Fullscreen HTML5Med Res (??MB)HTML4

HIGH POWER – VARIABLE LIN W e’ve been promising this project for a while! However, it has taken some time to get it just right. But it’s finally here. This power supply can deliver up to 45V at up to 8A, or up to 50V at lower currents. It has a fully adjustable output voltage down to 0V and an adjustable current limit. Its operating envelope is shown in Fig.1. That makes it suitable for many different tasks, including testing newly built or repaired equipment, temporar22 Silicon Chip ily running various devices, charging batteries etc. Its controls are simple. Two knobs set the voltage and current limits, and the power supply maintains its outputs within these constraints. It shows the actual output voltage, set voltage, actual current, set current and heatsink temperature on an LCD screen. These can be shown on an alphanumeric LCD, or if you prefer, you could use separate LED or LCD panel meters. It has a pair of internal high-speed Australia’s electronics magazine fans to keep it cool. These automatically spin up and down as required. If the Supply is operated in the orange shaded area shown in Fig.1, or at very high ambient temperatures, or the fans fail, a thermal current limit comes into play. This reduces the output current until the unit cools down, preventing damage to the Supply. While we originally planned for this power supply to be able to deliver 50V at 8A, it is difficult to achieve that with a practically sized transformer and a reasonable parts budget. It’s limited to siliconchip.com.au – 45V/8A NEAR SUPPLY Part 1 – by Tim Blythman This adjustable bench supply can deliver heaps of power, up to 360W in total, making it ideal for the test bench or just general purpose use. It can operate as a voltage or current source at 0-45V and 0-8A. It is an entirely linear, analog design. It’s fan-cooled with automatic fan speed control, short circuit/overload protection and thermal self-protection. It can even be used as a basic but powerful battery charger. Features Featu res & specifications • Up to 45V output at 8A, 50V output at 2A (see Fig.1) • Low ripple and noise • Adjustable output voltage, 0-50V • Adjustable output current, 0-8A • Constant voltage/constant current (automatic switching) • Shows set voltage/current, actual voltage/current and heatsink temperature • Fan cooling with automatic fan speed control • Thermal shutdown • Fits into a readily available vented metal instrument case • Switched and fused IEC mains input socket • Uses mostly commonly available through-hole components 45V at 8A because despite using a large 500VA transformer, its output voltage still sags significantly under load, meaning there isn’t enough headroom for regulation. However, if the transformer was upgraded (and possibly the filter capacitors too), it could be capable of delivering 50V at 8A. Design overview The basic design of the Bench Supply is shown in the simplified circuit diagram, Fig.2. It’s based around an siliconchip.com.au INTERMITTENT OPERATION (THERMAL LIMITING) Fig.1: the Bench Supply can deliver 8A but can only do so continuously with an output voltage of between 16V and 45V. Below 16V, internal dissipation is so high that the unit will go into thermal limiting after a few minutes. Above 45V, transformer regulation means that the DC supply voltage drops far enough that 100Hz ripple starts appearing at the output, so the actual voltage may be lower than the set voltage. Australia’s electronics magazine 8A LIMIT DUE TO TRANSFORMER VOLTAGE SAG & DC RIPPLE LIMITED BY DESIGN 7A 6A 5A CONTINUOUS OPERATION 4A 3A 2A 1A 10V 16V 20V 30V 40V 45V 50V October 2019  23 more current to the output of the supply. These transistors therefore supply virtually all of the maximum 8A output current Regulator control Here’s a teaser look inside our new Bench Supply, taken before we applied the dress panel. Full construction details will begin next month. As you might expect from its specifications, there’s a lot to this supply, dominated by the 500VA transformer at left. But the good news is that it uses mostly through-hole components so construction isn’t too difficult. LM317HV high-voltage adjustable regulator, REG3. The LM317HV variant can handle up to 60V between its input and output, at up to 1.5A. Clearly then, this regulator cannot pass the full 8A output current. And even if it could, it couldn’t dissipate the 400W that would be required (50V x 8A) as it’s in a TO-220 package. Therefore, the regulator itself only handles about 10mA of the load current, with the rest being delivered by four high-power current boosting transistors, Q4-Q7. Power is fed into the supply via the IEC input socket shown at upper left, and passes through the mains switch and fuse before reaching the primary 24 Silicon Chip of transformer T1. Its two 40V AC secondary windings are connected in parallel and then on to bridge rectifier BR1 and a filter capacitor bank, generating the nominally 57V DC main supply rail. This passes to the