Audio amplifier power supply design - Part 2: External supplies, inrush current & RF emissions
External Power Supplies
However much care is taken, it is very difficult to keep all traces of transformer-induced hum out of the signal circuitry. It is highly irritating to find that despite the cunning use of low-noise circuitry, the noise floor is defined by the deficiencies of a component – for the ideal transformer would obviously have no external field – rather than the laws of physics as articulated by Johnson.
The ultimate solution to the problem is to put the mains transformer in a separate box, which can be placed a meter or so away from the amplifier unit, and powering it through an umbilical lead.
• The transformer field hum problem is authoritatively solved.
• Will appeal to some potential customers as a 'serious' approach to high-end audio.
• The cost of an extra enclosure plus an extra cable and connectors, indicator lights, etc. The connectors will have to be multi-pole and capable of handling considerable voltages and currents. The transformer box must have fuses or other means of protection in case of short-circuits in the cable.
• A significant proportion of users will, exhortations to the contrary not withstanding, promptly place the amplifier box directly on top of the transformer box, immediately defeating the whole object. This is particularly likely if the two boxes have the same footprint, and so look as if they ought to be stacked together. However, all is not lost in this situation, as the transformer is still physically further away from the sensitive electronics (though if the transformer has a large field emerging from its ends things may actually be worse) and there are now two extra layers of steel interposed – assuming the boxes are made of steel, that is.
• The voltages involved will probably be above the limit set by the Low Voltage Directive, so it will be necessary to ensure that the connector contacts cannot be touched. If the cable has a connector at both ends then both must be checked for this. A cable that is captive at the power-supply end makes this issue simpler and will also save the cost of a mating pair of connectors, which may be considerable.
The most important design issue is the distribution of the power-supply components between the two boxes. One approach is to put just the mains transformer in the power-supply box. This has the disadvantage that the current in the umbilical cable consists of short charging pulses of large magnitude at a frequency of 100 or 120 Hz, and these will not only experience a greater voltage drop in the cable resistance than a steady current, but also give rise to much greater I2R heating.
The latter is unlikely to cause problems in the cable itself, but can easily be fatal to the contacts of connectors. Speaking from bitter experience, I can warn that connectors that appear to have a more than adequate safety margin can fail under these conditions, and it is best to keep connectors out of charging pulse circuits.
The alternative is to put not just the mains transformer but also the rectifiers and reservoir capacitors in the power-supply box. The current in the umbilical cable is now rectified and smoothed DC, and it is much easier to specify connectors to cope with it. The snag is that the reservoir capacitors have two functions; as well as smoothing the rectified DC, they also hold a store of energy that can be drawn on during output peaks.
The resistance of the cable between the reservoir capacitors and the power amplifier will cause unwanted voltage drops when there are sudden demands for load current, which can significantly reduce peak power outputs during tone-burst testing. Another worry is that the extra resistance in the supply rails could imperil the stability of the amplifier, though the use of generous local decoupling capacitors should be enough to deal with this problem.
A solution to both problems is the provision of significant amounts of capacitance at both ends; the capacitors in the power-supply box deal with the smoothing, while those at the amplifier end provide a ready reserve of electricity. In this case the current through the cable will still show some charging peaks, the size of which will depend on the proportion of the total capacitance at each end and the cable resistance. This could be artificially increased by adding series resistors of small ohmic value but high wattage, making an RC filter that will reduce the ripple seen by the amplifier. This is a bit of a doubtful remedy as it will reduce the power output on sustained signals, and it is a very poor way to reduce amplifier noise derived from the supply rails, as will be described later.
There you have some of the pros and cons of external amplifier power supplies. It is not quite the expensive but foolproof solution it first appears to be, and the design issues require careful thought.
When a transformer is abruptly connected to the mains supply it takes a large current that decays exponentially; this is called the inrush current (or sometimes the turn-on surge, or even the 'inductive surge') and it is highly inconvenient as it can be much greater than the normal current drawn, even at maximum output into the lowest rate load impedance. This inrush current is not a danger to the transformer, which has a big thermal mass, but it can and will blow primary fuses and trip house circuit-breakers. With small and medium-sized transformers the problem is not serious, but it does mean that you have to be very careful in sizing the fuse or fuses in the primary circuit, making sure that they have a high enough rating so their life is not impaired by repeated inrush currents.
With a large transformer (say bigger than 500 VA for a toroid) the inrush becomes large enough to trigger domestic overload protection. Since most houses now have magnetic circuit-breakers rather than wire fuses in the mains distribution panel, this is not as inconvenient as it used to be, but is still thoroughly annoying and will quickly earn you the enmity of your customers.
The inrush issue has to be taken very seriously as it can cause problems that only show up when the unit is out in the field. There is anecdotal evidence that circuit-breakers in Germany, while nominally rated the same as those in Britain, actually respond somewhat faster, so a design that has received careful checking in one country may cause serious trouble in another.
Inrush current is most conveniently measured with purpose-built instruments such as the Voltech power analyzer range. A cheaper method is to use a current transformer (typically of the 'giant clothes-peg' type) clamped around one of the primary connections, and connected to a digital oscilloscope; this is naturally only cheaper if you already have a digital oscilloscope.
It is characteristic of inrush current that its peak value varies widely from one switch-on to the next, as it depends crucially on the point of the mains cycle at which the transformer is connected. If you're unlucky the transformer core briefly saturates and a big peak current is drawn by the primary. For this reason repeated tests – possibly up to 50 – have to be done before you are confident you have experienced the worst case. This often has to be spread out over some time to avoid overtaxing inrush suppression components.
Toroidal transformers typically take greater inrush currents than frame types, due to their lower leakage reactance. There is a component of the inrush current that is due to the charging of the power-supply reservoir capacitors, but this is usually small compared with the transformer inrush. As a rough guideline, if your transformer is bigger than 500 VA you should consider using inrush suppression. If in doubt, then at least make provision for adding it to the design in the development phase.
The inrush current is controlled by making sure there is enough series resistance in the transformer primary circuit to keep the flood of amperes down to an acceptable level. The two main ways of doing this are to use an inrush suppressor component, or a relay that switches resistance into circuit for starting.
Inrush suppression by thermistor
An inrush suppressor component (sometimes called a surge limiter) is a giant thermistor whose resistance drops to a low value as it is heated by the current passing through it; they are usually of the disk type. The inrush suppressor is inserted in series with the transformer primary. The thermal inertia of its mass causes the resistance to drop relatively slowly, so the inrush current is restricted.
Because of their thermistor action, these components run very hot in the low-resistance state (about 200ºC) and must be mounted with caution to ensure they do not melt the plastic of adjacent components. The component leads must be left long enough to avoid thermal degradation of the solder joints with the PCB, and if these leads are insulated it must be with a high-temperature material such as fiberglass sleeving. They are also likely to burn the fingers of service personnel – it is only polite to put a HOT warning on the PCB silkscreen.
Inrush suppressors require a cool-down time after power is removed. This cool-down or 'recovery' time allows the resistance of the NTC thermistor to increase sufficiently to provide the required inrush current suppression the next time it is needed. The necessary time varies according to the particular device, the mounting method and the ambient temperature, but a typical cool-down time is about one minute. This is usually specified by the manufacturer as a thermal time-constant with values ranging from 30 to 150 seconds, the longer times being for the larger and more highly rated versions.
Inrush suppressors are available in many different sizes. The quickest design method is to select a few types that can handle the maximum current in the primary circuit, and try them out to see which is the most effective at controlling the peak inrush value.
Inrush suppression by relay
In this method there is a series resistance placed in series in the transformer primary circuit when mains is first applied. This limits the inrush current, and is then switched out by a relay after a suitable inrush control period, typically around one second. The basic circuit is shown in Figure 9.5a.
The inrush resistor has to sustain a very large short-term overload, so a chunky wire-wound type is appropriate, and it is vital to ensure that it can cope with this overload many times over the life of the amplifier.
However, resistor manufacturers are noticeably reluctant to specify how their products will cope with such conditions. It is therefore a good plan to use inrush suppression in its intended final form from the very start of the development process; by the time all other design issues have been addressed you will almost certainly have put the inrush suppression through enough operating cycles to have confidence in its durability. (Using inrush suppression from the start may well be essential anyway to prevent the workbench circuit-breakers from tripping.)
The inrush current is a complex phenomenon and the resistance value and power rating of the resistor is usually determined by experience rather than protracted mathematical analysis. Here are some typical values that I have used with success:
• 10 Ω, 10 W for an 800 VA toroid;
• 10 Ω, 20 W for a 1300 VA toroid.
Wire-wound resistors come in a limited number of types for sizes above 10 W, and it is often more convenient to use two 22Ω 10 W resistors in parallel when a 20 W capability is required. If the resistor is correctly sized, after a single inrush event it should be warm rather than hot. Repeated and rapid cycling of the power, as may occur in testing, can cause it to get very hot and could eventually lead to failure. Fortunately this is not likely to occur in service.
The circuitry must be arranged so that if the power is turned off then immediately turned on again, inrush suppression still operates for the full period. This situation is called a 'hot restart'.
Many amplifiers are not simply switched on and off, but have an on/standby system where the mains switch initially applies power only to a small transformer that energizes a small amount of control circuitry. A low-current switch, which can be more cosmetically attractive than something hefty enough to control the full mains power, activates the control circuitry and causes it to close a relay that energizes the main supply.
When this function is combined with inrush protection there are usually two identical relays in the primary circuit as shown in Figure 9.5a; at switch-on RLA closes and applies power to the transformer through the inrush resistor R1. After a second or so RLB closes and shorts out the resistor; RLA is now doing nothing so it is de-energized after a very short delay to make sure that RLB is fully closed. The alternative arrangement in Figure 9.5b should be avoided as now it is necessary to keep both relays energized all the time, which is a pointless waste of perfectly good electricity.
Fusing and Rectification
The rectifier (almost always a packaged bridge) must be generously rated to withstand the initial current surge as the reservoirs charge from empty on switch-on. Rectifier heat-sinking is definitely required for most sizes of amplifier; the voltage drop in a silicon rectifier may be low (1 V per diode is a good approximation for rough calculation) but the current pulses are large and the total dissipation is significant.
Reservoir capacitors must have the incoming wiring from the rectifier going directly to the capacitor terminals; likewise the outgoing wiring to the HT rails must leave from these terminals. In other words, do not run a tee off to the cap, because if you do its resistance combined with the high-current charging pulses adds narrow extra peaks to the ripple crests on the DC output and may worsen the hum/ripple level on the audio.
The cabling to and from the rectifiers carries charging pulses that have a considerably higher peak value than the DC output current. Conductor heating is therefore much greater due to the higher value of I2R. Heating is likely to be especially severe if connectors are involved. Fuseholders may also heat up and consideration should be given to using heavy-duty types. Keep an eye on the fuses; if the fusewire sags at turn-on, or during transients, the fuse will fail after a few dozen hours, and the rated value needs to be increased.
When selecting the value of the mains fuse in the transformer primary circuit, remember that toroidal transformers take a large current surge at switch-on. The fuse will definitely need to be of the slow-blow type.
The bridge rectifier must be adequately rated for long-term reliability, and it needs proper heat-sinking.
RF Emissions from Bridge Rectifiers
Bridge rectifiers, even the massive ones intended solely for 100 Hz power rectification, generate surprising quantities of RF. This happens when the bridge diodes turn off; the charge carriers are swept rapidly from the junction and the current flow stops with a sudden jolt that generates harmonics well into the RF bands.
The greater the current, the more RF produced, though it is not generally possible to predict how steep this increase will be. The effect can often be heard by placing a transistor radio (long or medium wave) near the amplifier mains cable. It is the only area in a conventional power amplifier likely to give trouble in EMC emissions testing .
Even if the amplifier is built into a solidly grounded metal case, and the mains transformer has a grounded electrostatic screen, RF will be emitted via the live and neutral mains connections. The first line of defense against this is usually four snubbing capacitors of approximately 100 nF across each diode of the bridge, to reduce the abruptness of the turn-off. If these are to do any good, it is vital that they are all as close as possible to the bridge rectifier connections. (Never forget that such capacitors must be of a type intended to withstand continuous AC stress.)
The second line of defense against RF egress is an X-capacitor wired between Live and Neutral, as near to the mains inlet as possible (see Figure 9.1). This is usually only required on larger power amplifiers of 300 W total and above. The capacitor must be of the special type that can withstand direct mains connection; 100 nF is usually effective (some safety standards set a maximum of 470 nF).
A drain resistor should be connected across the X-capacitor because if the equipment mains switch is open, and the mains lead is disconnected at the peak of the mains waveform, the X-capacitor can be left with enough charge to give a perceptible shock if the mains plug pins are touched. The resistor value should be low enough so that the X-capacitor is discharged to a safe voltage in a small fraction of a second, without being so low as to pointlessly dissipate heat. The voltage rating of the resistor should be watched; this is not usually a problem for 1/4 W sizes and above.
It is very often most economical to power relays from an unregulated supply. This is perfectly practical as most relays have a wide operating voltage range. Hum induced by electrostatic coupling from this supply rail can be sufficient to compromise the noise floor; clearly the likelihood of this depends on the physical layout, but it is inevitable that signal paths and the relay come into proximity at the relay itself. It is therefore desirable to give this line some degree of smoothing, without going to the expense of providing another regulator and heat-sink. (There should be no possibility of direct coupling between the signal ground and relay power ground; these must only join right back at the power supply.) This method of relay driving is more power efficient than a regulated supply rail as it does not require a voltage drop across a regulator that must be sufficient to prevent drop-out and consequent rail ripple in low-mains conditions.
Simple RC smoothing is quite adequate for this purpose and there is no need to consider the use of expensive chokes, which would probably cost more than a regulator, take up more space, radiate magnetic fields, and generally be a pain in the amp. Because relays draw relatively high currents, a low R and a high capacitance value for C are necessary to minimize voltage losses in R and changes in the rail voltage as different numbers of relays are energized.
Figure 9.6 shows a typical power-supply circuit giving a regulated +5 V rail to power a microcontroller, with the addition of an RC-smoothed +9 V rail to power relays. The RC-smoothing values shown are typical, but are likely to need adjustment depending on how many relays are powered and how much current they draw. The R is low at 2.2 Ω and the C high at 4700 µF.
Note the 10 nF capacitor across the transformer secondary; this part must be an X-capacitor or other type rated for continuous AC stress. This is typical of the extra components required to meet modern EMC standards.
 T. Williams, EMC for Product Designers, Newnes (Butterworth-Heinemann), 1992, p. 106.
Printed with permission from Focal Press, a division of Elsevier. Copyright 2009. "Audio Power Amplifier Design Handbook", 5th Edition, by Douglas Self. For more information about this book and other similar titles, please visit www.focalpress.com.
Audio amplifier power supply design - Part 1: Power supply types & transformer considerations
Understanding Class-D amplifier power supply requirements
PSRR: The Real Story about Closed- and Open-Loop Class-D Amplifiers
Distortion in power amplifiers, Part VIII: Class A amplifiers
Yet more on decoupling, Part 1: The regulator's interaction with capacitors
- Fourth-generation digital power controller negates need for loop compensation
- Wireless power: Accommodating new industry standards
- Power Tip 65: Power supply implications of envelope tracking
- Cooling without a fan
- Smaller chip architectures make digital power an imperative
- Helping board designers work with high-performance timing
- Selecting the correct level of DC-DC converter for your space applications
- Managing hold up /transparency time in high reliability power supplies
- Power Tip 64: Compensate for cable drop without remote sensing
- 3-phase MOSFET bridge module handles 2kW load
- Fast-charging supercapacitors using energy harvesters
- Simple solutions for a single-device PWM waveform generator
- Optimizing DSP power budget by adjusting regulator
- Power Tip 63: Testing high-di/dt power supplies
- Basic concepts of linear regulator and switching mode power supplies - Part Two
- Wearable textile battery is capable of being recharged by lightweight solar cells
- Cooling without a fan
- Texas Instruments releases the industry's most extensive reference design library
- OLEDS and solar cells to come straight from the printer in future
- The Top 10 Most Popular Power Management News Stories of 2013
- Digital DC/DC power controller's compensation free control loop reduces design risks and costs
- New primary side regulation constant voltage solution in LED driver
- World's smallest FM radio transmitter made from graphene
- Self healing lithium-ion battery electrode technology could prove disruptive
- Power amplifier solution for mobile base stations doubles efficiency
- Demo Bluetooth low energy keyboard is 10x thinner than any other
- Wireless power: Accommodating new industry standards
- Using film and aluminum electrolytic caps in power converters for voltage stabilization in HV grids
- Bio-based solar cell uses photosynthetic proteins to generate electricity
- Power Tip 65: Power supply implications of envelope tracking
- 3mm Ã— 3mm QFN IC Directly Monitors 0V to 80V Supplies
- Adaptive Cell Converter Topology Enables Constant Efficiency in PFC Applications
- Ultra-compact and innovative - new Schaffner IEC inlet filter series
- EMC Considerations for Switching Power Supplies
- How to Achieve Power Estimation, Reduction and Verification in Low Power Design
- Industrial-Grade Wide Temperature Assurance for Intelligent Systems
- High-voltage DC distribution is key to increased system efficiency and renewable-energy opportunities