Fixing the PR155G.. will it never end?

continued from my first encounter with the recalcitrant PR155G

At this point I'd thought everything was proceeding to plan, however I hadn't fully appreciated the differences between the modules used in the B and G versions of the receiver

A major critisism of the Plessey documentation that I've managed to find is the lack (or even accuracy) of basic information contained in them.

Nowhere can I find a basic block diagram of this complex receiver and in the case of the "G" variant, key information (carried in the documentation for the "B" version for example) is not included. This means I have to keep knocking off my investigation into faults to work out how exactly should it be working. To help understand this latter technical aspect I've included several drawings in my discourse below.

It's interesting to note that some areas of the circuit have been modified up to ten times (Module 10, the spectrum generator, was up to Mod 10 in the 1968 manual)i've yet see many of the waveforms associated with the workings of this receiver and adding these may help the odd restorer in the future.

 I got as far as cleaning up the signal from Module 12 (Fig 21 circuit further down this page) but it read as -62dbm on my Rigol (an arbitrary value due to the use of a 10:1 scope probe), which might be just about passable but when I tried the receiver.. no luck once again (dead on SSB). I unplugged the coax lead carrying the signal and it read -62dbm but plugging it into the mode switch lead it dropped slightly and this was definitely below the required threshhold so back to the drawing board.. out came the pcb and I cound find no way to raise the -62dbm. I decided to cheat and after 30 mins I'd added a couple of BC107 transistors to amplify the signal and to buffer it to provide a low impedance drive to the flip-flop. This should result in about 1.6 volts output and low and behold SSB worked well with background noise for the first time much the same as that for CW and AM.

At least I thought it worked well but I discovered both the LSB and the USB settings resolved 40m SSB.. perfectly. Either the signals I resolved were DSB or something isn't quite right. I checked on another receiver and the 10.6MHz and 10.8MHz were certainly being switched in accordance with the mode switch settings.. so what's going on?

 

 

 A simple amplifier to raise the SSB injection voltage to a level acceptable to Module 7B flip-flop (SN7272).

The amplification is approximately equal to the ratio of the 3.6K and 1K collector and emitter resistors. Output was about 420mV and should now be about 1.5v.

This will drop slightly due to base current effects in the above circuit and when loaded by the flip-flop circuit.

 As I understand it the 10.6MHz heterodyne oscillator mixes with the 10.7MHz 2nd IF to provide a "minus" 100KHz IF or the 10.8MHz oscillator mixes with the 10.7MHz 2nd IF to provide a "plus" 100KHz 3rd IF. The 200KHz signal from Module 12 is divided by two by the flip-flop in Module 7B to produce either LSB or USB in the product detector. Selection of the appropriate heterodyne oscillator is made by the mode switch S2AF in LSB or USB settings.

I tested the receiver on 40m SSB signals and either of the LSB or USB settings resolved signals perfectly without needing to tune the receiver which doesn't seem right. Maybe I should try the 20m band and see if USB is resolved properly?

However, I first looked at what the new amplifier is doing. I measured 978mV into Module 7B which maybe on the high side but more importantly the sine-wave has a flat bottom which will be the result of the emitter follower transistor ground resistor being too low. I'll change the 33K resistor to 68K and this will get rid of the bottoming and reduce the 3.6K to 2.7K which will drop the output voltage. Whether this will fix the LSB/USB oddity I'm not sure but maybe another way is to reduce the gain in the 3rd IF amplifier via a pot fitted in Module 7A for this purpose.

After modifying the amplifier as described above I didn't expect the end result, but it will have to do for now. I also backed off the IF gain pot within Module 7A and the audio output dropped but when I checked I could resolve perfectly LSB and USB on either LSB or USB settings so something's going on I haven't yet fathomed.

 

Module 12 amplified output

 

Improved Module 12 amplified output 

If I can't figure out and fix the USB/LSB conundrum I wonder if I could modify the BFO circuit to produce two off sets say 200+3KHz ad 200-3KHz, which would divide down by the flip flop to 100+/-1.5KHz 

 

 Looking at this circuit a varactor diode at VT1 collector could do the job with say three pots (CW, USB and LSB) selectable via the mode switch, however, switching out the BFO front panel tuning capacitor is awkward not to mention gaining access to the BFO module so I'm looking at an extra module (but based on the circuit above).

Firstly I'll need to consider switching and I reckon this could be achieved at the mode switch which could supply neg 15v to a varicap diode via potentiometers used to set the operating BFO frequencies. As the BFO is running at 200KHz nominal and the existing tuning capacitors are a 220pF fixed and a 10-420pF variable, plus C2 and C3 (in series) which work out to about 750pF the coil L1 could be say between 462uH and 633uH depending on the setting of trimmer C9 and this would give a range of adjustment consistent with the spec. To give an ample range of settings say 190KHz, and 210KHz would require 750uH plus 935pF and 765pF respectively. A "medium-wave" varicap diode having a typical swing of say 450pF is the easiest to configure. To handle selection of LSB and USB I've fitted a set of diodes. One pair is used to power the SSB oscillator and the other pair to select the frequency setting circuit. As always I'll need to experiment in order to get the desired requirement.

