Fullscreen HTML5Med Res (??MB)HTML4

D-200 RADIO TRANSMITTER /DVWPRQWK,JDYHVRPHEDFNJURXQGRQ6SXWQLN WKHƬUVWDUWLƬFLDOVDWHOOLWH DQGH[SODLQHGKRZ, UHFUHDWHGWKHUHOD\EDVHGq0DQLSXODWRUrWKDW VZLWFKHGWKHWZRUDGLRWUDQVPLWWHUVRQDQGRƪDW +]:HSLFNXSZKHUH,OHƱRƪP\QH[WMREZDV WRUHYHUVHHQJLQHHUDQGEXLOGRQHRIWKHUDGLR WUDQVPLWWHUVWKHQFUHDWHDVXLWDEOHSRZHUVXSSO\ A Vintage Radio Story, Part 2 By Dr Hugo Holden T he Manipulator is an oscillator based on two sensitive relays. It alternately switches off the output valves in the two transmitters by disconnecting the screen grids, stopping the transmitted carrier wave. Each transmitter is on for ~0.2s at a time, then silent for a similar period. Having gotten my Manipulator working like the original, mainly using period-authentic parts, I turned to the three-valve-based transmitters and the chassis they were built into. While I was not planning to go as far as to produce a complete D-200 unit with two transmitters and the Manipulator, I wanted to make a period-­correct recreation of one transmitter along with the Manipulator that I could put on display. I knew some parts would not be identical to the originals, but I was confident I could get very close. Transmitter details The two transmitters are based on 86 Silicon Chip small 2P19B pentode valves, which are still readily available. The data sheet extract shown in Fig.10 includes the customary bottom view of the valve’s base. Another 2P19B data sheet shows screen and suppressor grid connections reversed, as if viewed from the top. Still, it is easy to tell from the valve itself that this data sheet is correct. Before building the transmitters, I made a test jig to verify that the 2P19Bs I had bought (some shown in Photo 2) were functioning normally. They had been stored in corrugated cardboard rolls with a thin paper wrapping, which is not ideal, resulting in some corrosion on the tinned copper leads. I had to clean that off, initially by scraping and then smoothing the lead with 1000-grit sandpaper, being very careful not to bend the wires near where they enter the glass envelope. To determine their ‘normal behaviour’, I tested over 30 valves, Australia's electronics magazine a fair statistical sample. Three were defective: two had low gain, and the other had let in air. Fig.11 is the test jig circuit, while the actual device is shown in Photo 1. I took the sockets on the test jig that receive the wires from the 2P91B valve from some machined-pin IC sockets. I tied grid 3 to +12V rather than ground because it was tied to +10V in the Sputnik transmitter output stage. I added a 1kW series resistor to avoid an accidental short between the grid pin and adjacent heater pin from applying 12V to the heater. I used a 12V gel cell to power the filament circuit and my dual 0-60V CPX-200D bench power supply, connected in series, for the 120V test voltage. Sputnik-1 20.005MHz transmitter design The transmitter circuit is shown in Fig.12. Valve V1 is deployed as a crystal-­controlled oscillator while V2 & V3 (all 2P91Bs) form the push-pull power amplifier. The valves have a 1W plate dissipation, so a pair running in an output stage, sharing the load, will have no difficulty delivering a 1W RF output, provided there is adequate drive voltage (close to 40V peak) at the G1 grids. The circuits for the 20.005MHz and 40.002MHz transmitters are practically identical, aside from the coil and capacitor values. In the 40.002MHz unit, the main change was that they did not tap off the main tank circuit for an impedance match with the antenna, as they did for the 20MHz unit. They used a capacitive divider instead. siliconchip.com.au Photo 1: the test jig in action. The anode wire comes out the top of the valve envelope, hence the need for the clip lead. A detail not shown in the original circuit diagram is that the L5 and L6 coils are built into a rectangular can. Capacitor C28 is not visible in any historical photos, so it most likely was in the same shield can. C29 is visible in the photos, though (see Photo 3). In Photo 4, the shield around the glass-bodied crystal appears to project a little above the housing, but the shield can for L5 and L6 does not look that tall. I determined the transmitter chassis’ dimensions by studying the photos and scaling from the image details and the limited geometry data in the design document. I determined that the housing around the transmitter modules was 93mm wide, suggesting the chassis was 90mm wide, 180mm long and 60mm deep. It was OK that the crystal