Fullscreen HTML5Med Res (??MB)HTML4

Vintage Radio Restoring the Philco model 38-7 MW/SW radio from 1938 The Philco model 38-7 radio is one of the finest medium-wave and shortwave band domestic radio receivers to emerge from the USA before World War II. By Dr Hugo Holden T hese radios were shipped to many countries because they had an optional dual-voltage 110/220V power transformer. I acquired my radio in New Zealand during the early 1970s and restored it initially around 1976. I had some help at the time from John W. Stokes, the author of the famous book “70 Years of Radio Tubes and Valves”. John had an excellent radio repair store on Dominion Road in Auckland and helped me with the alignment and some of the repairs. At the time, he explained to me how the 6A8 pentagrid converter valve worked. I soaked up the information like a sponge, being an enthusiastic 18-year-old. John Stokes passed away in August 1999. I will never forget the way he helped me with radio repairs. This radio recently required more work to keep it running. The first restoration was just about 50 years ago now, and the radio is about 88 years old. It has six valves and covers the medium wave (MW) band of 5301720kHz, with a very wide range 90 Silicon Chip shortwave (SW) band specified as 5.718.2MHz but it can actually tune from 5.5MHz to 19.5MHz, as indicated on the dial. The cabinet is an attractive Art déco design, typical of the late 1930s. It has inlaid veneers but is not cluttered. However, from a user’s perspective, the most interesting aspects are the dial and the tuning arrangements. The tuning knob rotates around the outer dial’s perimeter and has a central shaft. The larger of the two knobs mounted on that central shaft and can be pushed inwards, slipping over the outer surface of the shaft. This engages the ridges on the dial glass’s brass retaining ring via a rubber ring on the rear of that larger knob. Rotating the larger knob then acts as a type of reduction gear for fine-tuning. The central shaft and the smaller knob on it can move in and out, providing additional functions. It can engage metal cones mounted on a rear vertical plate behind the dial drum that can be set to preset station positions to lock the variable capacitor’s position and hence the tuned station. This was called “Cone-­ Centric” tuning. This shaft also includes an electrical contact, which can mute the Photo 1: the dial is quite fancy. It’s hard to see here, but there’s a radial embossed pattern in the centre, from the spindle out to the barely visible dark ring that’s just visible inside the inner station labels. siliconchip.com.au audio between the selected cones and at different positions. If not using the cones, that feature is better disabled because if one is not expecting the audio muting, they can be caught out. The central shaft is attached to a diecast metal arm, which attaches to and rotates another shaft on the dial’s central axis that carries the dial pointer. This shaft also passes through the vertical plate carrying the cones and attaches via a flexible coupling to the input shaft and gears of the two-gang variable tuning capacitor. The extra machinery of that rotating arm and its physical mass is counter-­ balanced by a large mass on the central shaft. This imparts the tuning mechanism with a very nice feel in operation. Behind the dial is a closed, drumshaped chamber, painted white inside, into which the dial lamp shines. This lamp diffusely illuminates the dial material, for a welcoming orange glow. The dial is a good size at around 125mm in diameter. It has highly accurate calibration detail and more resembles the type of thing seen on a scientific instrument than any domestic radio dial from 1938. Up close, the dial’s plastic has an embossed surface with an elegant patterning effect. It is visible in Photo 1 but only just. Photo 2 shows the dial drum area and the adjustable cones on the vertical plate. In this case, the central shaft is not pushed in to engage a cone yet. The wire passing into the assembly detects the position of the shaft for muting. When the shaft engages a cone, a separate local mechanism in the rotating arm lifts a contact of a circular track so that the audio is always unmuted. Other features of the tuning mechanism include a variable capacitor with a spring-loaded twin-gear wheel arrangement to prevent backlash, and a special universal rubber coupling. This coupling recently required rebuilding. The laminated rubber discs had become warped under the force of the counter-balance’s mass, and the rubber had become stuck in a stiffened, deformed state. This resulted in a very irregular feel to the tuning mechanism. Also, since the first restoration, the variable capacitor’s rubber mounting feet had degraded. Back in the 1970s, with no parts available, I hand-­ fashioned these replacements