Fullscreen HTML5Med Res (??MB)HTML4

oscillator-driven mosfet vibrator replacement In this article, I present two more fully tested vibrator replacement designs, plus three additional circuits that readers may wish to experiment with. The first of the two designs is based on a pair of standard Mosfets and is the most efficient vibrator replacement I’ve made. It isn’t too complicated to build, either. Part 2: by Dr Hugo Holden L ast month, I presented a Mosfet-­ based vibrator replacement for older communications receivers and some vintage radios. While it works very well, it has a couple of drawbacks. One is the relatively large and obsolete TO-3 package Mosfets. The other is that it’s only about as efficient as the mechanical vibrator it replaces. This somewhat more complicated design also uses Mosfets, this time readily-available, low-cost types specified in TO-220 packages, so it’s a bit more compact. It also adds four small-signal Mosfets to form an oscillator to drive those power Mosfets. That makes it quite a bit more efficient and able to deliver a higher HT; I measured 72.7% efficiency at 289V DC output compared to 67% at 276V DC output for the self-oscillating Mosfet version and 66.6% efficiency at 267V DC output for the original mechanical vibrator. Most parts are available from local suppliers like Jaycar, Altronics, RS or element14. The brass plate and wire are available from Mr Toys in Australia, while the UX7 base is a standard American Amphenol part that can usually be found on eBay. Its circuit is shown in Fig.1. A multivibrator is formed by two BS270 Mosfets, Q3 & Q4. This zero-bias configuration gives more reliable starting from low voltages than biasing these Mosfets to an on condition, which would be analogous to the usual bipolar transistor multivibrator circuit. Due to the high impedance at the Mosfet gates, high-value gate resistors and low-value timing capacitors can be used (270kW & 10nF). This results in accurate timing and avoids the use of poor-tolerance electrolytic capacitors, as would typically be required for a low-frequency BJT-based multivibrator. Diodes D1 and D2 clamp the gate drive signals to -0.7V. The multivibrator runs close to 110Hz, similar to a V6295 vibrator that nominally operates at 100Hz. If anything stops the multivibrator, or it doesn’t start due to a very slowly rising supply voltage, the drain potentials of Q3 and Q4 would be high. That would be a problem if they drove the output Mosfets directly because both Mosfets would be on continuously, shorting out the transformer primaries. Therefore, an inverting buffer stage is included, made from identical Mosfets Q2 and Q5. These also help to isolate the multivibrator from the output stage. The 12V DC supply to the multivibrator is also heavily filtered with a 150W resistor and 15μF capacitor. These ensure that the significant voltage transients from pin 4 do not cause premature triggering of the multivibrator when it is in a vulnerable condition, about to change state. I used four BS270s rather than a CMOS IC here as they have much higher voltage ratings (60V) and are much more immune to damage from spikes and transients. They do not require as much protection on the power supply feed as a CMOS IC. This circuit will start from voltages as low as 6V. Mosfet switching times Fig.1: this vibrator replacement uses an oscillator built around signal Mosfets Q3 & Q4. They drive the gates of power Mosfets Q1 & Q6 via inverter stages Q2 and Q5, which prevent overheating in case the oscillator stops or can’t start. This is the most efficient of my vibrator replacement designs. 78 Silicon Chip Australia's electronics magazine It is standard practice in switchmode power supply design to drive the gates of the output Mosfets from a low impedance source, typically siliconchip.com.au Photo 1: the main physical structure of the Mosfet-based, oscillatordriven vibrator replacement is made from a 7-pin base, and a rectangle of 0.8mm-thick brass with a 15mm tapped metal spacer soldered to it. from 10W to 100W, for fast switching. The power Mosfet gates often have a significant capacitance of around 500-5000pF, depending on the Mosfet type. Suppose the gate series resistance is too high. In that case, it can slow the switching time down and decrease the efficiency (increasing the Mosfet heating) because it spends more time in