input of REG3 via a resistor, and also to the collectors of the NPN current-boosting transistors and the emitter of PNP control transistor Q3. As the current supplied by REG3 rises, Q3’s base-emitter junction becomes forward-biased, and it supplies current to the bases of Q4-Q7, switching them on. As REG3 draws more current, they switch on harder, providing more and Australia’s electronics magazine Like most adjustable regulators, REG3 operates by attempting to maintain a fixed voltage between its output (OUT) and adjust (ADJ) pins. In this case, around 1.2V. Usually, a resistor is connected between OUT and ADJ, and another resistor between ADJ and GND, forming a divider. As the same current flows through both resistors, the voltage between ADJ and GND is fixed, the regulator output voltage is that voltage plus the 1.2V between OUT and ADJ. But in this case, rather than having a fixed or variable resistor from ADJ to GND, we have transistors Q1 and Q2, connected in parallel. Their bases are driven from the outputs of op amps IC1a & IC1b. Their emitters go to -5V so that the ADJ pin can be pulled below ground, allowing the regulator OUT pin to reach 0V. This is important both to allow low output voltages and for the current limiting to be effective. Op amp IC1a compares the voltage from the wiper of the VOLTAGE SET potentiometer to a divided-down version of the output voltage. It provides negative feedback so that if the output voltage is higher than the setpoint, Q1 is driven harder, pulling the ADJ pin of REG3 down, reducing the output voltage. And if the output voltage is too low, Q1’s base drive is reduced, allowing REG3 to pull the output up. A capacitor from the ADJ pin of REG3 to the -5V rail helps to stabilise this arrangement. Current control op amp IC1b and its associated transistor Q2 work similarly, to regulate current. Because transistors Q1 and Q2 can only sink current, the output voltage will be determined by which is lower: the voltage setting, or the voltage required to achieve the desired current setting. The output current is monitored via a 15 milliohm shunt between the output of REG3/Q4-Q7 and the output terminal. Voltage feedback comes from the output side of this resistor, so the supply will automatically compensate for the shunt’s voltage drop (up to 120mV). siliconchip.com.au T1 S1 ~ 40V IEC MAINS PLUG 115V + – 115V 40V F1 +24V BR1 24V REGULATOR +57V 12V REGULATOR 5V REGULATOR Q3 Q4-Q7 +12V 0.015 OUT IN + ADJ OUTPUT + Q8 & Q9 CONSTANT LOAD – +24V _ –5V + 10k NTC _ OP AMP  siliconchip.com.au –5V (HEATSINK) REG3 Shunt monitor IC4, a form of differential/instrumentation amplifier, converts the voltage across the shunt to a ground-referred voltage so that IC1b can compare it to the voltage from the current set pot. By using control voltages to set the desired output voltage and current, we can easily show these on the front panel of the meter, so you can see what you’re doing. LM317-type regulators have a minimum output load current, which is provided by a constant current sink comprising transistors Q8 and Q9. Otherwise, the output of REG3 would rise of its own accord. The current sink dissipates a lot less power than a fixed resistor would, as the resistor would draw much more current at high output voltages. The NTC thermistor on the heatsink forms a divider with a resistor such that the voltage at their junction drops as the temperature increases. This voltage is fed to a PWM generator which increases the duty cycle fed to the gate of Mosfet Q10 as the temperature increases, speeding up the two 24V fans. The fans are connected in series and run from the 57V supply rail via a dropper resistor. This is a much more pow- -5V REGULATOR CURRENT BOOSTING TRANSISTORS 500VA Fig.2: a simplified circuit/block diagram showing how the Supply works. Four electrolytic capacitors filter the output of the bridge rectifier, which is regulated by REG3 in concert with current boosting transistors Q4-Q7. Op amps IC1a & IC1b monitor the output voltage and current (the latter via a 15mΩ Ω shunt and shunt monitor IC4) and compare it to the settings from potentiometers VR3 & VR4. They then control the voltage at REG3’s adjust pin to maintain the desired voltage and current levels. –9V +5V 4x 4700 F ~ VOLTAGE INVERTER +12V Q1 Q10 VOLTAGE SET IC1a –5V PWM GENERATOR OP AMP Q2 –5V TSENSE DIFF AMP IC1b IC4 –5V SHUTDOWN LOGIC CURRENT SET Q12 ISET ISENSE VSET VSENSE TO METER BUFFERS, CALIBRATION TRIMPOTS AND THEN ON TO PANEL METER(S) er-efficient arrangement than running the fans from one of the regulated rails. The temperature signal is also fed to control logic which biases NPN transistor Q12 on if the heatsink gets too hot, pulling the current control signal to ground and shutting down the supply. Several internal regulators are shown in Fig.2, at upper right. These are required to generate various internal control voltage and to power the control circuitry itself. The output of the +12V regulator is fed to a capacitor charge pump (IC3) which