from some 5mm-thick red rubber sheet siliconchip.com.au Photo 2: in this side view of the dial assembly, you can see the spring on the small shaft and the cones it engages when pressed to centre on a station. Photo 3: in this photo, the knob has been pressed in and it has engaged one of the cones. Photo 4: this view of the chassis includes the twin-gang variable capacitor and its somewhat unusual flexible coupling arrangement (after I had repaired it with new parts). The variable capacitor mounts and rubber coupling for the input shaft had hardened and were both replaced in this photo. using a scalpel. Over the last 50 years, this red rubber had hardened and cracked. Photo 4 shows an overview of the chassis. Earlier 1970s restoration The chassis underside held up very well over the years since the first restoration. At that time, I replaced all the original wax paper capacitors. The mica capacitors were and are still OK, even today. However, many mica parts become leaky by this age, Australia's electronics magazine and some types suffer from a form of silver migration disease. Back then, I also hollowed out the original wet electrolytic capacitor cans and placed modern electrolytic capacitors inside. I did that by machining a phenolic base and adding solder terminals taken from some 4mm panel banana plugs. The capacitor housings were then re-mounted using capacitor clamps, rather than the original screw and large nut arrangement that they once had. October 2026  91 Fig.1: the Philco model 38-7 radio circuit. It has six valves even though it’s a ‘five valve set’ because it uses an extra envelope for a particularly good delayed AGC implementation; that valve doesn’t improve the set’s sensitivity, unlike a six-valve set with an RF amplification stage. I also renewed the resistors then. Aside from the large wirewound 10kW part, I used modern (at the time) 1W resistors carefully spray-painted with the body-tip-spot colour code. I may have been the only person in the world back in the 1970s who was repainting resistors to make them look age-­ appropriate. I think this has become more of a fashion now with vintage radio and TV restorations. When John Stokes saw those repainted resistors back in the 1970s, he remarked that he could not see the point in doing that, but it made him smile nonetheless. The black rectangular box at upper left in Photo 5, where the mains wires terminate, contains two 15nF 92 Silicon Chip capacitors. These should always be replaced with what we now know as Y capacitors. In the 1930s, ‘Y capacitors’ with modern safety ratings did not exist. The designers simply used 1000V or 1500V rated parts to try to avoid failure. Note the unusual (for today) mains wire colours of red (Active), black (Neutral) & green (Earth). Those are not per today’s standards, but they were correct in the 1970s when I replaced the mains cord. At the time of the original restoration, I removed the valve socket rivets, IF transformers and upper chassis parts. The whole wiring assembly, largely complete, was then removed from the chassis. All the valve sockets Australia's electronics magazine are retained by screws and nuts now. Fortunately, Philco made good wafer sockets, so none required replacing. The chassis was re-plated in the 1970s. However, over the last 50 years or so, some fine pitting and corrosion on the top chassis surface has started to reappear. Electronic design The design here is outstanding, with a number of innovations. The circuit is shown in Fig.1. It uses a 6A8 pentagrid converter valve (Photo 6). The intermediate frequency (IF) amplifier valve is a 6K7, while the detector and AGC generator is a 6J5. The 6K5 audio preamplifier valve feeds a classic 6F6 3W Class-A siliconchip.com.au audio output stage. Finally, the 5Y4 rectifier furnishes the B+ supply. The valve lineup, aside from the 6J5, was very common in 1938. The 6K7 was a fairly standard IF valve used in multiple radios with very similar electrical specifications to the 6U7. These valves are ‘super control pentodes’, meaning that their gain can be well controlled, especially in an RF or IF stage, with the application of a negative AGC voltage. The 6J5 and 6K5 were also common general-utility triodes for detector and audio amplifier uses, as well as other applications. The 6A8 Pentagrid Converter had its origins in April 1933 when RCA released the 2A7, which is the same valve but with a different base and a lower heater voltage. The pentagrid converter valve in the early 1930s was a revolutionary method to combine a superhet radio’s local oscillator and mixer stage into one valve, while at the same time making the mixer amenable to variable mu (μ) for gain control by the AGC voltage. This meant that a very effective overall AGC could be created, with the same AGC voltage controlling both the converter and