shield projected a little above the chassis height in the original unit because this side of the transmitter module faced the Photo 2: some of the 30 2P19B valves I bought, of which three had failed. They had not been stored properly, so I had to clean the corrosion off the wire leads before testing them. siliconchip.com.au Fig.10: a page from a data sheet for the 2P19B pentode showing its pinout and critical parameters. Fig.11: a simple test circuit for the 2P19B pentodes that allowed me to weed out three faulty valves from the 30 I purchased. A test signal can be fed in, and the amplified output signal examined with various external load resistances. Australia's electronics magazine December 2023  87 Fig.12: the Sputnik-1 20.005MHz transmitter circuit. The two transmitters were very similar but had some slight differences besides the crystal frequency. Note that most versions of this circuit (including one we published previously) contained errors; this one should be accurate. interior of the D-200 housing, where there was clearance. The original document shows the width of the main unit that carries the two transmitter chassis as 132mm, more than enough to accommodate two 60mm-deep units with 12mm to spare, so a mid-line panel and wiring could run through the main body. Lead dress for the 2P19Bs Photo 5 shows how I insulated the bare valve leads with PVC tubing, although I later decided to use Teflon sleeves instead. Replicating the chassis When it comes to making replicas of a vintage electronic apparatus, the most difficult part is the mechanical engineering aspect of the project. If not done well, the final result does not represent how the unit actually worked and looked. It takes quite a while to examine the historical photos and figure out where the components were placed and the original geometry of the internal and external panel work. A good replica also requires tracking down most of the original parts; not just the valves but also resistors and capacitors, because they have a characteristic look, especially the Soviet chassis-mount and RF feed-through capacitors. Also, for RF apparatus operating above 5-10MHz, physical layout and shielding considerations become very important. This includes the mounting clips that attach the 2P19B valves to the module body. These serve as partial shields and conduct some heat away from the valves as well. Therefore, it is best to stick to the original physical layout closely. To make the transmitter module’s metal chassis precisely the same as the original would require the same tooling. The metalwork had been riveted and soldered together in places. Without the tooling, other methods exist to create a nearly identical-­ looking metal module of almost identical geometry. I decided to make the metalwork out of brass, which is easily soldered. I used 3mm-thick plates to replicate Common mode choke with glued slug C26 Capacitor missing from schematic – C47, 1.2nF 250V R10 C27 C29 C34 20.005MHz crystal in glass envelope Photo 3: a photo of the original transmitter with C27 and C29 visible, but C28 is nowhere to be seen. It makes sense that it was in the shielding can with L5 and L6 since it connects to both. 88 Silicon Chip Photo 4: this photo of the 20.005MHz transmission unit shows that the crystal shield was taller than the shield for L5/L6/C28 and even projected outside the chassis slightly. Australia's electronics magazine siliconchip.com.au Figs.13-20: these are the mechanical drawings that I provided to the machinist who made my reproduction transmitter metalwork. Fig.13 the top and bottom faces of the module, routed and engraved with a groove to fit the side panels, made of 0.8mm-thick brass. The three internal panels were also CNC machined. They are all soldered together too. This method avoided having to fold any metal panels, which can distort the material. I prepared Figs.13-20 to help with this task. Troy at Sunquest Industries in Warana, Maroochydore (Qld) did the CNC machining. The projections on the sides of the plate are 1.5mm tall and 5mm wide. The slots in the other panels that they pass into are 1.5mm wide and 6mm long. These are soldered together. I soldered them with the aid of a gas stove and the result is shown in Photo 6. I finished the chassis with 1000-grit sandpaper and spray painted it, using temporary screws to prevent paint from entering the threads and covering the Earth points. Very few paints stick to polished or shiny brass well. I have been experimenting with paints