjunction with a trasmitter for muting audio) with a set of exposed pins of which one carries HT of about 300 volts. I decided to fit an IEC mains socket after removing the plug carrying HT.

It's necessary to remove the side strenthening panel to get at the redundant muting connector which is held in place with 4BA screws and locking nuts. Also in the way is the HT block condenser used as a reservoir and for smoothing. With these parts removed I cut the muting circuit wires, tucked these out of the way and removed the muting plug. This left a hole in which I fitted an IEC plug on a small metal plate. This connects via a pair of leads to a couple of chokes below the chassis whose leads I disconnected from the old mains plug.

When I'd first tested the receiver it had several HT problems already covered above. Clearly the reservoir condenser wasn't working and I'd fitted a 1uF x 600V capacitor. This was chosen to lift the HT, now supplied from silicon diodes, to a level commensurate with the design value. Too much reservoir capacity and the HT would have risen to over 320 volts because the forward resistance of the new diodes is very much less than that of the original 6X5 rectifier. The new smoothing condenser is actually 470uF x 450V and is aimed not only at smoothing the HT, but to delay its rise a little.

As you can see on the right this old block condenser is marked "NOV 43" and its two 8uF sections measure 282pF and 95pF so useless as a reservoir or for smoothing, in fact useless for anything other than maybe a paperweight (if the R107 was being restored it would be worthwhile to remove its innards and fit a pair of new capacitors).

Below a view of the new mains lead and its fitting using the original 4BA screws. That second block condenser incidentally measured 4uF, precisely as its marking states and has an ESR of a mere 0.49 ohms.

How does the R107 handle now that it's been brought back to life? Sensitivity on a long wire is now excellent and the extra audio amplification can bring weak 40m and 80m SSB signals to good loudspeaker strength. The AGC produces about minus 4.5 volts on strong signals which isn't enough to dampen overall gain sufficiently so very strong broadcast signals can overload the output valve. This means that audio gain and RF gain controls need to be balanced in order to cope. However without the additional amplification ageing of coils in the front end would mean poor results on shortwave signals. To receive LSB the BFO is set right of centre and USB the same amount left of centre. Fidelity of audio is a trifle sharp but I found that using the narrow IF setting and/or the crash limiter results in a more mellow sound. The headphone and line output sockets can feed stereo headphones if the plug is not fully inserted and of course the audio level can be adjusted independently of the loudspeaker setting if either headphone socket is used. I found the balanced aerial input is insensitive. This might be a design characteristic or merely a problem due to ageing of the aerial coils. Image response is negligible due to the design of the set. The tuning control is very light and unusually the inner knob does not rotate in sympathy with the slow motion knob. Adjustments for smoothness or backlash can be made by tightening or loosening three 6BA screws on the rear of the drive or alternatively at three screws hidden by the inner knob.

The original front dial lamp is very uncommon, having a very low current consumption to keep its brightness low and I fitted an LED in its place.

As I dragged over the outer case to mate with the R107 I noticed it looked brand new but the rivets holding the inner extrusion to the outer case had mostly failed. Also with this set is the external cover.. this is certainly not new and looks like it's had a very rough life. Surprisingly the case and receiver both have identical serial numbers but the lid looks really battered so had it spent its life protecting the front of the receiver?

I drilled out a selection of the broken copper rivets and secured the inner extrusion with 6BA screws.

R107 MkI

Return to Reception