Advanced decoupling and bypass techniques for high quality audio performance
Due to the large gain of most audio amps, they can have significant current consumption. Regardless if the topology is class A, AB, or D, one thing is common and that is the need to be able to quickly and adequately respond to peaks in your music signal. The power output therefore is not constant, but rather is constantly varying depending on the input signal being amplified at that precise time.
If the ability of the amplification to immediately respond to transients Is compromised by inadequate supply bypassing your voltage will drop. This can result in clipping of the amplified output signal because it may not have the power it needs to amplify the input by whatever the amount of gain is.
This means when your voltage drops to any significant degree at a transient you will hear this present itself as distortion. Since voltage is directly proportional to the resistance x the current, to prevent voltage drop from occurring we must start by minimizing the resistance variable. In the case of our power supply rail, inductance is the very first factor we must minimize.
How can we minimize inductance in our supply rail?
First we can widen the supply trace as much as possible throughout its path. In addition we can use a thicker supply trace which equates to using a heavier copper weight. We may also opt to forgoe solder mask in certain areas of the supply trace and plate it like we would a components pad, so we can flood that plating with solder thereby making it even thicker. Minimizing the length of the supply trace as much as possible is also beneficial. You can think of it in terms of wires. Obviously using a lower gauge(thicker) wire decreases inductance, as does shortening said wire length. It is the same for the supply trace on your board.
Optimizing capacitor placement.
The capacitor(s) you use should be as close as physically possible to the amplifier without sacrificing other pertinent layout practices. Creativity in the layout is crucial to achieving the best performance. For example, on our own designed amplifier for https://givemebass.com/product/the-only/ we have mounted a polymer aluminum capacitor on the backside of the board opposed to the top side. Doing so allowed us to have the capacitor closer to the amp than we otherwise would have been able to. This is because mounting on the top side would have been impractical since the surface area which the aluminum can extends out to beyond the leads would be hitting the even closer ceramic capacitors we have laid out in parallel to the amp. Yes-mounting on the back of the board takes into account the distance through thickness of the PCB to the top layer-1.5mm in this case. However the distance we would have to move it back away from the ceramic capacitors so that it could sit flush mounted on the top of the board would exceed this 1.5mm. Therefore with this technique we have increased decoupling performance.
Additionally if adequate capacitance and voltage rating is available in a smaller case size then using the smallest case size possible could end up being very beneficial for placing the capacitor closer to the amp. Since there will usually be multiple decoupling capacitors used for high quality decoupling, this is even more paramount for the final one(s) in the chain(the ones closest to the amp). There are some other reasons it can be beneficial to use smaller cases sizes which will be addressed in the forthcoming sections.
Minimizing noise and impedance by using multiple capacitors in parallel
Most high quality amplifiers have a fairly high PSRR (power supply rejection ratio) which is the ability for the device to reject ripple on the power supply rail. However, it will always be true that having a cleaner voltage with less ripple is more beneficial. To maximize this reduction we must use a certain amount of high quality capacitors.
A capacitor will have an ability to filter noise predominantly at a certain frequency range and this being determined based on many factors such as the dielectric type, case size, and capacitance value. All capacitors inherently have a degree of ESR(equivalent series resistance) and ESL(equivalent series inductance) due to the construction. In a perfect capacitor such factors would not exist, however there are no perfect capacitors, so for purposes of illustration can be thought of as a capacitor in series with a resistor and an inductor or a series RLC circuit.
Ideally what you want to be able to plot is a capacitors impedance over a frequency range. One important thing to know is the self resonant frequency of the capacitor. It is at this frequency that the capacitor will exhibit its minimum impedance.
f = 1/2π√(LC)
While this formula is great to know we can break it down into some simple statements that are true. Impedance drops due to capacitance and increases due to Inductance or the ESL. The higher the capacitance value, the lower the impedance at a lower frequency range. The lower the capacitance value, the lower the impedance at a higher frequency range.