for this Photo 5: I added insulation tubing to the pentode leads; initially, I used PVC but changed to Teflon later. Photo 6: having received the CNC-machined chassis pieces, I soldered them together with a gas stove. The areas that were masked with screw heads are either chassis grounding points or where I didn’t want paint to get into threaded holes. siliconchip.com.au Australia's electronics magazine application for many years. One excellent product is the clear Dupli-Color spray number DS-117. It helps not to have any pigments or fillers, such as aluminium powder. After coating the brass with this clear coat, I waited 24 hours and applied silver DS-110 spray paint. Once that had dried, I applied a final clear coat. This makes for scratch-­ resistant paint with a good finish and maximum surface adhesion (similar to automotive paint). You can see the result in Photo 6. December 2023  89 Fig.14 Other options that give superior adhesion and scratch resistance are powder coating or electroplating. However, those would have meant sending it away to a factory, which I was reluctant to do. Note that while I used Phillips-head screws to keep the holes clear of paint, the final transmitter has slot-head screws to match the original. Photos 7 & 8 show the completed transmitter with the final 16:3 output coil. Terminal strips The original unit appeared to contain two side-by-side terminal strips with five tags each, each mounted with two screws & nuts and a thinner underlying insulating plate. I decided to make this myself as one 3mm-thick black fibreglass plate with four mounting holes and a rear 1.6mm insulating plate, as shown in Photo 9. It might have been done that way originally. I made a custom connector strip for the unit’s rear wiring connections (also shown in Photo 9). I used a six-row strip rather than eight (as in the original) as the extras were not required, and this way, it would be less crowded. Oscillator & output tank coils I searched for ceramic coil formers for several weeks. I determined the diameter of the original ceramic coils and the approximate number of turns from the photos in the design document. The formers have slots for the winding wire. Most likely, the originals would have been a pre-made part intended for amateur radio projects in the USSR. Generally, the wire used on these sorts of formers is silver-plated copper. I acquired the closest oscillator coil form I could find from the UK. It required a machined base, which I made out of Bramite, to help match the original appearance – see Photo 10. I wound this coil using 0.9mm-­ Fig.15 90 Silicon Chip Australia's electronics magazine siliconchip.com.au diameter silver-plated copper wire. My first attempt was a 12-turn coil with a five-turn centre-tapped secondary. An additional 10pF parallel capacitance was required to bring it to the correct frequency. It is possible that the original trimmer capacitance had a higher centre value than the one I selected. However, the photos of the original suggested a 13-turn coil, which would have given the option of a six-turn or four-turn CT secondary. Experiments showed that a four-turn secondary provided inadequate voltage to get the output stage to full power, so six turns were required. 40-42V peak was needed at each of the two output valve grids to attain the full power output of 1W. The closest ceramic former I could find for the output tank coil, which closely matched the geometry of the original coil, was from Surplus Sales Nebraska. It was close to the right diameter with the correct number of grooves, so the turns/inch (or turns/ cm) was correct, but it was too long. To solve this problem, I bought a diamond cutting disc from eBay and fitted it to my bench circular saw and removed 7mm of ceramic material from each end (see Photo 11). I machined the end mounting pieces from Phenolic rod, similar to Tufnol, and fitted threaded, machined brass inserts into those for the retaining screws. Because Sputnik-1’s antennas were bent dipoles straddling a 0.58m diameter ball, the antenna feed impedance would have been higher than the 72W typical of a straight dipole, possibly as high as 150W. It would be possible Fig.18 siliconchip.com.au Fig.16 Fig.17 to find the exact value by making a mock-up from a metal sphere and some antenna rods. Also, the antenna rods were a little shorter than ¼ of a wavelength each. When this is the case, for the basic dipole at least, the antenna behaves as a resistor with a capacitor in series and represents