Fullscreen HTML5Med Res (??MB)HTML4

Not quite vintage radio . . . or is it? by Dr Hugo Holden The Fetron . . . and the one and only all-Fetron radio You would probably be aware that there are some similarities between valves (aka vacuum tubes) and field-effect transistors, or FETs. You may also know that some people have created valveequivalent devices based on FETs. But did you know that there were commercially-made semiconductorbased triode and pentode equivalents known as “Fetrons”? I am fascinated by these, so I built a superhet using little else. 30 Silicon Chip Australia’s electronics magazine siliconchip.com.au T he Fetron, a unique combination of N-channel Junction Field Effect Transistors (JFETs), using the Cascode configuration, was a product of research and development in the Aerospace and Avionics industry (by the Teledyne Company in the USA) in the early 1970s. They were built primarily as a plugin valve or solid-state pentode replacement, although triode equivalents were also made. The basic idea behind the Fetron was to have the electrical properties of a pentode, but no microphony and no heater power consumption, along with the other advantages of semiconductors: greater efficiency and reliability, with lower noise and higher gain. Fetrons usually had a much higher amplification factor than the valve they replaced. Teledyne also produced a range of semiconductor devices such as high-voltage Junction FETs and they still produce beyond excellent-quality miniature RF relays. Every Teledyne product I have inspected and used has always impressed me with its innovative nature, outstanding manufacturing quality, excellent physical appearance and electrical performance. Because of this, I decided to engineer a multi-band radio composed of entirely Fetrons, powered by a single 90V battery or DC supply, and incorporating some of my other favouriteTeledyne devices. Replacing valves with semiconductors The idea of replacing a valve with a plug-in transistor substitute has occurred to many people since the invention of the transistor. Although there are mathematical models for transistors as voltage-to- Reproduced rather significantly larger than life size, this is the TS6AK5 used in the Fetron Receiver. The type number is designed to show its equivalence to the 6AK5 valve. current control devices, fundamentally, they are current-to-current control devices. I know that some people disagree with this (for example, audio guru Douglas Self), but it is generally accepted to be true. In most instances, the input (baseemitter) current controls the output (collector-emitter) current. Valves, on the other hand, are voltage-to-current control devices or transconductance amplifiers, where usually the grid-to-cathode voltage controls the anode-to-cathode current. Transistors in the grounded-emitter configuration have a much lower input resistance than valves in the groundedcathode configuration. When high-voltage JFETs arrived on the scene, they were possible substitutes for the triode valve. They had a similar transfer function of gate voltage versus drain current, compared to grid voltage versus anode current for the triode. Also, JFETs have a similarly high input impedance to a valve. In the grounded-source or grounded-cathode circuit, both the JFET and the triode are influenced by the effective amplification of the drain-togate (or anode-to-grid) capacitance – known as the Miller effect. This capacitance, which is intrinsic to the device, is multiplied by its amplification factor. This limits the high-frequency response and results in significant input to output feedback as the operating frequency increases. In triode circuits, if a tuned circuit with a similar resonant frequency is placed in both the grid and the anode circuit, oscillations occur due to the feedback capacitance and the two resonant circuits exchanging energy with each other. Historically, the Miller capacitance problem was solved with an added neutralisation capacitor feeding back an out-of-phase signal from a coil extension on the anode resonant circuit to the grid (or to the base in a transistor circuit) via a small adjustable capacitor. In early transistor radios, intermediate frequency (IF) amplifiers using devices such as the OC45, which had a sizeable internal feedback capacitance, required neutralisation. Later, better transistors such as the OC169, AF117 or AF127 had a much lower feedback capacitance and didn’t require neutralising in 455kHz IF stages. In vintage TRF radios based on triode valves, the added neutralising capacitor was called a Neutrodon Fig.1: four more-or-less equivalent inverting amplifier circuits. At left is the pentode valve, followed by a pair of triodes in a cascode configuration, two JFETs in the same configuration and the simplified scheme used in the Fetron (which requires specific JFET characteristics). siliconchip.com.au Australia’s electronics magazine March 2021  31 and the radios sometimes called Neutrodynes. Neutralisation is not necessary for grounded drain (collector or anode) or ‘follower’ circuits because the drain (collector or anode) voltage is pinned to a fixed potential, preventing signal feedback via the Miller capacitance to the input gate (base or grid). The pentode, however, has the unique property of high isolation between