Since we want the least impedance over the broadest frequency range we can parallel multiple capacitors. We may use something such as a bulk aluminum electrolytic or polymer capacitor that’s 1,000uF followed with a few ceramic capacitors of varying values. Ultimately the smallest case size possible would be best, but for the sake of having multiple values to get a further broadened impedance drop over frequencies we may end up needing to use a higher case size in 1 or 2 of them to get the capacitance or voltage specs we need for our design. We may for instance use a 1206 22uf, followed with 0201 0.1uf, 100nf, and 10nf placed as close as possible to the IC.
The bulk capacitor will impede low at low frequencies , the 22uf ceramic will do good at medium frequencies and the 0.1uF to 10nF will help with higher frequencies. The end result would be an overall impedance curve that is kept lower over a wider frequency range than what was possible with fewer capacitors.
Often times when we are looking at the data sheet of an electrolytic or polymer we have graphs that show a typical impedance over frequency curve which makes thing easy. This is rare with ceramics however.
One thing to note with the case of ceramics is that the voltage rating will also come into play here since they have a very high voltage coefficient meaning they lose some of their rated capacitance based on their applied voltage and how close that applied voltage is to their rating. All ratings equal, a larger case size will also have less voltage drop. While this should be understood and kept in mind it is not super critical to take into account here unless you are really shooting for precision.
Minimizing impedance by capacitor type
The type of capacitor used for this purpose is perhaps more important than any of the other previously mentioned factors since the wrong type of capacitor can greatly increase impedance and be very detrimental to all the benefits we are trying to achieve in the layout.
The lowest ESR capacitors will be ceramics. They will also have the lowest ESL and further decrease in ESL as you go down in case size. Some manufactures also have special ceramics where by they further minimize the ESL. The problem is they are only available in marginal capacitances. This is okay however because it makes them great for filtering mid and high frequencies supply noise.
Next up will be polymer aluminum. These vary a lot in their ESR values based on their other values such as voltage, capacitance, etc. Usually it is best to choose ones with the lowest ESR values that are also within your other desired/needed specs. These are available in higher values than ceramics but still may not be enough depending on the magnitude of your design.
Aluminum electrolytic are typically but not always higher ESR than polymer aluminum type. I recommend comparing and contrasting between the best polymer aluminum and aluminum electrolytic you can find in the other values that you need and choose the one that has lower ESR. Benefits with the aluminum electrolytic are that they are available in much higher capacitances and voltage ratings.
You will want to avoid tantalum like the plague. These are very high ESR and will cause massive voltage drop on your supply rail if used for this purpose. I remember when I was testing an “off the shelf” 5v amplifier years ago that had horrible distortion at high volume. I measured the voltage drop at the amps VCC pins and it was dipping from 5v all the way down to about 4.5v when a transient hit. They had a tantalum on the supply rail that was something like 800mohms ESR at 100hz. What were the designers thinking!?
-All of these capacitors on the supply rail must be connected directly to a solid ground plane to avoid additional series parasitics. Using via stitching to connect multiple ground layers together, if applicable in your design, can be beneficial for assuring an excellent ground connection.
-the smallest value and case size capacitor is always the one needed to be closest to supply pins of IC.
-If the supply trace is wide and the capacitor case size is small to where it is only covering a small amount of the width of the supply rail, you may consider adding the decoupling capacitors on each side of the supply rail. This would provide more total surface area being covered on the supply rail by the one side of the capacitor. Another benefit is the ability to get more capacitors overall closer to the amp. These are are also all techniques we did in our amp design for https://givemebass.com/product/the-only/
Decoupling/bypassing are extremely important aspects that go in to high quality audio design but are ones which often don’t get the attention to detail and care that they deserve. If you want to have extremely high quality audio performance you must not compromise this aspect of your design in any way. Even highly experienced engineers may not get things ideal on the first revision and therefore a design should have adequate testing and measurements taken along with additional steps to improve it as necessary.