WS 19 MkII Overhaul

 The purpose of the overhaul is to use the set on 80 meters and Top Band. The last time I operated one of these rigs was back in 1958 to 1960 before I got my callsign. Of course it was advantageous in those days to gain some prior experience of operating and to tinker with the innards of sets as the Amateur license was much more difficult to get. Not to mention attaining 12 wpm in the morse test. My first transceiver was a 17 Set with which I used to communicate with a pal from school. Those sets cost us 30/- each or £1.50 and the batteries were very cheap. We used 18 Set batteries as these were only 5/- (25p). The 18 Set wasn't as useful as it operated on 40 meters which was pretty narrow and extremely full, but at least let one listen to lots of broadcast stations.

I last overhauled a 19 Set in 2004 (a Mk3) and it worked OK. It was in excellent condition but had been dropped and had a bent front panel which was mighty difficult to straighten out.

I imagine it's still out there somewhere?

My MkII is more tricky to deal with as it has all the original circuitry present and it's had a hard life. Maybe left to rust in a damp shed or maybe it suffered when it was returned from Russia after WW2? I pulled out a 6K8 and it had a watermark indicating immersion up to half its width in a water rich with limescale or salt. At first I thought the white deposit was inside the valve envelope indicating it had lost its vacuum, but thankfully it cleaned off.

All the switches are seized although the wafers look fine. One of the first jobs will be to remove all the knobs and apply some freeing agent. Hopefully I'll be the first to attempt this and the grub screws are all in good order. Also it'll help to add a little oil to the indexing mechanisms as there's nothing worse than hunting for errant ball bearings.

Once the switches are working properly I'll attend to the tuning dials which have lots of rust on their rims. As the interior of the set has been tropicallised and any untreated components look to be in good shape the restoration should be easy enough. I'll have to remove wax or setting compound from the various coils so I can get at the slugs and free them up. This is a critical process as the slugs need to be intact and not split otherwise they'll need drilling out and replacing.

Most valved equipment is not too unhappy if their resistors drift unless the drifting is really bad. The worst offenders I've met are the carbon rod variety cased in small ceramic tubes as the carbon ends have brass sleeves which can corrode. The second problem is capacitor leakage. In a lot of cases a leaky capacitor won't do much other than reduce the voltage its decoupling but there are positions in sets where capacitors need to be free from leakage. The worst offender is the one connecting the anode of the LF amplifier to the control grid of the output valve. Because of the very high impedances in the area a leak of 1 megohm can have serious implications by placing a positive voltage on the output valve control grid. The set will generally work OK but the output valve will get very hot because its anode current is much higher than the designer intended.

Read this about valves in general...

A second serious problem can arise if there's a capacitor connecting the output valve anode to ground. This would have been used to make the audio sound more mellow but if the capacitor deteriorates too much the output transformer primary can be destroyed from excessive current. The transformer can also be destroyed if the grid coupling capacitor has a really bad leak. I'll add that AGC action will be reduced when leaky capacitors are in this circuit. Because of the large value resistors used the resistive capacitors act to produce potentiometer action which reduces the AGC voltage. This has the effect of making the receiver appear to be less lively and in some cases RF gain is too high resulting in overloading and distortion on strong signals.

Since overhauling the last 19 Set I've acquired some decent test gear and a couple of useful power supplies so results should be more rewarding.

Circuit diagram of a WS No19 MkIII

Keep an eye here to see how things progress...


 The first thing I decided to do was to measure the resistors. At this point I discovered that many of the available Internet copies of the various handbooks are not very useful (to say the least) as their definition is poor. Text is generally legible but circuit diagrams are virtually useless. Having found a copy that was useable the next thing I re-discovered was the methodology of numbering the parts in the 19 Set. This is quite clever as it groups resistors for example into exact part types, perhaps for logistics reasons or, in the case of valves it would permit various valves to be swapped around in fault finding?

