Howard 474 AM/FM Mantel Radio (No.3)

This is the third Howard 474 to enter my collection, purchased late 2015. It is the first of my three to have a walnut coloured cabinet. The significance of the 474 is that it uses a Fremodyne circuit for the FM section.
The Fremodyne is a single valve (using a 12AT7 or 7F8/14F8) super-regenerative superhet FM receiver developed by the Hazeltine Corporation. It appeared post war as a low cost option for FM reception.
Fremodyne circuits appeared in a few mantel radios around 1947-1949, and also stand alone FM converters, for use with existing AM receivers or audio amplifiers. Many of these have been described elsewhere on this site. The Fremodyne circuit has been extensively covered elsewhere on this site. See here for a detailed description of this unique method of FM reception.
My first Howard 474 was purchased almost 10 years ago, and it is described here. A second one was purchased a few years later, but has not been restored. This article describes my third acquisition.
It arrived in good cosmetic condition, except for a crack on the top of the cabinet. It is hard to say if this had been there for a long time, or was a result of the postage - it was rather tight in the box.
This set had the aerial connection instructions fairly well intact.

Aerial connection instructions on back of set.

Inside the set was something of a surprise. There was a radio shop repair ticket from 1949. That may well have been related to the interesting repair I found here:

The large dropping resistor at the side of the tuning gang is foreign and had been installed to power the 6.3V dial lamp from the 120V mains.

The No. 47 dial lamp (6.3V 150mA) had been rewired to run off the 120V mains via an 850 ohm 20W resistor. A piece of asbestos had been glued to the side of the cabinet to reduce the effects of heat dissipation. An attempt had been made to mount the resistor by securing it in a spring clip which was soldered to the back of the dial. However, spring steel does not solder, and needless to say the whole lot had become detached and was flopping around. Interestingly, the resistor measured open circuit. The 474 is different from most U.S. made mantel radios in that instead of a valve rectifier with a heater tapping to run the dial lamp, a selenium rectifier is used to obtain the B+. This necessitates including the dial lamp directly in series with the other valve heaters. To prevent the surge current blowing the dial lamp when the set is first switched on, a thermistor is included. When I looked underneath, I could see the thermistor had been replaced with a 200 ohm 10W resistor.

Asbestos had been glued to the cabinet to protect it from the dial lamp dropping resistor heat.

Evidently, the thermistor had failed, and with a replacement not available, an ordinary resistor was used. Since the dial lamp can no longer be included in the heater string, it was powered from its own dropping resistor. It's certainly a way out of the problem, but causes the power consumption of the set to be almost doubled, and adds 18W of heat to the inside of the c
The next strange thing I noticed appeared to be a manufacturing error. The earth connection isolating condenser had actually been connected to the FM aerial terminal.

The tubular condenser next to the 12BE6 is for chassis isolation. Above it, with the red band, is the 2.2uuF FM aerial coupling condenser. Wired thus, the FM input is shorted to earth, and there is no earth return for the AM aerial primary coil.

In this situation with the FM input effectively shorted to earth, reception would be limited to very strong local stations, and  any attempt to use the external AM aerial terminal would have worked poorly.
At this point, it is relevant to examine the circuit diagram.

Circuit of the Howard 474. it is a conventional AM receiver with a Fremodyne FM receiver added on.

The AM receiver.
The AM section is completely conventional, using a loop aerial built into the cabinet back to feed the input of a 12BE6 converter. The local oscillator section of the tuning condenser has specially shaped plates which avoids the use of a padder condenser, and should provide superior tracking. The loop aerial has a primary winding of a few turns which allows the connection of an external long wire aerial in weak signal areas. The external earth connection is via the previously mentioned .02uF condenser (17) to the chassis. This is required because of the live chassis design.

A remote cut off valve, type 12BA6, performs as a 455Kc/s IF amplifier in the usual way. The second IF transformer then feeds the diode plates of a 12AT6 to provide demodulated audio, and AGC voltage.
The 12AT6 is an earlier version of the 12AV6, and is plug in compatible with similar characteristics.  The audio is filtered by means of a 47K and two 100uuF condensers before proceeding to the volume control. The demodulated output is also filtered by the 2.2M (43) and .05uF (16) to remove the audio signal. This provides a negative going DC voltage which is proportional to signal strength, and thus becomes the AGC voltage. This is applied to both the 12BE6 and 12BA6 grids. There is no AGC delay, which means the set can be desensitised with noise, and that the gain starts to reduce even with weak signals. Nevertheless, this system of simple AGC works well enough and was standard in many sets.