generates a roughly -9V rail that is then regulated to -5V. As mentioned earlier, this is needed to allow the supply output to go down to 0V. Thermal considerations One of the biggest challenges when designing this supply was keeping it cool without needing a huge heatsink in a massive case. The worst case is when the output is short-circuited at 8A (or it’s delivering a very low output voltage at 8A). The required dissipation is then over 400W, and it should ideally handle this continuously. Three things became apparent during testing: 1) The current boosting transistors Australia’s electronics magazine needed to be mounted on the heatsink with as little thermal resistance as possible, to keep the devices themselves at a reasonable temperature when dissipating around 100W each. 2) To keep the heatsink and case size reasonable, powerful cooling fans are required. These should be thermally throttled to keep noise under control. 3) The case would need to be vented, with careful attention paid to the airflow paths. We also determined that the current boosting control transistor, Q3, would need to dissipate over 1W so it too would need to be mounted on the heatsink, along with REG3 and the bridge rectifier, which also dissipates a significant amount of heat at full power. Because the heatsink is connected to the collectors of Q4-Q7, which are sitting at 57V, it needs to be isolated from the Earthed case, so we came up with a mounting arrangement that achieved this, while still keeping the heavy heatsink nicely anchored. The fans are sandwiched between the rear of the case and the heatsink, so they draw air through large holes in the rear panel and blow it straight October 2019  25 over the heatsink fins. That air then turns 90° and exits via the pre-punched vent holes in the top and bottom of the case. This does an excellent job of getting all that heat out of the relatively small enclosure. Circuit details The full circuit of the Bench Supply is shown in Fig.3. 26 Silicon Chip While it’s considerably more complicated than the simplified diagram (Fig.2), you should be able to see how the various sections correspond. Starting where power enters the input, the 230V AC mains from the input socket/switch/fuse assembly is applied to the two 115V primary windings of 500VA transformer T1, which are connected in series. The 40V AC from its paralleled secondaries goes to BR1, Australia’s electronics magazine siliconchip.com.au Fig.3: the full circuit of the Bench Supply. The regulator, control circuitry and output current monitoring are in the upper right quadrant, while the panel meter display buffer circuitry is at lower right. At centre left is the PWM fan control, with the thermal shutdown and temperature monitoring circuitry below. The mains power supply, linear regulators and negative rail generator (IC3 & D1-D2) are at upper left. a 35A bridge rectifier and from there, to a bank of four 4700µF 63V electrolytic capacitors to carry the circuit over the troughs of the mains cycle. With no load, the main DC bus capacitors sit at around 57V. The diode drop across the bridge is offset by the transformer’s no-load voltage being slightly above nominal. In any case, it is just below the 60V limit of the LM317HV regulator (REG3). siliconchip.com.au Control circuitry As mentioned earlier, the LM317HV adjustable voltage regulator (REG3) is the core of the circuit. It maintains the output voltage steady in spite of changes in load impedance and current draw, as long as its ADJ pin voltage is held constant. The ADJ pin is pulled up by an internal current from the input. To regulate the output, the circuit sinks a variable current from the ADJ pin. Australia’s electronics magazine October 2019  27 This control is exerted by IC1, a dual op amp which runs from a 29V supply, between the +24V and -5V rails. The negative voltage is necessary because the LM317HV’s ADJ pin needs to be around 1.2V below the output to regulate correctly. To achieve 0V at the output means that the ADJ pin needs to be around -1.2V relative to GND. The voltage and current control sections of the circuit around IC1 are quite similar. The reference voltage from the potentiometers is fed into their respective op amp inverting inputs (pins 2 and 6) via 10kΩ resistors while feedback voltages from the output are fed into the non-inverting inputs (pins 3 and 5) via another pair of 10kΩ resistors. The user controls the Bench Supply via voltage set potentiometer VR3 and current set potentiometer VR4. One end of each is connected to ground so that when set to their minimums, their wipers are at 0V which corresponds to zero voltage and current at the output. These are set up as voltage dividers, and both have series 10kΩ trimpots (VR1 and VR2) connected as variable resistors on their high side. This allows you to adjust their full-scale ranges. The current setting pot also has a 27kΩ resistor in its divider chain, as the voltage and current adjustment have different scales. The supply’s output