an intermediate conduction state rather than on or off. The switching frequency is often in the range of 20-100kHz in switch-mode PSUs, so there are many switching events per unit of time, and these losses add up. In addition, switch-mode power supply transformers are generally wound with a low leakage inductance, often with bifilar wound primary windings. However, the ZC1 power transformer is not like this; it has a relatively high leakage inductance between the halves of the primary windings. It also operates at a much lower switching frequency than a modern SMPS. Therefore, the design rules for this application are different. Very rapid switching of the output Mosfets is disadvantageous because the transformer’s primary winding leakage inductance (and leakage reactance) is so high that this produces very high voltage transients on the contralateral Second diode down in hole Photo 2: the two series pairs of BY448 diodes are soldered directly to the base pins. siliconchip.com.au or fellow Mosfet’s drain at the moment one Mosfet switches on. These spikes are on the order of 70-100V with a resonant frequency of about 50kHz. This is ameliorated a little by the 1.5kW gate drive resistor network, which forms a mild LPF (low-pass filter) with the gate capacitances of the IRF540Ns. Also, the added 470nF ‘tuning capacitor’ lowers the resonant frequency of the leakage inductance-­ capacitance network to about 20kHz, and reduces the voltage transients on the Mosfet drains to an acceptable level when switching occurs. Fig.2: assembly on the doublesided PCB is straightforward, as shown here. The TO-220 package Mosfets are first attached to the brass plate, then the PCB mounts on the brass plate with the Mosfet leads bent up to meet their pads on the PCB. The three nuts in a triangle pattern are for spacers that attach the PCB to the brass plate and provide ground connections. Construction Start by populating the PCB sans the power Mosfets, Q1 and Q6. The PCB is coded 18106231 and measures 33 × 45.5.5mm, with the components mounting on it as shown in Fig.2. Fit all the resistors, using a DMM to check their values, then mount the diodes orientated as shown. Follow with the capacitors; only the tantalum type is polarised and should have a + marked on its body. Crank the leads of the four identical TO-92 package Mosfets out using small pliers, then solder them in place, as shown in Fig.2. A vibrator replacement requires a chassis or skeleton to support it, and preferably a metal heatsink for the output devices. The simplest way to do this is to start with a standard Amphenol UX7 plug and fit it with a structure composed of a brass spacer, brass plate and a ground wire from pin 7 of the UX7 socket. The basic parts are shown in Photo 1, and the ground wire details are in Fig.4. To ensure the 3mm diameter hole in the plug is drilled on-centre, a temporary 3mm spacer can be placed in Australia's electronics magazine Figs.3 & 4: details of the brass plate. Note how the tapped spacer is notched to slide onto the brass plate’s end so it can be soldered in place. The way to bend the 2mm-thick brass wire is shown adjacent to the brass plate, with the ground wire soldered to the plate (also see the photos). July 2023  79 Photo 3: the brass sheet has now been attached to the base via the spacer and the 2mm-thick ground wire has been soldered to it. Fibre washers around the ground wire help support the insulator. Photo 4: next, the Mosfets are mounted to the brass sheet with insulators in between, and the leads are bent up, ready for the PCB. Three wires from the base have also been bent and insulated to meet their PCB pads. Photo 5: with the PCB assembled and installed, the unit is now ready for operation. Note that some slight component placement differences exist between this prototype and the final PCB. the ¼in recess to guide the drill. The hole is then countersunk from the pin side of the plug. Next, solder the four BY448 rectifiers into the plug assembly, as shown in Photo 2. The brass plate can have its holes drilled before or after fitting to the spacer, but it might be easier to do it first because the plate sits flat. The required hole positions are shown in Fig.3. Cut a 2-2.5mm deep slot in the 15mm-long M3 nickel-plated hex brass spacer to accommodate the brass plate. To do this, I used a junior saw and a fine flat