Wireless Set No.19 Mk3

 These transceivers were used over a very long period starting in WW2 and at least well into the 1960s. This example is relatively late, having been rebuilt in 1960 and has had the B Set (the VHF transceiver) removed. Here's a rough MkIII S/No 122913. Later versions were manufactured without any provision for the B Set (see example of S/No.65861). As you can see below this example carries the serial number R/VL1/106. Click to see it's circuit diagram.


   The label indicating the set was been rebuilt in December 1960. The code "VL1 or VLI" with digit one or capital I is not definite. "R" will be "Rebuilt" and a new serial number.

   The rear of the label which was held in place by a couple of 6BA screws shows it used to be a MkIII with the serial number 102372, and the code P.C. perhaps Pye. But.. maybe this isn't the original label? Another clue is a couple of inked stampings on the chassis including "V.C.M.23".

 See a similar set which I overhauled in 2004.r

See an aticle from Practical Wireless 1960

 If you study the power unit below, you might notice that it isn't wired for HT2. This voltage shares an earth return in early WS19 models; but in the Mk3 the HT2 negative passes to the chassis via a grid bias resistor R19a (82Kohm) which helps the 807 power amplifier to stay linear in R/T mode,whilst saving around 10 to 30 volts or so HT which is wasted by the self-bias cathode resistor in the MkII. In CW mode the 807 operates in non-linear class C by shorting out the HT2 negative bias resistor which increases the RF output from around 6 watts to as much as 20 watts.

The transformer for HT2 fitted in the power unit is far from ideal. It has an HT secondary winding rated at 290-0-290 volts 130mA which either produces too low, or too high an HT voltage, depending on the way its configured, so I'll examine a few different options.

I could operate the winding by using a pair of rectifier diodes in full wave using the centre tap and a reservoir capacitor, choke and smoothing capacitor (this being the type of circuit used for HT1). There are two capacitors yet unwired. These are 30uF and rated at 600 volts DC. The centre tap method would give me a peak HT voltage of 290 x root 2, or 410 volts dropping somewhat under load to around 300 volts or so. This HT level is really too low for HT2, which is quoted at 550 volts when using the dynamotor in the original PSU.

To increase the voltage closer to the usual WS19 level of 550 volts I could use a full wave bridge rectifier connected across the full winding of 580 volts RMS. There are two options for this method. I could use a "swinging choke" circuit, dispensing with a reservoir capacitor and using the two 30uF x 600v capacitors wired in series for smoothing. The second option would be to use a reservoir capacitor. The latter option would give me a peak output of 820 volts. This is decidedly high for the WS19 so I'd rather go for the first option. Two desirable features would need to be added. First, I'd use a bleed resistor and secondly, by splitting the bleed resistance in half, using two resistors, I'd connect the resistor junction to the junction of the two smoothing capacitors. This will balance the voltage between the two capacitors and reduce the peak output voltage to a safer level for the WS19. The bleed resistor will also discharge the smoothing capacitor during testing. This choke input circuit has some slight advantages over the reservoir method in the second option. Firstly, the peak voltage will be lowered to a safer value, especially with a suitable bleed resistor load and secondly the output voltage level will be less dependent on the 807 current draw. I chose a convenient bleed resistance of (a measured) 66Kohm by using four 5 watt resistors marked 10Kohm and two marked 12Kohm from my junk box. If the off-load voltage heads towards the peak of 820 volts this might drop to something like 650 volts across the bleed resistance. If this is so, the bleed current will be 650volts/66Kohm=9.8mA and the total bleed resistor dissipation will be about 650volts x 9.8mA=6.3watts or approximately 1 watt per resistor. The transformer secondary winding resistance is about 200 ohms and the choke about 100 ohms so voltage loss here will be negligible at only 300ohms x 9.8mA = 2.9 volts.

If the operating voltage falls below optimum in practice, I could always add a little reservoir capacitance to nudge it up.