voltage is sampled by a 22kΩ/10kΩ voltage divider, with a 100nF capacitor across the upper resistor to give more feedback on transients, stabilising the feedback loop. The result is a 0-15.625V feedback voltage for a 0-50V output voltage. This divider is necessary to keep the feedback voltage within the input voltage range of op amp IC1a, which runs from the 24V supply. For the normal 0-50V output range, VR1 is adjusted to give 15.625V at TP1 with VR3 rotated fully clockwise (the voltage at TP5 should be similar). If you want to limit the voltage output to 45V, avoiding the loss of regulation at higher current settings, it can be adjusted to 14.04V instead. Current feedback from the 15mΩ shunt is via the INA282 shunt monitor, IC4, which has a gain of 50 times. That means that a 1A output current results in 750mV (1A x 15mΩ x 50) at output pin 5 of IC4. So at the maximum output current of 8A, we get 6V from IC4. Therefore, VR2 is adjusted to give 6V at TP3 with VR4 rotated fully clockwise (the voltage at TP6 will be similar). Under normal operation, it is expected that TP2 (“VSENSE”) will track TP1 (“VSET”) as the output voltage follows the control. If current limiting is occurring, then TP4 (“ISENSE”) will track TP3 (“ISET”), and the voltage at TP2 will be less than TP1. There are 100nF capacitors from the wipers of VR3 and VR4 to -5V, keeping the impedance of these control lines low, to minimise noise pickup which would otherwise make its way to the supply’s output. Getting back to the control circuitry, the output from each op amp stage in IC1 (pins 1 and 7) controls NPN transistors Q1 and Q2 via two 1MΩ base current-limiting resistors. We’re using BC546s because they have a 65V rating and they can see up to about 50V on their collectors. The LM317HV only sources about 10µA out of its ADJ pin, meaning its output can only rise by 1V per millisecond as this current must charge up the 100nF capacitor between the ADJ pin and -5V. However, Q1 and Q2 can discharge this capacitor more quickly, which is important in case the output is overloaded or short-circuited, as it means the supply’s voltage can be cut quickly. Op amp IC1 and transistors Q1 & Q2 combine to provide a phenomenal amount of gain in the control loop, which is handy to have for fast response, but needs to be carefully controlled to avoid oscillation due to overshooting. The minuscule base current through the 1MΩ resistors is one way the response of the loop has been tempered. Another is the use of the 1nF and 100nF capacitors between the op amp inputs and outputs, which dampen what would otherwise be a sharp response to a more gradual change, thus preventing oscillation. Scope1: the yellow trace is the clipped ‘triangle’ waveform at pin 5 of IC2b while the blue trace is the thermistor divider voltage at pin 6. Since the latter is above the former the whole time, the gate of Mosfet Q10 (green) is sitting at 0V, and so the fans are both switched off. Scope2: the thermistor temperature has now risen enough that the divider voltage (blue) is now just below the peaks of the clipped triangle waveform (yellow) and so the gate of Q10 (green) is now a 300Hz square wave with a duty cycle of 43%. The fans are now both running at a moderate speed. 28 Silicon Chip Power output stage As we noted earlier, the LM317HV does not carry most of Australia’s electronics magazine siliconchip.com.au The thermal equation 120 60 50 100 50 500 80 30 60 V Voltage oltage drop (left axis) 20 10 40 Output current (left axis) 0 10 20 350 40 300 30 250 Device dissipation (right axis) 20 20 10 0 0 200 150 Output current (left axis) 100 50 0 Fig.4(a) 400 V Voltage oltage drop (left axis) Dissipation (W) 40 Dissipation (W) Device dissipation (right axis) Voltage drop (V) / Current (A) 450 Voltage drop (V) / Current (A) You might notice some parallels between this High Power Bench Supply board and a power amplifier. Many of our power amplifiers, such as the Ultra-LD Mk.2-Mk.4 series and more recently, the SC200 (January-March 2017; siliconchip.com.au/Series/308) also use a 40V transformer to provide nominal 57V rails and use four power transistors in their output stages. While this circuit definitely has similarities with a power amplifier, the thermal and power considerations are significantly different. An audio amplifier only has to deal with a relatively small load impedance variation, delivering its power into 2-10Ω or so, depending on the speaker characteristics and frequency. The output current therefore varies more or less proportionally with the voltage. So the maximum power dissipation in the amplifier therefore occurs when the output voltage is half the supply voltage – see Fig.4(a). On the other hand, our High Power Bench Supply PSU cannot expect a fixed load impedance and must be capable of 60 30 40 50 Output Voltage (V) delivering the full load current with zero output voltage. So for the same maximum current, the maximum power is doubled, to over 400W – see Fig.4(b). Therefore, our design needs to