General Questions

What is a good power supply design?

As much as we'd like to tell you exactly what components to use, you need to choose them to meet your needs. Small size, low current? High current? High voltage? There are so many variables that we can't begin to tell you how to design your ultimate power supply. We do, however, provide some help on our transformer calculator page and in our data sheet, please consult those, and our tips page for information on keeping the regulator cool and preventing problems before they happen.

This is a power supply as we recommend

These are the components:

T1: Power transformer

BR: Bridge rectifier

D1-2: Reverse current protection diodes

C1: Filter capacitor

C2: Input compensation capacitor

SPX: Voltage regulator

R1: Pull down resistor (optional to improve performance)

C3: Regulator output capacitor

Z1: Load circuit (your DAC or mixer or whatever)

The function of these components

T1 converts 120V or 240V mains voltage to lower voltage.
BR converts AC to pulsating DC.
C1 filters the pulsating DC into much more stable "raw" DC, although under load it still contains substantial ripple voltage. If C1 is too small with respect to the load current and raw DC voltage, the ripple can dip below a value required by the regulator, causing it to momentarily stop regulating. This is called "drop out" and the voltage at which it occurs is called the "drop out voltage." There is no maximum value for this capacitor.
C2 prevents low level high frequency oscillation (~20mV at 2.5MHz) due to SPX high input impedance
D1,D2 are protection diodes. In normal operation it is reverse biased and does nothing. When input mains power is switched off, it gives a path for charge stored in the output capacitors to dissipate around the regulators rather than through them.
R1 is an optional fixed resistor that gives transient charge a place to go. Think of it as a "pull down" resistor (a negative regulator will have a "pull up" resistor.) An unipolar regulator is great for sourcing current but not as good at sinking it (the load does that), and the pull down resistor helps by draining charge at the regulator end of the connection.
Our measurements show, for a 100mA current step, a value that sinks about 20mA works well. Determining the optimum resistor is difficult because it depends not on the maximum average load current, but on the maximum change from average. However, a rule of thumb is a fixed load in the range of 10% to 20% of the average load.
C3 is the output capacitance. Electrically this capacitor improves loop stability of a Superpower and, when optimized for the load, allows the fastest possible transient response. Superpower has a 4.7uF tantalum already on the PCB and you can add whatever you wish. The optimum value depends on the regulator (SPM, SPJ, SPHP, etc.) and the load. Our recommendation is a minimum of 10uF at the load, preferably a low inductance capacitor such as tantalum, and we've also seen good results with 100µF Nichicon UVR1H101MPD electrolytic. Again there is no maximum and you can freely experiment with various brands and technologies. One word of caution—always power the circuit down before changing output capacitor, we've damaged SPJ in our lab by connecting a large (e.g. 10 000µF) capacitor while power was applied.
Z1 is the circuit that gets power from the regulator

Dual power supply example schematic

(Pull up/down resistors not shown)

Capacitors—where, how much, what kind?

We get a lot of questions about capacitors. One goal at Belleson is to have the optimum transient response, that is fast recovery from transients and minimum glitch energy, overshoot and ringing. An evaluation of output capacitors determined that neither total capacitance nor ESR have a dominant effect on transient performance. It's not clear yet but it seems the big difference is in the self-inductance of the capacitor. Tantalum caps have low internal inductance and give great transient response. Also a capacitor we found to give great performance is 100µF Nichicon UVR1H101MPD.

Where?

Input: Generally for a linear rectified source supply, we recommend as much filter capacitance as will fit, although there is a point of diminishing returns. See this page for an in-depth discussion of minimum filter capacitance (and a lot more).

Always put a local 0.22µF film bypass cap near the Superpower, from input to ground to prevent high frequency (>1MHz) oscillation at high current (>2A).

One Superpower user has made these specific observations about an input capacitor:

Anyway, I've now soldered a cap directly on the input of superpower. It didn't really improve the output noise, but the interesting thing to note is that it does alter the sound. I wasn't expecting this to happen, given superpower's very high noise rejection ratio. So far, I found that it really does take a good cap there, an audio cap preferably. I've tried Panasonic FM low ESR, Sanyo Oscon SP low ESR. The Sanyo sounded really bad. The Panasonic FM sounded almost as good as the Panasonic FC caps I've already got there for the filter caps. A cheap 330µF Rubycon cap also sounded bad (less ambient, more flat and dry). I ended up with a 1000µF Elna Silmic II cap which I have in my parts bin. It gives better detail, imaging, and it does make it sound special. So you may want to put in the application note that if user wants to further improve the sonic quality, an audio cap such as Elna Silmic / Silmic II is recommended.

Output: Given that Superpower have a tantalum cap on board (except SPM), the best response is obtained by adding additional capacitance at the load rather than at the regulator.

