Fullscreen HTML5Med Res (??MB)HTML4

Vintage Radio RCA Radiola 17 (AR-927) radio from 1927 The Radiola 17 is an interesting seven-valve ACpowered tuned radio frequency (TRF) set from 1927. By Jim Greig T he Radiola 17’s price on release was US$157.50 with valves, equivalent to around US$2900 or $4450 today. A TRF radio comprises one or more tuned RF amplifier stages in series. Each stage includes a bandpass filter tuned to the same frequency, which amplifies the desired signal while attenuating others. After several stages, the selected signal is significantly amplified, while out-of-band signals are progressively filtered out. The earliest such sets had individual tuning capacitors for each stage, making tuning to a station an art. By 1927, the use of a ganged tuning capacitor meant that only a single tuning knob was required. Radios were operated almost exclusively from batteries until around 1925. Early battery sets used directly heated valves, where the filament also served as the cathode. Later designs introduced indirectly heated valves with separate filaments and cathodes, solving several technical problems. One minor problem was uneven voltage distribution, both along the length of the filament in a single valve, and between the filaments of different valves. A more serious problem arose when sets began using mains power: if the filament also acted as the cathode, powering it with AC introduced 92 Silicon Chip significant hum. Various methods were developed to minimise this, including: 1. Using DC for the filament supply. This often meant using a battery in the 1920s, as high-current, low-­ voltage rectifiers were not yet available. 2. Centre-tapped hum-balancing potentiometers. A pot across the filament with its centre tap grounded made the ends of the filament equal and opposite in phase, helping to cancel the hum. 3. Special filament coatings. These may have reduced temperature variation along the filament over the AC cycle. 4. Lower filament voltages. Any AC ripple imposed on the signal would be less than it would be with a higher voltage. 5. Separating the cathode and filament. The filament then acted purely as a heater, with the cathode isolated electrically, eliminating the main source of hum. The first AC-powered set may have been the Rogers Batteryless from Canada. Rogers designed and produced their own AC valves (type 32) with a separate cathode. Powering the radio from AC saved the costs of batteries, making radio accessible to many more people. In 1927, RCA introduced a new tube, the UX-226 (also known as the type 226 Australia's electronics magazine or type 26), a triode that could be used for any stage of a receiver except the detector (unless the designer was willing to accept inferior performance). It had a low-­voltage (1.5V) coated filament that was optimised to produce very little hum when run on AC power. The same year, RCA announced the UY-227. The type 27 had the same shape as the type 26, but its filament was arranged to heat an oxide-plated cathode connected to a fifth pin in the base. Its filament was isolated from the electron path, and the new tube made an excellent detector (or audio amplifier). The RCA 17 was RCA’s first AC-­ powered receiver, and it uses the following valve types: UX226 (first RF), UX226 (second RF), UX226 (third RF), UY227 (detector), UX226 (audio preamp), UX171 (audio output) and UX280 (mains rectifier). Circuit details Much of the following information was derived from the RCA Service Notes and Service Data. The main part of the circuit, shown in Fig.1, is deceptively simple. The first RF stage is untuned, with the antenna connected to the grid through the volume control. Unlike in early battery sets, where the volume was siliconchip.com.au siliconchip.com.au Fig.1: at first glance, the RCA Radiola 17 circuit seems to be little more than seven valves connected in series with coupling transformers, some with tuning. often controlled by varying filament voltage, AC-powered sets required a different method. In this case, the volume control is simple but effective: a potentiometer in series with the antenna provides adjustable attenuation of the incoming RF signal before it reaches the first RF amplifier. The second and third stages are tuned RF triodes. The triode has significant anode-to-grid capacitance, and this, multiplied by the gain (Miller Effect), results in capacitances of tens of picofarads. A triode RF amplifier with high-Q tuned circuits at the grid and anode is especially susceptible to parasitic oscillation. Something must be done to ensure stability. This may be: 1. Neutralisation. Feed back an outof-phase signal from the anode circuit through a carefully selected capacitor to the grid. This was subject to a patent by the Hazeltine Company. 