McMichael 381

A second old family set arrived in June 2021 delivered by the owner of the Columbia C301. This example has been clumsily painted and a rather floppy garish speaker cloth fitted. It belonged to a late neighbour of the current owner who'd like to hear it working.

Click the picture above to see its data sheet from 1939.

The set is definitely not serviceman friendly because it's in two main sections.. the receiver chassis and the loudspeaker and push button tuning part and connected together by two sets of cables which must be carefully unscrewed to separate the sections before either can be extracted from the cabinet. I'd noticed the tuning knob didn't really do much and after removing the knobs and disconnecting the cable harnesses I found lots of dried grease jamming up the ancient slow motion drive. I also discovered the glass dial had dropped and was sitting at a slight angle hard up against the tuning mechanism. The markings on the glass were poor, particularly the station names. Either someone cleaned the glass without due care or the lettering was originally radioactive paint which has now faded. to attract customers there are special indicators for the tone control and waveband selection driven by levers coupled to the switch shafts (those white panels either side of the main dial, above).

Although dating from 1939 the valves it uses are generally older by a few years viz. RF Amp/Frequency changer AC/TH1 (B7); IF Amplifier AC/VP2 (B7), Detector/AVC/Audio amplifier HL41DD (MO), Audio output AC5/Pen (B7) and Rectifier UU4 (B4).

Below.. a collection of pictures, starting with a 1939 advertisement.

The dial has a cloudy look about it and the backing and dial markings are faded.

This card dated 1990 was lying on top of the receiver chassis.

It's in fact a greeting to a friend in Turkish.

Not very visible, being inside the case and partly covered by parts of the radio. Left.... presumably it was up to the customer to request the shopkeeper to set stations for the set of push buttons?

These pictures show the connections which need detaching before the chassis can be removed. One set (above) carried the six "mains energised" loudspeaker connections and left the six cables for the push buttons.

A new rather unsightly white cable has been fitted as a mains lead and for some reason the safety earth is not used and quite definitely insulated.

On the left is an early electrolytic condenser. Some of these early receivers have first class components still in perfect condition but until I've carried out investigative work who knows what I might find. The owner said it lit up and nothing else...

I removed the glass dial with a litte difficulty as it had glued itself in position with the four rubber mounting pads which had decomposed. Oddly the whole dial had slipped downwards where it was jammed against the tuning mechanism. The backplate has a strange finish which results in a cloudy looking dial. Was it originally matt black? You can partly see the mechanism for the indicators for tone setting and waveband, but the tone control switch idents feel very sloppy (this turned out to be just a loose screw).

When you're working on an old radio chassis it's essential to protect fragile parts such as dials and valves so I always add supports, usually a wooden frame but in this case it was more convenient to use some scrap metal.

Often the manufacturer would assume enough servicing could be done via a detachable panel under the case, but major works need the cabinet to be removed.

Later I removed that mains lead to gain access to the rear of the chassis where a couple of block condensers are fitted.

I haven't decided on what to do with the dial. The original glass lettering is so faded to be virtually useless unless it's luminous (the scan below is greatly enhanced with PhotoShop). The backplate is finished in a matt brown and I could stick a paper print onto this to aid tuning and looks. The problem is the pointers (three) are behind the metal plate and show through slots. It's possible that the dial lampps could project a shadow onto the paper, otherwise I'd have to cut slots in the paper (removing the numbered scales). There's also the other slots for the tone and wavechange indicators.

Another possibility is to print an enhanced scan of the dial onto a transparent film, but I think I've decided on the best solution which is to glue the negative picture (now PhotoShopped below) to the backplate with slots cut so you can see the three pointers and markers for the tone and wavechange switch.