its input(grid) and its output (anode) due to the screen grid. Pentode valves, for example, are excellent in radio frequency (RF) stages or intermediate frequency (IF) amplifiers as they are stable with a tuned circuit in both the grid and the anode circuit. Fig.1 shows several similar amplifying stages with ‘black box’ input and output circuits. No resistors or bias components are shown, to keep it simple. For the pentode, the screen grid voltage is held at a constant voltage K. This is usually done by connecting it to a resistive divider with a bypass capacitor, or connecting it to the HT supply. Two triodes arranged in Cascode work similarly, by clamping the upper triode’s grid to a fixed voltage K, which sets the upper triode’s cathode to another fixed potential (k). This stabilises the anode potential of the lower triode, and as a result, the Miller effect is eliminated. The JFET equivalent of the Cascode is also shown; to package this circuit in a single device would require four leads. Also, the ‘screen’ connection would require a different bias voltage compared to a valve circuit, so it could not be a direct replacement. The Fetron solves this problem by connecting the gate of the upper JFET to another voltage source; ingeniously, the source voltage of the lower JFET. This voltage is usually constant from an AC perspective in most valve circuits, as the cathode is typically bypassed. If it is not, it still does not matter, as any AC component coupled via the gate of the upper JFET via its source and the drain to the lower JFET is in phase with the input voltage on the gate of the lower JFET. Hence, there is no potential difference across the Miller capacitance (from gate to drain) of the lower JFET. Thus, the Miller effect is still eliminated. 32 Silicon Chip These pages, reproduced from the May 1973 issue of “Practical Wireless” magazine, show that Fetrons were more than a twinkle in an engineer’s eye The drain current properties of the two JFETs within the Fetron have to be carefully chosen for this configuration to work. Equivalent devices Reproduced above is a historical article (1953) on the TS6AK5 Fetron, which was designed to be equivalent to a 6AK5 pentode. There was also the TS12AT7, equivalent to the 12AT7 triode. Note the very high amplification factor of the TS6AK5 Fetron of 22,500, compared to the 2,500 for the 6AK5 valve, even though most of the other parameters are nearly identical. The drain resistance is very high at 5MΩ, as the JFET is an excellent constant-current source. The transconAustralia’s electronics magazine ductance (gm) or ratio of change in plate (drain) current to grid (gate) voltage is also the ratio of the amplification factor to the plate (drain) resistance. In this case, it is 4500μmhos (22,500 ÷ 5,000,000Ω); about the same as the 6AK5 valve. There are three “features” of the Fetron not alluded to in the data. The first is that the metal can must be Earthed if it is being used in a radiofrequency application. The second is that if the input terminal (gate of the lower JFET) is taken positive with respect to the source (cathode connection), the gate suddenly draws current. In the 6AK5 valve, this is a very gentle process, but the TS6AK5 suddenly conducts as the gate siliconchip.com.au runs from a single 90V battery, although later I built a 90V DC mains supply. It is a dual-band single conversion superhet with a tuned RF stage. The frequency coverage is 550kHz to 1650kHz (MW) and 5.7MHz to 18.2MHz (SW). The antenna is a 6-inch (150mm) long, 12.7mm diameter ferrite rod which also works well for shortwave up to about 10MHz. The MW coils are wound with 60-strand Litz wire. Above 10MHz, an external antenna is useful for the shortwave band. The 11 Fetrons are all TS6AK5s, used as follows: • one for the RF amplifier, • one for the local oscillator (LO), • one for the LO buffer, • one for the mixer, • two as IF amplifiers, • one for the audio preamplifier and • four for the audio output stage, wired in parallel for 1W undistorted Class-A output into a 3.2Ω, 4-inch (100mm) speaker. The LO buffer is needed to provide an output to drive a frequency counter. Two Teledyne 2N4886 high-voltage Nchannel JFETs are also used in a bridge circuit for a signal strength meter (Smeter). The detector, AGC and oscillator self-bias diodes are 1N663A silicon diodes (which were one of AMD’s first products). Band changing almost fifty years ago! In the early 1970s, many electronics hobbyists were still coming to grips with the relatively new transistors and other semiconductors. PN junction becomes forward-biased. In most circuits such as amplifiers, the grid (gate) always has a negative bias, so this is not a problem. However, in oscillator circuits that use grid current self-bias, if a Fetron is plugged in place of the 6AK5, the gate draws significant current and the oscillator malfunctions, producing a distorted output with multiple harmonics. This can be solved with a diode in the gate circuit to provide the self-bias function. The third is that practical experiments with the Fetron indicate that the input-to-output isolation is not quite as good as the 6AK5, in that when used in IF stages with identical tuned siliconchip.com.au circuits in the input and output, they are a little more prone to instability. The higher