Pedals obviously need to be powered and we are all familiar with the “standard” 9v arrangement associated with Boss pedals, and using a centre-negative 5.5mm/2.1mm barrel jack. Why 9v? Peak guitar signals are of the order of a volt so 9v provides enough headroom for amplifying and shaping the signal, and it is also the voltage of the ubiquitous PP3 battery. Note that 9v is not overly generous – almost any pedal that is not built solely from discrete transistors will use op amps, even if just to buffer inputs and outputs. Only specialised op amps get close to “rail to rail” operation, so 9v supply probably equates, at best, to a 7v total swing, or an AC signal of 3.5v peak to peak. Today it is much more likely that your pedals will be powered by a mains adapter than a battery – in many ways, batteries are a much better power source as they have low internal resistance and no potential for coupling in noise or hum, but they can run out at inconvenient times. As a consequence, I do not build my own pedals with battery support.
The challenge with any single rail supply for audio use is the fact that the AC signal needs to be referenced to half the supply voltage in order for the circuit to be able to swing positive and negative with equal headroom relative to this signal reference. This is termed “rail splitting” and can be thought of as creating a virtual ground that then makes the “real” ground look like a -4.5v rail and the +9v power input look like a +4.5v rail. In some circuits, there is no need for anything complex and a simple resistive divider will do perfectly. Consider the trivial buffer circuit below:
In this case, R1 and R2 form a simple divider its only purpose is to set the DC reference for the signal coupled through C1 to 4.5v through R4. R4 can be made arbitrarily large and the value of 1M ensures that the divider does not load the input and that the signal has negligible impact on the stability of the reference voltage. This latter point has no significance in this circuit as the virtual ground is only used once at the input but in slightly more complex circuits this won’t always be the case, but if DC bias is only being applied by suitably large resistors then this arrangement may still work well. Decoupling can also be applied to the voltage divider to ensure it remains rock steady.
Things start to get a little more tricky when the load on the virtual ground is more taxing than a couple of 1M resistors. In this case, we need to think a little more carefully about creating a virtual ground that can source and sink current while remaining completely stable. The resistive divider will still be needed to create the reference voltage but we need a way of buffering its output. Perhaps the most common approach is simply to use a “spare” op amp as a buffer thus:
This works well in practice, doesn’t take up much space and is cheap – potentially free if you do have a “spare” op amp. Note that it is important for any decoupling (to reduce any chance of noise – C1 in the above circuit) to be applied before the op amp – most amplifiers don’t like driving heavily capacitive loads and can oscillate if you decouple the virtual ground directly. Other options do exist, however, which might be attractive when you don’t have a “spare” op amp in your design.
Given that what we actually need after the resistive divider is a unity gain buffer rather than an amplifier (in the above circuit we are simply configuring the op amp to be a unity gain buffer by applying 100% negative feedback), the obvious answer is to use a unity gain buffer such as the BUF634. The BUF634 can source and sink up to 250mA and is open loop so will never oscillate – but it’s expensive and a pedal circuit should never require anything like that amount of current to be needed for a virtual earth. If your priority is to reduce cost then a discrete option can be used that only requires a pair of cheap, general purpose transistors. The simplest place to start when looking for a unity gain buffer is the humble emitter follower as shown on the left below (I have used this in the past without a problem). The voltage on the emitter of a bipolar transistor will always be one diode drop below the voltage on the base, so we add another diode into the resistive divider to compensate. The problem with this circuit is that the capacity to source current is determined by the transistor but the ability to sink current is determined by the emitter load resistor. To create a buffer with a symmetrical current capability we need to create something more akin to class AB output stage – the circuit on the right is by Sijosae.
The similarity between the two circuits should be apparent – the circuit on the right has essentially replaced the emitter follower load resistor of the circuit on the left with a complimentary PNP emitter follower that can sink current. The two diodes in combination with the emitter degeneration of R6 and R7 in the Sijosae circuit can also be seen as acting like the quiescent bias of a class AB output stage. The Sijosae circuit is cheap but does take up a bit more space on the board than other options – it is also the least accurate but, for guitar pedals, all we need is to be somewhere close to 4.5v and stable.
A final option to consider in the virtual ground department is the Texas Instruments “Rail Splitter” TLE2426. This provides a complete solution in a TO92 package and is reasonably priced from the right source (e.g. Farnell rather than eBay). The one point to watch is that the unity gain buffer in the TLE2426 is not open loop – that means that it is sensitive to capacitive loading. Fortunately, we can see from the datasheet (figure 17 on page 17) that it is unconditionally stable for load capacitances above 1uF (or below 820pF) so adding some electrolytic decoupling to the output will do the trick.
Note that Tangentsoft have an excellent discussion of this topic on their site, albeit more geared towards headphone amplifiers (with more demanding current needs) than guitar pedals. John Broskie’s blog also has another good article that covers a broad range of rail splitting scenarios.
So far, we have remained within the confines of the 9v handed down to us by Boss and battery manufacturers, but there is another avenue to explore when powering pedals which is particularly useful for creating clones of early Germanium fuzz circuits – using a power supply inverter to create a -9v rail from a +9v supply. The highly renowned Klon Centaur (of which the Local Planning Office is a clone) uses the ICL7660S (pin compatible with the more modern MAX1044) to create not only a -9v rail but also a +18v rail, thereby providing a total supply voltage of 27v. To put the icing on the cake, the Klon Centaur also combines this with a +4.5v virtual ground so that the supply behaves as though it is delivering symmetrical +/-13.5v rails.
The 7660S/MAX1044 can also be cascaded so a pair can provide -18v. This is useful if you want to use a regulator on the -ve rail as the regulator will need to drop a few volts and is a technique I have used for the aforementioned Germanium fuzz circuits because of their susceptibility to noise.
The 7660S/MAX1044 does have its limitations, however. Firstly, it is not rated beyond 10V (in terms of the voltage across its supply pins) and secondly, because switched capacitor circuits don’t have great current drive. Another option that can be used for creating larger positive supply rails is the boost converter, which utilises both an inductor and a capacitor for better filtering and current handling. There are many small ICs in SMT packages that only require a few external components to provide well-regulated outputs up to 36V @ 500mA. They typically operate in the MHz range and so only require inductors of around 10uH. Even with the external components, they occupy less PCB space than a comparable switched-capacitor circuit, but they cannot produce negative rails from a positive supply.