From the AM/FM switch, the audio then feeds the grid of the 12AT6 triode, functioning as the first audio amplifier. Contact bias, by means of the 10M resistor, simplifies the circuit by dispensing with the cathode resistor and bypass condenser that would otherwise be used. With high gain triodes, the grid will charge up to a negative voltage because of the electrons emitted from the cathode.
The grid resistor must be kept high for this scheme to work, otherwise the charge is drained away too rapidly, and the voltage is too low. Hence, grid resistors in this situation are usually 4.7M to 20M.
Extra hum filtering for the 12AT6 plate circuit is provided by the 47K (46) and .1uF (13).
The output valve is a 50L6 and operates under completely conventional conditions. There is no negative feedback, with the circuit simply relying on the 300uuF RF bypass condenser (21), and the 50L6 plate bypass of .002uF (11) to provide acceptable frequency response.

Power Supply.
Up to this point, the circuit is much the same as hundreds of thousands of what are these days called "All American Five" receivers. And the performance is the same. Like the AA5 circuits, a series heater live chassis power supply is used, which allows AC or DC operation from 105-125V AC or DC. One has to wonder just how often such sets were used on DC, but the real reason for using a transformerless power supply is for a much lower cost. Indeed, some such radios were really penny pinching minimalist affairs, with short cuts taken wherever possible.
The chassis in the Howard is not directly connected to the mains, but instead has a separate earth bus bar to which all the earthy connections are made. This is in turn connected to the chassis via a .22uF (18) isolating condenser. This is required for stability and hum reduction. However, the reactance of .22uF at 60 c/s is 12K. At 120V, this means a maximum current of 10mA can flow to earth. This is enough to give an unpleasant shock, although probably not lethal in most situations. The chassis and its securing screws are easily touched from behind the cabinet, so there is a risk of shock unless the set is run from an isolating transformer.
For this reason, the external earth connection is made via a much lower value condenser, of .02uF (17). This gives a maximum current to earth of closer to 1mA which is quite safe.

Unlike the usual AA5 circuit, there is no 35W4 or 35Z5 rectifier. With an extra valve added to the heater chain (the 12AT7 used for the FM section), there are not quite enough volts available to heat a valve rectifier, so a selenium rectifier is used here. The B+ power supply is a conventional half wave design, but actually does use a filter choke. In this regard, the set was not a penny pinching design.

All the valve heaters are connected in series, along with the 6.3V 150mA dial lamp, and a thermistor, across the mains. The thermistor is required because as the valves have a low cold resistance (about 100 ohms for the whole string in this set), a large surge current flows (a little over 1A) when the set is first switched on. Because the filament in the dial lamp runs white hot, it is much thinner (i.e. fragile) than the valve heaters which run at red heat. The effect is that the valves warm up much more slowly than the dial lamp. Unless a thermistor is included, the dial lamp would be subjected to this high surge current and immediately burn out.  The thermistor starts off with a cold resistance of 791 ohms and falls to 100 ohms once fully warmed up with 150mA flowing through it.
In most U.S. made series heater radios made post war, a rectifier valve is used with a tapped heater across which the dial lamp is connected. This means the dial lamp voltage rises slowly at the same rate of the valve heaters. However, the scheme has some disadvantages. I discuss it further in the Perco FM Converter article here.

FM Section - The Fremodyne.
The most important aspect of the Howard 474 is of course the Fremodyne FM receiver. Essentially, one section of the 12AT7 is a conventional local oscillator, operating at  23.75Mc/s above the carrier of the received station. The other 12AT7 triode operates as a mixer with the VHF signal fed into its grid, and as a self quenched super regenerative detector based around the plate and cathode circuits. It operates at the IF of 23.75Mc/s. The aerial coupling condenser is of low value (2.2uuF) to reduce aerial loading effects, and it also provides mains isolation. The input is unbalanced, although the instructions imply that a balanced transmission line can be directly connected. Of course, the noise reducing characteristic of a balanced line is negated when connected this way. A mains aerial is also included. This consists of chokes (63) in both mains supply leads to isolate the mains cable (at VHF) from the chassis. This allows the mains cable to function as a simple aerial in good reception areas. A 100uuF condenser (68) isolates the mains supply from the user and couples the signal into the FM aerial connection.
The detector has automatic stabilising circuits which eliminate the need for a regeneration control. Audio output is taken from the detector cathode circuit, with quench filtering and de-emphasis provided by the 100K (38) and the 1000uuF (25).