Many components are fitted vertically at terminal strips below the valveholders and at the rear of the chassis are a couple of waxed card blocks each containing some condensers. The smaller carries a 50uF and an 0.1uF whilst the larger one C32,100uF; C33, 16uF and C34, 8uF. The last being te smoothing and reservoir condensers. A replacement for one of the high values has been added. This measured 15uF and more than 20 ohms ESR. C29 meaured with a leak of greater than 20 ohms and zero capacitance. There are a total of 17 old condensers (excepting those used for RF circuits which are usually OK) that will need removing, testing and almost certainly swapping. The service data tells me I can buy replacement packs C29/C30 for 4/6d (=22.5p) and C32/C33/C34 for 10/6d (half a guinea no doubt! = 52.5p) .

Below you can see test results of the pair of block condensers.

The larger of the two had an open circuit 8uF with highish readings for the 1uF and the other 8uF. The case was bulging and the 16uF had already been swapped for the blue tubular condenser shown below.. also u/s. "ESR" means the internal resistance was miles too high for the condenser to work satisfactorily (and in the case of HT types they would get very warm).

The first task will be to swap bad condensers, starting with the blocks above. Numbering as per the service data sheets.

C12

C13

C20

C21

C22

C24

C25

C26

C29

C30

C32

C33

C34

Marked

0.1uF

0.02uF

0.002uF

0.1uF

0.001uF

0.06uF

0.005uF

50uF

0.01uF

100uF

16uF

8uF

Actual

14nF

ESR

10nF

open

ESR

I checked the most critical condenser C30 but found it was a Hunts moulded component rather than a waxed type. It was very difficult to get at (note the 10 minute fault-finding comment in the service sheets) but in fact most parts are touch soldered in place so I removed it and exchanged it for a modern 15nF capacitor. Much to my surprise it measured only 12mV volts in my test circuit, using 100Kohm and 300V, representing insignificant leakage. An adjacent wax condenser, C26 marked 0.005uF was 14nF and had a leakage of about 5Mohm so I replaced that although it would have been serviceable. At this point it was time to power the reiver and see if it worked... maybe some work on the tuniung arrangement though because dried grease in its works made it occasionally stop twiddling.

Here are some views of the second part of the radio carrying the mains energised speaker and preset coil circuitry.

The rear panel is held by four black wooden posts, two of which have become unglued so I fixed those and need to wait for the new glue to set. The speaker cloth is not ideal and certainly needs stretching in place because it's very loose. The choice of cloth echoes the choice of paint and really both should be swapped. This could be done by the owner as it would be a time-consuming task to remove the paint and restore the original finish.

I connected the speaker/push button assembly to the main chassis via its 12 leads.

The next step was to power the receiver to ensure it hasn't got a serious problem and if OK work out the jobs needing to be done to get it working.

As the dial glass isn't fitted I printed a paper dial, cutting slots so I could see the pointers and the waveband and tone indicators. Printing the dial was a little awkward but, by measuring the size of the glass dial, I was able to copy these measurements into the printer details. I used a scapel to cut the slots. The final dial could be improved... possibly using the negative version shown previously.

I connected a temporary mains lead and switched on. The dial lamps came on but after waiting whilst waggling the wavechange switch with the volume control advanced I couldn't hear anything from the speaker. The easiest way to proceed was to power the receiver from the mains transformer, unplug the rectifier valve and connect a variable power supply across the HT rail. Cranking the HT to 200 plus I heard some life and carefully tuning using a long wire pulled in Radio 4 at low strength. Monitoring the AVC line showed zero volts and in fact during further tests the AVC line didn't rise above 2mV. I connected my HP signal generator to the aerial socket and found a response at around 470KHz. Setting the HP to 465KHz AM, I tuned the pair of IFTs and after going around a few times using a wattmeter plugged into the external speaker sockets the receiver became very sensitive. Radio 4 was now strong and after some temporary alignment medium and long waves sounded pretty good. Shortwaves however proved a problem. I could hear a 15MHz test signal at the HF end of the band (=20m), but as I tuned the HP downwards whilst tracking with tuning, the receiver stopped working suddenly at around 11MHz. Either the frequency changer oscillator is packing up due to a bad resistor or bad condenser or the AC/TH1 itself has low emission. Fortunately I have a few spares so I can at least quickly check the valve which seems to work down to around 70 volts HT. (but not so easy to sort out.. see later)