 

I had to make several changes as I progressed through the development of the above circuit. Finding a suitable coil in my junk box was the key to the design, then getting it to tune to 200KHz by selecting the two feedback capacitors C2 and C3 was next. I ended up (without the tuning diode) with a tuning capacitance of 687pF and a coil of something like 720uH giving an oscillation of 226KHz. Adding a single varicap diode should in theory add about the right amount to let me tune over a range of say 190Khz to 210KHz. One puzzle was the voltage measured at the cathode of the varicap with my initial circuit which used an RFC of 100uH in the control lead. This was primarily rectified DC from the oscillator RF and the shape of the waveform was not too good. I removed the choke plus a 100nF decoupling capacitor and things improved dramatically although I still needed the 120K to provide switching current for correct operation of the diodes. I suppose I could add a series resistor in the varicap feed to minimise RF loss and a reduction in output voltage. This might also improve the capacitance swing which is fairly critical. A second option is to use two varicaps back-to-back but this would reduce the swing to 50% and may not be enough. As I had a large number of junk box CV8805 diodes I used 4 of these rather than the 1N4148.

As it stands I reckon the coil is 720uH and C2+C3 are 675pF giving me a frequency of about 228KHz. To produce 190KHz and 210KHz requires capacitance of 975pF and 800pF respectively, meaning the varicap has to provide 975-675=300pF and 800-675=125pF. I carried out tests using a ceiling of about 8 volts reverse voltage to the 1SV149 but I'll bump this up by swapping the 4.7K feed resistors as 15 volts is the minimum specied by the makers. Currently the control voltage is limited to about 10 volts minus the forward conduction voltage of the switching diode of say 0.7 volts.

Reducing the feed resistors to say 470 ohms will give me 14.3-0.7=13.6 volts (and as you'll read below these will serve for another purpose). The varicap spec shows a widish range of swings around 480pF to 25pF from 1 volt to 8 volts so my requirement of 300pF to 125pF seems reasonable. Note that the 47nF DC blocking capacitor reduces the swing slightly but only by about 3pF.

See the varicap data sheet.

The small diecast box carrying the oscillator plus other bits including a cable and two wires, LSB USB, fed to the mode switch plus a ground wire can be fitted into a suitable space within the PR155G and tuned to 197KHz and 203KHz which, when divided by two by the product detector flip-flop, will provide automatic SSB tuning at postions 5 and 6 on the mode switch without needing to tune the existing BFO.

Below shows progress to-date. I'm not sure about drift/stability just yet so I might have to swap the tuning capacitors with types better suited for this. Also I'll add a pair of 12 volt zener diodes to prevent changes in the internal control shifting the oscillator using the 470 ohm resistors.

 

This reflects the final circuit after changes to the "Modified BFO circuit for LSB & USB"

 A note on the method of construction. Assembly on a piece of tinplate typically cut from a biscuit tin is my favoured construction technique as it lends itself to rapid component changes whilst developing the end result. I generally use physically small decoupling capacitors as insulating anchor points. Those three multi-turn pots are held in place by brass 8BA screws soldered to the tinplate. As with most old components (resistors and capacitors) it's a good idea to scrape their leads to remove any oxide and expose shiny tin. I used my Peak coil/capacitor tester to check old coils. Below is the wiring diagram showing connections to the mode switch utilising spare contacts conveniently provided by Plessey for adding extra modes.

 

 Testing turned out to be a puzzle (because of my earlier attempts using the two 10.7Mz oscillators). Initially I connected the RF output to the PR155G and was able to hear perfect SSB using an external PSU. I then connected the LSB/USB power connections and oddly received USB on both settings 5 and 6. The same for LSB. I puzzled over this for a while then noticed the 10.7MHz local oscillator was set to 10.6MHz and 10.8MHz on 5 and 6 respectively. I removed the 10.8MHz connection and re-connected 10.6MHz and all was well with perfect USB from my new oscillator set to 197KHz on 20m and perfect LSB set to 203KHz on 40m. Next I finished off the mechanical work.. screwing the tinplate to the underside of the box lid adding feedthroughs for the USB/LSB selection wires and the coax lead carrying theRF output, then conveniently locating the box etc.. I did notice that if the RF was adjusted downwards when below about 500mV the product detector stopped working. At about 650mV it works perfectly. Below the two RF outputs. A key component was C4 which acts as a filter to clean up any distortion in the sine waves.

  This is the 203 KHz LSB local oscillator signal

 This is the 197 KHz USB local oscillator signal

 

 
 

 

 A picture of the oscillator being powered from an external neg 15 volt supply and with a coax cable linked to the PR155G mode switch so that ranges 5 and 6 can be selected.

The black wire on the right is connected to the LSB selection input. The metal chassis will be screwed to the lid of the diecast box so it can be lifted off if any adjustments are necessary.

During testing I made some extra changes which are included in the circuit below. I found rectified RF was present at the varicap diode cathode which prevented the minimum setting of zero volts (= max capacitance) so added a second varicap as shown. This does reduce the maximum swing by about half but with the extra 3.5 volts of DC which was available and the addition of a small trimmer, a decent tuning range was achievable.

Note the odd junk box components viz. switching diodes made by Lucas and ITT zeners.. both long obsolete.