Peering into the underchassis reveals a couple of points on which it's worth commenting. First, because the thing is conformally coated it's difficult to measure resistors because the meter probe must penetrate the coating. Second, although the majority of the resistors are the classic WW2 US versions, some resistors are the familiar British types so might have been UK supply.

Although resistor measurement is easy enough, actually identifying which is which is more tricky and needs some intimate knowledge of the set. For example I measured an 82Kohm resistor at 121kohms and identified it as R7C which is designated 100kohms. This, being a US-made set, has a few variations in parts.

I mentioned earlier that many resistor values are not all that critical in practice, but it's easy enough to experiment by bridging a high value part with a parallel resistor suitably chosen to bring down the value to the correct resistance. This can be done (with care!) whilst watching say the AGC line voltage. If necessary either the old resistor can be cut and a new correct value inserted or you could just add a parallel part and hope that further deterioration or changes won't occur.

Before I could finish resistor measurement, I had to mark up a line drawing of the underchassis view with resistor values so I could be sure of correct identification. My set has single 22Kohm resistors fitted where the parts list has pairs of 47Kohm in parallel and a meter shunt resistor which is different plus a few others. Measuring resistors in-situ is OK except in a few cases expect variations due to circuit connectivity. Almost without exception, I found that resistors had shifted up in value, but as most have a 10% tolerance some could have started life already high in value. Strangely, I've always found that 47kohm and 470kohm resistors are the worst for drifting and some of those in this set don't disappoint. Examples are a 470kohm= 590kohm (+25% high) and 47kohm=71kohm (+51% high) although one 270 ohm (=R3B) cathode resistor was noted to be 750 ohms (+177% high). See results

I had an idea to measure these old resistors using a high voltage power supply to see if there's any difference to the measurement using a low voltage as I'm not entirely happy that a modern digital meter is altogether reliable for valve equipment components. Certainly this is the case with capacitors. Leakage degradation, I suspect gets much worse as the voltage across an old capacitor increases. Maybe someone has experimented along these lines, for example measuring the DC voltage wrt ground at the disconnected end of an infamous audio coupling capacitor. Click to see this experiment


Switches & Tuning

 Freeing the switches wasn't too bad. The only one that gave real trouble was the mode switch stuck on CW. After removing the knob, oil and a pair of pliers eventually freed the spindle which was oxidised. All the grub screws were OK and knobs came off with a little help from levers. I detached the main tuning dial. To do this I had to remove the plates with the markers, the tuning knob mechanism and the four bolts for presetting frequencies together with the square knob.

The dial edge was bad and I had to file away rust deposits which had etched the smooth surface in places. New dials would be nice if I can find a couple as the tuning is slightly rough in places. I found the main tuning capacitor was very tight but after applying switch cleaner to its end bearings it improved.

For some reason I couldn't get the flick tuning arrangement working. I'm missing the threads into which two of the preset screws locate, but I'll sort it out later. Also the tuning capacitor and dial aren't connecting together properly giving me backlash. I'll have to figure out what's going on.


Initial powering
 Today I decided to power the set. After getting muddled up over the two connectors, using crocodile clips, I applied 12 volts DC and 250-300 volts HT (with current consumption standing at 45mA and varying slightly with audio). Connecting headphones to pin 4 of the second connector and ground revealed a high level of noise. I turned off my network camera and the noise abated. Connecting a long wire brought in lots of stations on the higher frequency band and a few signals on the other. There are obviously a few things to sort out, other than the tuning knob, viz. there seems to be no BFO beat and the lower frequency band needs tweaking plus the front panel meter has a rather odd looking pointer and the meter isn't working. I'm pretty sure I have a 19 set meter sculling around somewhere in the workshop and I might have to fit that if I can't ressurrect the original. Then I'll rig up a loudspeaker so I don't have to use headphones and then attempt to align the set.