2. Including a resistor, often in the tuned circuits, to reduce the Q. However, Don Sutherland argues it is the phasing of transformers and the layout that have the most effect. In this radio, the coils are all at right-angles to reduce mutual coupling and improve stability. There is also a resistance (800W in the circuit, but my resistors were actually 1kW) added in series with the grids. There is a section in the Service Notes on “Uncontrolled Oscillation”, with a number of possible remedies, including dropping the filament voltage (and therefore gain) of the type 26 valves. The third stage is coupled to the grid-leak detector through a mica capacitor of around 150pF. A gridleak resistor bleeds off any charge that might accumulate on the grid from rectification of the incoming RF signal, preventing charge build-up and allowing the grid to operate at a slightly negative voltage. In a grid leak detector, the grid/ cathode of the tube acts as a rectifier, albeit an inefficient one. While the grid voltage to plate current transfer relationship may not be perfectly linear, non-linearity is not required for detection, unlike in an anode bend detector. The RF is filtered off at the plate, and only the average voltage remains; the audio interstage transformer does this from its limited bandwidth and any residual RF is filtered out by the capacitor in parallel with the transformer Australia's electronics magazine March 2026  93 Fig.2: the power supply circuit is similarly quite simple. The connector strip on the left corresponds to the one on the right in Fig.1. That’s how they are physically connected in the radio. 94 Silicon Chip Australia's electronics magazine output. If no audio interstage transformer is used, the RF can be removed by a capacitor in combination with the valve’s anode resistor and its plate resistance. The rectified signal is fed via an audio transformer to the first AF stage. Its output is transformer-coupled to the output stage, which has a 1:1 transformer to the speaker. AF transformers were used in early radio for coupling as the valves had little gain, and the (typically) 1:3 voltage ratio offered by a transformer was basically free (sometimes capacitive coupling was used too). The amplification factor of the type 26 is around eight times; the two audio transformers add another nine times, or as much as an additional 26 valve. Today, the amplification factor of a 12AX7 (for example) is around 100 times, and the losses from RC-coupling a 12AX7 are easily covered. The transformers are just two windings with no interleaving and minimal iron, resulting in high leakage inductance and limited bandwidth, limiting this radio’s audio frequency range. The expected loudspeaker response at the time was also limited, so this was not a concern. Audio transformer properties Audio interstage transformers in vintage radios are special parts; they cannot be analysed with the usual equations for a transformer because no power is transferred. The grid of the following stage (operating in Class-A) draws no significant current. The analysis that applies to them is that of a damped, coupled resonant circuit. The damping is provided by the anode resistance of the valve driving the transformer. The resonant frequency is defined mainly by the self-capacitance of the secondary winding, with contributions from the primary. The mutual inductance (and therefore leakage inductance) is very important in the balance of factors that result in a flat frequency response. When the manufacturer gets the balance of factors just right, the transformer behaves as an astonishingly effective bandpass filter in the audio spectrum, with a flat response. They should always be preserved where possible. Their band-pass response typically extends from as low as 100Hz, up to around 7-10kHz. They provide improved dynamic siliconchip.com.au range compared to anode resistor loads as they hold the anode voltage closer to the B+ voltage. They can provide passive voltage amplification up to around three times, or at high as five times; the higher the gain, the more difficult it is to attain a flat response. A block of five wax/paper capacitors is included to bypass the type 226 filaments (both sides) to the chassis, bypass the 135V line to the chassis, and bypass the ‘ground’ to the 135V and -9V rails. There are some unexpected connections in both this and the main filter block in the power supply. I expect the engineers were minimising the component count and making the best use of the space available. The three RF and first audio stage valves (type 26) have a grid bias of -9V. While it is easy today to generate any DC voltage, in 1927, ingenuity was required to keep the design simple. Cost and the availability of skilled repair people required that these early mass market sets were straightforward. Power Supply The AC is rectified with the UX280 valve, new for 1927. It was followed by a choke input filter, with the inductors in the 0V path, probably to minimise flash-over to the frame and electrolysis of the wires. A resistive divider and wax paper filter capacitors provide the four supply voltages and the ‘cathode bypass’ for the output. The connections to the series dropping resistors in the power supply were arranged so that the chassis is at -9V with respect to the filaments of the type 26 valves. Their grids connect to the chassis through transformer primary windings and the volume control to provide the required -9V grid bias. The detector has zero bias, and the output valve is self-biased with a 1690W cathode-bias resistor from the tap on the balancing pot across the UX171 filaments to the -9V rail. The power supply, shown in Fig.2, generates DC voltages of -9V, 45V, 135V and around 170V for the output stage, as well as 1.5V AC, 