Looking at the AR88LF


 This very fine example of an AR88LF was given to me by my pal, and ex-Plessey colleague, John McGowan many years ago. Since then I'm sure it's doubled in weight, and having finished the refurbishment (or as much as I wanted to do) of an R109A, I decided to investigate this receiver. My XYL and I struggled to get it from the rack in my workshop into our conservatory then after another struggle removed the receiver from its case. Before the case came off I'd plugged it into the mains supply and connected a loudspeaker (I have the proper AR88 one) and switched on. A short aerial wire was already fastened to it and I was quite amazed when Radio 4 on 198KHz boomed from the loudspeaker, presumably as it had been doing at the moment it was last switched off at least ten years ago?

 Click to see the circuit for the AR88LF

Having had previous experience repairing an AR88, I'd been somewhat apprehensive turning on the receiver on, but having removed the case, I found I needn't have worried because the flat brown condenser between anode and ground at the output valve had been snipped off. I wonder how many deaf AR88s are out there with open circuit output transformers? It's reported that failure of that specific condenser results from stress when the receiver is used without a loudspeaker. There's sufficient power developed in the output circuit that speech can be clearly audible from transformer vibration.

The AR88 models have a good selection of black metal valves and all these cleaned up nicely. Sometimes left in the damp these valves can get very rusty but all were gleaming after a rub down. In fact all the valves, including the glass 6V6GT, 5Y3 and the 0D3, have very clear markings and seem to be either very lightly used or have been replaced. I noticed the output valve is the correct 6V6GT instead of a 6K6GT (fitted in the AR88D) but the former although consuming a heater current 50mA more and having an anode load slightly higher can provide an extra watt of power (5.5W) at 3% less distortion (12%) so is a bonus... but it does get extremely hot.

Recently, I was given a couple of boxes of surplus meters and one of these will make a good S-Meter for the AR88 as this example was never fitted with such a thing. As I twiddled the dial on the latest receiver to be receiving some TLC I noticed that the tuning, although mechanically very precise, the same couldn't be said for Radio 4, whose signal seemed to be spread over far too much of the spectrum. It was hard to define the best place to listen to the broadcast so I pondered over fitting an S-Meter that would help.


 When I first tested the AR88LF it was very crackly and the audio kept dropping to a low level and back.. really annoying.

The RF gain control has a newish 6.8Kohm resistor grounding one end of the pot. Here you can see a lot of solder and nicely wetted steel chassis but unfortunately the solder hadn't quite reached the resistor wire so the joint is dry and hence intermittent..

 The commonly available document on fitting an S-Meter to an AR88 can be simplified by adding extra detail and some pictures for the LF version. The wiring is already in place but the hole through which the new sensitivity pot should be mounted is not drilled in my chassis.. Why is this I wonder? Before starting I'd recommend anyone planning to add a meter to first identify V5 (1st IF amplifier) valveholder and familiarise with the connections to the 6SG7.

Many AR88s are not fitted with an S-Meter and its position is occupied instead by a yellowed plastic panel screwed in place by 4 short US-threaded screws (coarser than 6BA) which penetrate some way into the rear of the front panel but not all the way through. The top two are slightly longer because they hold in place a bracket holding a dial lamp whose holder is pushed through a hole in the metal bracket. The first step is to remove this bracket and the lower screws... the latter are tricky due to limited access but can be removed with a bit of fiddling and perhaps with the help of pliers to loosen them. The yellow plastic panel comes off leaving a clear plastic aperture. To fit a meter you need to make a panel on which to mount it, then fasten the panel using the 4 original tapped holes. My meter had mounting studs facing rearwards and, as I didn't want to modify the meter case, had to be sandwiched between the mounting plate and front panel. This required a set of 4 new screws having the correct thread and long enough to penetrate through the new mounting panel into the front panel. The fitting instructions suggest removal of IF cans (they must be joking) but an easier method, if you have correct nuts, is to fit two locked nuts to the shafts of the lower screws so that a suitable spanner can be used to drive the threads into the front panel. To do this press your finger on the head of each lower screw, position the start of the thread in the hole with tweezers then gently use a spanner to rotate the screw until its secure.