Meter & metering problems

 A couple of steps forward. I removed the meter which turned out to be made by Westinghouse and inscribed "W.S. NO.19 VOLTMETER" in very pale writing. I dismantled it because it had a very odd needle, which turned out to have been made from a piece of biscuit tin and glued in place. It looks dreadful and it's far too heavy. Also there were lots of rust particles inside the moving coil air gap that were causing the pointer to stick. The meter came apart easily and I cleaned the magnet, removing lots of particles sticking to it, then used a soldering iron to melt the glue and remove the replacement pointer. I thought for a few minutes about a suitable new pointer, then when a blackbird flew past, I thought I'd try a feather. Lots of those around and I picked up a couple and after a few minutes with scissors and a scalpel followed by emery cloth had a new pointer. A dab of superglue and it was in place. I checked the movement and it read 482uA full scale deflection, so its an 0.5mA meter confirming the pale writing. The new needle is as light as a feather.

Following are pictures of the meter restoration, starting with the carefully shaped piece of biscuit tin...

A clue to poor operation are the rust particles seen on the dial. Some of these have been attracted by the magnet and were jamming the coil.

Their removal meant disassembly.

Some manufacturers use sticky tape glued to internal surfaces to catch stray bits and pieces within meters.

The new pointer has to be made of material as light as a feather. So why not use a feather....

 After refitting the meter which has 11mm nickel (or cadmium?) plated brass nuts (clearly no metal shortages in North America during WW2) I turned on the power and the meter read 5 for the LT rail. Seems decidedly pessimistic so I'll look into that next.

The meter is used to measure the precise current through one or more resistors connected to the rail being selected.

First the sort of resistances one would expect to see to match the two scales (0-15 volts and 0-600 volts)

As the movement consumes 500uA to produce full scale deflection (fsd) there will be a resistor of R=V/I or (15/0.5)Kohm=30Kohm for the 15 volts scale and (600/0.5)Kohm=1.2Mohm for the 600 volt scale. Bolted to the back of the meter is a square paxolin board carrying six resistors. These are marked 100Kohm, 27Kohm (R21C), 2 x 58Kohm in parallel (R26A), 1.2Mohm (R24A) and a second larger 1.2Mohm (R25A) on the opposite side. These are related to the circuit diagram as indicated by the codes in brackets. I haven't traced the 100Kohm resistor yet.

Measuring these resistors I found R21C (AVC) was OK, the two comprising R26A (LT) were 80Kohm and 117Kohm = 47.5Kohm or 63% high, R24A (HT1)was 1.5Mohm or 21% high and R25A (HT2) was 1.42Mohm or 18% high. So R26A will indicate an LT reading of about 4.5 volts hence the reading of 5. When I checked the actual current to produce fsd it measured as 482uA which means the readings should be 3.6% high, meaning the expected 4.5 volts will be nearer 4.7 volts.

I fitted a set of new modern resistors but using for R24A and R25A a number of small resistors in series. The reason being to cope with high voltages and wattages. Note that some modern resistors will have a voltage rating that may be quite low. Each comprised 2 x 470Kohm + 270Kohm + 15Kohm= 1.225Mohm. Oddly I found the HT of 275 volts to register too low on the meter so I'll need to discover the reason.


Start of alignment

 Before describing alignment I'll mention how I monitor output. I favour, at least in initial alignment, simple test methods. It's useful to hear what's going on so a loudspeaker should be connected. However, as the ear is relatively insensitive to small changes in output the measurement of audio must be done with a meter. A digital meter is much less than ideal so you should use an analogue meter set to AC volts. I actually have in my collection some audio wattmeters so I use one of those. This lets one make adjustments to deal with the large increase in output (if you're lucky) as alignment proceeds. An oscilloscope can be used or something more exotic but initially a meter is best.

I turned the set to the HF waveband and found it tuned from over 8 MHz down to 3.2MHz, responding well to a test signal of 1mV and with the meter on AVC it was reading a small voltage which went downwards when tuning a strong signal. Next the LF waveband. Not too good as the gain was way down on the HF band, but able to receive test signals from 2MHz to about 4.5MHz. Signals on both bands were improved by twiddling the centre tuning dial which seems to be calibrated correctly. Listening on headphones I could sense something wasn't quite right. The gain control, which is extremely stiff, tends to make the set oscillate at maximum setting, and there's a background noise indicative of narrow band reception, almost like that from a CW filter.