2.5V AC and 5V AC for the filaments. These DC voltages are measured from the ‘ground’ tap on the resistor chain, not the chassis, which is at -9V. Restoration At around 17kg, the radio is heavy. It was in good condition when I received it, with no major rust visible, and the cabinet had no major damage. The circuitry is in two parts: a power supply, with the radio linked to it via a multicore cable and secured with metal screws through the cabinet. I was able to remove these sections easily, separating them by undoing the screws connecting the cable to a tag strip on the radio. I have worked on many transistor and valve radios, but this was the dirtiest job so far. 100 years of dust and fine dirt on the surface with wax and pitch to be encountered later. I first checked the mains transformer. If it was damaged, I felt it would be best to preserve the radio as-is. I replaced the mains cabling with a three-core cotton-covered cable, with a small fuse added in the Active line, and the Earth connected to the chassis. I applied 50V AC from a variac to the mains input and left it for several hours. The 220V/240V switch was set to 240. I measured 62V a side for the HT, 1-2V for the filaments and the transformer The original waxed-paper capacitors were all packed into a small metal box. Not surprisingly, 100 years later, they have become leaky. did not heat up. It was left to run for a while on 120V AC, then 240V AC. I estimated the HT current requirement as being 15mA for the output valve and 5mA for the others, giving 40mA in total. To emulate this, I connected an 18kW across the HT windings. With 230V AC at the input, the outputs were 586V AC across the whole HT winding (~293-0-293V AC centre-tapped), 1.49V AC, 2.21V AC, and 5.05V AC, and the transformer stayed cool. So, the mains transformer was functional, and the restoration could continue. The two chokes (reactors) were next. One measured around 11H, the other 0.6W, a likely short circuit as the insulation on the wires to it was crumbling. They were encapsulated in pitch. Two removal procedures are suggested: warming with a heat gun, or dissolving with paint thinners. I used paint thinner; the heat gun approach would probably have been less messy. With both inductors out, The rear of the Radiola 17 with the back taken off. The power supply circuitry is contained in the enclosure on the far left of the interior and its circuit is shown in Fig.2 opposite. siliconchip.com.au Australia's electronics magazine March 2026  95 Table 1: replacement resistors Original resistance Nominal New resistor voltage 410W 135V 390W 5W 3750W 45V 3.9kW 5W 2140W 45V 2.2kW 5W 205W -9V 200W 5W 1690W 30V 3.3kW 1W || 3.3kW 1W they looked alright, so I checked them with a bridge. The second one measured 18H; my simple RLC transistor checker gives up at 15H, so it read as a resistor. The wires were in poor condition and heavily oxidised. I cut them near the choke and cleaned them with many passes over the ends with folded emery paper. I then applied flux paste and soldered new wiring to both, then insulated the joints with heatshrink tubing. More of the original wiring was in poor condition, so I replaced it with segments from a length of HRSA seven-­ conductor battery cable. The cabling was laced to keep it tidy and preserve some of the original appearance. All filter capacitors showed significant leakage, so I replaced them. They are housed in a large metal case and held in with extra wax. It’s a beautiful design that filters six different voltages within a limited space. Unfortunately, heating the wax did not release them, so I had to cut a few out so the remainder could be extracted. I replaced them with multiple 1μF 200V Mylar capacitors. As there is plenty of room, I doubled the value of each capacitor to provide additional filtering. The multi-tap wire-wound resistor The volume control had degraded over time and needed a delicate repair. showed signs of overheating; most segments were open, and the rest intermittent. Ideally, new wire would be wound on the ceramic former and tapped, but I am not that skilled. Using the voltages and other data from the service notes, I calculated the power for each segment. Assuming the original values would be accurate to within 10% at best, I selected a combination of standard resistors, as shown in Table 1. The remaining working segments on the original resistor were permanently broken and new resistors placed under the tags and connected to them. If a future owner wishes to restore the original resistor, it is still there. Seized tuning capacitor The tuning capacitor would not move. I oiled the shaft near the tuning wheel and left it for days. Eventually, it could be moved slightly, but it was still very stiff. The tuning wheel is made of pot metal (a brittle zincbased alloy), which was cracking, so I had to treat it carefully. The only way to free the movement was to remove the wheel. This required knocking out the pin on the shaft and, given the state of the wheel, it may have disintegrated. I took measurements and photos so a new wheel could be created if needed. Once the pin was out, the tuning wheel slid off, and the shaft rotated easily. The pot metal had expanded and was binding to the screw locating the shaft. I filed the binding end