I noticed several problems.. the AVC line is dead, the tuning needs an overhaul, the tone switch has too much attenuation in any setting other than the "Fidelity" position. Switching to the presets, these work but the alignment between oscillator and RF coils is hopeless, also I suspect there's a fault between the aerial socket and the RF amplifier grid as putting a finger on the top cap of the frequency changer increases the audio quite dramatically. This could be due to a tracking problem in the RF circuits or a fault. I need to study the circuit diagram to see what's involved.

Some of this section of the circuitry is carried on the loudspeaker unit.

This also carries L5, L6 (part of the loudspeaker) and the loudspeaker which means the HT line is carried on fairly long cables between the chassis and loudspeaker assembly.

Note the link between the tone control and the RF circuits. Is this designed to reduce any chance of overloading the set from local broadcasts? In fact it's to broaden the response to improve fidelity.

Connections "X" and "Y" are used to supply the frequency changer with completely alternative RF and oscillator coils.. perhaps a poor way to arrange presets because of the extra expense involved?

A similar method was used on the Murphy A72 also made in 1939, but in that case the preset module was sold as an extra.

Alignment is straightforward except for the note on push button RF tracking. Shortwave setting is midway between 20m and scale top. Mediumwave setting has a mark on the dial but represents 214m. Longwave setting is 1125m although with the shortage of broadcasts it may be better to use say a weaker broadcast than Radio 4 such as Luxembourg? Use test signals below AVC level.

As the AVC circuit uses 500Kohm resistors it's important that the decoupling condensers are leak-free. These are all 0.1uF; C4 and C5 (essentially in parallel) in the RF assembly and C20 at the 1st IF transformer. During initial testing the AVC voltage was only a few mV.

After some more testing the AVC voltage worked properly. This was due to fitting three new capacitors replacing C4 and C5 plus C8 (0.002uF) and adjusting the trimmers. The push buttons worked but the audio was far less than direct tuning. I traced the problem to a small preset T13 which had no effect. Replacing it with another plus a fixed 25pF enabled the RF to be boosted to roughly the same level as direct tuning. I found a few anomalies though.. One is that Radio 4 appeared twice within the LW tuning range. Assuming the LW coverage is 150 to 300KHz the local oscillator tunes 615 to 765KHz with 663KHz tuning Radio 4 on 198KHz. I can't see how Radio 4 appears twice unless there's a bad IF response.. for example if one transformer coil has a peak at say 480KHz there would be a response at 185KHz which is broadly what I'm seeing. The easiest way to discover whether this is what's happening is to use my spectrum analyser hooked up the the detector diode of the HL41DD. I can also check on the shortwave lack of local oscillator. Below is the result of IF alignment.

This scan was taken by removing the HL41DD so there was no audio output or AVC action. The set was tuned to the strongest 465KHz breakthrough which of course is the long waveband. The tracking generator is running from 415KHz to 515KHz and the response is a bit peculiar.

There are two clear peaks and what you're seeing is the response of the IF at both 465KHz and 495KHz. Is this why I could hear a second Radio 4 higher up the band?

In the following picture I set the waveband to the short waveband making it easier to align the IF amplfier.

The curve above must be displaying a peak due to 465KHz plus the local oscillator mixing with 960KHz to produce 960-465=495KHz. Why is the peak at 495KHz? Because the receiver dial is tuned to that frequency and the set of tuned circuits is proving the response shown. If I tuned the set lower in frequency the peak would follow. The size of the peak of course indicating the tracking accuracy of the RF and local oscillator settings.