a reactive load where the current leads the voltage. This may have helped to tune out the inductive reactance of the three-turn coupling coil on the 20.005MHz unit. From the original document images, I saw that the output coil had close to 15 turns. The centre tap supplying 130V to the coil being on the same side as the end connections suggested an even number of turns. I initially wound an experimental 15:3 coil and later moved to a 16:3 for the final output coil (Photo 12). Fig.19 Australia's electronics magazine December 2023  91 Fig.20 Fig.20: this is the last of the seven mechanical drawings for the chassis. To conveniently measure the output power into a 50W load, I made several coupling baluns that presented the transmitter output with a range of loads, with the results shown in Fig.21. The transmitter was tolerant of load resistances from 70W to around 240W, delivering at least 1W into that range of loads. Output power peaked at 1.32W with a load close to 138.8W, with the plate-to-plate load resistance for the 2P19B valve pair close to 4kW. The applied load resistance affects the exact tuning of the tank coil with the butterfly capacitor. If the output were peaked with a low-range load resistance (around 70W), it would tend to down-shift the graph of load resistance versus power output. If the tuning were peaked with a higher load resistance (around 300W), it would tend to up-shift the graph. Presumably, the D-200 transmitter modules were tuned for maximum power output when connected to the actual antennas in the Sputnik-1 spacecraft. Also, at full power, the plate voltage of the 2P19b with the 138.8W load fell lower than its screen voltage. The RMS voltage swing across the 16:3 output coil primary is 72V, while the peak voltage from plate to plate, across the coil primary, is close to 102V. Each plate sees half of this, so the plate dips to around 79V (51V below the 130V HT voltage), ie, 11V below the 90V screen. This is not a concern for most pentodes unless the plate voltage is much lower than the screen voltage; then, there can be excessive screen-grid current. I measured the screen-grid current under all output loading conditions, even when the plate voltage dipped to 23V below the screen voltage with the 312.5W load, and the screen current altered very little. Also, the output waveform remained normal. With lower load resistances than 138.8W, the plate voltage swing is less. With the 78.1W load, the plate voltage dips only 35V below the 130V HT and stays 5V above the screen voltage. Replica air-variable capacitors The transmitter contains two air-variable capacitors. To help match these as best possible, I machined a matching-looking nut for a Johnson-­ Viking butterfly capacitor (Photo 13) and attached it to a white Bramite plate, which resembles ceramic. I also machined a shroud around the original adjusting nut for the oscillator trimmer capacitor and painted that black to resemble the original parts. It was made from a vintage germanium transistor mounting clamp and ◀ Photos 7 & 8: the completed and operational replica 20.005MHz transmitter. Photo 9 (above): this is the tag strip I made (shown at the top). I wasn’t sure if the original had two parallel 5-terminal strips or a single arrangement like this. Regardless, it was easier to make it as a single unit. I then made the connector strip with six terminals (shown at the bottom) rather than the eight of the original, as only six were used. 92 Silicon Chip Australia's electronics magazine siliconchip.com.au a machined brass insert – see Photos 14 & 15. Replica common-mode choke Photo 17 shows the relative heights of the crystal socket and shield and the common-mode choke in the replica. Coils L5 & L6 were likely wound as a common-mode choke on the one ferrite core; the photos show a single ferrite slug. I think they made this choke tuneable to allow a small amount of fine adjustment of the exact frequency provided by the crystal. The idea behind the choke was to ensure the cathode (filament) of V1 (2P91B) had a very high impedance with respect to ground so the oscillator could work correctly. In a typical Colpitts-style crystal oscillator for medium wave frequencies up to 2MHz, the cathode (or filament, in this case) choke is typically chosen to be around 1mH, with an inductive reactance at that frequency of about 12.5kW. In the case of the 20MHz oscillator, a choke of 100μH or thereabouts is satisfactory, giving about the same reactance. One thing about making an RFC (radio