amplification factor might be the reason, as this tendency can be eliminated with a small amount of degeneration to lower the stage gain. So despite the Fetrons being marketed as plug-in valve substitutes, they were not always a suitable direct replacement, depending on the specific circuit. Designing & building an all-Fetron radio I built the radio shown in the photos, which has some unusual features. Its complete circuit is shown in Fig.2. As the Fetrons have no heaters, it Australia’s electronics magazine Band changing is via three miniature Teledyne latching RF relays. These are controlled by a band change switch on the front panel, which is an industrialgrade motor switch from Telemecanique, so it will not wear out in a hurry, and it has a good feel to it. The main three-gang tuning capacitor is driven by an Eddystone ball-epicyclic reduction drive knob and dial assembly. Incandescent lamps are used to illuminate the dial. I also placed lamps inside the battery voltmeter and the Smeter. These meters are moving-coil types which were intended for use in helicopter avionics. I repainted and labelled the faces for voltage and Sunits, respectively. These days, LEDs might be used with a consequent reduction in current. The radio-frequency trimming capacitors are metal vane ceramic variable types, and chassis-mounted. March 2021  33 SC Ó FETRON DUAL BAND RADIO RECEIVER Fig.2: the full circuit of my Fetron-based radio, a superhet with an RF stage, two IF stages and a Class-A audio output. It uses 11 Fetrons (four in parallel in the audio output stage), two JFETs and three silicon diodes. The MW/ SW band switching is achieved using three latching RF relays in metal cans, also manufactured by Teledyne. The RF coils were wound on formers and then placed inside military spec shielding cans with high permeability adjustable powdered iron cores. The IF transformers are 465KHz American-made Miller units. The audio output transformer is made by Hammond in the USA and supplied by AES. Two of the 12V lamps are in the meters, with the remaining six on a stripline PCB added into the base of the Eddystone dial. 34 Silicon Chip Note the 1N663A diode in the gate circuit of the local oscillator for self-bias, to prevent the Fetron gate conduction problem described above. The input is fuse- and diode-protected. Unlike a valve, a Fetron could be damaged by the application of reverse polarity DC. Earthing the Fetrons To Earth the Fetron bodies, I modified the ceramic valve sockets. I did this by removing the phosphor bronze Australia’s electronics magazine and spring assembly from some standard miniature test laboratory clips and fitting them into the centre metal ring of the valve socket using a small machined bush. The phosphor bronze wire is slipped through the spring and then through the centre of the socket from the top. The bush is soldered into the valve section on the socket base, and the bronze wire is folded over and cut off after it passes through the clearance hole in the bush. This results in the siliconchip.com.au flat-top section of the phosphor bronze wire projecting a little above the top of the socket. When the Fetron is plugged into the socket, the bronze wire springs against the Fetron’s base, securing the Earth connection to the Fetron body without having to make a soldered connection. Mechanical construction The chassis is grey painted steel. It was supplied by AES (Antique Electronic Supply, USA). After making all siliconchip.com.au the holes, I painted the bare edges. To prevent any surface damage, I heavily coated the chassis and panel in plastic tape while cutting the holes, so that they remained scratch-free. The front panel was crafted from 3mm thick stainless steel and treated to create an engine turning finish (also known as jeweling or guilloché). All the hardware used in the radio, mostly 6-32 and 4-40 UNC machine screws, is stainless steel. These were supplied by PSME (PreAustralia’s electronics magazine cision Scale Model Engineering) in the USA. The Fetron sockets are ceramic with gold-plated pins. The wiring in the unit is with highquality Teflon multi-coloured hookup wire from a submarine parts supplier. The front panel handles are chromeplated brass. The switch labels, for the most part, are pre-made items which came from the electronic markets in Akihabara, Japan. The tag boards used on the radio underside also came from there. March 2021  35 No-one is expecting you to be able to build your own Fetron Radio from these photos . . . but just in case (!) you can get a very good idea of both the above-chassis layout and the under-chassis wiring. The Speaker mesh is perforated aluminium with a clear lacquer applied. Captive pressed stainless steel 4-40 nuts were fitted to the chassis base to allow repeated removal of the base plate. The three Teledyne RF relays (in TO-5 cases) have spring clips to Earth their metal bodies. 