The shortwave setting reduced the overall response by about 5dB but the gain of the RF tuned circuits isn't messing up the response.

That little spike is puzzling.. maybe ringing?

The tone control is now set to "Fidelity" which gives a broader response.

Each IF trimmer had a different effect with the first shifting the centre of the curve from 465KHz as one would expect.

While adjusting the RF trimmers it was very noticeable that the bandpass circuits had a dramatic effect on gain and I found that some trimmers needed a parallel fixed capacitor in order to tune across a peak. In the case of some a false peak was seen at the maximum trimmer capacity. For correct settings a distinct peak (both rise and fall) should be noticed before maximum or minimum trimmer setting. In this receiver there is no provision for adjusting the coils, with alignment solely concerned with peaking at the HF end of each waveband. Even the preset coils use a fixed ganged method of adjustment with twin cores moving through the RF and oscillator coils. The single alignment trimmer T13 has to be set to a compromise position.

After re-aligning with the HL41DD plugged in with a short wire poked in socket pin corresponding to the detector diode anode and with the tone control at "Normal" switched to short waves the response looked clean with a bandwidth of around 3KHz.

With the tone control set at "Fidelity" after a tiny tweak at one of the trimmers at the second IF transformer which lifted a drooping left edge upwards a nice flat top resulted.

It appears to have a bandwidth of about 18KHz.

Switching back to "Normal" produced, with no more twiddling, this nice IF response.

If you look at the circuit diagram above the difference in the two response curves is achieved by having two extra coils on top of T4 primary winding selectable by the tone control switch.

I've seen much more complex arrangements that give results nothing like as good as this.

I then went back to aligning the three wavebands. This was very difficult but was easier once I'd grounded the AVC line at the RF unit which was quite easy as the top of the assembly had the AVC line connections exposed. This left the IF amplifier AVC connected. Because of drifting with age in the coils, which are completely inaccessible, I needed to wire small capacitors across the trimmers in order to average out the receiver gain across the tuning ranges. Medium waves were not too bad and, as very few broadcasts are heard at the low frequency end of the band during the day, I optimised the range from 200m to 500m. Long waves had only two usable broadcasts and was impossible to optimise across the whole band so I trimmed to receive Luxembourg leaving Radio 4 at the RF amplifier settings for the former.

I know the main coils hidden away in their enclosure are not ideal because the preset coils perform better. The design of the set, especially the presence of the bandpass circuitry and absence of coil adjusters makes alignment difficult. Whether the coils can be adjusted or not I don't know because its not possible to get inside the RF assembly with the receiver working and too much work would be involved in dismantling. I'm happy with receiver performance except for shortwaves where reception only covers wavelengths corresponding to the local oscillator range 11 to 18MHz. AVC action is very good and of course fidelity is excellent with its class A audio amplifier.

Read on to see more on the shortwave problem.

These are all frequency changers dating in design from the mid 1930s. Left to right AC/TH1, FC4, MS4B, TH4 and a different shaped AC/TH1. As you can see, three are new old stock (NOS).

Click the valves to see the Mazda datasheet.

Tackling the shortwave oscillation problem isn't easy. The chassis must be upended together with the cumbersome speaker unit to access the frequency changer. The main resistors viz. screen oscillator anode and cathode measured about right. The oscillator circuit is a bit odd because the amplitude of the output seems to be governed by a resistor in series with the triode grid (in fact the valve datasheet explains the reason for these resistors which is to limit amplitude to avoid harmonics... whose presence would result in ghost signals appearing on the dial). For long waves this is R4 plus R8, medium waves R4 plus R9 and for shortwaves just R4. R4 is called a "regeneration modifier" and is 100 ohms (or 150 ohms in brackets.. whatever that means?) R8 and R9 are called "het.voltage control" and are 5.5Kohm for long waves and 2.5Kohm for medium waves. This suggests the oscillator amplitude is quite frequency sensitive so is R4 possibly too high (or indeed too low) for maintaining the output across the whole of the band? The fact that oscillation is taking place at the high end and stopping lower than 11MHz is puzzling. It means that loop gain is too low below 11MHz. The grid condenser is 100pF (call it Rc which is essentially150 ohms at 11MHz and 90 ohms at 18MHz) R4 plus Rc at 11MHz is 250 ohms and at 18MHz 190 ohms. At the low end of the band, say 6MHz Rc is 265 ohms making 365 ohms total. Maybe shorting R4 making the total at 6MHz 265 ohms which equates to the 11MHz value is the answer? The problem may be ageing of the parts in the shortwave circuitry or something completely different?