file. Make the plate a push-fit into the spacer, then solder them together by holding the assembly with grips over the flame of a gas stove or with a suitably powerful soldering iron. The spacer’s end needs to be rounded off a little to fit into the deep hole in the UX7 plug. You can temporarily fit the brass plate and spacer to the plug to align it correctly, with a brass wire positioned to pass from pin 7 (Earth) of the plug to the brass plate, as shown in Photo 3. Once it’s aligned, solder the brass wire to the plate. The thick (2mm diameter) brass wire ensures that the plate cannot rotate easily even if its fixing screw becomes loose. I put masking tape on the plate where the power Mosfets and PCB spacers will go to allow a good connection, then sprayed it with lacquer to prevent future oxidation. Once the lacquer is dry, you can assemble the hardware ready to receive the PCB, as shown in Photo 4. Make a 25mm washer from insulating material like Presspahn or similar to cover the rectifier connections. The other wires can be made from 0.7mm diameter tinned copper, covered in silicone rubber or PVC insulation, or small diameter heatshrink tubing. Add the ‘tuning’ capacitor, C1, between the drain connections of the IRF540N power Mosfets. Photo 5 shows the PCB fitted over the Mosfet leads and the output wires soldered to it. This prototype board differs slightly in layout from the final version shown in Fig.2 but has the same circuit. Photo 6 shows how the Fig.5: due to the design of the transformer the vibrator drives creates a leakage inductance (XL) in series with the currently undriven primary, which resonates with Ct, generating voltage spikes at the transitions. Resistance R of the transformer windings slightly dampens the ringing. 80 Silicon Chip Australia's electronics magazine siliconchip.com.au Photo 6 (left): three short spacers between the brass plate and PCB hold them together and make the ground connections. The tuning capacitor connects across the insulated Mosfet mounting screws that connect to the Mosfet drains. Photo 7 (right): another view of the completed vibrator replacement unit. tuning capacitor mounts between the Mosfet tabs and how one of the output wires, plus the 12V supply wire, pass through holes in the brass plate. The PCB mounts onto the brass plate using three 5mm-long M2-tapped metal spacers. These also make the GND connections between the PCB tracks and the brass plate. Photo 7 shows the finished assembly, while Photo 8 depicts it being tested in the ZC1 communications receiver via an extension socket. It is a good performer, and there is no significant RFI, unlike the original mechanical vibrator: A metal can is not required as there are no contacts to protect, but if you want to hide the electronics, you could use just about any metal tube with an inner diameter of at least 34mm. It’s safe for the can to rest on the brass plate as it’s at ground potential. the transformer windings or tuning capacitor(s). Fig.5 shows the centre-tapped primary of a transformer driven from only one side, as it would be half the time in a push-pull scenario. In this case, the two halves of the primary are labelled primary (P) and secondary (S); even though they are both part of the primary in actual use, one acts as a secondary in this particular example. XL is the transformer’s leakage reactance, an inductance acting in series with the windings, which Photo 8: the vibrator replacement undergoing testing in a ZC1 Mk2 communications receiver. It’s plugged in via an extension that allows the connections to be probed during operation. Leakage reactance It is worth looking at the leakage reactance problem and why the vibrator transformer primaries have a tendency for voltage overshoot. If these overshoots (oscillations) are too large, they can exceed the drain-source voltage of the Mosfet (or collector-­ emitter rating if a bipolar transistor is being used) and are a potential source for insulation breakdown of siliconchip.com.au arises because not all of the magnetic field links both windings P & S. The leakage reactance appears in series with the primary, or the secondary winding when the other winding is shorted out or has a fixed voltage applied to it. The tuning capacitor, Ct, is the inter-winding capacitance plus any externally added capacitance. The resistance (R) is mainly that of the