to allow the shaft to move sideways, to centre the variable plates within the fixed portion. The wheel had to be stabilised, or it would break in future. A local company suggested using epoxy glue, so I purchased some E-143 metal epoxy from Technicqll in Poland. I washed the wheel to remove grease and oil, then forced the glue into the cracks, small sections at a time. I blocked the shaft and tension screw holes with Blu-Tack so I wouldn’t accidentally get glue in them. Once the wheel was repaired, I refitted it with thin stainless washers added to ease its movement where the shaft was binding. I connected the dial wire to the tuning shaft, stretched it over the grooves in the wheel and tightened it with the screw and clamp. The radio had resistor/capacitor coupling added between the first audio valve and the output type 171, suggesting the coupling transformer was broken. Resistance checks showed the primary was open circuit. The transformers are located in a metal shell clipped to the chassis, so I unsoldered the wires from both transformers and unclipped the transformer case. It was filled with wax – better than pitch, I suppose; I used a heat gun to melt it. Both transformers are roughly made, with no clamping of the laminations; they are ‘glued’ together with wax. A suitable transformer (very small, with a 3:1 turns ratio) was donated by an HRSA collector. I connected the wires from the failed The tuning capacitor wheel had seized. Care was required in dismantling and fixing it as it was made of pot metal that had become brittle. The series wirewound resistors (encased in black, at top) were bad so I bypassed them with a string of modern resistors. 96 Silicon Chip Australia's electronics magazine siliconchip.com.au transformer to it and refitted both in the case. Ideally, they would be installed at right-angles to each other, but the new unit was slightly larger, so I had to make them parallel. The antenna volume control was another casualty of time; the resistive element was corroded in places, and the resistance wire had broken. Advice from another HRSA member was to bridge the broken sections by finding the ends, gently scraping off the insulation with fine emery paper and twisting them together. The resistive wire does not solder, so the twisted section was covered in conductive silver paint. There will be tiny (1-2 wire) sections of the control that are missing, but for its use as a volume control, it will not matter. This repair worked perfectly. The grid stopper resistors are wire wound and tested alright. The grid leak resistor is supposed to be 4MW but measured as 3MW. I replaced it with a 3.9MW ½W resistor hidden under the original red 4MW resistor. The bypass capacitors are held in metal cases clipped to the chassis. I tagged and desoldered the connecting wires, then removed the assembly. I then heated the cases and removed the capacitors. The cases were washed with vegetable oil (to dissolve the wax) and detergent, glued together and the capacitors replaced with 1μF and 2.2μF 250V polyester types. I emission tested the valves and replaced any that failed. With everything tested, I figured the radio should work. The output transformer is a 1:1 type and the recommended anode load for the UX171 (type 71) is 4.8kW at 180V. For initial testing, I used a 5kW to 3.5W transformer and 4W speaker. Fig.3: the frequency response of the set, either injecting a signal directly into the detector output transformer (cyan) or directly to the antenna (red). Expectations were lower in those days! Having attached the multi-core cable from the power supply to the radio, I powered it slowly through a variac with the UX280 out of circuit. All valve heaters were operational. Plugging in the rectifier gave the voltages shown in Table 2. All were a bit high, so I needed to swap in a ‘worse’ type 80. For further testing, I employed the HRSA mini transmitter and a 1m wire antenna. The signal was tuned in easily, and the volume control had to be wound down to minimise distortion. With no signal, there was some 100Hz buzz from the medium-fidelity test speaker. Examining the output with an oscilloscope, there are noise spikes of around 1V peak-to-peak at 100Hz. Tracing back through the circuit, they were present on the detector anode but not on the grid. The +45V rail (and other voltages) had some ripple, but no noise. I suspect it is coupling between the two audio transformers causing the problem. 50Hz hum was also visible, the minimum residual after adjusting the three filament potentiometers. I tested the receiver audio bandwidth but there is no specification for this in the original documents. Removing the type 27 detector, I connected a signal generator to the detector audio coupling transformer through 10kW to simulate the anode resistance. Audio bandwidth is often specified between -3dB points on the amplifier response curve, but today’s amplifiers have broad flat responses, and I feel applying -3dB to this radio is not fair. -10dB, or half the perceived loudness, would be easily detectable but not prevent the signal from being heard. Here, the -10dB bandwidth is around 4507500Hz (see the cyan curve in Fig.3). The frequency response from the antenna to the output is a similar shape, with a more useful -10dB