I started alignment and made 8MHz line up with the dial by tweaking the local oscillator, also tuning the RF stage, and after removing wax from the IF coils trimmed these to 465KHz. A 2uV test signal is now clear. The LF band is still deaf, but I then found C11A a 6.5PF-140pF capacitor that was faulty. It looks like a small nut should be crimped to the shaft and because it's not the movable vanes are free to move in and out (longitudinally) so turning the adjuster makes the capacitor short out. As the capacitor was in the fully open state, I have a suspicion it was left like this at the manufacturing stage, maybe on a Friday afternoon? It's also possible the set was returned to depot and rejected and subsequently demobbed because of this fault.


A couple of faults resolved

 I returned to the set after checking the value of the faulty trimmer capacitor and compared the sensitivities of the HF and LF receive wavebands. To do this I first removed the tops of the screening covers from the 6K7 RF amp and 6K8 mixer so I could connect the signal generator to either valve top cap. An 8MHz signal producing a low comfortable audio level was 10uV at the 6K8 top cap and 2uV at the 6K7 top cap. On the LF band I found at 4MHz, 25uV at the 6K8 but to produce an equal strength sound in the headphones needed a whopping 100mV at the 6K7 top cap. If I'm doing my sums correctly this represents a stage gain in the RF amplifier for the LF band of minus 70dB. By this point I'd already switched the 6K7 for an alternative and so ruled out a faulty valve. I removed the 6K7 and measured the voltages at the socket pins and found:- Pin 1 was 265V (probably an HT tie off point), Pin 2 (heater) 7V, Pin 3 (anode) 4.8V, Pin 4 (screen)160V, Pin 5 zero, Pin 6 0.3V and Pins 7 & 8 zero. Pin 8 (cathode) was 237 ohms to ground. The problem voltage is clearly that at the anode which is less than 5 volts.

The 6K7 RF stage connects via the bandswitch to two coils, L22B and L23B commoned to R5A (a 2.2Kohm measuring 3.1Kohm) and a decoupling capacitor C4C. For C4C to have failed it must have a DC resistance of around 50 ohms, however the HF band is working seemingly normally. This leaves two possible faults..

(1) The wavechange switch S11A/5 or S11A6. As one contact set earths the unwanted coil this is a possible fault scenario.

(2) Coil L23B, its coupling coil L23A, or its trimmer C10D.. for example an open coil or a short-circuit trimmer.

To be bourne in mind however is the low anode voltage. This seems to indicate that switch S11A/5 is not switching properly or L23B is open circuit or disconnected.

Thinking about it, I did notice one or two resistors had lumps of verdigris on their wire ends so L23B may be open circuit due to corrosion?

See the results of the investigation

There are several sets of coils (relating to the Canadian MkII that I'm working on) viz.

Receiver RF (V1A 6K7) HF coils=L22B and L22A tuned by C10A (common to LF)+ LF coils=L23B and L23A tuned by C10D

Receiver Osc (V2A 6K8) HF coil= L24B tuned by C35A (common to LF) + LF coil=L25B tuned by C35B; C11A is for tracking at the low end of the band ie.around 2 MHz.

LC103A is a tiny coil for adjusting netting errors introduced primarily by mechanical differences in the preset frequency arrangements

Then there are the transmitter local oscillator and mixer output coils.

The key information for anyone aligning any superhet is whether the oscillator tracks below or above the signal frequency. In this set the oscillator frequency is higher than the signal frequency so images will be higher by twice the IF. As the IF is set at 465KHz you'll find the image, say for 7MHz, at 7.930MHz. In a badly adjusted set it would be easy to misalign it by setting ones signal generator to 7MHz, tuning to 7.93 MHz then adjusting the oscillator coil to make the 7MHz dial reading correspond to this. I'll attempt to demonstrate this later using a spectrum analyser


Where's the BFO gone?