 The power amplifier in the WS19 is an 807 beam tetrode. This valve can be operated in a variety of ways. In AM the options are to use high level modulation or grid modulation. The WS19 uses the latter which then relies on the 807 to operate linearly, at least when in R/T or MCW modes. In CW this restriction is lifted and the 807 operates in non-linear class C mode. Looking at the 807 circuit diagram you'll see the anode of the valve is powered from HT2, but its screen grid is powered from HT1 (the 275 volt feed). This is slightly unusual and requires the latter voltage to be present before switching on HT2. This is handled in the PSU above by wiring the HT2 on/off switch to the HT1 "on" connections thus preventing HT2 from working without HT1. I imagine the WS19 includes circuitry to prevent the screen grid of the 807 from consuming loads of power and destroying the valve.

 Testing the output from HT2. Mains was 241 volts RMS and the mains transformer gave 650 volts RMS across the secondary. That would put the theoretical peak DC output at 919 volts. The first test was a puzzle because the 500mA fuse was open circuit which made me recheck the wiring; however all was well and a new fuse allowed HT2 to establish at the ballast resistor.

The ballast resistor is 11.55Kohm centre-tapped and will get reasonably hot so its balanced on a brasso can.

I measured three voltages. Off-load using only the 66.4Kohm bleeder gave 710 volts; with a total load of 9.83Kohm it was 555volts and with a total of 5.33Kohm it measured 534 volts. These voltages varied somewhat, depending on the temperature of the bleeder and ballast resistors so I've quoted the minimum readings. Regulation is quite good.

The bleeder drain was 710 volts/66.4Kohm=10.7mA, with the ballast + bleeder at 9.83Kohm the current was 555 volts/9.83Kohm=56.5mA and with ballast + bleeder at 5.33Kohm the current was 534 volts/5.33Kohm=100.2mA.

The latter two readings give power outputs of around 555v x 56.5mA=31.4watts and 534 x 100.2mA=53.5 watts. Excluding power lost within the PSU these provide 7.8watts less or around 24 watts and 45 watts.

 In summary, the choke input circuit provides almost the normal 807 HT2 voltage of 550 volts. Under load this will be between 534 and 555 volts, although with HT2 applied and only a very small load the voltage could rise towards 700 volts.

I finished the wiring on the PSU,connecting HT2 negative and positive wires. Plugging in the WS19 revealed the HT reading was low (circa 450 volts) on the meter, but this type of problem is usually is due to a high value resistor (in this case R25a) in the meter circuit. In CW mode the HT2 negative is grounded and in R/T it sits at -23 volts. Because the meter negative terminal connects to ground, the reading for HT2 will change between CW and R/T by the amount I measured between HT2 negative and ground.. 23 volts. So if HT2 is 550 volts it will read 527 volts in R/T and 550 volts in CW. The PSU has a small digital display wired into the heater circuit. When I got the PSU a few days ago the display was working but for some reason it now doesn't come on. Its fed from the 12 volt supply and monitors the LT current flowing into the receiver when the WS19 is plugged in. The WS19 LT reading is OK at the meter, but I'll check R25a (1.2Mohm 1 watt) and also R24a (1.2Mohm 0.5 watt) used for HT1 which reads about 250 volts instead of the expected 275 volts.

After trying out receive on both wavebands I noticed the AVC meter reading varied with the position of the RF gain control but didn't move at all when strong broadcast signals were tuned in. This "rebuilt" example has no AVC on/off switch which would usually be fitted below the RF gain control. Most likely C38a is leaky or the AVC diode in V3a is poor, although a prime candidate is C16a, the cathode decoupler at V3a which is a 12uF electrolytic with a +100% tolerance. Now to remove the outer case...