be able to dissipate much more power than a typical audio amplifier module under worst-case conditions. We initially mounted our power transistors on the heatsink using insulating pads but found that even at modest power outputs, the transistors tended to overheat, even though the heatsink was not that hot. Even switching to a thin layer of polyimide tape did not help significantly. It was only when we directly mounted the transistors on the heatsink that we were 0 0 Fig.4(b) 10 20 30 40 50 Output Voltage (V) able to keep them at a reasonable temperature when dissipating close to 100W per device. The thermal resistance of the heatsink (with natural convection only) is quoted as 0.72°C/W, meaning that we would expect a temperature rise of 288°C above ambient with 400W total dissipation. As the maximum operating temperature of the transistors is specified as 150°C, forced cooling is necessary. The final solution of mounting the output transistors to the heatsink, insulating it from the chassis and having two high-power fans blowing directly over its fins is necessary for correct operation of the unit under heavy load. the load current. It is supplemented by four power FJA4313 power transistors, Q4-Q7. These are controlled by a 68Ω pass resistor on the LM317HV’s input. As its output current rises above 10mA and the voltage across the 68Ω resistor exceeds 0.6V, Q3 switches on and so do Q4-Q7, supplementing the output current. This situation is stable in that if the output current through REG1 drops due to the output transistors sourcing more current than necessary, the base current through Q3 is automatically reduced and so transistors Q4-Q7 start to switch off. Each of these transistors has a 0.1Ω emitter resistor to improve current sharing even if the device characteristics are not identical. Scope3: the thermistor temperature has increased significantly, and the divider voltage (blue) has fallen, so the duty cycle at the gate of Mosfet Q10 has risen to 90%. Scope4: the thermistor divider voltage has now fallen further as the thermistor is very hot (above 80°C) and so the gate of Mosfet Q10 is permanently high, with the fans running continuously at full speed. siliconchip.com.au Australia’s electronics magazine October 2019  29 At the maximum 8A output current, each of these transistors only passes about 2A, so the loss across these emitter resistors is only about 200mV. This transistor current booster stage again provides a tremendous amount of gain which needs to be dealt with carefully. A 100nF capacitor connects from the junction of the current sharing resistors back to the base of Q3. This provides negative feedback at high frequencies, preventing oscillation. Transistors Q3-Q7 and REG1 (the LM317HV) are mounted on the main heatsink. As we noted, REG1 does not dissipate much power, but it is capable of thermal shutdown. It should not get hot enough for this to occur, but it does form a ‘last-ditch’ safeguard. The 15mΩ high-side current shunt is monitored by IC4, an INA282 high side shunt monitor. IC4 and the shunt are the only two surface-mount devices used in the circuit. IC4 takes the difference between its two input voltages (the voltage across the shunt) and multiplies it by 50 before shifting it to be relative to the average voltage on its REF pins, which in this case are both connected to GND. Thus we have a voltage proportional to the current and referred to GND, which we can compare to the voltage on the current set potentiometer (VR4). A 10µF capacitor from the output of REG3 to ground provides some smoothing and stability. It is purposefully a small value to limit the current in case the output is short-circuited and to ensure a fast response to voltage and current changes when the supply’s load is light. It’s paralleled with a 100nF capacitor for better high-frequency performance. Minimum load The LM317HV requires a minimum output current of around 3.5mA to maintain regulation. Otherwise, the output voltage will rise. Scope5: the yellow trace shows the Supply’s output voltage, and the green trace shows its current delivery, at around 2.5A/div. It’s delivering 4A at 24V into a 6Ω Ω load but the load impedance then suddenly drops to 3.5Ω Ω, increasing the current to nearly 7A. The current limit has been set to around 5A, so the supply reacts within a few milliseconds to reduce the output voltage. The load current settles at the set value around 10ms later. 