36 Silicon Chip Australia’s electronics magazine siliconchip.com.au SC Ó DC-DC CONVERTER FOR FETRON RADIO The latching relays save battery power and are driven by a simple RC network, which provides a current pulse to execute band changing. The two TO-5 cased JFETs for the S-meter can be seen in the chassis underside view with the red, green and black sleeving on their leads. The three-gang variable capacitor is mounted with posts within rubber grommets to prevent acoustic feedback to the capacitor’s plates. As all of the trimmer capacitors and adjustment potentiometers for the S-meter are chassis-mounted (just above the 2N4886 JFETs), all adjustments can be made from above the chassis top. I created the dial artwork in a photo editor and made it as a transpar- Fig.3: the circuit of my 12V-to-90V step-up supply which I use to power my Fetron radio when I don’t want to use the 90V Nicad battery! It’s designed to bring NPN transistors Q1 and Q2 (which drive transformer T1) into and out of conduction slowly, at 40Hz, eliminating EMI which would otherwise affect radio reception. ent sticker, which I then applied to the metal Eddystone dial plate. I very carefully cut the kidney-shaped meter holes in the dial plate and front panel by hand. Power supply The radio itself draws about 47mA <at> 90V, making its power consumption around 4W. That is significantly less than a valve radio employing 6AK5s because there is no heater demand. The current consumption with the dial lamp string running is 75mA. About 2.5W is consumed by the Class-A audio output stage, which has a current drain of 28mA. A Class-AB output stage would draw significantly less, but calculations showed that it would have been My home-made power supply PCB is pleasingly simple. It is dominated by the PCB-mounting transformer, two TO-66 package driving transistors and high-voltage output filter capacitors. siliconchip.com.au Australia’s electronics magazine March 2021  37 These scope grabs just how gentle the switching waveforms of transistors Q1 and Q2 are. Even at the longer timebase used in the left-hand scope grab, you can see that they are not vertical lines but rather smooth ramps, reducing the higher-order harmonics that are typical of square waves and this minimising high-frequency EMI. more difficult to attain the 1W output with two paralleled Fetrons per side. Also, a phase inverter circuit or transformer would have been needed to drive them. The Class-A output stage, although a little more powerhungry than Class-AB, does give very good results with pleasant-sounding audio reminiscent of a typical valve radio. I made the 90V battery from many 2000mAh AA-sized NiCad cells and stuck Eveready logo on it for a bit of fun. Step-up supply Ideally, the radio would be powered by a rechargeable 12V battery or 12V DC plugpack. This would require a 12V-to-90V switch-mode converter. Many enthusiasts of valve radios have attempted this sort of converter, but RFI or radio frequency interference (affectionately referred to as “hash”) is a significant problem. This can result in buzzing signals being detected by the radio. A medium-wave or shortwave radio makes a very sensitive detector of radiated electromagnetic fields! Most people would be surprised by the high levels of RFI I modified the sockets by soldering in a brass bush and using it to hold a spring-loaded bronze wire which contacts the Fetron case when it is inserted. This means that I can Earth the Fetron case to provide adequate shielding, without affecting their pluggability or having an ugly solder joint on the case. 38 Silicon Chip emitted by appliances like computers and flat-panel TV sets. These signals can not only cause interference on shortwave reception, but they can also desensitise RF receivers in home automation systems. Some folks have had solar systems with switchmode inverters installed, only to find that their garage door controllers stop working! So I set about creating an RFI-free step-up circuit to power my radio. The result, shown in Fig.3, is somewhat similar to Ken Kranz’s Battery Vintage Radio Power Supply from the December 2020 issue (siliconchip.com.au/ Article/14670), although there are some important differences. It delivers 90V <at> 50mA with an input of 12.6V <at> 550mA, giving an overall efficiency of 65%. There is no detectable RFI above 150kHz; I didn’t even bother shielding it. It uses a Jaycar PCB-mounting toroidal transformer, driven in push-pull mode at around 40Hz. Its low operating frequency, combined with the ironcored transformer reduces the switching events per unit time, and this helps compensate for the deliberately slow switching transitions. The slower transition time contains lower HF spectrum components. The switching time and transition shape were controlled by tuning the primary of the transformer with a large capacitor and RC snubber networks on the transistor’s collectors. Also, the drive to the switching transistors is adjusted to be enough to gain saturation of the collector-emitter voltage to 380-400mV and no lower. Experimentation shows that all other things being equal, the RFI increases significantly the more heavily the transistor is saturated. RFI is produced when the transistor suddenly comes out of heavy saturation The two scope grabs above show the collector waveform from one of the 2N3054A transistors at two different timebases. You can see that the transistor switches slowly between being in and out of conduction, over about 0.8ms each time. While this reduces the efficiency, this is offset by the slow switching speed, so the number of switching events per unit time is relatively low compared to most switch-mode PSUs. SC Australia’s electronics magazine siliconchip.com.au