Superpower = Low Noise

Scale adjusted RMS noise voltage measurement

Total unweighted noise across the audio band of 20Hz to 20KHz measures 12µVrms, an excellent specification. Note that the units show mV because of a gain of 1000 (see below). Due to circuit variations for different output voltages, noise varies for different output voltages. This RMS value is noise measured for a +12V out standard Superpower, which equates to 1PPM of noise per volts out.

A noise spectrum shown below is from a Superpower with 12V out, sampled at 44.1K samples per second, 16384 samples using a Blackman Exact window function. Load is approximately 10mA. Five samples are averaged to smooth the graph.

Superpower noise spectrum (click to see full image)

First note the relatively flat spectrum across the audio band. Second, the spectrum is at about -140dB with respect to 2V full scale. How can this be? First, by the laws of FFT, the total noise is divided across the number of bins, in this case 16384. Thus the noise is less per frequency. Second, because the measured noise is very low, so low that it is much better than the 24 bit ADC used to measure it.

To overcome noise floor limitations of the data capture instrument, a low noise preamp with a gain of 1000 is used to boost the noise to be measured:

Power to the amplifier and the regulator is delivered with batteries to eliminate any possible contamination from AC power supplies.

The test fixture also performs a bandpass function, limiting the measurement to the audio bandwidth of interest:

Bandwidth of noise test fixture

Superpower noise spectrum at 240mA load (click to see full image)

Above is the same noise measurement made with the Superpower delivering a 240mA load. The input supply ripple increases linearly with load current and is 350mVpp at the input. Some of this ripple can be seen at 120Hz and its harmonic components. Notice the supply rejection is better than 104dB below 2V full scale at the fundamental.

With a gain of 60dB due to the test setup, the measured values were all scaled down by 60dB (= divide by 1000), so, e.g. a value at -90dB became -150dB. As a comparison, the spectrum of the fixture alone is measured to see the noise floor:

Spectrum of measurement system

Battery powered with gain = 1000, noise in the measurement system is almost non-existent, seen here in the -170 to -180dB range relative to 2V.