 Having been used to the work of Plessey wiremen back in cold war days, I was surprised to see the weak standard of workmanship revealed under the chassis, but nice to see an absence of wax condensers. One electrolytic seemed to be leaking into its sleeve and a quick check gave an ESR of >20ohms, meaning its probably close to open circuit. click the picture above to see a LARGE circuit diagram of a complete MkIII. Sure enough the duff condenser is C16a. The spec should have read +/-100%. Then again.. reading the circuit details a huge amout of delay is applied to the AVC line.. in fact around 27 volts, so that it takes a very strong signal to reduce gain.
 I decided to check a few resistors. Surprisingly most were within tolerance. In-circuit testing is not always clear cut however, because various switches and pots might affect the readings. I thought this when I tested an 82Kohm connected to the 807 cathode. It measured 11 ohms no matter where the various switches were turned. Looking at the circuit diagram showed a few connections to the 807 cathode including a contact at the transmit/receive relay that bypassed the 82Kohm resistor,. Sure enough, wedging a screwdiver behind the top of the relay changed 11 ohms to near zero. Something is definitely not right. The area close to the 807 base has a crowded tagstrip whose end contact was soldered to a green wire connected also to the 807 cathode. At the same tagstrip point is a purple wire which connects to a diode of the 6H6. In addition a high voltage condenser C17b is wired to the cathode. I snipped this without much conviction and of course it was blameless. I unplugged the 6H6 just in case it had an internal short but that was also blameless. Another wire goes to a large condenser, C4j (0.1uF) which goes to a coil (L25a) at V2b. At this point I unplugged the 807 which naturally was blameless. During tests I noticed the resistance had changed to something closer to 82Kohm, but after some tapping and prodding 11 ohms reappeared. I suppose that resistance value must provide a clue? L25a wired through L5c goes to chassis and this pair might be 11 ohms if C4j was short-circuit? I poked and prodded the tagstrip next to the 807 valveholder and the 11 ohms shot up, wiggled around and returned. I also poked the wiring going to some transformers above the chassis adjacent to the 807 and this had a similar effect. Maybe a loose wire or damaged insulation is to blame? Further scrutiny revealed a bunch of rubber covered wires squashed together at the T/R relay.. maybe a blob of solder is also squashed in there?


 Scruffy wiring behind the 807. Much of the new plastic covered wire is roughly soldered, with strands left unsoldered and melted insulation.


 Rubber covered wiring replaced by plastic covered wiring, with joints (as in this case) insulated by extra rubber. The reason for replacing the rubber was for the very reason you see here. I guess a mitigating factor is the expected lifetime. Sets built say in 1944, updated in say 1959 would have been around for 15 years. Then the 1959 example given 15 years for the extra rubber would make 1974, beyond the expected life of the WS19.

The squashed appearance is because V1c metal screening can was pressed against the rubber boots.

Inside the metal case is T3a, the microphone transformer.

Strangely, when the 6K7 was unplugged the 807 cathode short went away. The 6K7 rattled when shaken but I couldn't see any shorts between electrodes. I'll plug in a new 6K7 and see if that short reappears.

 Still no luck with the dratted short. Turned on the ohmeter and it read 250Kohm at the brown wired pulled off the 807 socket. Then, without warning the reading started changing and ended up at 9.5ohms. Then, tapping virtually anywhere, changed this to around 30Kohm. Maybe a loose nut or washer caught up in the bare wiring? The set's a bit heavy to shake around but this didn't produce any rattling sounds. Shining a strong torch on the wiring and donning my magnifying headset revealed that each length of coax was fitted with rubber at the end to insulate the braid and in all cases the rubber had perished, going sticky. Maybe there's one of these touching something and allowing the braid to cut through the perished rubber? The relay has all its connections sheathed in rubber, some with different colours, but the black variety are all perished. I suppose a good clue is the minimum resistance reading which seemed to be either 9.5 or 11 ohms?

I decided to power up the set and make DC measurement at the brown wire now disconnected from the 807 cathode. The voltage varied, starting at 17 volts but dropping to circa 12 volts. I was able to draw about 0.9mA. Unplugging the 6H6 proved that it wasn't from that path. When I was checking without power the resistance was usually high and dropping when the chassis was tapped, however, if the brown wire was grounded (to show zero ohms to ground), lifting off the short left the resistance sitting at circa 12 ohms or so, which was then pretty stable. I think this effect is caused by whetting ie. once a small current is persuaded to flow (by grounding the brown wire) the contact is made stable.


 I hope I finally cracked the problem. I'd discovered that plugging in V1c re-introduced the short-circuit, so I found a new 6K7 and tried that; with exactly the same results. The DC resistance dropped from circa 380Kohm to a wildly variable value from a few Kohm upwards... but not 9 to 11 ohms as initially seen.