 

 

 

 I fitted the diecast box onto the inside of the rear panel and wired it to the mode switch. It worked perfectly and I was able to tune 20m, 40m and 160m SSB using a random wire and an 80m Inverted Vee. 80m was dead although local noise was pretty bad and might have been masking signals.

 

 Now that the receiver is working fairly reliably I decided to look at the anomalies. Listening to SSB stations was OK.. nice and stable but with the occasional loss of lock. This is maybe due to component drift in the phase lock loop and although annoying may not be a serious fault. On the other hand if I use a scope to look at say the 100KHz external output (ie. the 3rd IF) the signal is jumping up and down all the time so much so I was inclined to think it was the scope playing up as received signals sounded OK.

Time to follow the maintenance manual. It said to fit a 75 ohm load resistor across the 100KHz output and as I intended using a scope rather than their VTVM I fitted a 50 ohm 10dB attenuator at the 100KHz output socket but forgot I'd done this when I swapped to using the DSA815TG.

The first step is to check the 100KHz IF output and see if it follows the guidelines. The bandwidth switch is set to 6KHz and the AGC line measured with a multimeter. This was neg 3.25 volts and I could duplicate this figure with the manual gain control. With an input at the aerial socket of 2uV at 60KHz the 100KHz output (with a 50 ohm termination) measured at the high end 120 to 130mV RMS, but switching to the second test at 29.9MHz proved that the loss of lock wasn't intermittent but a permanent feature above 23MHz. I carried out the test at 22.9MHz and found that overall sensitivity had dropped such that the IF output was now less than 25mV. This is something I'd already noted from listening to broadcasts.

The oddness of the manual gain was revealed when I used the spectrum analyser instead of the scope. Usually say on 40m, turning down the RF gain resulted in a surge of noise, whilst turning it clockwise reduced both noise and signal. Below are examples with the gain at different settings.

Read on to discover why the RF gain control was behaving oddly.

 

 

 

 

 These pictures show the 100KHz IF output with (top left) AGC ON, above with the manual gain setting at nominal and lastly with RF gain set fully anticlockwise at minimum.

The picture on the left corresponds to the setting resulting in excessive noise although with a weak SSB signal buried but strangely at its loudest. You'll note the 100KHz signal is at -18dBm, -8.4dBm and -7.9dBm respectively.

Clearly the receiver gain is far too high resulting in severe overloading somewhere in the RF/IF amplifiers.

The 100KHz picture on the scope looked OK with no obvious distortion but somewhere in the various amplifiers there must be a sawtooth shaped wave as this would account for both even and odd harmonics. You can see opposite that even the 13th/14th harmonics are present. Each module in the PR155G is indendently driven by the AGC/Manual gain control so it could be any, or even several modules, where the associated AGC preset is wrongly adjusted.

 The way the AGC works is that a voltage proportional to the level of the 100KHz signal, close circuitwise to that at the 100KHz output socket (the signal shown in the pictures above), is sent to the AGC amplifier (Module 8) where the master AGC controlling signal is developed. A "reference" voltage is also developed (actually from a 4.7 volt zener diode) which is used by the attenuator diodes used for setting the amplification in the following modules.

I checked the type of diodes used in the AGC system and found a mixture. In Module 4 is either a 1N4148 or a BAY36 (conflicting info in the manual) but four of those in the RF Amplifier (Module 1) are Plessey 21SG, with other modules using the BAY36 from STC. The BAY36 is similar to the 1N4148 but so far I haven't seen a spec for the "21SG".

(1) RF amplifier (Module 1) which includes a preset gain control pot; (2) 1st IF amplifier (Module 4), (3) 2nd IF amplifier (Module 5) which has an internal gain control pot; (4) third IF amplifier (Module 7a) which has an internal gain control pot

Module 8 includes an AGC threshhold setting pot which dictates the signal level at which AGC commences.

 

 This is possibly the best setting for the AGC system. An incoming test signal is producing this 100KHz IF output with a minor 2nd harmonic.

I was able to see this curve with aerial inputs from microvolts to a full volt. Similarly when manual RF/IF gain is selected this output level shouldn't be exceeded otherwise you'd see the results shown in the three pictures above.

The figure of about -28dBm is measured with a 10dB attenuator at the IF socket so would be -18dBm at the receiver or circa 30mV into 50 ohms.

 

 As I investigated the various preset pots in the modules, and adjusted these to reduce gain, the penny dropped. The Plessey designers used 1N4148 diodes as attenuators throughout their modules and, for all diode cathodes, established a common reference voltage from a 4.7 volt zener diode located in Module 8. All modules employing AGC carry AGC and Ref at the same numbered pins which is useful. What I hadn't noticed was the orientation of the diodes and the true (negative) reference voltage when I had been puzzling over the RF gain pot wiring. The attenuator diodes are more forward biassed to establish more attenuation and as their cathodes are at the 4.7 volt reference voltage you'd expect the RF gain pot to subtend a control voltage of something in excess of 4.7 volts. Wrong... the reference voltage is neg 4.7 volts with respect to ground because the power supply is neg 15 volts with pos 15 volts grounded. To make the diodes attenuate/conduct requires the RF gain pot to supply a more positive voltage than the reference voltage. That means as the RF gain control is rotated clockwise (to increase gain) the wiper voltage needs to approach neg 4.7 volts from the direction of zero volts (= ground) and anti-clockwise to increase diode attenuation (hence an increase in voltage with respect to neg 4.7 volts which means a lower number) so for example neg 3.5 volts results in more attenuation then neg 4 volts. Spot the differences below!