Once the meter is fitted fasten the leads provided to the meter terminals, including any shunt required to give you 5mA full scale deflection. The leads may be darkened through age but the black/red coloured lead goes to meter negative (see later for advice).

There are now three jobs remaining. Firstly, fit a 100 ohm potentiometer to the rear panel so it's accessible once the case is in place. Secondly, identify the grounded lead intended for connection to the pot. This is to be found soldered to the ground pin of V13, the 150 volt regulator valve. Lift this off and connect to the pot. In my chassis a spare hole, which I suspect was added by a previous owner, for a now-missing coax socket was located on the chassis rear apron. This proved suitable for mounting the pot, and a few inches of wire soldered to the freed wire enabled me to reach the pot centre pin. A second wire to one end of the pot track returned to the regulator ground pin completed pot wiring.

The final task is to identify two wires soldered to the first IF amplifier valve base ground connection. One of these wires goes to the positive terminal on the meter and the second routed to the rear apron where it's soldered to the regulator ground pin. Lift off these two wires and tidy up their ends prior to resoldering. Cut the ground connection of the 100 ohm resistor fitted to the 1st IF amplifier cathode pin, pull the resistor so it's sitting vertically then tin the free end ready for soldering. Find a suitably-sized decoupling capacitor (the official screed says 4700pF but I chose a 47nF x 250v capacitor because it was a perfect physical size for the job) and solder one end to a suitable ground connection leaving the capacitor mounted vertically. Solder together the free end of the 100 ohm resistor, the free end of the new capacitor and the two wires freed from their grounding point.

You now have in place the S-Meter and its wiring. If you're not confident or if you're not able to recognise or identify the correct wires involved in the exercise (for example if they're discoloured with age), don't connect the meter until the wires are checked out. Once you've unsoldered the three wires that were in the original harness you can buzz out their ends. One freed from V5 Pin 1 goes to the meter positive terminal, whilst the other goes to the end of the wire freed from V13 Pin which should be connected to the wiper of the new pot. The remaining wire at the meter is grounded and connects to the meter negative terminal. Turning on the receiver should result in a meter deflection but it's a good precaution to have previously wound the new pot to minimum resistance before switching on. Gradually increasing the shunting resistance should give you a mid-scale reading before testing with a tuned signal.







 Above.. fitting a new S-Meter with a replica off-white tinted scale. The fixing plate was made from a piece of scrap fairly flexible non-brittle plastic which has the benefit of flexing and giving more leeway in fitting. Positioning this non-official meter was tricky and I might move its position downwards slightly by enlarging its mounting hole to make it more symmetrical at a later date. The published scale (thanks to N3FRQ) was made to fit the meter by measuring the needle pivot point to the scale edge (=25.4mm) then setting the printing size factor to make the print match this measurement. I glued the paper to the reverse of the original meter plate using a very thin coating of spray mount adhesive to keep it dead flat and to avoid contact with the needle.

My choice of meter was one from my collection. It was a centre-zero 400uA fsd instrument having sufficient pointer adjustment, by sliding both upper and lower positioning levers, to make it line up with the right hand scale limit. I added a 68 ohm shunt resistor to make the fsd roughly equal to about 5mA to match the official AR88 meter.






 Left... the wiring around V5 after modifications and above... the new potentiometer fitted in a surplus hole. Avoid using any hole provided for reaching trimmers etc.

Now that I've fitted a working S-Meter I can clearly see several major problems with the receiver. Tuning across a strong broadcast it occupies far too much bandwidth with several woolly peaks due to poor IF alignment and meter deflection seems to have little bearing on tuning. The RF gain control setting has an odd effect on both recovered audio and the S-Meter reading. With a long wire aerial, broadcasts are roughly on the correct dial markings but, as the control is rotated to increase gain, an initial increase in volume is followed by a quietening then blanking out of audio. The S-Meter reading rises then reverses, dropping back again. Clearly AVC action is almost completely absent on one or more stages and the huge gain from the AR88 circuitry massively overloads the later stages. First I'll examine resistor values, then replace most if not all decoupling condensers. A pound to a penny most of the bathtub condensers (eg above left) are duff!