30 Silicon Chip As we cannot guarantee that there will be a load connected to the supply, we have to provide one. In a fixed voltage application, a resistor would be adequate, but not in this case. To ensure a minimum current is sunk across the full voltage range, a constant current configuration with a pair of BC546 transistors (Q8 and Q9) is used, with the current set by a 100Ω resistor to around 6mA. Again BC546s have been chosen to withstand the output voltage of up to 50V. This circuit does not work unless there is more than 1.2V between its top and bottom due to the forward voltage of the two base-emitter junctions. The current is therefore sunk into the -5V rail, to ensure that regulation is maintained, even at low output voltages. At high voltages on the output, this part of the circuit can dissipate a few hundred milliwatts. Fan control A thermistor-controlled fan circuit is provided so that the powerful cooling fans only operate as needed. The thermistor is also used to reduce the output current in case the heatsink gets too hot despite the fans running at full blast. Dual op amp IC2 is powered from the 12V rail. One half of the op amp (IC2a) is a triangle waveform generator, with the 1µF capacitor alternately charged and discharged between around 3V and 9V. The triangle waveform does not have linear ramps (they’re exponential), but that doesn’t matter for our application. With timing components of 1kΩ and 1µF, the circuit oscillates at around 280Hz. The triangle wave from pin 1 of IC2a is fed to the cathode of zener diode ZD1 via a 10kΩ resistor. This creates a truncated triangle wave (see Scope1), which is fed to the non-inverting input (pin 5) of the second half of the op amp, IC2b. Due to the limited current applied to ZD1, the peak voltage is around 6.5V. Scope6: this shows a 4A resistive load being connected to the Supply while it is delivering 25V. The output is never more than 200mV from the setpoint and settles in much less than 1ms. A load with any amount of capacitance will see even less deviation than this. Australia’s electronics magazine siliconchip.com.au The 10kΩ NTC thermistor is connected in series with a 9.1kΩ resistor, to form a voltage divider across the 12V rail. The thermistor is connected at the bottom of the divider, so that as its temperature rises, the voltage at the divider junction decreases. At 20°C, the voltage is around 7V, dropping to around 2V at 60°C. This voltage is fed into IC2’s pin 6, the inverting input. When the truncated triangle waveform voltage is above the thermistor voltage, output pin goes high and when the triangle voltage is below the thermistor voltage, that output is low. Thus pin 7 of IC2b produces a square wave at 280Hz with a duty cycle that increases as the thermistor temperature increases. This drives the gate of N-channel Mosfet Q10 (IRF540) via a 1kΩ resistor, which powers the two fans. A 10kΩ pull-down on the Mosfet gate ensures it switches off when power is removed. We have two 24V DC fans wired in series and connected via CON4 and CON5. When Q10 is on, about 9V appears across the 33Ω 5W ballast resistor, reducing the ~57V DC supply voltage to around 48-49V so they each run off about 24V. The powerful fans we have chosen draw about 280mA at 24V. If you use different fans, you will need to alter the resistor value to suit. When the temperature at the thermistor is near ambient, the thermistor divider is at around 7V and is above the 6.5V peak set by the zener diode. Thus output pin of IC2b remains low and Q10, and the fans are off. When the divider voltage drops below the voltage set by ZD1, the fan quickly jumps up to a duty cycle of approximately 40%. This ensures that the fans start reliably, and is the reason for the presence of ZD1. The duty cycle increases as the temperature rises until the thermistor divider voltage is below the trough of the triangle waveform, in which case Q10 and the fans are switched on 100% of the time. Thus the fans can dynamically respond to changes in temperature. Scope1-Scope4 show how the duty cycle varies in response to changes in temperature. Scope7: the green trace shows around 2V of ripple on the pre-regulator 4 x 4700µF capacitor bank with the Supply delivering 4A into 25V. The yellow trace is the Supply’s output. The scope measures 3mV of ripple, but this comparable in magnitude to the noise that the scope probes pick up when grounded. Scope8: This is the reverse of the scenario seen in Scope6, with a 4A resistive load being disconnected from the Supply at 25V. There is around half a volt of overshoot followed by a lesser amount of undershoot and the output settles completely within 2ms. siliconchip.com.au Thermal shutdown The thermistor voltage is also fed to NPN transistor Q11 via a 100kΩ base resistor and diode D4. This means that Q11 switches off if the thermistor voltage drops below 1.2V. The high resistor value means that this part of the circuit does not affect the thermistor voltage significantly. If the thermistor temperature rises above 80°C, the divider voltage drops below 1.2V and Q11 switches off. Its collector voltage rises enough to allow current to flow through D3, charging the following 1µF capacitor. This eventually provides enough base current for NPN transistors Q12 and Q13 to switch on, lighting LED1 and pulling down the wiper voltage of current set potentiometer VR4. In practice, the current limit setpoint does not reach exactly zero when this happens, but stabilises at around 100mA, reducing the maximum dissipation in the output devices to below 10W. The 1µF capacitor can