 

 

 

 So, correcting that error in wiring should finally resolve the oddness regarding the RF gain control. As it stood turning the control anti-clockwise made received signals suddenly temporarily increase in strength before massive overload due to badly adjusted module AGC pots before vanishing in a lot of noise. Turning clockwise hardly changed the signal. The change in gain was masked by a huge amount of overloading resulting in a huge amount of background noise. Using the spectrum analyser, whilst tuned to Radio 4, showed the 100KHz IF signal went from the clamped level (governed by the 4.7 volt reference zener) of circa -28dBm progressively to -3dBm as the RF gain was turned fully anti-clockwise (which is clearly wrong). In other words, turning the gain pot clockwise weakened the signal whilst the audio level control masked this by making the output constant whilst allowing the background noise to rise. Essentially, turning the pot clockwise just increased the overall noise level. Very confusing.

From the resistor values shown in the drawings above you can work out the voltage at the pot wiper applied to the (manually controlled) AGC line. This isn't a simple matter because current is drawn by the AGC circuitry. In my previous experiments I'd worked out that the pot wiper voltage is reduced by about 2 volts once wired to the AGC line and, because of this, the designers had made a change to the resistor values. During my initial experiments things were completely upset because the pot was open circuit across some of its range.

When you're designing transistor circuitry you'll frequently come across the problem of calculating voltages due to current variations and so, in the case of the RF gain circuit, the designers decided to increase the current through the control circuit to minimise any unwanted feedback. Normally one uses a worst case of a ten to one rule so if the AGC line drain is 2mA one would make the minimum current through the RF gain circuit 20mA. In practice this may not be practical or enough.

Originally the chain was 680/359/120 ohms with the 359 figure representing the 1K pot in parallel with 560 ohms (resulting in 13mA through the chain or a ratio of 13/2 = 6.5) where the AGC circuit draws 2mA.

This was apparently changed to 68/91/142 ohms with the 91 figure representing the 1K pot in parallel with 100 ohms (resulting in 50mA through chain or a ratio of 50/2=25) again where the AGC line draws 2mA.

One of the errors I made earlier was due to a strange coincidence ie. reading the 68 ohm resistor as 680 ohms and a 12 ohm resistor (obscurely in series with a second hidden resistor of 120 ohms) as 120 ohms (because both were hidden away and the digits were the same as the details in the manual). Calculating AGC line voltages using the actual (amended) resistor values now makes sense and fortunately I can easily change these parts as I'd soldered them in full view (and reverse the wiring to the pot as shown above).

After (randomly!) changing the module gain settings to what I hope are close to acceptable I air-tested the receiver, but just as I began (around 5pm) an enormous noise erupted across the entire broadcast and amateur bands. I traced this using a portable radio to our next door neighbour's mains wiring as he runs a pair of easily accessible and unprotected cables across the corner of my garden to his garage.

The next day I swapped around the RF gain pot connections and changed the resistors to 68/100 and 120 ohms, however because signals became distorted at max gain, I had to find the optimum resistor connected in the power feed by inserting a test pot, setting max gain to a little under neg 4.7 volts at the AGC line (this turned out to be 310 ohms) then substituting fixed resistors of 240 and 70 ohms to make up the value. The S-meter then moved smoothly from a low reading to max without distorting signals. I have a strong suspicion that the preset gain settings in the various modules influence the value of that resistor so 310 ohms may well need changing once alignment is complete.

The next task is to fix the PLL lock problem so I'm going to have to study an entirely different area of the receiver. The circuit boards shown below which are inside a metal box. Incidentally it's very easy to blow the 15v fuse located on the rear panel and also it's too easy to come into contact with mains wiring. The latter I resolved by fitting a plastic panel of the underside of the transformer wiring (just visible at the bottom of the picture below). Definitely a good idea for anyone opening up a PR155!

 

 

 Above shows the fairly clever way the Plessey engineers dealt with servicing the phase locked loop area of the receiver. Three circuit boards called "G", "H" and "J" are located in a screened box where each is held in place by a pair of springy metal locating strips which can be eased outwards allowing one or more of the boards to be slid upwards for adjustments. Two are shown in their service positions here.

All the interconnecting coax cables are numbered with plastic clips and any can be unplugged to carry out investigation or to make preset adjustments. Number 14 for example is visible lower right. I need to familiarise with what's going on in my receiver, as it's all very well showing the odd waveform in the maintenance manual, but if there's a fault it can be quite difficult to follow the instructions. The waveforms I can see on my scope bear little relationship to the diagrams and twiddling the pots on the boards has only a limited correspondence with what's supposed to happen.

The set works happily from zero to the 14MHz range but cannot lock beyond this. The manual suggests checking the MHz input to board G for correct selection. I tried this and I believe all is well except the waveforms on my scope are very messy. The 0-29MHz switch does sort of produce 35MHz to 64MHz sinewaves of the right amplitude (250mV RMS) at the correct switch settings, but large interfering spikes upset the scope's measurements. Are these normally present or is this a fault?

