We all know how important voltage rail stability is for the behaviour and audible qualities of an amplifier circuit (or ADC, DAC, or clock generator, for the matter). Consequently, no audio circuit design can be deemed complete if the required attention has not gone to its power supply. Sadly, just as there are numerous amplifier topologies there are a myryad of voltage regulator circuits. But what is lacking in the literate is an objective comparison of them. This is what this article series is about, and I got prompted to start this by three main events.
One: in the past year or so there has been a lot of talk on the excellent DIY forum DIY Audio, a result of which was the study and design of ever more complex, and expensive, regulators by some of the forum members. I wanted to try the same for simple, and cheap, regulators.
Two: I found the time to build a practical breadboard that contains a copy of my regular line preamp, the circuitry of which can be fed from any of five or six voltahe regulator types on the same board, jumper-selectable.
Three: having acquired a decent sound card (actually a USB-connected outboard box) for my laptop, I am now able to do measurements.
A voltage supply should be "perfect", which any electronics textbook will tell you in a none-too-subtle way means that it has zero output impedance and no current limits. This is a far cry from reality, hence the following list of requirements that are desirable in audio power supplies:
I start this series with this article on measured output noise of various regulators (and two battery types!). In future installments we will look at regulator output impedance, and the way this reacts with capacitors on the board, and if time and resources permit, regulator linearity/distortion, and finally, listening tests.
As said, the device under test is a stereo line amp, using opamps in the classical Jung AD744/LM6181 configuration giving 10dB of gain, each channel running off its own +/-12V regs. Up to five different styles of voltage regulator are present, all of them fed from a +/-21V preregulated outboard supply (in fact a Trichord Dino+).
One positive and one negative rail of this preamplifier are permanently connected to a sense preamplifier and ADC, which together make up the measurement system. The sense amplifier (actually an NE5532 rather than the TL072 in the schematic) gives 40dB gain to the voltage rail "signal" so as to lift any noise on it above the system noise of the measurement setup. The ADC is a Terratec Phase 26 USB digital recording system, connected to my PC, and running at 44.1kHz 16 bit, even though it is a 96kHz "24" bit ADC.
Above plot is the system noise with inputs shorted. The noise peak at 10kHz is an artefact of the Terratec ADC which may be related to the fact that it is permanently fed from the laptop's USB bus (I plan to build it its own quality power supply). Initially I was very disappointed by this property of the Phase 26, as quite contrarious to the well-spread Soundblaster-type of soundcards this one ought to be a quality recording ADC fit for use in project studios: its build standard is on a par with for instance Behringer's digital room equalizers. Then again, this noise is generally not audible, and it may act nicely as dither while recording, so I'm cool about it now, thank you.
The '0dB' mark in the spectra corresponds to a level of 2100 mV RMS at the ADC's input, or, taking into account the 40dB gain of the sense amplifier, 21 mV as detected on the supply rails under test. A level of -60dB then corresponds with 21 uV measured noise at a given frequency.
© Copyright 2004 Werner Ogiers for www.tnt-audio.com