Hopefully.. now that the S-Meter is present I shall be able to monitor the results of repairs and tweaks. Using my high impedance meter and tuned to Radio 4 on 198KHz and looking at the AVC line.. starting at V8 where the incoming IF signal is fed into a diode cathode and rectified at the diode anode Pin 8, I can see minus 15.3 volts, which is decoupled at C48 and routed through a pair of 560Kohm resistors, R23 & R27 to the 1st and 2nd IF amplifiers V5 & V6. The AVC bias is also routed via a 100Kohm resistor R9 to the pair of RF amplifier stages V1 & V2, where it's further decoupled by C47, 4700pF. Checking the grid bias voltages at V1 & V2 showed around minus 13 volts which isn't too bad. However, checking the control grids of V5 & V6 indicated all was not well as these were sitting at minus 3.9 volts and minus 2.9 volts respectively showing a leak equivalent to some 12 volts for both.V5 & V6 grids are decoupled by C76 and C93, which oddly are the top and bottom condensers in the bathtub in the picture above left, and only millimeters from the new S-Meter wiring.

I unsoldered C76 & C93 and temporarily soldered in place a pair of new 0.1uF plastic capacitors then switched on the receiver. Radio 4 came on smoothly and remained perfectly clean no matter what the RF gain setting. With the RF gain at maximum, the S-Meter moved smoothly up to a maximum of almost 80dB (I am using a 350 foot long wire) and back down again as I tuned across the station and the bandwidth control roughly worked when I switched it to its different settings. The receiver was transformed, with the extra wide bandwidth, multiple tuning humps, backwards reading S-Meter and audio distortion/blanking gone. I again tested the AVC voltage and found it's now within about a volt at all the points in its circuit and the S-Meter works perfectly. But.. there are a further 16 bathtub condensers left....



 This is a picture showing the main components in the AVC circuitry, missing out RF components etc.

V1 and V2 are RF amplifiers, V5 and V6 IF amplifiers and V8 the AVC rectifier.

The condensers are located within two 3-section bathtubs. Resistor values are not particularly critical, neither are condenser values as long as these do not leak current.

Note that original AR88 schematics use the letter "M" for Kohm with "Meg" for Mohms.

 Now for a computer problem. I decided to tackle the IF alignment of the AR88LF so moved my Wavetek signal generator into the conservatory where I'd decided to work during the spell of nice weather and not shut myself away in the workshop and my XYL. I'll just check the IF before I get underway... but my XYL is busy using our computer and my battery charger for the laptop disappeared after a tidy up of the office... but no worries as I have the AR88D instruction manual and 455KHz looks fine. I connected the Wavetek to the aerial tag and set it to 10mV so it would get through any input rejector. I heard a slight hiss but no sign of the modulation tone so I upped the input to 100mV and the hiss increased in volume and a slight tone in Waveband 2 which seemed to tune which didn't feel so good as the IF input signal might change slightly but shouldn't tune sharply as this did, so I switched to Waveband 1, the LF range... the hiss was now present and no tone so I increased the input to 1000mV.. then tuned the Wavetek up and down. This takes ages because you really need to push the up/down buttons a lot and moving in my TF2008 would be awkward as space is limited.. No luck so I moved the input lead to the mixer grid. I could now hear a tone but nothing like as strong as it should be. Eventually my XYL vacated our computer and I checked my own website in the listing for IFs which I uploaded back in 2000 and discovered the IF was 735KHz. http://www.radiomuseum.co.uk/commsifs.html

Reconnecting the Wavetek to the aerial and reducing the input to 1mV at 735KHz worked a treat and showed me the IF alignment was hopeless with the various bandwidth settings giving widely varying results. After some twiddling I realised alignment is not straightforward so I'll leave that to the spectrum analyser after moving the receiver back into the workshop. The reason for the complication is that some bandwidth settings use a crystal to help shape narrow IF filtering whilst wider settings do not so that it's important to base overall IF alignment on the crystal characteristics rather than a precise figure (in this case)of 735KHz. If this isn't done you'll find that changing bandwidth settings results in signals being off-tune.