frequency choke) is that it is vital to keep the self-capacitance low. The self-capacitance is in parallel with capacitor C27 (20pF). This means that the construction of the choke must either be a single-­ layer coil, or a wave-wound Photo 10: I was lucky to find this coil former in the UK as it’s very close to the original. I just had to add the base. Fig.21: the reproduction transmitter’s output power vs load resistance. It peaks around 138.8W; we don’t know the exact impedance of Sputnik-1’s antennas but expect they were in the 70-150W range. low-­capacitance coil, to keep the self-­ capacitance below a few picofarads. I could have used two 100µH axial chokes, but that would not make for a good-­looking replica. I therefore made a single-layer coil Photo 12: after some experimentation, this is the configuration I came up with for the output coil. It’s a 16:3 coil, with a 3/4-inch diameter, 3in length, 8 turns per inch using 1mm diameter silver-plated wire. (bifilar wound) with an inductance of 85μH and a self-capacitance of 3pF, determined by a self-resonance test – see Photo 16. I fitted C28 (a Soviet-­made 1200pF capacitor) inside the can, as shown in Photo 16. This arrangement is probably similar to the original part. The choke also provides some of the DC resistance required in the heater chain. Each valve has a 2.2V heater, accounting for 6.6V in total, while the battery supply is 7.5V. The DC resistance of each coil is 4W, and the filament current is close to 100mA. The value of R2, a resistor in series with the filament string, was not specified in the design document. The total voltage drop due to the choke is 0.8V, which would make the value of R2 close to 1W. However, it’s possible they ran the filament chain 15% ‘over voltage’ with fresh batteries. The 2P19B data sheet says the filament should be in the range of 1.8-2.5V, so that should be OK. Replica crystal Photo 11: this former was also almost perfect, but I had to cut the ends off to make it the right length, then machine some end pieces. siliconchip.com.au Photo 13: the shroud (made from the transistor mounting clip) was painted black & can be seen fitted in Photo 14. Australia's electronics magazine The crystal was an interesting challenge. The original crystal was in a 7-pin glass envelope, typical of many of the late 1950s era. While these December 2023  93 Photo 14 & 15: the shroud around the original adjusting nut for the oscillator trimmer capacitor made from a vintage germanium transistor mounting clamp. crystals are still sometimes available from Ukraine, I could not find one at 20.005MHz. A typical 1MHz crystal is shown in Photo 18. To make a replica, I cut the top off a 7-pin valve using a diamond file in the lathe and made a 7-pin base for it, initially only fitting three pins as a trial. The closest crystal I could find was 20.004864MHz. After I cut the glass valve, I heated the cut glass edge to red heat with a blowtorch. This helps to ensure that microscopic cracks in the cut edge don’t start to spread through the glass wall later. Also, to get the modern smaller crystal to operate properly in the circuit, I had to add 12pF of parallel capacitance. I hid that inside the base of the replica crystal – see Photo 19. RF output connectors The photo of the original unit shows what appear to be two round RF connectors. To help replicate them, I used F connectors. When the module was finished, it was time to combine it with the Manipulator. I had considered replicating the entire D-200 housing that contained the two transmitters and the Manipulator but decided against it. The main reason is that it is impossible to inspect one side when the transmitter module is mounted inside the D-200 casing. A better move would be to mount the transmitter module on a rectangular plate, visible on both sides, along with the Manipulator relays and the timing capacitors. This way, all the parts are readily seen. The achieve this, I had a natural anodised 3mm-thick aluminium plate CNC machined and engraved, then 94 Silicon Chip Photo 16: the common-mode L5/L6 choke and their shield can. The photo on the right is with the capacitor C28 added. filled with black paint. This plate mounts on top of an insulated base. The transmitter is fixed on one side of the engraved plate, and the plate is fitted to a Phenolic baseboard – see Photo 20. You can see videos of the replica operating, including reception on a shortwave radio, at https://youtu. be/9N26pkGGPew and https://youtu. be/_rq2yrdeGK8 