only discharge via the two 100kΩ base resistors, giving around a one-second delay between the thermistor voltage dropping and the current limit returning to normal. This, in combination with the thermal mass of the heatsink, prevents the thermal limiting from switching on and off rapidly. Monitoring voltages and currents To avoid the need for hooking multiple multimeters up to the Bench Supply to see what it’s doing, it incorporates five read-outs. These can be shown on a single LCD screen or multiple panel meters. Regardless, the Bench Supply board has to provide analog voltages to feed to these displays. These voltages are buffered by dual op amps IC5 and IC6, Australia’s electronics magazine October 2019  31 which are powered from the same +24V and -5V rails as control op amp IC1. They form four unity gain amplifiers. Their non-inverting inputs are connected to TP1, TP2, TP3 and TP4.The output from each buffer is fed into a 10kΩ trimpot (VR5-VR8) to allow you to adjust the voltage scaling to suit the display(s). These trimpots effectively allow any fraction of the reference voltage to be fed to the panel meters. The thermistor voltage is scaled down by a pair of 1MΩ resistors to provide a 0-5V signal suitable for feeding to a microcontroller. A 100nF bypass capacitor provides a low-impedance source for whatever is connected to sample it. The time constant of the 1MΩ/100nF low-pass filter is not a problem because the thermistor temperature does not change rapidly. All the buffered signals are fed to DIL header CON6, along with ground connections and a 5V supply to run the LCD screen or panel meters. As an example, when the Bench Supply is delivering 50V, there will be 15.6V at TP2. IC5b buffers this, and VR6 can be set so that 5V is fed to pin 5 of CON6 in this condition, ie, one-tenth of the actual output voltage. The panel meter just needs its decimal point set so that it reads 50.0 when receiving a 5V signal. Similarly, the current values can be displayed on a voltmeter, with the range appropriately set by scaling and placement of the decimal point. A similar scaling by a factor of 10 is appropriate here too. pin 11 of CON6 to an analog meter and draw an appropriate scale, calibrated to match the thermistor temperature. Five-way Panel Meter While we don’t know of any panel meters that will be able to directly read the thermistor voltage and convert it into a temperature, our microcontroller-based Five-way Panel Meter design can interpret it, as well as displaying the two voltage and two current values. The details of this low-cost Five-way Panel Meter will be in next month’s issue, coinciding with the PCB construction and testing details for the Bench Supply. If you don’t want to use that Panel Meter board, but you want a temperature read-out, you could feed the voltage from There are three heatsinks in this design, small flag heatsinks for the 12V and 24V regulators (REG4 and REG1) and the main heatsink for REG3, Q3-Q7 and BR1. Due to the high voltages present, regulators REG4 and REG1 have significant dissipation, despite the series ballast resistors which reduce their input voltage. The 24V regulator is key to setting the voltage and current references, so keeping this device at a uniform temperature will help with the stability of the output. As mentioned earlier, to efficiently get heat out of transistors Q4-Q7, they are not insulated from the main heatsink Scope9: here we have simulated a step-change in the voltage control input by shorting the VSET point to ground and then releasing it. The output voltage drop is much quicker than the rise, ensuring that the chance of overshoot is minimised under dynamic conditions. Scope10: this current control step-change test shows a similar response as in Scope9. Again, the fall is faster, indicating that the Bench Supply is designed to respond to over-current conditions quickly. There is no visible overshoot. 32 Silicon Chip Internal power supply 24V linear regulator REG1 is fed from the 57V rail via a 220Ω 5W dropper (ballast) resistor. This reduces dissipation in the regulator while its 100µF input bypass capacitor prevents that resistor from affecting regulation. The 24V rail powers the output control op amps (IC1), the sense buffer op amps (IC5 & IC6) and is the reference voltage for the output voltage and current adjustment potentiometers (VR3 & VR4). The 24V rail also feeds into 12V regulator REG4 via another ballast resistor, this time 68Ω 1W. The 12V supply feeds the negative voltage generator, the current shunt monitor IC, the thermistor and fan control, and the 5V regulator (REG5). The resulting 5V rail is for powering the panel meter/display(s). The negative voltage generator consists of a 555 timer (IC3) operating in astable mode at around 60kHz, with a near 50% duty cycle. Its output is connected to 1N4148 diodes D1 and D2 via a 100µF capacitor, forming a charge pump. The 100µF capacitor at pin 3 of IC3 charges up through D2 when pin 3 is high. When pin 3 goes low, D2 is reversebiased and current instead flows through D1, charging up the 100µF capacitor at REG2’s input. This results in around -9V at the input of REG2, resulting in a regulated -5V rail at its output. Heatsinking Australia’s electronics magazine siliconchip.com.au and it is therefore at around +57V DC potential. 