ohmic losses of the windings. Initially, no current flows. When switch S1 (which could be a transistor) closes, 12V DC is applied to the primary winding P, effectively shorting it out from the AC perspective (until the core saturates). The leakage reactance XL appears in series with the secondary winding S and induces a voltage that attempts to raise V2 to 24V, as one side of secondary winding S is connected to +12V. To achieve this, Ct must be charged; it forms a resonant circuit with the leakage reactance XL, with some damping by R. Therefore, oscillations (spikes or ringing) occur on terminal V2. The frequency of this resonance is primarily determined by the leakage inductance XL and the tuning capacitance Ct. Resistance R also plays a part in the frequency, as the damping is pretty heavy, but there can often be four or five cycles of oscillation or ringing before they dampen out. This is why increasing the tuning capacitance lowers both the frequency and the amplitude of these oscillations or ringing. To look at it another way, the Q of this resonant circuit comprising R, XL Australia's electronics magazine July 2023  81 Fig.6: a similar arrangement to Fig.5 but showing both halves of the push-pull configuration, which results in a burst of oscillation each time one of the Mosfets switches on. and Ct is lowered with a larger tuning capacitor because the resonant frequency is shifted down, and the inductive reactance of XL is lower at that lower resonant frequency. In the case of the push-pull rather than the single-ended example above, the same situation occurs, as shown in Fig.6; the resistance is omitted for clarity. When Mosfet Q1 switches on (red drive waveform high), voltage V1 goes rapidly to zero in a few microseconds or less. XL1 vanishes when Q1 is conducting as a fixed voltage +V is applied to winding P1, and all the leakage reactance then appears as XL2 in Q2’s drain circuit. Q2 is also off at this time. Ct is in resonance with XL2, so the leading edge of the voltage V2 has ringing and overshoot. The situation is reversed when Q2 conducts, making XL2 vanish and placing all the leakage reactance XL1 into Q1’s drain circuit. The peak amplitude is around twice the supply voltage, which holds true until the magnetic core of the power transformer starts to saturate. For the ZC1 radio transformer, this takes about 8ms, so a 100Hz drive waveform does not take it near core saturation. However, in a future issue I will present a different vibrator replacement using bipolar transistors that relies on core saturation to sustain oscillation. Scope 1 shows a ZC1 Mk2 radio’s primary winding voltages with the vibrator replacement unit presented here. The oscillations are visible on the drain connections (transformer primary) immediately after one Mosfet comes out of conduction and the fellow Mosfet goes into conduction. They switch quickly, over less than a few microseconds, even with the 1.5kW gate resistors. Scope 2 gives a closer look at the oscillations. With the 470nF tuning capacitor, the ringing frequency is about 20kHz: Without the added tuning capacitor, as shown in Scope 3, the ringing frequency is about 50kHz, and the peaks are much higher. Other smaller oscillations are superimposed due to the transformer’s high-voltage secondary windings, their leakage reactance and associated capacitance. The initial peak is very high at around 70V, and on its negative half-cycle, causes the Mosfet’s internal drain-source diode to conduct, clamping the negative half-cycle. Scope 4 shows the timing of this transient, which occurs just after the Mosfet switches on and its fellow turns off. Therefore, that 470nF tuning capacitor is important with this Mosfet version, or any version using silicon transistors driven by an independent oscillator (like commercial transistorised units). With the mechanical vibrator, this first peak is lower at around 30-40V. That’s because, with the reduced duty cycle, the transformer’s primary voltage falls from 24V to about 16V before the next switchover as the energy transfer to the circuit comprising Ct and XL is a little lower. Another potential method to solve the leakage reactance/voltage spike issue is to snub off the high-voltage transients with a TVS (transient voltage suppressor) to around 30V. However, there is a little more chance of RFI with this method versus tuning