Transmission test I also built a power supply for the replica of the Sputnik-1 Manipulator and its 20.005MHz radio transmitter module. In the absence of batteries, the standard method to power a valve radio or amplifier in a home or laboratory setting was from a line voltage power supply. These were called “battery eliminators”. Sputnik-1’s silver-zinc batteries (not available to the public at the time) were specially manufactured for the task. The high-tension battery was tapped at +10V, +21V, +90V and +130V. The 10V One does not simply transmit a 1W carrier at 20.005MHz because it might cause some interference. Instead, I fed the transmitter output into a dummy load to absorb the power but, by adding some small whip antennas, the leakage was enough that I could receive the signal on a shortwave radio in the next room. I assembled a 5:3 balun to attach to the transmitter and used a 50W dummy load to present the transmitter with the ideal 138.8W output load (see Photo 21). A battery eliminator Photo 17: the heights of the crystal socket in its shield and the common-mode choke shield can. 47mm 39mm 9mm Australia's electronics magazine 11mm siliconchip.com.au Photo 18: an original glass envelope crystal (right) and my reproduction 20.005MHz unit (left). Replica crystal ◀ Photo 19: the replica crystal and its matching shield can. Original format crystal supply was used for the suppressor grids in the two 2P19B output valves in the transmitter, for which the current draw is negligible. The 21V tap powered the Manipulator relays. The Sputnik design document referred to the common (negative) connection of the B battery and 7.5V filament battery as “-A”. I decided to stick to that on the front panel labelling of this battery eliminator. The version shown here is based on four 15W MEAN WELL RS-15 switchmode power supplies. These supplies are compact, their outputs are isolated and they have become quite inexpensive. They are also overload protected and are available with an output voltage of 3.3V, 5V, 12V, 24V or 48V. These voltages are adjustable to an extent using an onboard potentiometer; a very helpful feature. Since the output of each one is isolated, they can perform the same job as an adjustable battery. The battery eliminator circuit is shown in Fig.23. A large range of output voltages can be provided by selecting these supplies appropriately. The 12V unit has a higher output current, so that is what I used to power the valve filaments. Three 48V units in series provide the B+ voltages. Since the +10V and +21V supplies don’t need to deliver much current, I used zener diodes with a 1.2kW 2W current-limiting resistor to derive them from the output of the first 48V supply. When the replica transmitter unit was running with the Manipulator, loading the supply, I adjusted the 90V and 130V levels to be exactly correct at the supply’s output, aided by some built-in series resistors. The +7.5V, +10V and +21V supply outputs required no adjustments. The power supply module outputs are floating (aside from 2nF of capacitance to the unit’s housing), which to some extent makes them safer because a one-handed contact to the +90V or +130V rail won’t result in a significant current through the body to ground. It is still better to tie the outputs to ground electrostatically so they don’t float up to some unknown value. I did that using a 100kW anti-float resistor. That value limits the one-handed contact current from the +130V terminal to around 1mA, which is reasonably safe. I decided to use robust 5W-rated zener diodes, which require a modest current to get their terminal voltage to the labelled value. Photo 20: the completed, fully functional replica. The Manipulator and transmitter module can both be examined in detail. Photo 21 (upper left): this dummy load plugs into the transmitter’s output socket. A tiny amount of the signal makes it to the antennas and can be picked up by a nearby radio. siliconchip.com.au Australia's electronics magazine 95 Fig.22: the front panel drilling details and artwork for my battery eliminator. There is a 0.24W loss in the 1.2kW resistor and a 1.25W loss in the 18W 2W resistor. 