the IF amplification stages. The 6A8 is a single-cathode valve. All the grid elements are placed concentrically around the cathode, and the electron stream passes by all of them to the anode (plate). The first grid, closest to the cathode, acts as the grid for the oscillator circuit, while the second grid acts as the plate for the oscillator. The G2 grid is a pair of rod-like structures – see Fig.2. The trick to the design of the 6A8 is that G3 acts as a space-charge grid, or a virtual cathode, an effective electron source for G4. This fourth grid acts as the signal input grid, where the received radio station signal is injected from the tuned antenna circuit. The G4 connection is on the 6A8’s top cap. Generally, in most radios, the local oscillator runs above the received frequency by the intermediate frequency. For example, if the received frequency coming into G4 of the 6A8 is 1000kHz, the oscillator will be running at 1470kHz. When these two signals get multiplied by the 6A8, one of the signal components is the difference frequency at 470kHz. Everything but this component of the output is filtered out by the first IF transformer and then passed to the 6K7 IF valve for further amplification and filtering. Although the function of the 6A8 was designed to be multiplication, there will be some non-­linearity, which also aids its mixer function. Shortwave oscillator problems I found that the sensitivity of the radio dropped fairly significantly on the SW band above 12MHz. Some of this is expected because the performance of the 6A8 at the higher Photo 5 (left): the underside of the chassis is neat for a point-to-point wired set. The cotton wire insulation still appears to be in good shape nearly 100 years later! Photo 6 (right): these three versions of the 6A8 share the same base but have different envelope styles. From left-to-right: 6A8GT, metal 6A8 and a 6A8G. siliconchip.com.au Australia's electronics magazine October 2026  93 frequency end in known to be somewhat deficient. Later converter valves such as the 6K8 triode-hexode (released a little late for the 38-7) improved the high SW reception and AGC behaviour. Investigation revealed third harmonic content in the oscillator signal; eg, in the region of 16MHz, the spurious signal was at around 48MHz. This harmonic was very high in level, almost swamping the fundamental wave. It took me a while to discover that this was due to a resonance in the primary of the SW oscillator coil with the capacitance of the G2 grid of the 6A8. This was not a problem with the 6A8 itself, but more related to the proportions of the primary of the oscillator coil and the level of damping in the primary circuit. This was affected by the fact that the 4μF bypass capacitor (11) on the B+ supply is an electrolytic type and poor at bypassing high frequency RF. Bypassing that electrolytic with a 330nF film capacitor suppressed most of the 48MHz oscillation (Photo 7), but then, interestingly a high level fifth harmonic appeared, at around 80MHz. Unrelated to these problems, I also added a 27pF capacitor to trim the value of the fixed 250pF capacitor in the ceramic block (7B) as it was a little low. To damp the very high frequency resonances, I replaced the link wire between the oscillator coil primary and the 6A8 socket pin 6 (grid G2) with a 150W resistor – see Photo 8. This removed nearly all the harmonics and gave a clean oscillator fundamental over the whole tuning range on the Fig.2: the configuration of the 6A8 pentagrid converter valve. It works well, but because the two stages share an electron stream, there is some interaction between them, which can make tuning in to strong SW stations a bit fiddly. SW band, with no practical effect on the MW band. This improved the high-frequency shortwave reception. Probably all the 38-7 radios were affected by this. It wasn’t until I had a 100MHz-capable scope with a very low input capacitance ×100 probe (the Tektronix P6009 – only 2.5pF) that I was able to detect and remedy this problem. In general, it is fair to say that some higher-order harmonics in the local oscillator of a typical radio are not a problem, especially because in the G1 grid circuit, the resonance of the tuned circuit there dominates and a clean sinewave is nearly always seen at that location, even if the tickler winding shows some harmonics and some distortion. Mixer electron stream concern A quirk of the 6A8 is that the AGC control of its RF gain affects the frequency of the oscillator section. Ideally it would not, but it does. The effect of this is more noticeable on the high end of the SW band with a strong enough signal to get the AGC to shift from its normal inactive state of -2V to -6V or more. An increasingly negative AGC voltage increases the oscillator frequency. Since the oscillator part of the 6A8 and the mixer part of the valve