I looked under the chassis and noticed a few pins of V1c valveholder were very close to the lower of two condensers clipped to the rear of the chassis. Plugging in the valve pushed the valveholder solder tags closer to the lower condenser. Without the valve plugged in there was a minute clearance, but with the valve in place there was definite contact. Although the condenser is fitted with a plastic sleeve, I guess its been punctured. Unscrewing the double clip removed the short. I'm now concerned that this may not be the only intermittent short because previously I'd measured only around 9 ohms, and I still haven't worked out the connection between V1c pins and the 807 cathode, unless via R18b? R18b is 270Kohm so I should have seen a drop from 380K to perhaps 150K certainly not 9 ohms.

But read on below...


 What was the voltage on the brown wire which is disconnected from the 807 cathode? This read 238 volts and dropped to 95 volts when reconnected. I checked the current and saw about 0.9mA. Because R18b connects to HT1 I would expect around 0.9mA so we're in business. But why only about 9 ohms across the 82K 807 cathode resistor. The answer was revealed when I checked later. For some reason, yet unexplained, V1c Pin 1 is used as a tie-off point for the circuit connected to the 807 cathode, so shorting this to ground results in a low resistance across the 82Kohm resistor R19a (the resistance will be a combination of grounding resistances, which happened to be about 9 ohms). Shorting V1c Pin 1 to ground results in a current of 0.9mA because we're shorting HT to ground via R18b a 270Kohm resistor.


So in summary.. the short across the 82K resistor, R19a, was a short-circuit between V1c Pin 1 and the case of C4k (cathode decoupler for V1c). The case of C4k was, in turn, shorting to the condenser clamp holding it to the rear chassis. But what is the purpose of routing the 807 cathode connection to V1c Pin 1? I detached the clip and looked under the lower condenser... R18b, the 270Kohm HT feed resistor has been moved here where it shares the HT supply to V1a and V1c screen grid. Because the set is a transceiver, with large areas of the circuit being used in both receive and transmit, the HT1 rail is split to enable it to select certain parts of the circuit to operate solely in receive or combined in transmit mode.

Interestingly, in receive with HT2 switched off, grounding V1c Pin 1 has no effect, but because HT2 is normally applied when the set is in receive/transmit (ie normal) mode, the 807 cannot draw much current through the high resistance of R19a, but because of the short, the 807 cathode is grounded and allows the valve to draw significant current and, most likely, adversely affect reception. I checked to confirm this and the background noise level in receive increased by something like 10dB when I grounded the 807 cathode when HT2 was present.

Below is a listing of condensers and resistors.


 Cct Ref


 Cct Ref


 Cct Ref


 Cct Ref

 Cct Ref

 Cct Ref  Value  Cct Ref  Value


















530pF var












540pF var


50pF trim














750pF trim


































 110pF trim














 15pF trim





 Cct Ref


 Cct Ref


 Cct Ref


 Cct Ref

 Cct Ref

 Cct Ref

 Cct Ref

























































































20 CT











 The aim now, is to align the IF amplifier and check the BFO can be varied equally about 465KHz,then align the two wavebands to line up with the dial markings in receive. The transmitter alignment is only really important on the 40 and 80m amateur bands although I'll check that top band is feasible also. Below are views of the chassis with parts identification.



 The first task with WS19 alignment is to understand the design of the equipment and identify exactly the location and purpose of the various trimmers. With a standard receiver one must ensure the local oscillator tracks precisely with RF tuning. It's common to adjust trimmers at the HF end of a waveband and coils (using dust cores) at the LF end. I recently aligned a Moreton Cheyney receiver and this used a combination of dust cores and padder trimmers to maintain precise tracking so I wasn't surprised to discover the WS19 uses only padder trimmers. Because it's usual to find a ganged tuning condenser with sections having matched capacity-swing an oscillator padder is needed to modify the tuning range of the local oscillator to match that of the RF stage. This is not always so.. for example the R1155 uses a completely different oscillator section to that of the RF tuners.

WS19 not only uses oscillator padder trimmers rather than dust cores but alignment of its two wavebands is interactive, meaning the correct sequence of adjustments has to be followed. For example there's a 50pF trimmer hard-wired across each of the 4-gangs of main tuning condenser and these are adjusted only with the higher frequency band selected. When it comes to the LF band there are separate trimmers which are adjusted AFTER the HF band has been aligned, and you'll discover that the correct method of completing alignment of the LF dial calibrations to the correct frequency is to slacken securing screws and move the plates carrying the cursors.