Similarly when I look at the sweeping waveform it bears little relationship to the clean picture in the manual. I did notice however that the sweep oscillator (a multivibrator) has a strange waveform and the sweep frequency which is pretty low (in the order of a second) can be heard on a (locked) demodulated carrier as a sort of breathing noise. This could be due to the absence of the top and bottom screening plates but of course it could be the result of a fault or even a design weakness. I measured the capacitors in the multivibrator and surprisingly both were pretty close to 12uF but with their associated timing capacitors a bit high. The theoretical square wave frequency output should be double 0.69 x 12uF x 47kohm or 0.8 Hz. Following the multivibrator is a simple differentiator which is said in the manual to produce a sawtoth waveform. Looking at the manual for the "B" version I note that the sweep oscillator frequency has been changed. This may have been done to avoid the 12uF capacitors and instead use more common 15uF values, but the difference is not too significant once the 10% component tolerances have been taken into account.

 

 

 

 Here are three interesting pictures, all of which were difficult to display on my oscilloscope due to the very low frequencies involved combined with much faster noise (note: "auto" fails to work due to this combination of signal and noise). Above is the sweep oscillator which has a frequency of 0.8 Hz (exactly as calculated), the timings of waveform edges look fine although the actual shape could be influenced by the probe and scope characteristics.

Above right is the differentiated output described in the manual.

Right is a view of the tuning signal fed to the local oscillator when the receiver is working and the PLL is locked. The scope measures the noise at about 9mV RMS and looks to be based on 99Hz and this latter may well result in modulation of received signals. If I strap a 100nF capacitor across this signal to ground the noise drops by half but, as this messes up the lock, something cleverer could be incorporated such as a very narrowband filter set to the key interfering frequency (likely to be induced 100Hz from the power supply. This modulation effect is likely to be a design weakness and from previous experience decades ago a T notch filter might resolve the problem.

 

 By carefully adjusting the various pots on board J I was able to get the receiver to lock up to 16MHz but above this I could find no combination of pot settings that gave reliable locking.

I looked at the local oscillator drive when the receiver is almost locking at 17MHz and this is the signal opposite.

The feeling I get is one of the PLL RF signals is perhaps low in amplitude** as this was the exact problem I was having with signals passed to the demodulator. In that case 500mV failed but 600mV worked fine. A design weakness perhaps which has shown up from component ageing?

Looking at the waveform I believe the horizontal sections represent locked RF output with the remainder unlocked.

** In fact this turned out not to be the case
  
 

I need to measure each RF signal at each of the three PLL boards and if one is low test the components looking for one that may have drifted or even gone bad. Working on the three PLL boards is slightly confusing because the pictures in the manual are upside down (fortunately... below.. I found more information which shows these the right way up). I'm not the first to be looking at this area of the receiver because, pencilled on the box are board letters, and I noted that pot J-RV4 is set fully anti-clockwise and, oddly at first sight, appears to have no effect when turned the other way (I found the voltage at the wiper wasn't changing so, suspecting the pot was broken, I removed it from the board. It turned out to be a serious design weakness in the pot design such that if the locking screw is turned too far the peg that engages with the wiper becomes dislodged and the pot no longer works. I fixed this and replaced it, noting to only turn the locking screw only just sufficiently in future. After a lot of experimenting I found the pot used for setting lock (J-RV3) works OK up to 16MHz (having cranked this up from 14MHz) but cannot be persuaded to lock anything higher. Maybe the signal here is the one that's too low?

I then found a module test document and below are the factory tests for the three PLL circuit boards G, H & J.

I can see there are a few (tech author) errors in these pages

 

 

  I don't necessarily need to unsolder all the wiring (just yet) because the tests indicate the type of DC levels and RF signals at important points in the circuitry. I reckon I need to find an alternative scope that works at the low frequencies and the higher frequencies involved because my GDS1102 seems to easily get confused, or if not, then of course I may be looking at a PR155 problem. I have a Tektronix and a rather large LeCroy (relative of the "million dollar" scope) that might be suitable. I'm still of the opinion though that failure of the PLL to lock is due to one of the RF signals being too low. My suspect is either the MHz feed from the turret as I'd expect the amplitude of the RF to fall as the higher frequencies are developed and if these fall too low may not provide enough drive via PCB H input amplifier VT1 etc.. or, if the turret outputs are good, then VT1 and its associated parts may have degraded.

Below.. Modules G, H and J.

 

The PCB is fed (top left) from the "Interpolating Oscillator" together with the "Turret" output (=MHz IN) (top right) and has what looks like a balancing control connected to the outputs of the pair of ring modulators which feeds the input to PCB H.

 

 This PCB is connected to the local oscillator (top left) and the output of PCB G (bottom left). The LO is rectified by the bridge rectifier (=phase detector) and the DC output goes to Module G above..

 

 PCB J has lots of pots and has connections to the Turret (RV3) the very low frequency multivibrator, Local Oscillator (RV4) and Phase Detector (RV1 & RV3).

 

 At this point something else needed attention, and after this had started yet another item required sorting out so the PR155G is still sitting up-ended on the bench until I get around to it... I finally finished overhauling my Eddystone 358X receiver and got back to the Plessey PR155G..

 

 Picking up the thread wasn't easy but I recalled wanting to see the MHz signal emanating from the turret so printed off several drawings. The MHz signal enters Module H at Input 1.