depend on the same electron stream from the cathode, they interact. When a strong SW station is tuned in, with the oscillator initially running below the required value to tune in a station, a problem becomes evident on tuning the station. Say the station was on 16MHz, requiring an oscillator frequency of 16.470MHz, and the oscillator is sitting at around 16.300MHz and increasing as the user moves the tuning knob toward the station. As the signal is received and the AGC voltage becomes more negative, there is a rapid increase in the oscillator frequency and an avalanche Added bypass capacitor 150W resistor to replace wire link 27pF added to trimmer and fixed 250pF Photo 7: adding this 330nF capacitor fixed a stability problem at the upper end of the shortwave band that probably affected all of these sets. 94 Silicon Chip Photo 8: this added 150W damping resistor was also required to totally eliminate the HF instability. Australia's electronics magazine siliconchip.com.au Fig.3: RCA’s AN-87 document from 1938, showing how much lower the interaction between the two stages is in the 6K8 compared to the 6A8 due to the shared cathode being in the middle. increase in the negative AGC voltage. The oscillator frequency jumps up, often above 16.470MHz. Therefore, the station appears to abruptly jump into tune and overshoot, requiring extra fine-tuning effort. This undesirable effect is not too noticeable on the low SW band below 10MHz or the MW band at all, as the tuning is less critical for any rotational angle of the variable capacitor. On weak stations (most SW stations are in my locality), the effect is not evident. This effect is more significant with a pentagrid converter such as the 6A8, rather than a triode-hexode converter such as the 6K8 – see Fig.3. The 6A8’s large AGC-dependent frequency shift is because the AGC voltage has more of an effect on the transconductance of the oscillator siliconchip.com.au section as a consequence of the oscillator and mixer sections of the valve sharing the same electron stream. RCA chose to place the 6K8’s anodes on opposite sides of the same cathode, so each plate receives electrons from opposite sides of the cathode – see Figs.4 & 5. This keeps the electron streams separate. Another interesting feature of the 6K8 is that at high frequencies, its G3 input resistance is negative. This acts as a Q multiplier for the resonant circuit feeding it. This was discovered at RCA, where an improvement in selectivity and image rejection was noticed upon testing it. There are other mixer valves with higher conversion transconductances than the 6K8, including the Philips ECH35, introduced in 1939. It appears Australia's electronics magazine to be a further development of the British X41 converter, released in 1936. These valves have an internal architecture where the triode is physically separated from the hexode to reduce interactions between the two sections. Due to the superior performance of converter valves such as the 6K8, ECH35 and X41 compared to the 6A8, I considered fitting one to the Philco radio instead. I selected the ECH35 because it has a higher transconductance than the 6K8 and it can plug directly into the octal socket in place of the 6A8 without having to rewire the socket. However, because there was no pin 1 tag fitted in the socket (to Earth the conductive coating on the ECH35), I linked pins 1 & 8 (the cathode) on the valve base. October 2026  95 Scope 1: the 6J5 anode voltage with no antenna signal. No other significant changes were made except altering the wiring of one resistor (part 10 in Fig.1). I simply moved this 5kW resistor from the junction of part 12 and part 16 to the junction of part 16 and part 22. This was to reduce the anode voltage of the triode section of the ECH35 to remain within its maximum specification. I also shorted out the 22W cathode resistor as the ECH35 is designed to run with a grounded cathode to minimise triode/hexode interactions. I re-aligned the radio because the capacitances of the two converter valves are a little different. The oscillator behaved normally on both bands with the ECH35. On testing the radio, I found improved sensitivity and less noise. From the low MW band up to the 14MHz mark, the sensitivity had improved from 10μV to 5μV. The 5μV figure is almost starting to rival a set with an RF stage, which would have Scope 2: the 6J5 anode voltage with a 50μV RMS signal fed into the antenna. a sensitivity of around 2μV, although with less selectivity. The AGC operation on stronger signals was also improved, with less frequency pulling. Interestingly, the noise with the ECH35 has a different sound to the 6A8, with more low-frequency components. The 6A8 has a preponderance of high-frequency noise. This is likely due to the higher level of phase noise with the 6A8 compared to the ECH35 or 6K8. The