In the meantime I'll sort out the bathtub condensers. A quick check with an ohmeter told me that in most cases the outer condensers are around 1Mohm and the centre several Mohms so I'll need to swap the lot. I'm a little unsure whether the 0.01uF examples can be changed to 0.1uF without degrading or altering performance. For example, changing the AVC response? A resistance of 2Mohm and a decoupler of 0.01uF gives 20mS and 0.1uF will give 200mS so in practice, with a swing of around 10 volts it shouldn't really matter.





 The picture above shows three pairs of 100nF capacitors and on the left three single capacitors. These are rated at 500v working. Initial tests showed resistance reading around 1.5Mohm each and with the new capacitors these resistances were basically not measurable or equal to the resistance of surrounding circuitry. Common earth points are made by adding a solder tag under a bathtub screw.

Here's a couple of Canadian bathtub condensers all of whose sections are used for decoupling . Top left a 3 x 0.25uF x 400v working (C99=audio amp bias, C112=audio amp screen, C113= audio amp anode supply) and below this a 3 x 0.1uF x 400v working (C56=mixer screen, C76=1st IF amp AVC, C93=2nd IF amp AVC). See the simplified sketch above.

I was going to open these and replace the innards but they may be filled with a dodgy type of oil and the larger started to leak when I applied a large soldering iron to the back where a plate is soldered in position.so I decided to instead use surface mount capacitors. I used two different methods as you can see. It wasn't easy because the wiring which is cotton over rubber is perished. Luckily not to the extent it needs replacing, but its very stiff because the rubber has hardened. Removing the end of a wire reveals corroded strands which fortunately respond to wetting with a hot iron and tin/lead solder. During this process the rubber melts and the results looks messy but is perfectly sound. I used a junkbox bathtub whose three section all read about 250pF in place of the badly leaking one (to keep some semblance of authenticity!)


A quick test then showed Radio 4 LW now at +84dB on the S-Meter, a 6dB improvement.


 I treated this 3 x 0.1uF 400v leaky bathtub like the first, using its tags to mount the grounded ends of the capacitors (C79=1st IF amp anode supply, C84=BFO anode supply, C92=2nd IFT bandpass).

Manipulating wiring is difficult because it's very stiff and brittle with hardened and cracked rubber insulation. Fortunately the cotton covering keeps the wiring from developing shorts.

Radio 4 is now at nearly 90dB on the S-Meter and the shortwave broadcast bands are full of very strong stations. The BFO centre position has moved slightly from 735KHz due to the new C84 capacitor.


 I completed the swapping of new surface mount chips for the bathtub condensers. As I bought a vast number of 100nF x 500v of these it seemed logical to change everything up to 0.1uF with single chips and use a couple in parallel for the few 0.25uF condensers. The first three bathtub swaps were fine then I swapped out the third and everything was still fine, but after changing the final couple the IF strip burst into oscillation. One set was 3 x 0.1uF so I left these but the other was 3 x 0.05uF so I added a second capacitor in series with each and the oscillation stopped. Presumably the overall gain has oustripped shielding with the extra decoupling?

The set was working after swapping the block containing C68=? , C109 and C110= in parallel, decoupling noise limiter circuitry.

Oscillation started after the final pair had been swapped. One block carries 3 x 0.1uF but the other 3 x 0.05uF includes two decouplers that explain the oscillation, being a cathode decoupler for V7, the third IF amplifier and its screen grid. Both would contribute to extra IF gain and to make matters worse this stage is self-biased. C71= g2 for V5 & V6, C95= anode supply V6, and C102 decouples the band-pass filter between V6 and V7. The other has C103=g2 for V7, C106= cathode of V7 and C107=anode supply for V7.