A further complication is dealing with the designer's choice of whether the local oscillator tracks higher or lower than the signal. In this case it varies.. on the LF band the oscillator is aways higher than the RF input and lower on the HF band. If one calculates the oscillator range for the LF band, if it were lower than the RF input, the results would be difficult to implement. In fact for all receivers covering medium waves, or the band immediately above this in frequency, the oscillator tracks higher than the RF input. This also simplifies choice of oscillator padders.

It's important during alignment to check for the image (the drawback in superhet designs) to ensure it's present at its correct frequency, and indeed a lot weaker than the desired RF setting. Not only is this essential to check, but it's also vital one doesn't inadvertently swap over from tuning the correct RF input at one end of a band and the image at the other (easy to do if one's signal generator level is set too high). The pictures below show the locations of the trimmers and variances that probably occurred in different factories, or in the rebuild. Another point one must consider when looking at circuit diagrams and drawings for WS19 is that there are differences between Canadian and British manufactured sets, and it is not impossible for errors to crop up. For example R43a is both a potentiometer and a 3.3Mohm fixed resistor in the same document (EMER F224)



 Now that all the trimmers have been identified, alignment seems straightforward. Initially the IF amplifier needs to be checked to see if it has a decent characteristic shape and centred on the correct frequency of 465KHz, then to set the BFO to this same frequency. There are several methods of aligning IF strips but I favour the use of a spectrum analyser. One could use an FM signal of suitable width, a wobbulator or even manual plotting, but interaction of adjustments, and a common problem of adjustments relying on maximum audio output, usually means you end up several KHz off-frequency is avoided if a spectrum analyser is available.

Before using the spectrum analyser I did some basic checks. First the BFO which was a bit iffy in receive mode. Setting the signal generator to 465KHz and feeding a high enough level, around 10mV into the aerial socket I turned the receiver to CW mode so I could beat the BFO control against 465KHz. The BFO setting fully clockwise read 460.0KHz and fully anti-clock read 465.4KHz, an average of 462.7KHz, so I reset the BFO core to 465KHz with the setting at mid-way.

Next a rough check of the IF response...Using an audio wattmeter and a signal generator I measured the response when the wattmeter read 7dB either side of its maximum reading (I did this because the latter was too broad to ee a well defined peak). The average reading gave me the centre of the IF response as 463.5KHz. The next step was to check the IF transformer tuning cores for adjustability. Two IFTs are original and have plastic fittings on the end of the dust cores. These were free to move... but the centre IFT was not original and had a note about a modification (a resistor disconnected) and uses sunken dust cores set in wax so I removed most of this with a small screwdiver, then heated the screwdriver with a soldering iron and freed the core by gently turning it forward and back a few times while the wax was hot. Once the core moved reasonably well I used a plastic adjuster to set it to 465KHz. This initial check was to roughly centre the IF response at 465KHz by peaking an AM test signal on an audio wattmeter. Now that the IF and the BFO are roughly correct I proceeded to check tuning alignment.

As the WS19 will be used on 40m and 80m I checked the dial and it wasn't far out at 7.000MHz and 3.500MHz. The two receiver trimmers on the top of the tuning condenser were pretty close to optimum on both the HF and LF band and little needed to be done.

Next was a check on transmit. Rather than just turn on HT2 I decided caution was better so I used my Solartron variable HT supply. This enables me to gradually inrease the voltage whilst monitoring HT current. Intending only low power tests, I connected a 5 watt dummy load, which measured 65 ohms (which is why a label says it's u/s), across the aerial plug together with my oscilloscope via a x10 probe. First, with HT2 at zero volts, and with a monitor receiver tuned to 7.000MHz. Grounding the relay pin caused the RX/TX relay to click on and a carrier appeared on the monitor receiver. Adjusting the two Tx trimmers on the main tuning condenser lifted the signal nicely. Turning on HT2 I noticed the HT current was very high, but switching to MCW tamed it and with 120 volts the current measured 55mA. PA Drive read 3 volts and the scope showed 7 volts RMS. That seems to be about 3/4 watt. Turning to RT though gave me 100mA at 100 volts with Drive 2 volts and output 4 volts RMS or about 1/5 watt. Switching to 80m I could see RF output in MCW but in RT the HT current was way too high and only by reducing the voltage did it come down to 150mA.