sensitivity toward 19MHz improved to around 12μV, about twice as good as it was with the 6A8 at around 24μV in this part of the band. Therefore, I elected to leave the ECH35 in the radio, as the performance improvement was hard to ignore. Delayed AGC circuit with active clamp With the 6J5, the radio had one more valve than the typical five-valve radio. Often, five-valve radios used a 6Q7 or similar, which incorporated two diodes along with the audio amplifier triode. The valve count might have been a marketing feature, but the use of the 6J5 in this case is clever; the valve performs a useful function. To get the AGC to work well, this circuit deployed the 6J5 valve in an unconventional manner. The grid-cathode was used as the AM detector diode, while the anode was tied to a negative potential derived from the “Candohm” resistor (part 43 on the diagram). The Candohm resistor is a three-part wire-wound resistor from the power supply’s negative terminal (the transformer’s centre tap) to ground. A negative bias voltage is developed across it. This is to bias the AGC line, the anode of the 6J5, the grid of the 6K5 and the G1 grid of the 6F6 audio output valve. Fig.4: this shows how the 6K8’s cathode shields the oscillator’s G1 grid to minimise the effect of the AGC voltage on oscillator frequency. Fig.5: the physical configuration of the 6K8 converter valve. Compare this to the Fig.4 schematic. 96 Silicon Chip Australia's electronics magazine siliconchip.com.au Anode voltage 6J5 0V Av = -10V -20Vpp Scope 3: the 6J5 anode voltage with a 50mV RMS signal fed into the antenna. This is fixed bias, rather than having to rely on self-bias with cathode resistors or other bias methods. It is not until the peak positive going IF voltage, coupled by the 110pF capacitor (part 23 on the diagram) to the anode of the 6J5 exceeds the applied negative voltage value of about -2.5V that the 6J5 develops anode current on the IF carrier peaks. This solidly clamps the carrier peaks on the anode close to ground potential, as shown in Scopes 1-3. The initial negative bias on the anode of the 6J5 delays the development of increasing negative AGC voltage until the carrier voltage at the antenna input exceeds about 50μV. Below that, the AGC is kept inactive. As can be seen from Scope 3, the actively driven 6J5 makes for an excellent peak carrier clamp. Any increase in the IF carrier amplitude results in an increasing average Scope 4: the 6J5 anode voltage on a longer time scale, with an unmodulated signal producing a -10V AGC voltage. negative shift in the AGC voltage. Scope 4 shows the 6J5 anode voltage with an unmodulated 470kHz IF signal, coming out of the IF transformer at 20V peak-to-peak. When this signal is time-averaged by the 1MW resistor (part 15) and the 50nF AGC filter capacitor (part 3), this provides about -10V of AGC voltage. Scope 5 shows the situation with modulation added. This does not affect the average AGC voltage. This anode clamp circuit cannot be used as the radio’s detector because of the initial negative voltage applied to the anode, which causes it to stop its action if the voltage from the IF transformer is below about 2.5V peak. Stability problems and IF neutralisation Philco found that in some cases, there could be instability in the 6A8 converter stage of the original Scope 5: the 6J5 anode voltage on a longer time scale, with a modulated signal, still producing a -10V AGC voltage. Anode voltage 6J5 0V Av = -10V -20V siliconchip.com.au Australia's electronics magazine production models at the top end of the SW band. To ameliorate this, they added a 22W resistor between the 6A8’s cathode and ground. This resistor is present in nearly all under-chassis photos of the 38-7 I have seen, but is not on their original schematic. It is handy to have because it is a useful place to monitor the valve’s cathode current and also to detect the oscillator frequency without significantly pulling it. Philco produced very high-quality IF transformers, and the gain of their IF stages was relatively high. Instability could occasionally be a problem there, too. The fundamental problem of instability in an IF stage relates to the fact that the IF valve’s (or transistor’s) input and its output both have high-Q tuned circuits at the same operating frequency. Any electric, magnetic or capacitive interaction between the two resonant circuits results in energy transfer between them, so instability or oscillation can occur. All active amplifying devices have an output-to-input feedback capacitance. Early transistors with high feedback capacitances, in the order of 10pF, always required neutralisation to cancel this effect. Later transistors with feedback capacitances around 1pF or less seldom needed it in a 455kHz IF stage. A screened grid pentode, such as the 6K7, would seldom require neutralisation because the screen grid provides near-total anode to G1 isolation. Still, occasionally, even screened grid pentode IF stages can be unstable. October 2026  97 Close inspection of the 38-7’s schematic shows that there is an additional coil, not often seen, inside the first IF transformer. This is merely a few turns of insulated wire wrapped around the secondary coil, feeding the grid of the 6K7. It took many years and having a good oscilloscope, such as the Tektronix 464, until I was able to examine the effect of this small coil and observe the relative coil polarities, not shown on the schematic. The suppressor grid of the 6K7 that the small coil is connected to acts as a small coupling capacitor to the 6K7’s anode. It is actually a brilliant idea because its capacitance with respect to the anode is a highly controlled parameter due to the valve’s construction, rather than having to create a more temperamental small gimmick capacitance. The phase relationship of the small coil to the coil driving the grid of the 6K7 is such that the feedback provides a small degree of neutralisation to the 6K7 stage. Audio output stage The 6F6 audio output stage can provide 3W, and the sound is enhanced by the well-sized timber cabinet and 8-inch (203mm) electromagnetic speaker. Tone Control and the Fletcher-Munson effect The radio has a three-position tone switch, which is combined with the power switch. Stray hum injection can occur when the on/off switch is mounted on the rear of a volume control; this can be a real problem in some radios. It is much better where Philco put it. With this combined power-on/tone switch, the hum injected into the loudness tap via the 51kW resistor is heavily attenuated. Assuming there was a capacitance between the power and tone switch contacts as high a 60pF, the presence of the 6nF capacitor in two of the tone switch positions would attenuate the hum voltage by about 100:1. In one position, the 6nF capacitor is shorted out, eliminating hum injection by this route completely. The volume control has a loudness tap. This idea was invented in the 1930s to account for a property of human hearing known as 98 Silicon Chip the Fletcher-­ Munson effect. Harvey Fletcher and Wilden Munson published their paper on loudness curves in 1933, only five years before the Philco 38-7 radio was made. It is interesting how quickly this concept got into the electronics designs of the time. Later, more precise filters were added to audio systems with the loudness button. It is hard to determine who first modified the volume control potentiometer in this manner with the loudness tap. Both RCA and Philco were doing it in the late 1930s. In essence, at low volume levels, we perceive reduced bass. The tap normally has a shunt-to-ground RC filter. In the 38-7 radio, it consists of a series 51kW resistor (#32) and a 6nF capacitor (#38). This bypasses the higher-range audio frequencies to ground, resulting in a relative bass boost at low volumes. It is a form of frequency-­shaping that is dependent on the volume control position. In the first position of the tone switch, the 6nF capacitor is shorted out, giving a flat response. Since the 51kW resistor shunts all frequencies equally, it sounds like the bass is cut. In the second position, no frequency shaping is used, but the loudness circuit operates. And in the third, most clockwise position on the control, additional capacitance is added to the anode circuit of the 6F6, cutting the treble. Rewinding the speaker’s field coil When I first got the radio, the field coil was open-circuit and needed rewinding. This required driving out some spiral pins that attached the coil assembly to the speaker’s frame, and some machining to release the coil. I replaced the pins with screws and nuts – see Photo 9. A new bobbin was made of discs of Formica sheet and other materials. This is a common problem with vintage electromagnetic speakers. Other areas it crops up in are vintage valve interstage transformers with very fine wire. This is invariably due to ‘green spot disease’, which is corrosion of the fine copper wire. The vintage enamel did not protect the copper as well as modern enamels. Apart from making the coil go open circuit, this makes it very difficult to count the original number of turns Australia's electronics magazine because the wire breaks so easily in the corroded areas. Re-winding field coils revolves around filling a fixed-volume bobbin with fine wire, similar to the original, to about the same winding height. The DC resistance of the coil that you end up with is inversely proportional to the fourth power of the radius (or diameter) of the wire. This is because the resistance of the wire drops by the square of the radius/diameter, and the number of turns you can fit in the same volume goes up with the inverse square of the