the Scope 1: the drain voltages of the Mosfets during operation. Scope 2: a close-up of the drain voltage of the Mosfets with a They switch pretty fast and the oscillation and ringing due 470nF tuning capacitor at transition, showing the oscillation to the transformer’s leakage reactance is well damped. and ringing at about 20kHz. 82 Silicon Chip Australia's electronics magazine siliconchip.com.au resonant frequency downwards with the added tuning capacitor. A bidirectional 30-40V TVS between the output Mosfet drains would work. The makers of commercial vibrator replacements with electronic driver circuits do not seem to consider the leakage inductance of the primary of the vibrator transformers. The tuning capacitors they specify do not suit an electronic driver with an independent oscillator; more capacitance is needed, or the voltage transients threaten the output devices and the transformer insulation. A safe design One thing that bothered me about the commercial designs, which have gates and logic or other ICs as oscillators, is what would happen if that clock stopped or did not start. This can occur if the power supply ramps up too slowly and is common with circuits that use logic gates. It leaves one transistor switched on and the other off, applying full voltage to one half of the primary and that will blow the fuse, if there is one, or overheat the device or the transformer. With this design, the output devices remain off if the multivibrator stops and/or doesn’t start, thanks to the two extra BS270 signal Mosfets. Darlington-based alternative Another vibrator replacement I came up with is based on Darlington transistors, and this one is simple enough that it doesn’t need a PCB, although the metalwork is a bit more complex. Fig.7: a simple self-oscillating, Darlington-based vibrator replacement. There are more efficient arrangements than this but it is simple and reliable. Darlingtons have a low input threshold voltage of around 1.4V, so the circuit will start (oscillate) from low power supply voltages. The circuit described here operates with a supply voltage as low as 3V. Darlington power transistors also have the advantages of internal base resistors and collector-­ emitter diodes, saving on parts. Frequency limiting and stable self-switching can be obtained with 47nF Miller integrator capacitors between the collector and base of each Darlington transistor. Without this negative feedback, the oscillator circuit is highly unstable and oscillates at a high frequency corresponding to the power transformer’s primary leakage reactance and associated capacitances in resonance. If this persists, the transistors can overheat and be destroyed. 20V/cm 10V/cm Scope 3: the overshoot is much faster and reaches higher voltages without the tuning capacitor. This could cause insulation breakdown or damage to the Mosfets. siliconchip.com.au Using Darlington transistors as switches results in a base drive power about 10-20 times lower than BJTs (bipolar junction transistors). The positive feedback capacitors to sustain oscillation from the collector of one transistor to the base of the fellow transistor can be a modest value of 4.7μF, meaning non-electrolytic types can be used. Electrolytic capacitors are best avoided where the values are responsible for setting time constants, due to their lax tolerances. The circuit, shown in Fig.7, is based on MJ3001 or MJ11016 NPN Darlington transistors, oscillating at close to 62Hz. Scope 5 shows the resulting transformer drive waveform (at one end of the primary). The collector-emitter Scope 4: the two Mosfet drain voltages at a short timebase (without tuning capacitor) shows a large spike at the switched-off Mosfet’s drain, after the other Mosfet turns on. Australia's electronics magazine July 2023  83 Scope 5: the Darlington collector waveforms are clean square waves with rounded edges due to the Miller capacitor slowing switch-on/switch-off. There’s little sign of ringing. saturation voltage drop of the Darlingtons in this application with a peak collector current of 2A is about 0.9V. Therefore, this Darlington unit results in an output power about 6% lower than the Mosfet version. However, the output voltage and efficiency are very similar to the original electromechanical V6295. The advantage is that the Darlington unit is relatively simple for the home constructor