Inspection of the valve datasheet reveals McMichael have faithfully followed Mazda's recommended circuitry which essentially tunes the anode circuit leaving feedback to the grid circuit. Poor overall gain will affect the ease of oscillation and that means the tuned circuit needs to have a high impedance at resonance. In the past problems have been met when, for example, a coil degrades and its Q or quality falls to the extent that a sharp resonance isn't achieved. Numerous techniques can be used to restore sufficient gain to restore oscillator operation. I used a few of these viz. increasing the operating voltages around the valve electrodes and increasing the value of the feedback condenser. These together with a selected valve restored operation.

The frequency changer heptode is fed from the HT line, but the triode from a reduced HT which it shares with the audio amplifier. As the HT rectifier is probably a lttle aged and not up to scratch I tried shunting the subsidiary HT line dropper resistor of 5K with a handy 91 ohm resistor. It was easy to monitor the local oscillator with a nearby shortwave receiver and I found the increased HT improved the cut off point from 11MHz to 10MHz. At this point I plugged in a new AC/TH1. Next I reduced the regeneration resistor by shunting it with 100 ohms. This improved things so cut off was now 9MHz. I was now using my variable HT supply with the UU4 unplugged, that red lead clipped to the heater of the UU4 base carries around 300 volts. The components for the AC/TH1 are mostly mounted vertically under the tagboard above. I checked the two decoupling condensers (screen and cathode.. both very good) and values of the screen resistor and triode anode load (both within spec of 40K) but I reduced the triode resistor to 27K and this improved things a little more.. down to 7.2MHz before swapping C19 from 100pF to 150pF.

I removed the new AC/TH1 and fitted an ancient TH4B. Much to my surprise it worked better than the new valve and the cut off point improved again to 7MHz. I put back the original AC/TH1 and it worked.. shortwaves now went down to the band end and broadcasts were loud and clear from 49 to 19 meters. What's the answer? Well it seems that more than one design of the AC/TH1 was produced.. it's reported as a triode hexode or a triode heptode. The valve was said to be a breakthrough for shortwave radio reception because it no longer resulted in pulling (that's when a strong signal, producing a high AVC voltage inadvertently moved the local oscillator frequency). Some designs used pulling to produce a kind of AFC such that, as one approached a strong signal its presence locked the local oscillator.. automatically tuning the set for you. Pulling was not a very nice thing to cope with generally and the AC/TH1 fixed the problem.. but made shortwave tuning a more precise task than previously.

Because of ageing of RF components I had to add extra capacitance at the RF amplier and bandpass cols with a beehive 3-30pF trimmer and a ceramic trimmer to replace failed trimmers.

Left is long wave, centre medium wave and right short waves trimmers plus the orange trimmer which sets the long/medium wave preset coils.