radius/diameter. So if you doubled the diameter of the new wire compared to the original, the DC resistance would be about 16 times lower than the original coil! To drop the resistance by a factor of two, the wire size only has to increase by a factor of 1.2 or 20%. The main target value for a field coil rewind on a vintage speaker is the original coil’s DC resistance. The exact inductance is less important, because one wants the B+ voltage to be about the same as it was with the old field coil, and thus a similar voltage drop across the coil with the average supply current. However, the inductance is also affected in a similar manner to resistance with a change in wire size. In a nutshell, it pays to measure the winding height and the original wire diameter very carefully and stick to both those parameters for the rewind. This is not to say that other parameters do not affect the DC resistance of the final result; for example, how Photo 9: the re-wound Philco 38-7 speaker field coil. The spiral rivets were replaced with screws/nuts. siliconchip.com.au Photo 11: the flexible coupling for the tuning shaft had seen better days. Photos 12 & 13: I cut these pieces of Teflon with a circular punch to make a new flexible coupling. The existing rivets were converted into nuts. neatly or scrambled the winding architecture might be. The enamel thickness of the particular wire will also have some effect. If the original bobbin was layer-­ wound with thin paper between the layers, which the re-wound bobbin does not have, the DC resistance could end up 10-15% higher than the original. But this is nothing as extreme as changing the wire’s size. If the new winding turns out to have a slightly higher resistance than the original, some wire can be unwound from the outer surface of the bobbin, so that is not a drama. Replacing cracked and damaged rubber parts Fortunately these days, with interest in vintage radio restorations increasing, reproduction parts are available for this radio. For example, the cloth seen over the speaker is an exact reproduction woven to match Philco’s original cloth, which has long since disintegrated. I also replaced the rubber mounts for the variable capacitor, along with the rubber feet that sit on the chassis corners, used on many other Philco chassis models – see Photo 10. The tricky part was replacing the two hardened rubber discs in the variable capacitor’s coupler. They were riveted in place, and because the casting is a type of pot metal, one must remove them slowly and carefully to avoid damaging it. Photo 11 shows this coupler. The original rubber material was close to 1/8-inch (3.175mm) in thickness. It was a three-layer laminate with a twin fabric layer embedded in it. To make new discs, I used a piece of 1/8-inch thick Teflon plate I had in my hardware collection. It was firm enough for the task but also flexible enough too. It came out of some disassembled medical machine. It was just big enough to cut two discs from; there was no room for error. I bought a hole punch kit specifically for the job. I made a scale drawing on a computer, printed it and used thin double-­ sided sticky paper (called JAC Paper) to stick the paper diagram to the surface of the Teflon plate. The punching was done in a large vise, using an acrylic sheet as the backing for the Teflon, because it was stiff enough to support it for a clean cut, but not so hard that it damaged the sharp cutting edge of the metal punch. Photos 12 & 13 show the reassembly of the coupler. Over time, the counterweight had apparently lost some mass for a perfect balance with the other rotating parts. I cured that by attaching two 5g iron weights (designed for balancing car wheels) to either side of the counterweight – see Photos 10 & 14. I managed to preserve two of the long rivets by drilling them and threading those with 4-40 UNC threads so the removed end could be replaced by screws. I had to preserve these because of their very low-profile heads, required for clearance on the vertical plate holding the cones. The bulge in their body prevented them from sliding through the hole in the casting after the peened-over part was removed. These pins cannot be driven out with that bulge present, or the casting would crack. I applied a thin protective washer and patiently hand-filed them down for clearance, then smoothed them in a lathe. I found four suitable replacement eyelets for the other parts. Alignment & sensitivity testing Photo 10: I replaced the perished rubber variable capacitor mounts with new reproduction types and added some more mass to the counterbalance weight (lower left) as it needed it. The red parts at upper left are the handmade parts from approximately 1976, the black parts below it are the new reproductions. siliconchip.com.au Australia's electronics magazine I am a great believer in having a