The frequency of this signal is dependent on a couple of things.. firstly the setting of the MHz knob (the test manual say this should vary from 35 to 64MHz equating to 0 to 29MHz) and the main tuning dial. This registered 1000KHz for my tests. In other words with the MHz set at zero the received frequency was 1000KHz and at "29" this would be 30MHz.

Two scans are shown below.

 

 

 

Fig 1

 

Fig 2

 

Fig 3

 Fig 1: The receiver tuned to 12MHz + 1000KHz = 13MHz

Fig 2: The receiver tuned to 11MHz + 1000KHz = 12MHz

Fig 3: The receiver tuned to 11MHz dead

I could see that the MHz signal was established fairly quickly as the MHz knob was turned from zero to "14" but from "15" to about "23" the signal was intermittent going in and out of lock and above this it was hopeless.

Ignore the signal levels because my probe batteries were completely flat necessitating use of the internal front-end amplifier in the Rigol.

I also tried setting the signal at precisely 49MHz and the receiver tuning read 11.697MHz.

The test procedure suggests with the MHz knob at "11" the MHz output should tune 48.2 to 49.4MHz. The differences between the local oscillator and the tuning dial represent the IF offsets.

 

 How do the various oscillators and mixers operate?

The 1st IF is 37.3MHz. A 10.8MHz or 10.6MHz oscillator is mixed with a 100KHz signal to produce 10.7MHz. Tuning over each one MHz range is carried out by a 2.2 to 3.4 MHz VFO giving a 1200KHz spread.

In Fig 1: Receiver tuned to 13.000.

50.291766-37.3= 12.991766MHz (37.3MHz 1st IF)

12.991766-10.7=2.291766MHz (10.7MHz 2nd IF)

2.29176-0.100=2.191766MHz (100KHz 3rd IF)

This represents the VFO frequency of 2.2MHz.

The discrepancy is merely the cursor setting on the tuning dial.

Time to recap on the details.

 

 

 Opening the top lid reveals this complicated view of screened boxes interconnected by coax cables.

The turret assembly lid has been removed so you can see what's in the screened box.

Comparing with the view of the underside below you can see that the phase lock loop circuitry is directly underneath the turret.

Click the picture below to see it in full.

 

 
 

 

 

 And here's a view of the underside which gives access to the wiring between the various modules.

This includes unprotected mains wiring (T2) which predates modern health & safety rules (when people used common sense).

Click to make it bigger.

 

 

 
 

 When this receiver was being designed and developed the technique of phase locking the local oscillator was a relatively new thing and, although the use of one integrated circuit (SN7470N) is employed, this is limited to what I believe was a later modification.

 

At the time, other parts of the Plessey Company were actively working on digital methods of achieving the same thing so basically anyone working on the PR155 has to relearn its analogue technique.

 

Years ago (circa 1978) I was in touch with the chap that was working on digital PLLs and managed to get hold of some early chips that enabled me to construct a 2 meter transceiver.

I discovered the remains of this in my garden shed recently and having some rare free time decided to dig it out and see if it could be persuaded to work. Over the last 45 years it's been getting a bit worse for wear so it may never delight 2m band users again.

 

 

The job of the turret is twofold. One is to provide the correct RF input circuits for the select waveband and the second to select the correct VHF signal (in the range 35 to 64MHz) for production of the local oscillator. The latter uses the three wafers E, F1 & F2.

 

 Here are pictures of the turret and its three companion boards located underneath it.

It was an extremely tricky task to design an analogue phase lock loop that worked over a range of 30MHz using components available at the time.

When was this you might ask?

The date on the PR155G document is hidden away but is 1968 so the design must have been implemented, perhaps over the years 1965 to 1967. I started work at Plessey in 1965, and at the time we were still using germanium transistors. Although silicon types were available the British MoD were not convinced these were reliable enough to trust in the "Defence of the Realm".

Designing circuits with germanium transistors wasn't easy and, thankfully the PR155 designers datewise just managed to get to use silicon transistors (bar four). Even so, the PLL circuitry makes use of loads of pots and even the odd wiring change to make it work reliably (I counted a total of 24 potentiometers).

Also... 134 transistors and 77 diodes not including bridges and an SN7470N.
  

 The test specs rely on testing boards G, J & H using signals from external generators but maybe it's possible to test these as a set using locally generated signals as long as these meet much the same standards as described in the test specs?

As the PLL system is a closed-loop affair where does one start as many signals will be dependent on others? One firm area is Module 10 in which carries a 1MHz crystal oscillator. This is used to develop a comb of VHF signals by driving a transistor amplifier into saturation. I wonder how reliable is this circuit? Looking at the Module 10 circuits for the "B" and "G" variants I see there have been several changes so my first task is to see if these have been embodied.

Interestingly the inconsistency in the waveforms below may be associated with early circuitry?

Change #1: (Board A) The 1MHz feed to the harmonic generator is decoupled with a 1000pF capacitor in place of a 150 ohm resistor which will increase the amplitude of the 1MHz signal and filter any harmonics.

Change #2: (Board A) Transistor amplifiers VT3 and VT4 have some bias alterations.