to manufacture. Scope 6 shows a close-up of the collector waveform for the Darlington unit. This shows only a small resonance during the switching event, with no significant collector voltage overshoot, due to the 47nF Miller capacitors and the switching frequency of just 60Hz. A 470nF tuning capacitor is not required here. Construction Prepare four brass plates, two of each type shown in Fig.8. When working with 0.8mm-thick brass plate, it is best to mark and drill 1mm pilot holes Scope 6: the Darlington collector voltage during switch-off with a short timebase. A tiny bit of oscillation is visible here, but nothing to worry about. first, then drill the holes out one size step at a time to get to the final size. 0.8mm (0.032in) thick brass plate is made by K&S Engineering and is stocked in Australia by companies selling models, such as Mr Toys. The results are shown in Photo 9. The machined brass base and top are shown in Photo 10. I had them turned by a local machine shop, then added the 7mm-deep threaded holes myself. The reason for the groove in the base is that my ZC1 Mk2 communications receiver has clips around the base of the vibrator to retain it, and they fit into this groove (see Photo 11). If your application is different, you may need to change the details of the groove, or eliminate it and simplify the machining if your device lacks such clips. When tapping into blind holes, use a tapered tap first and lubricate with WD40 (or the recommended lubricant for your metal) during the process. Then wash all the swarf out of the hole with a jet of contact cleaner from the applicator tube. After that, you can tap to the base of the holes with a bottom tap to ensure the thread runs to each hole’s base. Then wash out the swarf again with contact cleaner. It is critical to be patient and careful when marking, centring and drilling the holes, which are all 9mm from the edges of the square section, as per Fig.9. The Amphenol 7-pin plug base is prepared with the BY448 rectifiers, just like the Mosfet version described earlier. Only three wires (the two collector wires and ground) are required as the +12V connection (pin 4) is not used – see Photo 13. Glue this plug arrangement into the brass base. This is best done as a twostep procedure; use a small amount of 24-hour epoxy to attach it and align it on the correct axis when the unit is plugged in. Once cured, add more epoxy to the well created by the edges of the plug and the inside of the brass housing. There’s no risk of it draining out before it sets because the first bond has sealed it – see Photos 12 & 13. Photo 9: these four brass plates form the four larger sides of the housing. Two have holes drilled for the TO-3 mounting screws & leads. Photo 10 (right): I drilled and tapped 7mm-deep holes with 4-40 UNC threads in the lid and base to attach the sheets shown in Photo 9. A local machine shop made these pieces as I don’t have the required tools. 84 Silicon Chip Australia's electronics magazine siliconchip.com.au Photo 11: the groove in the base is designed to engage these retention tabs in the ZC1 Mk2 transceiver. Photo 14 is of the Augat TO-3 transistor sockets I used, usually available on eBay, plus an insulated standoff (it is a bit like a single-point tag strip). Both create convenient tie points for components, obviate the need for insulators, nuts/washers & lugs for the collector terminals, and the transistors are easily removed for testing or replacement. You also don’t have to solder to the transistor leads. These single insulated mounting posts are becoming rare. Surplus Sales of Nebraska still stock a range of mounting posts like this. Another option is a phenolic tag strip with a single lug. If TO-3 sockets are not used, and the transistors are instead mounted with the usual insulator set, reduce the 5.5mm holes in the brass plates to 4mm in diameter. Photo 15 shows the device partially assembled, with the capacitors and diodes mounted to the socket and post. Both sides are identical. Each transistor base has two capacitors and one diode connected to it. No resistors are connected to the bases because the base resistor is internal to the Darlington transistor. I scribed marks for the holes on the inside surfaces of the brass plates so they would not be visible from the outside of the assembled unit. Note that I sprayed the brass