controlled RF signal source with a good attenuator and accurate frequency counter to test and align vintage radios. It pays to follow the manufacturer’s alignment instructions faithfully. I use the excellent Philips PM5326 solid-state RF signal generator for this task. It has an inbuilt digital frequency counter and was designed for aligning AM and FM radios. It covers 100kHz to 125MHz and has sweep functions October 2026  99 interesting 6J5 detector/AGC stage which adds no gain) and a pentagrid converter. Most of the system noise is in the 6A8 converter valve, which sets the limitation on sensitivity and noise. A tuned RF stage will improve this, and moving to a triode/hexode converter such as the 6K8 or ECH35, as noted earlier, will also improve the S/N ratio. Many communications radios with an RF stage and a threegang variable capacitor have around a 2μV sensitivity for 50mW output. Summary Photo 14: the new flexible coupling installed in the set. to help with the alignment of TV IF stages and AM and FM detectors too. It is always better when the frequency is measured prior to the attenuator inside the generator, because at very low output levels, external frequency counters can struggle even with added amplification. Many manufacturers of radios after the 1960s provided a specific RF input level at the antenna and the associated audio power output level, such as 50mW, into a dummy speaker load. This is very useful for checking the radio’s sensitivity. However, for many domestic valve radios of yesteryear, not only was the sensitivity figure absent, but the alignment instructions were often brief with poor detail. The maximum audio output power was often quoted; for example, for the 38-7 model they stated “Undistorted output: 3 Watts” on the schematic. Of course, the distortion at full output power is not insignificant, and typically in cases like this, on the order of 10% second harmonic distortion. In the case of the radio’s performance with unspecified sensitivity data, I am interested in what the radio can do and what the radio can’t do, at 100 Silicon Chip the threshold of usability of the radio with weak signals. Weak signal performance is where the system noise and the recovered modulation content of the carrier become subjectively equal on a listening test. Any signal level lower than this and the noise starts to corrupt the recovered modulation to the extent that the modulation starts to become unintelligible. In the absence of other specific data, this is the test I deploy after a full alignment to check the radio’s performance. I determine what RF input voltage level at the antenna, using a 30% modulated carrier, results in a 50:50 noise versus recovered modulation on a subjective listening test. Because any RF signal level much lower than the modulation won’t be intelligible. In the properly aligned Philco 38-7 with its usual 6A8 converter, the 50:50 signal & noise equivalence occurs at an RF input voltage close to 10μV at frequencies below 12MHz. In the high SW band area of 14-19MHz, the apparent S/N equivalence is around 24μV. This is a typical value for a radio with a similar five-valve count without an RF stage (not counting the Australia's electronics magazine The Philco model 38-7 radio ticks all of the boxes for a valve radio from the late 1930s. Its industrial design is outstanding for a home appliance, and the mechanical engineering is well above average for any domestic valve radio. The dial and the tuning arrangements are works of art and science. The radio also provides wide-range SW coverage, in one band, plus the standard MW band. The sound quality is excellent, and radio is a sensitive and selective receiver on both the MW and SW bands, where it easily resolves many weak stations with a relatively short wire antenna. There is little evidence of any significant practical defects in the 38-7’s performance in use, except for the AGC pulling problem on strong shortwave stations and somewhat reduced performance in the high SW band area. Using the ECH35 en lieu of the 6A8 improves those problems, but that is not a must. Philco incorporated new ideas and circuit innovations, such as the loudness tap on the volume control. They also incorporated a creative delayedAGC circuit and a unique IF neutralisation method, while avoiding hum coupling problems from the power switch. They were sensible enough to power the radio via a transformer, which avoided any ‘hot chassis’ problems. Importantly, the chassis can be Earthed with a three-wire mains cord. The radio could be supplied in 110V/220V options, making it suitable for export. Overall, this radio has a blend of both form and function, which is very difficult to beat. They appear on eBay from time to time, mainly in the USA, but with the occasional one showing up in New Zealand & Australia. SC siliconchip.com.au