How much?

Superpower are compensated with enough output capacitance for stability under any load condition. You may add as much or as little as you wish. Best dynamic response is achieved with a low inductance 100µF capacitor near the load. Capacitors distributed across a PCB, including decoupling caps near digital components are also good. If you are replacing regulators on a commercial PCB, this will likely be the case already, and Superpower works well in these applications.

What kind?

Feel free to experiment with any brand or type of capacitor. As always, we do not recommend anything that will "sound better", we only measure the electrical performance and you decide what you like.

Should I use a pre-regulator?

Superpower works best without a pre-regulator. The best source of energy for a Superpower is a full wave rectified and filtered power supply. A pre-regulator increases source impedance and limits dynamic current to the Superpower.

Why is Superpower used in audio circuits?

Clean, clear audio needs reserve power for bass and low midrange frequencies, fast response for mid and high frequencies and low noise across the audio spectrum and beyond. These requirements are not independent but must be met simultaneously.

The step responses shown in the load regulation page illustrates how Superpower can provide a great deal of current in a very short time. The spectra in the line regulation page show great noise performance across the audio spectrum.

So no matter what your audio requirements may be, a Superpower will contribute to clean, clear dynamics in any audio subsection.

What can affect dynamic performance measurements?

Delivery of high current means a small amount of resistance can cause relatively large voltage changes. One amp times 10 milliOhms is 10 mV, a housand times more voltage than the noise at the output of a Superpower. To achieve the measurements as seen in the graphs and photos shown on this web site, extreme care must be taken to minimize wire and PC trace lengths, maximize wire gauge and carefully place measurement probes.

How is dynamic performance measured?

Good question. A standard power supply has limitations on output current that affect dynamic performance of Superpower, so what can be used for a Vin source? The only solution we found so far is to use a Superpower regulated source or an unregulated tranny/rectifier/filter cap as Vin to test another Superpower. We use a Superpower regulated supply to test both dynamic performance and ripple rejection.

Dynamic performance is tested by switching a resistor from Vout to ground through a high speed transistor.

Ripple rejection testing is described on the ripple rejection page.

Does Superpower have an output current limit?

To provide minimum output impedance and maximum current, our regulators have no current limit and no thermal protection. Careful design and heat sinking must be used to prevent damage to the Superpower and your circuit.

See the protection circuit here for a way to shut down SPHP when current exceeds the maximum.

Can noise performance be improved?

Good input bypassing as described above decreases noise, as does a large electrolytic capacitor across the output. Some portion of the noise comes from current through Rset, the resistor that sets regulator output voltage. The higher is Vout, the higher is output noise (thus the reason noise is given as µV/Vout).

All our regulators have a 0.1µF noise suppression capacitor in parallel with Rset to reduce reference generated noise. This cap also increases start-up time because it is charged with a fixed current. Its value of 0.1µF is what we consider an optimum compromise for best start-up time and lowest noise.

For SPX, if your application can tolerate a longer start-up time, you can add a capacitor from the left-most SPX pin to GND to reduce noise. We recommend a maximum value of 4.7µF, which gives a start-up time of about 250msec.

For SPHV, Rset can reach high values, for example 180kΩ. Using a large capacitance in parallel with Rset and slowing the start-up voltage for B+ can be advantageous for extending the life of vacuum tubes (valves). The trade-off in this case is, the slower Vout rises, the longer there is high voltage across the regulator. This can thermally stress the SPHV output transistor depending on the load current, so do a simulation of regulator power dissipation for your specific circuit to find an optimum Rset capacitor value.

SPHP and SPLV do not have this node available.

Editorial on the noise about noise

Audiophiles are all over voltage regulator noise. However, it's way down the list of what's important for a voltage regulator in an audio circuit. Consider:

Most audio systems are differential in the critical input stage, which nulls power supply noise.
Noise matters most in the very first stage of a preamp or DAC
Other power supply parameters such as ripple rejection and transient load response can result in milliVolt (or tens to hundreds of mV in some poor regulators) voltage change whereas noise is typically microVolts.

The most important power supply parameters are, in order,

transient step response, so power supply does not change due to digital pulse load of the clock
ripple rejection, to prevent input supply noise from affecting output voltage
noise

If regulator output changes by milliVolts due to load current spikes or input ripple, less microVolts noise will not help.

Voltage regulator noise is important for

Single ended preamp input stage such as SET triode
Phono preamp input stage, even differential, because gain is so high
Microphone preamp
DAC current to voltage (I to V) converter
Direct heated triode filament supply

Nonetheless, we have gone to considerable design effort to make our regulators very quiet! Here is a video of a noise comparison measurement of SPZ78@12V and LM7812: