My mentor in the hifi industry, John Oakman, once explained to me that
when we listen to hifi, really we're just listening to modulated mains. The point he was trying to make to me was that ultimately the source of the sounds we hear comes fundamentally from what we get out of the wall in our houses - mains electricity. This is why audiophiles and the engineers who design hifi care so much about power supplies.
It's not uncommon for people to invest in additional or improved power supplies in their systems. Two famous brand that have made a fortune selling additional power supplies are Naim and Cyrus, for example the iCyrus PSX R2 pictured above, but the pattern can be found throughout the industry.
This article explains a little about how power supplies work, and compares the relative advantages and disadvantages of the two most common designs of power supply, so you know the difference between them, and you can understand a little more about why some people like to change or upgrade the power supplies in their systems. Let's start with a quick primer on mains electricity.
When you connect a battery to a load, e.g. a bulb, an electrical current flows. In the case of a bulb connected to a battery, this current is stable, and flows in one direction only. This is what is meant by a direct current (DC). The battery also has a voltage, which in simple terms explains how much current it has the potential to generate. A small cylindrical battery might be a 1.5 volt battery.
The electricity that comes out of the wall is different - it alternates (AC). Half the time the signal is positive and half the time it is negative. As it alternates, we also need to know how often it alternates - the frequency of the changes from positive to negative. In the UK this happens about 50 times a second, or at a frequency of
50 Hertz. The change isn't abrupt - suddenly positive, suddenly negative - it follows a wave-like pattern (a sine wave), like this:
You can think of this as being a bit like the tide, slowly going out, then coming back in again, from 0 to +230V then back to 0, and down to -230V.
The electronic components inside hifi equipment, and computers etc., generally do not require anything like 240V, and won't work with an alternating current, and so we need some way of converting the power from the mains outlet into something we can use in our amplifiers, DACs, etc. This is the function of a power supply - to go from a high voltage, alternating current to a low voltage, direct current that we can use.
There are two common approaches to building a power supply - linear (LPS), and switch mode (SMPS). I'll describe how both work, and touch on the relative advantages and disadvantages of each. But first let me cover a few common components that both designs use: transformers, rectifiers, and filters.
When an electric current flows through a cable, it generates a magnetic field. If we make a coil of wire instead of it being in a line, that magnetic field will get stronger. You can test this at home by winding some insulated wire around a screw or a bolt a number of times, and then connecting both ends to a battery. You'll be able to pick up paper clips with the bolt - you created a magnet.
When an alternating current flows backwards and forwards around a circuit, this also creates a magnetic field, but one which - you guessed it - alternates. The polarity and intensity of the magnetic field changes in line with the sine wave.
Electricity and magnetism have a close link to each other - you can make a magnetic field with a current, and you can also create a current with a magnetic field. This is the principle behind a transformer:
In a transformer we have some wire wrapped around one side of a core. When an alternating current is applied, this creates a magnetic field. Because the current is alternating, the magnetic field moves too. If we wrap some wire around the other side of the core, the opposite phenomenon occurs: the moving magnetic field induces a voltage.
Depending on the ratio of the turns of wire on each side of the transformer, the transformer either steps up or steps down. If the primary wiring has fewer turns than the secondary, the voltage on the secondary side will be greater (and the current less), and conversely, if the number of turns on the primary winding is greater, the induced voltage will be smaller, but the current higher.
This, incidentally, is how power is moved between power stations and homes. The electricity coming out of the power station goes through a massive step up transformer to create the huge voltage needed to push the current along long-distance wires. Then in a local substation, there's step-down transformer that drops the voltage to something more useful in a domestic setting - 230V.
If transformers can change the voltage of an alternating current from low to high or high to low, we are half way to our desired state - low voltage DC. We now know how to get the voltage down... but how do we stop the current from oscillating? This is done with a rectifier. In essence this is very simple - we use a set of four components called diodes to route the current in a uniform direction regardless of whether the phase of the wave is positive or negative. Diodes are components that permit electricity to flow in only one direction. The characteristic diamond shape of the bridge rectifier is shown here:
The input to a rectifier is an alternating current - a sine wave. When the signal is on the negative side of the wave, the electrons are moving one way, and diodes 1 and 3 permit the current to flow. When the wave crosses zero and goes positive the direction changes, and diodes 1 and 3 stop conducting, and diodes 2 and 4 conduct instead. The end result is that regardless of the direction of current flow, the output will be positive, and slightly less than the voltage that entered the rectifier. What the rectifier effectively does is effectively take the bottom half of the signal that entered the rectifier and inverts it - so we have a pulsating direct current - it never goes below zero, but it is still oscillating somewhat, as shown in the DC output in the diagram above.
So, the output of the rectifier doesn't alternate in phase - every value is positive - however, it's still very bumpy rather than the steady, flat level that a battery would give. To solve this problem, we use filter capacitors.
Capacitors are devices which store energy and then release it again. If you've ever switched your hifi off and noticed that the light stays on for a short while, before gradually dimming to off, you've watched a capacitor discharging.
By picking a capacitor that discharges more slowly that the rate at which the voltage decays on our wobbly output, we can rather cleverly fill in the gaps. When the signal rises, the capacitor charges, and when the signal falls, the capacitor discharges, but as long as the capacitor is discharging more slowly than the signal falls, the effect will be to flatten the curve. We'll never get it completely flat, but it'll be pretty good. Now we have something near-as-darn-it equivalent to DC.
Both linear and switch mode power supplies use transformers, rectifiers and filters, but in significantly different ways, which helps us to understand their different characteristics and relative advantages and disadvantages.
Linear Power Supplies
A linear power supply is fundamentally a very simple device. It takes mains power straight in, straight into a step-down transformer to a much lower target voltage. This is then passed through a rectifier to move the negative part of the signal to the positive side, and then through a filter to flatten it. The final stage is to go through a regulator. This is a circuit which takes in a voltage that may not be consistently the same, since it will vary according to the accuracy of the transformer, and what comes in from the mains, which can certainly vary by up to 5-10% at any time. It then emits a very steady and "regulated" voltage according to the design and requirements of the system.
That's basically it!
So why would anyone go to the trouble of doing anything else? Well, the biggest problem with linear power supplies is that the frequency of the signal entering the transformer is low - 50 Hz in the UK. You'll remember I mentioned the property of coils to induce a voltage. Well, this is called "self-induction". When designing a transformer, there's a relationship between self-inductance, the impedance of the coil, and frequency. This means that when operating at low frequencies we need a lot of wire, and metal cores in our transformers. Depending on the use case, the transformers can get very big and very heavy, with many windings of thick copper wire. Naturally this also makes them very expensive.
Secondly, linear power supplies are inherently wasteful. Power is lost as heat in the transformer and the regulator, by design. This means that if you need your power supply to be powerful - i.e. if it has a lot of work to do, it needs to be even bigger, even heavier, and even more expensive. It's also going to get very hot!
Switch Mode Power Supplies
Switch mode power supplies are radically different. Compared to linear power supplies, they are complex, however they are much lighter, much more efficient, and consequently much more affordable. For an example, a 30W SMPS will weigh about 100 grams. Just the transformer needed for an 18W linear power supply weighs four times as much, and with a weight budget of 400g, you could make a 260W 24V switch mode power supply.
As the name suggests, a switch mode power supply works using switching devices that go on and off at high frequencies. Here's a block diagram of how one works, at a simple level:
There are five basic phases in a switch mode power supply:
1) EMC filter and fuse
2) Rectifier and primary filter
3) Driver and mosfets (chopper)
4) Transformer and rectifier
5) Secondary filter
Unlike an LPS, an SMPS does not take the low frequency high voltage AC and feed it straight into a transformer. Instead the first stage of an SMPS is a filter - slightly different from the filter we described above. This filter removes removes high frequency interference from the input signal, to give a clean 50Hz ~230V sine wave.
Rectifier and Primary Filter
The second stage is exactly the same as in an LPS - we have a full wave bridge rectifier and filter. The design is the same - four diodes, ensuring only the positive side of the signal passes through. And, just as in a linear power supply, we use capacitors to smooth the output to give a DC value. The difference here is that the voltage is still very high - in fact by the time we've left the filter the signal
is around 320V DC.
The next phase is where things get radically different. The signal is now chopped up into pulses by very fast switching transistors, controlled by an integrated circuit called a PWM controller (pulse width modulation). The result of the interplay between these transistors and the controller is a series of pulses of varying width - so a square wave looking a bit like this:
Importantly, the frequency of these pulses is very high - 20-50KHz - so up to a thousand times faster than the frequency of the input signal from the mains.
Transformer and Rectifier
This pulsing signal is then fed to a transformer, and as we know, the higher the frequency, the less wire and less metal we need, so the transformer can be much much smaller and can have a ferrite core, reducing weight significantly.
Because the pulses are square in form, and in high frequency, the effect in the transformer is for it to charge and discharge generating an alternating current on the other side of the transformer, only at a lower voltage, and still with a square shaped waveform, the width of which is controlled by the PWM chip.
This signal is then passed through another rectifier - similar to the one we discussed earlier, only this time a "half-wave" rectifier - throwing away the negative part, but not moving it to the positive part.
The final stage is again somewhat familiar - we have a filter including a capacitor which charges up and discharges more slowly than the frequency of the pulses, to flatten out the signal. All that remains is one clever last bit - the feedback. The signal that comes out of the final filter is fed back to the PWM chip to compare with a reference voltage. Depending on that value, the controller will increase or decrease the switching speed, widening or narrowing the width of the pulses, which in turn will raise or lower the final output voltage.
Advantages and Disadvantages
Both PSU designs are popular and widely used, but is there anything to choose between them for audio purposes? This is a hotly debated topic, but a few general pointers can help us. In broad terms, it is clear that an SMPS is going to be cheaper, lighter and smaller. This means if you're building something to a budget, or want it to fit in a small chassis, a switch mode power supply is probably a good idea.
However, if you're not constrained by size, have a large case, can fit big heatsinks etc, and your target price is a little higher, these issues may not be such a concern.
In addition, in general compared to SMPS, linear power supplies tend to be much quieter, less susceptible to both radio and electromagnetic interference, have less ripple, and be able to deliver a much quicker transient response (ie deliver large amounts of power when needed). It's for these reasons that it's common to see manufacturers who offer additional power supplies tending to use hefty linear designs.
Does it make any difference?
I've been in the industry long enough to know that the best answer to this is always to try it out for yourself. It's very common to find people on forums insisting that if you can't measure something you can't hear something, but I can tell you without any doubt that I've experienced differences between designs that thus far we're unable to explain by science. This doesn't mean it's imagination, or snake oil - science is expanding its boundaries continuously, but there are many things that our current models don't fully explain, and I would prefer to be willing to accept that one day we will understand why, rather than take the slightly arrogant view that because we can't explain it now, it must not exist.
Power supplies aren't so much in this category. In most cases linear power supplies are simply measurably less noisy. However you can also hear the difference - back in the days when I used to sell Naim, I always looked forward to being able to demonstrate the difference adding a hicap or flatcap made to a system - it really can make a good system sound great.
These days my interest is primarily in the power supplies we use to control computer-based systems - Raspberry Pi, Intel NUC, Mac Minis etc. Almost invariably these will we supplied with budget SMPS, as you'd expect - for general purpose use it's a cheap and efficient approach, and absolutely fit for purpose.
However if you're using a Raspberry Pi or a NUC as part of a high quality hifi system in which you've spent hundreds or thousands of pounds on DACs, amplifiers and speakers, you might want to consider whether it would be worth making an investment in the power supply of the machine supplying the source music.
As ever, if you have any questions, drop us a note in the comments, email us, or give us a call - we'd be happy to chat to you, give you advice, and help you in any way we can.