Clearly.. all things being equal.. a new AC/TH1 fails to receive the bands from say 31 to 49 meters in this McMichael, but the old one works fine after some fiddling with component values. So what's the reason? As a matter of academic interest I checked the interelectrode capacitances for the various valves shown above. The original valve measures as follows (with figures for the NOS valve in brackets).. triode anode to cathode 29pF (7pF); triode grid to cathode 10pF (12pF) and triode anode to triode grid 10pF (9pF). The anode figure is a lot higher.. so, does this explain the problem? I also noticed something else which looks a bit odd and doesn't figure in the receiver datasheet. It's clear in the picture above but is easily missed in the schematic because it's shown on a different page to the AC/TH1.. there's an RF choke in the screen grid decoupling circuit. It looks like it would be around 0.6uH with a little capacitance, so will have an impedance ranging from less than an ohm at long waves up to say 60 ohms at 18HMz so will have an effect within the troublesome frequency range. Why did McMichael fit it.. perhaps to keep the oscillator amplitude more or less constant across short waves? Also if you look at the resistors within the oscillator tuned circuit.. the "regeneration modifier" has twin resistance values, so was there a production problem with the design?

One problem you get with shortwave receivers is image reception (a failing associated with superhet receivers) and another, reception of ghost broadcasts tuned from harmonics of the local oscillator. A combination of these effects can produce hundreds of signals. McMichael's design is better than some because it uses a set of bandpass coils forcing the receiver to tune to real broadcasts whilst minimising false responses. A drawback of the technique is aging of components which causes receiver deafness.

I did check all the valves on my AVO tester and all showed up as serviceable, so maybe the valve inter-electode capacitances may have a bearing on things.. especially because the McMichael mechanical design requires really long leads between coils, frequency changer valve base and the tuning condenser. Does this mean there's some kind of choking effect that's interfering with the local oscillator?

Now that the oscillator is more or less reliable I was able to remove the HT dropper resistor and found I could reduce the HT to around 75 volts before it cut out on short waves. I also tried bridging the RF choke and that seemed to have no effect. Using my spectrum anayser hooked onto the oscillator section of the tuning condenser (with a high impedance probe) I could see it running from 6.25 (-63dBm) to 15MHz (-51dBm). Note the amplitude figures are for comparison only. The medium wave oscillator tuned from 1030-1905KHz at roughly -66dBm. See the table below for what these mean.

Measured Oscillator	RF True Frequency	RF True Wavelength	RF Image Frequency	RF Image Wavelength	Dial	Correct Oscillator
6.25MHz	5.785MHz	51.9m	6.715MHz	44.7m	51m	5.88MHz
15MHz	14.435MHz	20.8m	15.465MHz	19.4m	18m	15.93MHz **
1.030MHz	565KHz	531m	1495KHz	200m	560m	1.001MHz
1.905MHz	1440KHz	208m	2370KHz	127m	199m	1.973MHz **

In order for the dial to read correctly the local oscillators need to be slightly adjusted. The only way to do this (as coils are not accessible) is to tweak trimmers to line up the HF end of the medium and shortwave bands with the dial. These are the two frequencies marked **. In reality this means setting the dial to 20m on SW and a suitable broadcast between 200 and 250m on MW. You'll notice images are not a problem on medium waves but are exceedingly tricky on shortwaves, especially with powerful AVC.

As far as I can judge from circuit analysis the LO is higher than tuned signals on shortwaves. The main factor in this is the 3500pF padder condenser C16 which determines the effect of the tuning condenser. Usually any variation in the padder condenser can be more or less counteracted by tweaking the oscillaor coil inductance but in this set that isn't a practical proposition. The drawback might be loss of dial accuracy and receive sensitivity at the low frequency end of the dial.

Above with the darker dial in place and the speaker (upside down to protect the coils from damage) and below assembled in their cabinet after strething the speaker cloth.

The only difficulty in reassembly was the small tuning knob which had a broken securing screw which I needed to drill out then tap the hole for 6BA and fit with a special screw.

The missing knob has been replaced. The reflections from the glass makes the dial look lighter than it is.

I replaced the original mains lead but added a grommet where it passes through the chassis and grounded the safety earth wire, then fitted a modern 13A mains plug with a suitable fuse.

The final steps are to check everything works, set the preset station buttons and replace the pair of rear covers.

pending

McMichael 381

Return to Reception