 The next fault I tackled was CW reception. When the mode switch is turned to the CW position you can hear a change in the set's filtering but there's no sign of any heterodynes. The BFO would normally be tuned to the IF which is 465KHz and I would have expected to hear some sort of whistle as the tuning dial is rotated through AM stations, but nothing. The BFO is the triode section of the 6K8 (V2B) in the front right corner and I looked on the anode (pin 5) with a voltmeter and turned to CW and 82 volts appeared. Also the cathode (pin 8) went to 0.8 volts so the valve looks like it's working. I switched on my oscilloscope and low and behold there's a decent sinusoidal signal present at the 6K8 anode. This scope is really useful as it simultaneously displays signal amplitude and frequency, and there's the answer... the scope showed 485KHz. The tuning core is set behind a mass of wax which I scraped out and put on one side. The core could then be twiddled inwards to show around 465KHz on the scope. This reading is likely to be out by up to a few KHz so the next step was to tune the BFO to exactly 465KHz. To do this I tuned my signal generator to 465KHz and placed the output lead on the top cap of the 6K7 next to the BFO valve. This connection bypasses any attenuation in the tuning arrangements of the RF amplifier and gives a cleaner signal in the speaker/headphones. As my 19 set has a trimming control for shifting the BFO frequency up or down by a KHz or so, I set this to roughly the centre tuning point so I can hopefully resolve upper and lower sideband by offsetting one way or the other. Having got a decent unmodulated 465KHz signal getting into the IF amplifier I turned to CW and carefully zero-beat it by turning the BFO tuning slug. Once done, I put back the wax and heated it to set the core in place.

This was a very odd fault, maybe due to a drifted capacitor somewhere or another Friday afternoon in the factory lapse? I can now resolve CW and sideband transmissions.


Initial transmit tests

 The next task is to find out what happens if I select transmit. The 19 set should transmit on the same frequency as it's receiving because it mixes a 465KHz fixed frequency local oscillator with the superhet local oscillator. This mixing process uses the BFO which I've just sorted out, mixing it in V2B hexode with the receiver local oscillator signal generated by the triode portion of another 6K8 (V2A), the receiver mixer. The output of V2B passes to V5A, an EF50 where it's amplified before driving the 807 final amplifier. Bearing in mind the initial frequency of the BFO, goodness knows what I'll find in terms of alignment. Clearly, this 19 set was transmitting 20KHz away from it's receive setting and if used in a net would have resulted in lots of dial twiddling by its members.

Here goes. I worked out that the transmit relay which handles the changes to circuitry when moving from receive to transmit is activated from Pin 7 of PL2B, normally connected to the microphone pressel switch.

I grounded Pin 7 and heard an awful squealing in the loudspeaker. This problem had an easy solution. I removed the scope probe from the 807 grid, which helped a little...

The meter showed some drive when this position was selected so I set the tuning dial and waveband to 7MHz and peaked the trimmers on the main tuning gang. The drive increased slightly. I then bridged the HT connection between the receiver and the 807 supply. The meter showed 200 on the 807 HT position and the current on the meter on my bench HT power supply increased.

I connected the aerial socket to a 50 ohm wattmeter with a built-in dummy load and saw some RF power indicated. By carefully tuning the transmit dial the indicated power increased to about 4.5watts. Temporarily cranking up the HT voltage to around 350 volts increased this to over 6watts, so the transmitter seems to work, at least on CW. I reset the HT to a level where the power was 4.5 watts and checked the waveform on my scope. It read 7MHz and 1 volt RMS, with the 10:1 setting on the probe. This equates to 10 volts RMS, and assuming a 50 ohm load, indicates around 2 watts RMS output. Lots of sources of error however.. I was using croc clip leads and the scope probe is not rated too well at HF.