As well as switching the B+ to either the AM section, or the Fremodyne circuit, the AM/FM switch also selects the audio source fed into the volume control. A problem that can occur here is that if one section is not used for long periods, the relevant cathodes become poisoned because of no cathode current flowing while they are heated. Indeed, a ruined 12AT7 has turned out to be not uncommon with these AM/FM mantel radios using Fremodyne FM receivers, when the set was only ever used on AM. One way around the problem is to supply a low current to the section not in use, as I did with this receiver.

Restoration of the Howard 474.
For the most part, like any radio restoration, it's a matter of replacing paper condensers and high resistors. In many U.S. made radios I have also found electrolytics shorted or open circuit. In the present day, mica capacitors are now longer not above suspicion either, particularly where intermittent faults are involved.
Starting with the power supply, I found the main filter electrolytics OK (6), although they did need some reforming to get the leakage current to an acceptable level. However, the 40uF filter (7) for the Fremodyne was short circuit. The non standard dial light wiring was removed.
All paper condensers (or all I thought were paper condensers) were replaced except for the 12BA6 cathode bypass. Here the leakage is completely swamped by the 270R cathode resistor. While I operate my U.S. made sets from a double wound step-down transformer, I still thought it wise to replace the chassis isolation and earth connection condensers. There is always the possibility of these sets being plugged into a non isolated auto-transformer, and I wouldn't be happy about someone being electrocuted should the mains wiring be incorrect, when replacing two components would go a long way to avoid that.

The unrestored chassis. The replacement 200R heater dropper can be seen. The grey and red wires leading to the bottom corner are not original and supplied the dial lamp.

With a few resistors then replaced, that completed the B+ power supply, audio, and AM sections. Next was to deal with the Fremodyne. Here, service access is poor because the circuitry is assembled onto a sub chassis with the tuning gang. This is mounted on top of the chassis, and it is awkward getting into the depths. Furthermore, some of the components are mounted inside the FM IF transformer and this would have to be opened up.

Inside the Fremodyne's IF transformer. The actual 23.75Mc coil is the one closer to the adjustment slug. Nearer the resistors is the 12AT7 cathode choke winding. Although mounted on the same former, any inductive coupling is merely coincidental.

By desoldering a few connections, and a couple of earth straps, the sub chassis is removable. Most of the resistors were found to be high, including the 150K grid stabilising resistor. In this set it is mounted inside the IF transformer. With wires wrapped around tags it was a pain to extract, and I broke a tag as a result, but was easily fixed. The 10uF electrolytic (9
Upon reinstalling the tuner sub chassis, I could now turn to the heater and dial lamp wiring.

Restoring the heater and dial light circuits.
I thought about the options here; ideally using a thermistor to restore the original circuit. The only types I had were CZ9A Brimistors, made by STC. These looked suitable at first with a cold resistance of about 1000 ohms, but alas when I tried one the dial lamp promptly blew. The trouble was the warm up time was way too fast. Once I actually looked at the data I could see why. The CZ9A is totally unsuited to valve heater strings, as it has a very rapid warm up time. It's intended function is for things like reducing transformer inrush current and incandescent lamp surge protection. The Brimistor data did show types which would be ideal, such as the CZ3, but alas there's no way I would find any in the modern day. In Australia, series heater circuits are few and far between, so the demand was always minimal.

Dial lamp shunt resistor method unsuitable.
So, that meant using a normal dropping resistor for the valve heaters. The dial lamp was then the remaining problem. I considered (and tried) the shunt resistor method as I used in the Perco and Emerson CF255 receivers. However, the shunt resistor had to be such that when the valves had actually warmed up, there was only about 3V across the dial lamp. This was just too dim.
In the Howard, the valve heaters add up to 100V. The dropping resistor should then be 130R. As the valves have a cold resistance of 100 ohms in total, when the set is first switched on, the surge current is about 500mA. With a dial lamp of 150mA, it's obvious it has to be shunted by quite a low value resistor to keep its voltage below about 7V when the set is cold. Thus, once the valve heater string current has reduced down to the normal 150mA, the dial lamp is now grossly underpowered.
The reason for this method's acceptability in the Perco and Emerson receivers is that the valve heater dropping resistors are of much higher value, because of the lower heater string voltages (483 ohms and 250 ohm respectively). This in turn means the surge current at switch on is much reduced, and the dial lamp shunt resistor doesn't have to be as low.
(A variation of this method is to also have the B+ current flow through the dial lamp and shunt resistor. As before the lamp will run dim after the initial cold surge, but once the valves warm up, an extra 60mA or thereabouts, flows through the dial lamp and shunt resistor, restoring some of the brightness. When I restored this 474, I was not aware of this method so did not try it).