plates with DS117 clear automotive lacquer to prevent oxidation. Mount the transistors with the usual mica insulating washer, with thermal paste on both sides. Clear silicone siliconchip.com.au Fig.8: drilling details for the four brass plates (two of each) that make up the sides of the rectangular Darlington-based vibrator replacement. Fig.9: details of the machined base and top of the rectangular case. The groove in the round base is for the retaining clips in the radio to engage; not all radios with vibrators will have this feature. Second Bond First Bond Photo 12: start assembling the base by gluing the plug into the machined brass piece, sealing all around the perimeter with 24-hour epoxy. Australia's electronics magazine Photo 13: once the first lot of epoxy has set, you can add more around the perimeter at the top edge of the plug to make it really solid. July 2023  85 grease is less messy than the white compound, and the extra is easily wiped away. In this instance, each transistor’s dissipation is only 1-1.5W, so they only run warm; still, it is better to have some thermal coupling to the brass plate. Screw the transistors down with 12mm or ½in 6-32 UNC screws that fit the threads in the Augat sockets. Each screw has a split-spring lock washer under its head. Photo 16 shows the transistors installed, while Photo 17 shows the internals assembled. The 560W resistors pass from one side to the other, connecting the mounting post connection to the collector terminal lug on the opposite transistor. The screws used to attach the brass panels to the top and base are stainless steel 4-40 UNC, ¼in long with a Binder style head, similar but slightly different to a pan head. These are available from PSME (Precision Scale Model Engineering in the USA). Performance The Darlington version is almost a dead-ringer in performance to the electromechanical unit, but of course, with no reliability or wear problems. The output voltage is a little lower than the other electronic units due to the collector-emitter voltage drops of about 0.9V for the Darlingtons. The similarly low output voltage of the mechanical unit is due to the reduced duty cycle compared to the electronic units. So the two devices have about the same performance parameters for different reasons. Logic IC based vibrators Fig.10: a vibrator replacement circuit based on a pair of Mosfets & SN7400 quad NAND gate IC. Note the zener diode to protect the IC from voltage spikes, and the use of logic-level Mosfets, as their gates are only driven to 5V. AUGAT TO-3 SOCKET STANDOFF POST Photo 14: I mounted the TO-3 transistors via sockets to make construction and servicing easier. The insulated standoff post mounted near the socket also makes the wiring easier. Photo 15: the two identical halves of the circuit are fully assembled and ready to be merged. 86 Silicon Chip Australia's electronics magazine In reference to the Mosfet vibrator replacement described above, I mentioned in passing devices that use logic ICs for the oscillator. Figs.10-12 are circuits of unusual variants you will not see elsewhere. Fig.10 shows a 6V-powered unit I designed using a TTL logic gate. I built some of these using a beam lead style 7474, mil-spec 5474, or the 5400 NAND gate in ceramic packages, like those used in the Apollo 11 computers. These are incredibly robust parts, able to survive re-entry into the atmosphere in a satellite and still function! They are the most robust ICs ever created. The circuit of Fig.11 is an oddball arrangement that enables one flip-flop to be used as an oscillator and the other as a 2:1 frequency divider (both in the same IC) to give a spectacularly perfect square wave. If the wave duty cycle is not exactly 50%, the current consumption increases, and the efficiency drops as the transformer core develops a net flux. One of the problems I had with commercial electronic vibrator substitutes was that they used somewhat fragile CMOS ICs with an imperfect duty cycle. On top of that, the designers didn’t understand that in the case of replacing the secondary contacts of the synchronous vibrator, you need an extremely high PIV rated diode. And they ignored the requirement for additional tuning capacitance as well. Fig.12 is a 12V-powered design that uses a 7400 (or 5400) logic IC. The zener diode protects the IC from voltage transients on the +12V rail. If