During testing after bathtub condenser replacement the S-Meter stopped working. The reason was that Radio 4 was now so strong that the meter reached over 90dB on its scale and the needle jammed because the extra paper thickness had cancelled the minute clearance between needle and scale. I had to detach the meter, open it up and bend the pointer away from the scale. Once done and refitted I changed the 68 ohm shunt and fitted an 82 ohm shunt which allowed me to set the pointer at zero reading for noise-free reception (I selected a frequency near 30MHz for this.

Now, tuning Radio 4 gives me something over 80dB (In terms of the AR88 S-Meter 80dB over 1uV, or if S9=100uV +80 represents S9+60dB) which represents almost zero current through V5, the 1st IF amplifier.

The next stage will be to accurately align the IF amplifier which means carrying the AR88 back to the workshop, although very slightly lighter less its outer case.







 Results of IF alignment are shown here; BW1 to BW5. I'd suspected the rough method using an audio power meter had resulted in a dodgy response and sure enough the wider bandwidths were skewed about 5KHz HF of the crystal frequency. It took a lot of adjusting to move the response downwards to match the crystal, but now the bandwidth can be narrowed without having to retune the signal.

These scans of BW1 to BW5 have a 200KHz span (20KHz per vertical division) so that you can see the skirts of the responses close to -70dB.

Below I've shown the curves for BW1 and BW2 (which I use for comfortable listening) with a narrower span where each vertical division represents 5KHz.

The user handbook has curves shown to -30dB and for the widest setting (BW1) has +/-21KHz with the narrowest (BW5) +/-6KHz. The curves which I measured show +/-12KHz and +/-5KHz respectively so are nicely in spec.





 Click to see a larger picture.

I don't know why this official picture is upside down from convention. Presumably the graph needs to have a zero reference although the explanation is simple: the axis title "Times Normal Input" means the same thing as attenuation in terms of the resultant curves as long as AVC is disabled.

The scales shown on the vertical axis are logarithmic with wider spacing for low numbers and smaller for high numbers. Usually these sort of scales are rationalised as in the spectrum analyser pictures above to show even spacing. 10, 100, 1,000 and 10,000 are the same as -10, -20, -30 and -40dB and early test equipment would not have been able to work at levels that are common nowadays. The DSA815TG shows noise at -70dB indicating the huge improvements made since the 1940s.

It's interesting to note that the official spec here is a drawing and as such will be a somewhat theoretical indication, hence the perfect symmetry compared with the real curves shown above.

The notes about the crystal indicate that this is not in circuit for the two wider bandwidth ranges and does not mean the crystal should be unplugged.

 I've noticed that the audio sounds very clean no doubt helped by the 6V6 operating in Class A. Listening to Radio 4 on the AR88 and on a digital radio there's no comparison. The AR88 sounds a lot nicer than the rather muffled digital receiver. The circuitry around the audio amplifier is quite interesting. The designers used oil-paper decoupling condensers throughout, but using lozenge-shaped mica condesers elsewhere. There are two 4700pF of these mica condensers in parallel connecting the audio amplifier to the 6V6GT output valve and appear to be leak-free as the 6V6 control grid is solidly negative at minus 16 volts. If you look at the impedance of this pair of condensers and compare this with the 6V6 330Kohm grid leak there's a considerable loss of fidelity of some 30% at 50Hz. This however is dealt with by shunting the 6V6 grid with a 560pF mica condenser and, a tone control using a 4700pF mica. These additional parts tend to restore some audio fidelity. Another component was also added, this is C119, a 0.003uF-1000v oil-filled, or 2700pF-mica (depending on which issue of the documentation you have) which again acts to top cut the audio and help with fidelity, but because this is prone to failure it is often cut, as I determined when I first began this exercise. Without this extra top cut you will not experience the designer's intended results, however.. as one gets older your ears will compensate and give you the desired top cut automatically.. so you don't really need C119 after all once you turn a certain age...

 Before replacing the RF screens and putting the chassis back into its case I decided to remove the new decoupling capacitors that seemed to be the source of IF instability and put back the connections to the bathtub. That done the IF instability went away.


 Return to Reception