Next option was to operate the dial lamp separately again, as our previous repairer had done, but with a bit more efficiency. 18W just to run a dial light was not acceptable, especially if the set was to operate off the 12V house supply via an inverter. A capacitive dropper was considered, but it would be frequency dependent and cause a low power factor, both of which can be problematic with inverters.
And then there's the switch on surge possible with capacitive droppers. A diode dropper with a resistor is another option, but the light would produce a visible flicker operating at half the supply frequency.
A miniature transformer perhaps? Yes, that too was considered, and would be the best option so far, although it meant no DC operation, not that I'd ever need that.

130V dial lamp.
But then, I remembered 130V dial light bulbs. These have the same miniature bayonet BA9s base as the No. 47 dial light, have the same size glass envelope, and I had a small quantity in my collection. They're used as indicator lamps in industrial control applications. Current consumption is only 20mA. Trying one out proved the point - this was the solution I was looking for! The lamp socket was connected across the 120V supply to the heater string, and the 130V 20mA bulb inserted. Only someone noticing the longer filament inside the bulb would realise it wasn't original. The light output is the same colour and only slightly brighter than the No. 47 bulb. The bulb socket is far enough away from the celluloid dial glass for the slight extra heat not to be a problem. All in all, this was a most elegant way out of the problem.
The data on these bulbs shows a life of 1000-2000 hours, depending on whose data you're looking at. This is not a lot better than a household light bulb. However, remember that's at 130V. On 120V or lower, the bulb life is extended considerably - it should be at least doubled. While not quite the same life as the No.47 bulb, 4000 hours is still a lot of listening time. One disadvantage with these bulbs is the longer, more fragile, filament. For a portable radio that's bumped around in use I wouldn't recommend them, but for a radio that just sits on a shelf this would not be a problem. I would recommend anyone looking to purchase these bulbs to shop around because the prices vary from reasonable to quite ridiculous.

The 130V 20mA dial lamp looks like it was always there. It was the most elegant solution with no thermistor available.

With the dial light sorted out, next was the valve heater string. Now that the dial lamp was not to be included, I could calculate a suitable dropper resistor. The 200 ohm resistor that the previous serviceman had installed was a little on the high side. I calculated a value of 130R for 120V mains. In practice, I found 150R to be better than 120R. (130R is not a preferred value). Power rating was 5W.

Not long after the radio had been powered up, I could smell something burning, and then the smoke came out. The culprit was the .05uF mains filter (10). The original looked like it was mica so I had left it in situ. Alas, it turned out to be paper, hence its failure. The receiver worked in that stations could be received. However, it was now time to check the alignment.
This all seemed to peak up OK and the AM reception seemed very good. On FM things weren't quite right. Pressing on the tuner sub chassis would shift the stations and cause reception at the high end of the band to disappear.

Restoration complete. The new silicon rectifier can be see just to the bottom right of the filter choke.

Aside from this, the B+ was a bit low. Remembering my first 474, and selenium rectifiers in general, I decided to replace the one in this set too. I removed the rectifier and installed a four lug tagstrip in its place. Mounted on this was a 1N4007 diode and a 22R 1W surge resistor, chosen to produce 140V B+ at the 50L6 plate. I became aware of a hum on the upper end of the FM band which hadn't been there before. Placing a .0047uF across the 1N4007 fixed that completely. Obviously, the sharper switching of the 1N4007 had introduced modulation hum.
I also took advantage of the 22R resistor with regards to the dial light. With the 22R resistor connected to the 120V AC supply before the diode, I connected the dial light to the other side of the 22R.
There are two reasons for this. First, the 22R will act as a fuse if the lamp socket insulation fails, and secondly my theory is the surge current will be lessened through the filament when first switched on. This is because with the filter condensers discharged, there will be a high voltage drop across the 22R for a very short period until they charge. Thus the lamp voltage is less until the condensers have charged.

FM problems.
Finally, it was time to deal with the erratic FM behaviour. First I found the metal shield on the side of the tuning condenser was only just touching the side of the tuning condenser. When it did so, the stations shifted. The cure was to bend it further back so it didn't touch. There was still the disappearing of the stations at the 108Mc/s end. Pressing on the sub chassis still affected this. The sub chassis is supported by three rubber grommets. When it was pushed down enough to make contact with the main chassis, all was well. But when isolated by the three rubber grommets and connected only by the one piece of braid, the signals disappeared. Out with the spectrum analyser, and sure enough the local oscillator dropped out. Two extra pieces of braid to the chassis fixed that. I am curious why most sets isolate the tuning gangs with rubber grommets. They do not seem to serve any useful purpose.

These two pics show the extra copper braid added. The original tinned braid has too much inductance at 108Mc/s.