a reversed polarity is applied, the siliconchip.com.au Fig.11: another vibrator replacement circuit, this time based on two NPN Darlingtons and a dual flip-flop IC. The first flip-flop is the oscillator, while the second halves the frequency for perfect waveform symmetry. Fig.12: a similar circuit to Fig.10, only using Darlingtons instead of logic-level Mosfets, and with values changed to suit a 12V battery supply. All of these circuits (Figs.10-12) also need the diodes shown at right. Photo 16 (left): an outside view of the two halves showing how the TO-3 transistors are retained. Photo 17 (right): once the two halves are attached to the base, the wiring can be finalised by adding the two resistors that go from one side to the other, plus the two collector (blue) and two ground connections (black sheathed wire). siliconchip.com.au Australia's electronics magazine July 2023  87 Parts List – Vibrator Replacements Mosfet version 1 double-sided PCB coded 18106231, 33 × 45.5mm 1 Amphenol 7-pin base [www.ebay.com.au/itm/115461595962] 1 brass plate, 65 × 34 × 0.8mm (0.032in) 1 50mm length of 2mm diameter brass wire 1 200mm length of 0.7mm diameter tinned copper wire 1 200mm length of 1.5mm diameter heatshrink or spaghetti tubing 2 TO-220 transistor insulating kits (washers + bushes) 2 M3 × 6mm panhead machine screws and nuts 3 M2 × 12mm panhead machine screws and nuts 1 25mm disc of insulating material (phenolic, FR-4, Presspahn etc) 3 metal spacers (4mm diameter, 5mm tall) with matching screws and nuts 2 3mm solder lugs hardware etc (available from K & S Engineering USA) Photo 18: the rectangular prism brass case of the Darlington vibrator replacement forms the structure and provides heatsinking for the TO-3 metal can encapsulated transistors. Semiconductors 2 IRF540N 100V 33A N-channel Mosfets, TO-220 (Q1, Q6) 4 BS270 60V 400mA N-channel Mosfets, TO-92 (Q2-Q5) 2 1N4148 75V 200mA diodes, DO-35 (D1, D2) 4 BY448 1.5kV 2A axial diodes (D3-D6) Capacitors 1 15μF 35V tantalum 1 470nF 250V polyester or polypropylene axial 2 10nF 100V MKT polyester or greencap Resistors (all ¼W or ⅛W 1% axial) 2 270kW 2 10kW 4 1.5kW 1 150W 2 100W Photo 19: with the lid and four sides held together and to the base by screws, the vibrator replacement is ready for testing and use! Darlington version 1 Amphenol 7-pin base [www.ebay.com.au/itm/115461595962] 2 Augat or similar TO-3 sockets [www.ebay.com.au/itm/144066503423] 2 TO-3 mica insulating washers 4 brass plates, 68 × 42 × 0.8mm (0.032in) each (see Fig.8) 1 machined brass base, 40 × 40 × 14mm (see Fig.9) 1 machined brass lid, 40 × 40 × 7mm (see Fig.9) 1 200mm length of 0.7mm diameter tinned copper wire 1 200mm length of 1.5mm diameter heatshrink or spaghetti tubing 1 6mm or ¼in long stainless steel 4-40 UNC screws, panhead or Binder-style [PSME] 4 12mm or ½in long 6-32 UNC panhead machine screws 2 6-32 UNC split spring washers 2 insulated standoff posts with matching panhead machine screws Semiconductors & passives 2 MJ11016G 120V 30A NPN Darlington transistors, TO-3 (Q1, Q2) [RS Cat 463-000] OR 2 MJ3001 80V 10A NPN Darlington transistors, TO-3 (Q1, Q2) [www.ebay.com.au/itm/303226250083] 2 1N4004 400V 1A diodes (D1, D2) 4 BY448 1.5kV 2A axial diodes (D3-D6) 2 47nF 400V axial plastic film capacitors 2 4.7μF 63V axial plastic film capacitors 2 560W 1W axial resistors 88 Silicon Chip Australia's electronics magazine zener conducts in the forward direction, protecting the IC. In that case, the collector-­emitter diodes intrinsic to the Darlington transistors conduct, blowing the fuse (hopefully, there is one). I’m not presenting construction details for any of these because I believe the three discrete designs I’ve published so far (with one more to come) are more robust and generally better. Coming up I have built one more vibrator replacement design that is quite a bit more difficult than any of the versions described so far. It is based on two bipolar transistors, a custom transformer, and a few passive components. It is a design that could have appeared in the early days of transistors, when they were expensive, as it uses them sparingly. Despite this, it works just as well as the Darlington-­ based version described in this article, with similar efficiency and delivering a similar output voltage. You can expect to see that article within the next few months. SC siliconchip.com.au