Change #3: (Board A) VT4 has a 6.4uF electrolytic capacitor clamping its base and VT5 has a change of clipping diode from a BAY36 to a 1N4148.

Change #4: (Board B) The level of signal fed to the 48MHz selector circuit seems to have been provided with a "select-on-test" feature and again a BAY36 changed to a 1N4148.

Change #5: (Board B) VT3 bias change and a filter capacitor (C15) at the 48MHz output presumably to reduce unwanted harmonic signals.

 

I looked at the signal into Module G from the turret. These were combs with the "correct" MHz peak varying from 5 to 20dB stronger than the remainder as you can see below. Oddly (because it's a natural thing to do with low voltage circuitry) when you poked a finger into the turret circular circuit boards the comb nearly vanished leaving the prime MHz signal unchanged in amplitude. I unplugged the adjacent coax lead going to Module 10 and again the comb disappeared leaving the prime signal unchanged in amplitude which is to be expected. Note the spike about 2.8MHz HF of the prime signal.

Continue reading and all will be revealed...

 

Output 35MHz. MHz set to "0" (=0 to 1MHz)

 

Output 42MHz. MHz set to "7" (=7 to 8MHz)

 

Output 47MHz. MHz set to "12" (=12 to 13MHz)

Note the missing HF part of the comb.

 

Output 48MHz. MHz set to "13" (=13 to 14MHz)

 

Output 55MHz. MHz set to "20" (=20 to 21MHz)

Note the spurious half meg spikes!

 

Output 64MHz. MHz set to "29" (=29 to 30MHz)

 

 Tackling the loss of lock was a bit puzzling until I discovered Page 30 was followed by Page 32. This wasn't obvious because Page 30 and 31 had both ended with para (d) and Page 32 started with (e) (ie. a perfect example of Sod's Law)

At this point (in retrospect) I'll mention that the Plessey spec is too vague in its description of a working receiver. One reason for this was the limited test equipment suggested for adjustments. I bet a pound to a penny that the factory specs were infinitely better otherwise most of the receivers would have been rejected by purchasers.

After some laborious fiddling with pots I got the receiver working right up to 14MHz and beyond but a few days later it had lost lock over all its bands so I decided to remove boards G, H & J and test the components. Lots of the resistors were more than 30% out of spec so I guess I'll have to change these. I also found a small black component that is identified as a 10 volt zener diode but tests in-situ with my new "T7 Multi-function tester" as back to back diodes so I'll need to apply power and test it properly (it turned out to be fine). The receiver test spec does provide stand-alone board test specs so I'll do these and at least set some of the pots correctly.

I noticed during previous tests that one particular range (7MHz) had a cleaner VHF oscillator selection with better than 20dB to the next adjacent signal and this was similar to other ranges "with finger applied". I wonder if the turret either has a faulty part of maybe the poor definition is due to ageing? The various miniature coils have a setting compound so may not easily succumb to adjustment.

 

 Having got nowhere with twiddling pots to get any sort of reliable local oscillator lock I removed the three circuit boards G, H & J (see these below) from their housing and tested their components. Loads of resistors, especially on J were miles too high so I swapped all the types with similar values. Typically 10K were 13K and 1K were 1.5K. I'm pretty sure this in itself doesn't matter too much but where there are critical circuits, say around balanced modulators it may be important to have near identical resistors to aid "balancing". I hadn't really studied the transistors except for checking base/collector/emitter readings looked OK, but on looking at their codes I spotted a 2N708 (at J-VT4), a code I've not seen before. It seems this must have been substituted for a BSY95A by the guy that stuck little labels all over the boards. Looking it up I could see it's a switching transistor rather than an RF amplifier so I'm wondering if it's slugging the loop locking time?

 

 

PCB G with new 10K and 4.7K resistors

 

PCB H

 

PCB J with new resistors and a BSY95A swapped for an incorrect 2N708

In the centre, unusual 2S3030+2N1507+2S3030

I did solve one puzzle whilst checking the current into one of the boards. I read 47mA but as the transistors were running warmer than this sugested, I inserted a resistor in series with the 15 volt supply and is seems 100mA minimum is being drawn, so I need to take a look at my multi-meter.

Sure enough it had a few faulty parts!

Now, after finding no improvement, I wonder if the PLL boards can't cope with the poor filtering of the local oscillator spectrum? The table below indicates the level of the worst adjacent (=next) MHz and sure enough the worst differences seem to follow the bands where lock is particularly unreliable. The poor figures may be so bad the diode mixers used can't cope reliably. Unfortunately there isn't a "proper" definition for the MHz spectrum filter outputs, but the fact I can see a huge improvement to the difference whilst hardly affecting the primary signal when I press my finger on the selected coils seems to support my theory. I suppose, with all the potentiometers adjusted spot-on, the receiver may work but if any are slightly out then the MHz selection could be too poor for reliable locking?

The turret has loads of coils used for MHz filtering and all have their cores glued, but interestingly, clipped to the rear of the turret box is a tool (it was actually broken) for adjusting the cores. So... the presence of this tool suggests that one might have to make periodic adjustments and maybe therefore a spec for this? Below are the details given in the receiver documentation, apart from a sentence that reads that only the "required" MHz signal is sent to the PLL, and as you can read the only check is for a frequency counter to display the desired MHz signal. There's also a voltmeter reading which to me is bad because the comb of signals produces a total RF power which is read as a voltage and this is certainly not the voltage of the desired MHz signal. Carry on and lower down you can see more on this topic.