As the WS19 is not in its case and without its chassis screening plate I wonder if RF feedback is the problem, and if so, or otherwise, why is RT behaving so differently to MCW? I read that official testing is carried out with a dummy 807 missing its screen grid pin.

Below is a picture showing the area around the 807 base.


 After much head scratching and puzzling over the first transmit tests I'd found some anomalies. Despite concerns about the way HT2 should be connected to provide "free" 807 bias, the negative supply was correctly connected and the 807 cathode was correctly being grounded in transmit. The puzzle was the difference I'd noted between RT and MCW. In the latter case, I could advance HT2 to 400 volts without a problem, but in RT there seemed to be a bad short, or the 807 was drawing loads of current. Advancing HT2 allowed only tens of volts before the HT current went beserk. The clue was the 807 grid voltage, sitting at plus 2 volts in RT Transmit and Minus 12 volts in MCW transmit. Here's an echo of the leaky coupling condenser between any audio amplifier anode and an output valve grid, of which most are familiar. So, I convinced myself the problem was a bad leak in condenser C22b which connects V3a anode to the 807 grid circuit via R7g (100Kohm), but where was it? It didn't appear on the parts layout drawings (above), so I traced the circuit around V3a (6B8). There are several candidates clamped to the chassis near V3a, mostly not shown on my parts drawings so I looked for their connections. Some have thin coloured wires disappearing into the odd harness but by checking with a buzzer I found where they ended up and ruled them all out. Eventually I spotted four candidates for C22b near the 807. The most likely was easy to get at so I unsoldered one leg and connected it to an HT supply via a 100Kohm resistor with HT neg to its other end. I found 4 volts across the resistor at an HT setting of 400 volts. Slightly leaky so I left it disconnected and switched on the WS19. No change.. the current drawn by HT2 in RT mode was still not sensible. What's V3a's supply voltage situation? In MCW the anode of V3a is about 250 volts, but oddly in RT mode it read minus 11 volts. Then I checked again.. the set was upside down and in MCW V3a anode is only powered when the key is pressed.. so that was a red herring.

I eventually found C22b (below). It was hidden under R7d which I had to cut off in order to get to it. This is a tiny Metalmite component mounted vertically and connected to the chassis end of R7g, itself hidden away, where it's joined by the decoupling condenser C5e. By unsoldering the top of C5e and detaching the top of C22b I found it could be pulled upwards to reveal enough of its lower lead to waggle it free. Here it is below. To replace it, I found a suitably sized modern 33nF high voltage capacitor which I could only fit on the opposite side of the tagboard (above). I also fitted a new 100Kohm, R7d and a new 100ohm, R20a which had perished in experiments (could the problem be too high a voltage on g2 compared with a low anode voltage?.. No).



 All the new condensers fitted in the rebuild were rated at either 500 or 600 volts, but unfortunately the available space for C23b could only accommodate this tiny 350 volt component. Apart from an old 12uF electrolytic condenser, this is the only part that's failed, so far.

But the parts lists have two other Metalmites of this exact type viz. C22a and C22c. The former (for the MCW oscillator?) is hidden away close to where C22b was fitted and C22c isn't relevant, being in the power supply unit.



 Just what has happened to this component?

I understand mine is not the only Metalmite to bite the dust.

This LCR meter had one view and the multimeter another. My Peak ESR meter which reads capacitors from about 0.5uF upwards told me it was open circuit.


 With everything soldered back in place I cranked up HT2. Currentwise RT and MCW were much the same. I could use the maximum voltage from the PSU, about 550 volts with the current measuring about 30mA. Now I can proceed with transmit tests.

I traced what I thought was C22a to the place I could just see a condenser and following connections proved it was indeed C22a and spent an hour extracting and replacing it. Of course Sod's Law ruled and it was in pretty good order, measuring lots of Mohms and 30nF. MCW oscillation still wasn't happening so it must be something other than C22a to blame.