This is obviously a design fault because I had this exact same problem with the first 474. How many Howard 474's are out there unable to receive stations at the 108Mc/s end of the band?
How many of the owners know what the cause is? I suspect most simply blame the Fremodyne concept as cheap and nasty instead. It's situations like this that a spectrum analyser is an extremely useful instrument.
Having dealt with that little problem, I could now align the FM section correctly.
Finally, the crack was glued, the cabinet polished, and the dial glass re-glued.

As I've mentioned elsewhere in other articles, the Howard is the best of the AM/FM sets using the Fremodyne circuit, both in sensitivity and sound quality. Indeed, the AM section of this set is very sensitive.
Not only could I pick up 2ZB from Wellington NZ, but also 2YC. This was just with the internal loop aerial.

Aerial and earth connections can be seen here.  Upper Fahenstock clip is for the FM input. The wire connected to it goes to the mains cable via a 100uuF condenser. The lower Fahenstock clip is the earth connection for both AM and FM. For the external AM aerial, connection is made to the coiled up piece of wire (why not a Fahenstock clip for this too?). Notice that the back of the chassis and its securing screws can easily be contacted by the user.

The FM sound quality is very good with a decent signal, and is certainly up to mantel radio standards. The power line aerial works very well with the local station about 5km away, and acceptably enough with the major Sydney stations located about 80km away. Anything weaker than that requires a better aerial for entertainment quality reception.

Ad from September 1947 in "Radio & Television Retailing".

Postscript August 2020:  A strange instability in the audio had started to appear, in that there would be a violent oscillation or crackling, as the volume control was rotated over the 'dead spots' that exist with most consumer grade potentiometers. Normally, crackling as a pot is rotated indicates DC flowing through it. Since the coupling capacitors had already been replaced this should not be a problem, but just to go through the routine to test, I disconnected the .01uF grid coupler to the 12AT6, and it burst into oscillation. The audio amplifier is a simple affair being a 12AT6 triode feeding a 50L6 beam tetrode. The gain is not excessively high, so why the 12AT6 grid had to have a low impedance to earth was unusual. The cause turned out to be simply the second B+ filter electrolytic was open circuit. With the filter choke forming a common impedance to the 12AT6 and 50L6, it was clear why the stage oscillated.

The other two sections of the filter electro were tested with the Electronics Australia ESR meter, developed especially for testing electrolytic capacitors. Both were excellent, so a 30uF was squeezed in and connected across the defective one. Perfect stability! Next, a test of the B+ voltage was done. Although the ESR was good, I wondered if there might be leakage dragging down the voltage. As it turned out, the volts were down to about 80 at the third filter, whereas it should be 100. However, it wasn't the capacitor at fault, but the 2k filter resistor having drifted to around 3.2k. Replacing this brought the voltage back up. Low value resistors drifting high is rare which is why I had not tested this resistor previously.

Finally, it had been noted the FM was not performing as it should be. Sensitivity was down, and tuning was more critical than normal. Something was not right with the Fremodyne circuit. Remembering the cathode poisoning problem in AM/FM sets, and that the 12AT7 in this one was original, I tried another and all was as it should be. I had been using this set for my evening entertainment for a few years, mostly on AM, so it was perhaps not surprising. As mentioned previously, if a valve is run with the heater powered up but with no cathode current flowing, the cathode eventually becomes poisoned. This incidentally was a problem with early valve computers, where valves might be in a cut-off state for long periods, and new special types were developed to overcome this.

To prevent this happening again, I connected a 100k resistor between the AM and FM B+ rails. When switched to FM, the AM valves receive about 16V, enough to prevent cathode poisoning, and similarly, when switched to AM, the 12AT7 receives about 26V. These voltages are low enough so the relevant stages are inoperative to the point where there is no risk of audio breakthrough.
So far so good, but it was found that when switched to FM, there was some kind of interference which wasn't there before. Sure enough, connecting a CRO to the AM B+ showed it oscillating. The cure was simply a .01uF across the B+.

Modification prevents cathode poisoning. Resistor was later changed to 180k.

March 2024: It had been noticed the AM performance was not what it should be. About every 20kHz across the AM band there was a hetrodyne whistle. Suspicion fell upon the 12AT7 oscillating because of the 100k resistor I'd added to prevent cathode poisoning. Shorting out the FM B+ got rid of the whistles. Evidently, even at 26V, it seemed the super-regenerator was not only oscillating, but sufficiently so that it quenched - hence the interference being harmonics of the quench frequency. Increasing the resistor to 180k dropped the FM standby B+ to around 12V which fixed the problem.

Fremodyne FM receivers

My first Howard 474

Home Page