It's quite possible to adjust one of these cores to maximise a voltage believing that you're tuning the desired signal to a peak, when in reality you're tuning to a comb of MHz harmonic signals. In other words, with the wrong test gear you'll tune the coils to the worst possible setting for locking the PLL. The broken trimming tool does make me think that this has happened.

Below.. I rest my case!

 See the rest of the pictures above

 BAND

 RF Level Primary

 RF Level next

 Difference

 0-1MHz

 -57dBm

 -72dBm

 15dB

 7-8MHz

 -58dBm

 -80dBm

 22dB

 12-13MHz

 -59dBm

 -79dBm

 20dB

 13-14MHz

 -59dBm

 -69dBm

 10dB

 20-21MHz

 -59dBm

 -64dBm

 5dB

 29-30MHz

 -59dBm

 -64dBm

 5dB

 

One modification is apparent from comparison with the "G" version document, being the addition of R11 in the base of VT3.

 

 After changing the resistors for new accurate ones I tried to get the set going but got nowhere at first. Eventually I found a length of black wire that I'd missed. In order to remove the three boards their wiring needed to be unsoldered and an earth wire carrying the positive voltage from the 15 volt supply was unsoldered at both ends and I'd missed it. Next, I discovered one end of a coax lead was wired live to ground. I fixed this and started to get promising results. There are several miniature pots (5 on these boards plus 4 in the turret filter) and all need to be adjusted but, in only a unique setting for each, will the local oscillator lock correctly. Looking at waveforms is not easy due to the wide differences between oscilloscope settings (the PLL scan rate is circa 1 second). Eventually, by setting the receiver to Range 0 and 198KHz (ie. Radio 4 long wave) I was able to get lock and prove the three boards were capable of working. For some reason what I thought was correct reception seemed to be a false response (the wrong MHz signal?) but by very very slowly making adjustments first with one pot then a second and a third I was able to hear Radio 4. It was a case of gradually lessening the out-of-lock noises. The end result wasn't ideal because even switching to CW from AM lost lock indicating at least one pot isn't set properly (maybe in the turret). A picture of the three boards is shown below.

 

 

 I suppose the next step might be to adjust the turret pots although those MHz selection combs are still worrying me. Each of the four pots caters for multiple ranges and are apparently a "fiddle factor" setting different DC levels for increasing blocks of frequencies.

The adjustment of these is helped by setting a sync position on a waveform produced by the PLL scanning 1-second multivibrator. Doing this requires simultaneous access to the top and bottom of the turret box and maybe I can investigate the peculiar MHz output comb at the same time.

To be honest I find the instructions in the test spec to be puzzling. I can see that the VHF LO needs two signals to get it to find lock and remain in lock as the tuning dial and MHz selection switch are adjusted. It's a case of matching the control signals to the requirements of the LO so it's allowed to meet its upper and lower frequency outputs with enough slack for reliable operation. It looks like Plessey also prepared a second set of instructions because of confusion amongst the test people.

 

 

 

 These scope traces are measured by setting what they call the "interpolating oscillator", to 2.5MHz (it covers 2.2 to 3.4MHz), inserting 100Kohm between 1st local oscillator (Module 3) Pin 2 and its connection, then monitoring Module 3 pin 3 with a scope after breaking coax cable No.9.

It also says to monitor the reactor control on Board J to see these three pictures. Maybe it means summating two DC traces at Pins 2 and 3?

For the moment, I chose to investigate those MHz "combs" before the 4 turret pots.

Below is a picture of the turret wafers carrying the 60 filter coils for MHz selection

 

 

 This is the RF signal relating to the selection of the MHz input for the 29-30MHz tuning range. Because of the frequencies used in the various mixers it's 64MHz. For tuning 0-1MHz the MHz selection would be 35MHz.

 

 Amazingly this is the same signal after tuning one filter coil.

As you can see the unwanted harmonics have dropped in power by about 30dB which is a huge reduction from the previous 5dB.

 

 
 

 Adjustment of the filter coils is made through the holes in the side of the turret enclosure. You can just see the top of the core for the selected range.

Of course the process of adjustment may not be easy because a red-coloured locking compound has been used. In addition each slug has been treated with wax to make it stay in place. I found about a third of the cores could be adjusted but the rest needed to be extracted and their locking compound scratched off (particularly from in the adjusting slot)

A plastic trimming tool was clipped to the rear of the enclosure but this needed filing as it was quite soft and rounded off (the last owner??). The tool is particularly thin to get through the access holes.

I managed to get at least 30dB between all selected MHz signals and their neighbours, but be warned as its very easy in the case of the higher frequency signals to pick the wrong MHz signal (eg. 65MHz instead of 64MHz)

 

 Although all 30 right-hand cores needed tuning those on the left (the rear of the turret) seemed to be OK, maybe because they're more damped than the others? As I made the adjustments I could hear short-wave broadcasts suddenly start appearing, presumably as the PLL locked, so, cross fingers, the main reason for the poor locking may now be resolved.
 
 
 
 

 so..... "watch this space"

Go back to the first page about the PR155 puzzles 

 Return to Reception