I then checked transmit again. Ths time I hooked the oscilloscope to both the 807 grid and the 50 ohm dummy load. I found that on 80m the two drive trimmers did work OK by tweaking the 807 drive voltage although not making as much change at the output. I recorded the following best results... 807 g1 = 20 volts RMS and O/P = 6.4 volts RMS with HT2 bias at -38 volts representing 0.8 watts output. In CW these figures improved because HT2 bias was shorted, giving 11 volts RMS out for 22 volts drive representing 2.4 watts output. Thesefigures are worse than pevious tests when I managed 10.5 volts RMS or 2.2 watts output in R/T. HT2 dropped from 660 volts to 525 volts under load, but I found by adding only 0.47uF as a reservoir for HT2 the HT rose to to 613 volts under load. The transmit stages are not yet fully functional.

 More testing. This time using an ATU. I was able to get the output up to 10.2 volts RMS into 50 ohms. That works out to 2.08 watts in R/T mode. I then tried an 80m inverted vee and using the ATU got 11.4 volts RMS = 2.6 watts with perfect matching.

I then tried running at 5MHz into the inverted vee. It wouldn't match and I saw 23 volts RMS at the antenna feed point, but as the impedance isn't well defined that voltage is pretty meaningless. Maybe a long wire using the ATU might get me out on 5MHz?

The next step was to see if the microphone circuitry was OK. I no longer have a WS19 microphone but recall it was a low impedance moving coil type so hunted around for something similar in my collection of microphones. I found two that seemed to have the right sort of DC resistance, but neither could be persuaded to work. The only other that might work is the carbon mic I use for the T1154. Using a variable DC PSU set to current limit at a lowish value and set to 6 volts produced decent modulation monitored initially on a nearby receiver, then on my Andrus SDR with my XYL listening in. The WS19 mic transformer is said not to favour DC current so I intend to not continue along those lines, instead to use the intercom amplifier which remains intact on the rebuilt chassis. The circuit is shown below. To ease the problem of wiring, I'll remove the two blanking plugs from the WS19 front panel and fit a Tx/Rx switch plus a socket for a microphone. I'm advised that this was a method used to improve modulation depth and requires only a few passive components.


I experimented with the intercom amplifier above, to add gain to the system, initially connecting its output via a pi attenuator 4.7Kohm with 100 ohm in and out to the A set mic input. Using a small Japanese dynamic microphone having a DC resistance of 340 ohms connected to the intercom transformer T4b input I didn't have much luck.. the modulation was minimal, although disconnecting the speaker improved this a littlte. As I was looking for the connection to V1f grid the screened lead fell off at R23q so I connected the lead to ground via 1Mohm and connected the dynamic mic to the grid via 1uF. Modulation shot up and unplugging the loudspeaker improved the level significantly.

Up to now I haven't checked to see if any resistors in the audio areas have drifted high or whether any condensers are bad. Also, looking at the intercom amplifier component values I can see they have been selected for optimum performance with the standard WS19 ancillaries, so using a microphone (or a low impedance loudspeaker) will degrade the design performance. Looking at the circuit above I can see that disconnecting R2d and R2e or finding a dynamic mic with a higher output might improve matters; also it's important that I check C29c in case it's leaky, although when I tested the amplifier using a 1KHz sinewave the 6V6 anode easily reached 230 volts RMS. Since I decided to operate the amplifier without T4b I added a 150pF capacitor at V1f grid (like C14b) to reduce any RF pickup.


 The WS19 is part of a communications system and as such uses a set of matching ancillaries designed to interface with the receiver and transmitter. For the casual user it's handy to connect a loudspeaker across PL2a Pin 4 and ground... and a low impedance speaker works OK, but has the disadvantage of shunting most of the power when the same transformer is used for modulating the 807. This means that the modulation level is too low. It was found in practice to be already a bit low with the correct headphones and a service modification was introduced to wire in the Intercom Amplifier to boost the mic output. This included an attenuator across a couple of pins of PL2a to limit the boost to the mic input of the original circuit. This mod is OK if you're using the correct microphone, but not necessarily OK if you're using a non-matching mic. In my case I rewired the input to the Intercom amp, dispensing with the attenuator and feeding the grid of V1f directly rather than using the original mic transformer.
 The first thing I've done is to wire the coil of a small 12 volt relay in parallel with the Tx/Rx relay coil and having traced the wire to Pin 4 of PL2a cut this and wired it to the normally closed contacts. This disconnects the loudspeaker loading from T2a secondary and considerably enhances the modulation. If I wanted a little sidetone I could wire a resistor across the NC contacts.. by trial and error, but maybe something like 220 ohms. I also changed R12a, which measured about 92Kohm for a correct 68Kohm. This may have accounted for some of the lack of modulation, because further tests showed this had increased somewhat with pronounced modulation dips in transmit aerial current. Now that I've proved my small dynamic mic works OK, I've fitted a standard 1/4 inch jack plug to it and added a jack socket to a spare hole left from the B set parts. Whilst trying to make decent RF connections I tried to make up a coax lead with a mathing Pye plug, but I found this too fiddly and gave up after a dropped the tiny grubscrew that holds the centre tag in place. Once dropped on the floor of the workshop there's little chance of finding anything that small. I tried a PL259 socket but the hole was too small so I fitted a BNC socket held in place by a nut.