57V DC is considered ‘low voltage’, but of course there are also mains voltage present around the transformer, so it doesn’t hurt to use caution while working on the supply when it’s powered. The LM317HV regulator has a live tab connected to its output, which can vary anywhere between 0V and near the DC rail voltage, so it must be insulated from the main heatsink. We used a silicone pad and an insulating bush. Similarly, the tab of Q3 is connected to its collector. If the collector were connected to the DC rail, then the output transistors would turn on hard, so this must be avoided. It too is mounted with a silicone pad and insulated bush. We have purposefully mounted Q3 reversed on the PCB, with its pin 3 on the left, so that its metal tab faces away from the heatsink. That’s because, despite an insulating washer, we found it was still shorting to the heatsink via the screw. Reversing the device solved that. Its dissipation is not that high, so the added thermal resistance is not a big problem. Of course, the thermistor is also mounted on the heatsink and must be insulated too. We used a stud-type thermistor which has the active element potted, so that is already taken care of. Performance Scope grabs Scope5-10 demonstrate some of the performance characteristics of the Supply. These grabs demonstrate the effects of sudden ‘step’ changes in the operating conditions. In reality, most changes won’t occur so suddenly. Importantly, the Bench Supply can respond quickly to changes in load without excessive overshoots, including switching into current limiting when necessary. The scope grabs demonstrate that it typically responds within milliseconds to these sort of changes. See the details of the individual tests underneath the scope grabs. We also did some thermal tests to determine how well the Bench Supply handles heat dissipation. As noted in our panel about “The Thermal Equation”, the Bench Supply works hardest when the output voltage is low, but the current is high. In these cases, the full supply voltage appears across the output transistors. For example, dumping 8A into a short circuit means that the Bench Supply is delivering around 400W into the heatsink. During our scope grab tests, at 25V and 4A, it is dissipating around 100W. Under the latter condition, the thermistor registers around 20°C above ambient, and the fans run at around half speed. One of our more severe tests involved connecting a 2Ω dummy load. With the output set to 8A, the voltage reaches 16V, and the Supply is dissipating around 300W. Under these conditions, the thermistor reached 77°C (around 55°C above ambient) after around 10 minutes and then held steady. Contrary to what you might think, delivering 45V at 8A is not that stressful to the supply, as there is only about 10V across the output devices and thus a dissipation of around 80W. Delivering 8A into a short circuit is more difficult; the supply can manage for this, but only for a few minutes at a time before it enters thermal current limiting. SC NEXT MONTH: As promised earlier in this article, our November issue will commence the full construction details, including the parts list. If you want to be sure not to miss that issue, why not subscribe to SILICON CHIP? (See page 97). siliconchip.com.au KCAB ISSUES First the good news: Did you know . . . that back issues of SILICON CHIP magazine are still available, with only a few exceptions, for the LAST TWENTY YEARS +? Check out the following list to see if the issue you want is still in stock. Order any of these issues online or by phone for just $13.00 INCLUDING p&p in Australia! See the website address below or call (02) 9939 3295 9-4, Mon-Fri 1997: all except August & September 1998: all except March 1999: all except February 2000: all except April 2001: all except October & December 2002: all except June & July 2003-2005: all available 2006: all except January & October 2007-2010: all available 2011: all except November & December 2012: all except December 2013: all except February 2014: all except January 2015: all except January 2016-2019: all available And the even better news: Did you know . . . that if the issue you want is out of stock, we can supply a copy of any article from any issue (nominate which article you require. Price is the same as a back issue due to the extra work required). And now the best news: Did you know . . . that you can also view most articles on the SILICON CHIP website! And if you’re a subscriber, (print or online) there’s even a discount on the price (as there is with any other SILICON CHIP ONLINE SHOP merchandise). Log on today for all the details: www.siliconchip.com.au Australia’s electronics magazine October 2019  33