1.875W are lost in the 5W-rated 7.5V zener, dropping to 1.125W under load. There is a combined loss of only 0.3W in the 10V & 11V zeners. This makes the total zener regulator losses in use close to a modest 3W. The shunt zener method is highly beneficial for another reason. The switching supplies have significant noise on their outputs, around 80mV peak-to-peak on measurement. This noise is sourced from a very low output resistance. For example, adding 100µF directly to the supply output terminals does little to reduce this noise. However, the series resistance and the low dynamic resistance of a shunt zener regulator create a voltage divider that flattens most of the noise out, even without significant filter capacitors added, especially for the +7.5V, +10V and +21V outputs. The 90V and 130V output required RC low-pass filters to get the switching ripple low and under 3mV peakto-peak. The finished unit is shown in Photo 22. Line power safety I built the battery eliminator into a very high-quality Takachi MS66-2123G extruded and cast aluminium enclosure that I got by mail from Japan. It has the internal chassis option and the tilt feet option. A switched and fused IEC connector on the rear panel avoids a cord dangling from the instrument when not in use – see Photo 22. It also means running mains power to a front-panel switch is unnecessary. The IEC connector contains a very short physical link between the live pin and the fuse; the link is easily protected with an added insulation sheet with slots punched for three pins. Fig.23: the circuit for my battery eliminator that powers the Manipulator and transmitter. It’s based on four MEAN WELL mains to DC switch-mode power supplies plus some zener diodes, power resistors and capacitors to help filter out the switch-mode noise. 96 Silicon Chip Australia's electronics magazine siliconchip.com.au Photo 22: the ‘battery eliminator’ mains supply is built into a very nice instrument case. All the voltages needed to run the Manipulator and transmitter are available at the front panel banana sockets. The IEC mains input socket, power switch and fuse are all in an integrated unit on the rear panel. Some constructors put silicone rubber over this metal link, but I don’t subscribe to that as it can fall off. Another option is an insulating boot, but they are somewhat bulky. Two Earth wires attach to the Earth pin of the IEC connector. One goes directly to the metal housing with a shakeproof internal star lug. The other Earth wire connects to all the Earths on the RS-15 switch-mode supply terminal strip, which are all also grounded to the case by their mounting screws. This double-Earthing makes the Earth wiring a lower resistance with a higher current carrying capability and more electrically robust than the single wire connections comprising the Active/Live and neutral wiring. I soldered the wires to flat circular lugs to suit the screws on the RS-15 units and applied heatshrink insulation. Putting stranded wire directly under the screw connections is a bad idea, as single strands can break. I retained the plastic covers over the RS-15 screw connections. This helps prevent finger contact with the mains terminals while probing inside the powered unit. The RS-15 supplies can be screwed directly to the metal surface of the internal chassis. However, I added an insulating black FR4 fibreglass sheet in the region of the connectors, as seen in Photo 23. The bodies of the units are still double-Earthed to the chassis by their pairs of fixing screws and their individual Earth wires. The front panel dimensions and panel artwork are shown in Fig.22. It was made as a transparent Sticker by Stickerman. The holes for the 4mm banana plug connectors (made by Hirschmann) are not round but have flats to prevent the connector from rotating when it is tightened. So I had to drill the holes to about 7mm, file the flats out to 7.4mm and then finish the holes on the opposite axis with a round file to create the shape. The 11 solder terminals are single 3mm screw-mount Teflon insulated types. One is a solid 10mm tall threaded hex Earth post for the 100kW anti-float resistor. A good aspect of the enclosure and sub-chassis system by Takachi is that you can assemble everything, including the sub-chassis, front and rear panel assembly, before you drop them SC into the main housing. Photo 23 (left): the four switch-mode power supplies just fit into the case with a small amount of space left for the resistors, capacitors and zener diodes. The circuit is simple enough that a PCB is not required. Photo 24 (right): the wiring on the underside of the baseplate (which is separate from Photo 23). Note the power zener across the 7.5V supply of the baseplate, this was added to protect the tube filaments from accidents. siliconchip.com.au Australia's electronics magazine December 2023  97