bandwidth of 200-4500Hz (the red curve in The transformer between the audio preamp and audio output stages had gone open-circuit. Table 2: voltages on initial power-up The tuning capacitor wheel after filling the cracks with glue. siliconchip.com.au Australia's electronics magazine Rail Reading Raw HT 175V 135V 151.3V 45V 52.8V 9V 9.1V UX171 filament (5V) 5.1V UX226 filament (1.5V) 1.5V UY227 filament (2.5V) 2.2V Bias on UX171 filament (-30V) -26.8V March 2026  97 A better look at the chassis of the Radiola 17. Fig.3). The RF bandwidth of the TRF is generally much wider than a superheterodyne set, but in this set, it tapers off from about 1kHz. A possible cause is that there are no trimmer capacitors on the tuning gang, and it is quite likely the three sections are not aligned, leading to unpredictable bandwidth of the tuned signal. The lack of trimmer capacitors made factory alignment critical and limited user retuning. The audio from a music CD played over the mini transmitter was quite intelligible on the bench test speaker and not (to my tin ears) greatly distorted. Cabinet restoration The timber cabinet for this radio is in very good condition for its age. There are scratches and a probable burn mark on the top, but the other surfaces are reasonably clear. It is missing the hood over the dial light, a common problem with these radios. I considered refinishing the cabinet, but it is 100 years old and you can’t expect it to be in mint condition. So I simply cleaned the timber and rubbed it down many times inside and out with a mixture of 50/50 white spirits and linseed oil. The appearance remains consistent with its age. Over time, the brass escutcheons have oxidised and discoloured. I washed them but didn’t polish them, so that they and the cabinet look right with each other. RCA Speaker Model 100A The RCA 100A speaker was sold with this radio, and this one was bought at an HRSA auction. The case is made of pot metal and it is breaking up in parts; small sections have flaked off. I will clean it up and repaint it in the future. The speaker is a high-impedance device, as were the horn speakers and headphones of the time. When connected to a (late model) radio through ◀ Fig.4: the unusual construction of the RCA Model 100A loudspeaker. 98 Silicon Chip Australia's electronics magazine a step-up transformer, it works with no grating or scratching noises. It is a moving armature design (sometimes called ‘balanced armature’). Fig.4, from the RCA Model 100A Service Notes, shows the mechanism. Not shown is the armature sitting close to the pole pieces of a horseshoe magnet. Changes in the magnetism of the armature from the coils will cause it to move to one or other pole piece at one end and move the drive pin at the other. The motor is small, at 40 × 25 × 40mm, and sits within the magnet. There is a low-pass π filter between the input and the drive coils; possibly the speaker mechanism rattles if driven with high-frequency signals, so they are filtered out. At the low-­ frequency end, the cone is very stiff, which would limit the low-­frequency response. I checked the speaker frequency response with the PC-based AUDio MEasurement System (AUDMES). The RCA 100A loudspeaker is housed in this early Art déco style pot metal and fabric case. siliconchip.com.au Table 3: model 100A speaker Frequency Input impedance 100Hz 4.36kW 525Hz 11.82kW 2730Hz 2.0kW 4700Hz 8.7kW A good-quality line transformer was connected between the PC to the speaker for impedance matching and to increase the drive voltage. The -10dB response is around 200-3500Hz – see Fig.5. The speaker resistance is reflected back to the output valve, so I thought it would be interesting to see how close it was to the desired 4.8kW load specified for the type 71 valve. I estimated the speaker resistance by connecting it to a signal generator through a variable resistance. When the voltage across the speaker equalled the drop across the resistor, the resistances would be equal, and the variable resistor could be measured. With the π filter and the speaker coils, a complex impedance was likely. I took measurements from 100Hz to 8kHz and recorded the highs and lows in Table 3. The input impedance drops below the desired 4.8kW over the range of 2-3kHz, but generally it is well above it. Listening to a CD received from the mini transmitter via the radio and RCA speaker, it is not HiFi, but the music was clear. I expect at the time it was released, people would have been very impressed. Fig.5: the frequency response of the RCA Model 100A loudspeaker. Screen 1: the repaired radio set picked up some mains hum and buzz. Partly this could be due to its unshielded TRF construction, but my replacement audio coupling transformer having to be reoriented might have also negatively impacted its EMI rejection. References • Deeth Williams Wall: siliconchip. au/link/ac7v • RCA Victor Service Notes: 1923 to 1928 • RCA Service Data: 1923-1932 Vol. A • RCA Loudspeaker Model 100A Service Notes, June 1927 • Don Sutherland NZVRS Bulletin, Vol 5 Number 2, August 1984 • RCA Radiola 17, Eric L. Santanen, Bucknell University • https://sourceforge.net/projects/ SC audmes/ siliconchip.com.au A close-up of the loudspeaker moving armature motor. Australia's electronics magazine March 2026  99