Once I've got a little further with alignment I'll attempt to measure the power output more accurately.


Another fault
 I vaguely wondered why CW didn't seem to respond to a morse key so I peered inside the chassis and saw the morse key jack socket had bits dangling from it. I'm aware that the morse key wiring is not entirely straightforward and as I looked at the loose parts I saw a spark, so I switched off the power so I could investigate further. The socket is insulated from the front panel by pieces of ebonite and held in place by a pair of countersunk screws which had very rusty heads and virtually no slots left. In fact the only way to detach the socket was to drill out the screw heads. This done I noted the wire colours (all much the same muddy colour) and cut them off. The socket was clearly missing some parts and the two screws holding the springy bits together had no heads, hence the reason for it disintegrating.



 The socket comprises two contacts which make when a plug is inserted and two prongs which connect to the jack plug. These parts slide over a pair of screws (possibly 8BA?) each carrying a tubular insulator preventing the parts from touching the screws. Each part is insulated from its neighbour by pairs of bakelite spacers. The end spacers and final insulating washers were missing. Gone the same way as the screw heads. I refitted the parts using two new screws. As I couldn't find any with proper threads I used slightly thinner screws and secured these with small nuts. In place of the missing spacers I just built up the gaps to the correct depth using some red fibre washers originally used for assembling computers.

I refitted the jack socket taking care to fit the insulating washer between the front panel and the body of the socket. The socket securing screws need to be sunk sufficiently into the outer piece of ebonite so that if a metal jack plug is used it cannot touch their heads and hence the chassis.

CW & MCW tests

 Powering the 19 set showed the repaired jack socket worked OK. On CW and MCW I was able to see power of around 6 watts or so in the dummy load. Next I need to see if the microphone input works. I poked each of the pins in turn with a wire connected to an audio generator but no signs of modulation whilst listening on a monitor receiver. Lots of carrier on R/T but no audio, whilst on MCW sidetone on the speaker and audio on the monitored signal. Time to study the circuit diagram....

The circuit diagram shows a transformer for the microphone input connected to pin 1 of the lower plug.

Modulation tests on R/T

 I found a picture of how to test the 19 set using an audio signal generator. Using an attenuator pad constructed to match the microphone input various audio frequencies and levels are input between PL2A Pin 1 and ground.

The attenuator uses a Pi network of three resistors. 3.3 kohm shunting the audio output, 820 ohms T resistor and 52 ohms shunting the mike input (Pin 1 to ground). Then a modulation depth of 75% should be achieved by inputting 1 VRMS at 400Hz, 5 VRMS at 150Hz and 7 VRMS at 5 KHz. Essentially the pad is a potentiometer feeding 6% of the generator output to the mike input, hence 60mV at 400Hz, 300mV at 150Hz and 420mV at 5 KHz. Tests will commence when I get a lull in lift repairs...

Today I spent 5 minutes looking at the lack of modulation problem. It was easy and has explained a couple of other problems too. Having identified PL2A Pin 1 as the mike input I went to connect my audio generator to it. Where's Pin 1? There's a croc clip with a green lead going to the loudspeaker on Pin 1. Oh dear that explains why the audio sounded odd and the gain control was peculiar AND why I got feedback on R/T transmit. The loudspeaker earth return was going through the mike transformer to ground. I unclipped it and clipped it to the edge of PL2A. The audio sounded much mellower and the crystal filter echo has gone away. The volume control works properly, the feedback on transmit has gone AND a spot of audio on Pin 1 produces sidetone and a modulated signal.

Power supply

I'll soon be pausing the overhaul to build a power supply for the 19 set.


Here's a document describing the MkIII

 Once things are sorted out, as I've got a spectrum analyser I might as well use that to indicate the frequencies and relative amplitude of the receiver local oscillator. On the other hand I might also test the valves next in case their submersion in water has oxidised their pins or old age has lowered their emissions.

I recently bought a battered example for spares with seemingly good dials ...

some pictures later...

Return to Reception