 Don't fret...purely a temporary solution to matching my supply of RF cables, and right.. filling useful holes left from the old B Set with a Tx/Rx switch and mic jack socket (both fitting the original holes!)

 Next, I decided to tackle excessive noise in receive. The symptom suggested weak AVC action. The obvious culprit is C38a, which, if leaky will degrade AVC performance. That particular condenser is located in the rear corner of the WS19 chassis, some distance from the AVC diode within the 6B8 valve. This example of the WS19 was rebuilt in 1960 and uses metal-cased condensers which probably replaced the earlier wax covered variety well known for failing, but are physically larger and are crammed into place. C38a is marked Sprague 0.1uF 600vw and after extracting it I found it wasn't too leaky but when later reconnected did seem to shunt the AVC line by nearly 50%. The new capacitor is a good quality 0.15uF 250v AC working.

 The new C38a tucked away in the corner of the chassis. The four condensers in the corner needed to be unclipped for access to C38a and it's possible that there was a short in the area due to the jamming in of parts. There was a short-circuit at another pair of clamped condensers further along the rear of the chassis that was affecting the bias conditions of the 807.

After the old condenser had been snipped the receiver was completely unstable with a whistle replacing mush.

After the new capacitor had been fitted I shunted it with the old one to measure its effect. The AVC voltage with an RF input of 2uV was -1.3 volts and this dropped to -0.86 when the old condenser was paralleled up. The table below shows the AVC performance with the new capacitor in place and coincidentally indicates the huge overall amplification within the receiver. Comparing the British and Canadian circuits the latter includes 2 controlled IF stages whilst the former only one (excluding V1b). Some versions of WS19 have an AVCon/off switch, but this Mk3 does not.

 RF In






















 During tests I'd noticed that whilst plugging in an aerial resulted in a tremendous increase in noise from the speaker, switching to CW to read SSB resulted in a much quieter background. The BFO tuning control, which is a small wirewound affair, is in poor condition and the audio crackles as the kob is turned. It's also awkward resolving SSB. The usual technique for resolving SSB on old receivers is to tune to the centre of the signal using AM, then switch on the BFO and turn the tuning control to resolve decent-sounding audio. I've already checked that the control is centred on 465KHz so there shouldn't be a problem. I hadn't noticed though (in this Mk3 model) that there's an audio filter selected when the mode switch is set to CW and this affects reception of SSB. Looking at the filter which is designed to peak audio response at 900Hz, it's highly likely that drifing resistors have altered the performance. Below centre, hidden in the switch and relay wiring, you can see the filter. This is a twin T notch filter and connected across an audio circuit will attenuate everything except 900Hz.




 Dead easy to calculate the response of this filter because the Net has calculators for virtually anything. Here you can see the filter notch is at precisely 900Hz.



 The response using measured values of resistors.

 Because old ceramic bodied resistors of the type used in this WS19 Mk3 drift upwards I measured their actual values to see how the response had been affected. The silver mica condensers are usually OK but the resistors were:

R47a:1.336Mohm, R47b:1.298Mohm and R48a: 171.5Kohm

As you can see, the calculator now gives the notch to be shifted down to about 720Hz. SSB will sound a bit odd but reception can be via RT with the Net switch turned on if the RF gain is adjusted to provide a signal roughly the same level as the feed from the BFO.

The filter is matched into the circuit via R7e (100Kohm) and R8c (1Mohm) which add to the overall audio attenuation.



 Return to Reception