c y b e n e t i c s

LAMBDA (Power Supplies)

Along with efficiency, we are also able to calculate a PSU's overall output noise, by averaging the noise measurements that we take from our test results.

The noise data averaging process isn’t straightforward since decibels (dB) are based on logarithmic scales. For example, if you have a PSU producing 40 dB, a doubling of the acoustic power sound intensity would not be 80 dB but 43 dB (and 50 dB for the volume loudness). This means that we must convert first the dB values to sound pressure units (Pa) before we average them, and then convert the result to dB again, to make it familiar to most of you. We call our noise measurements program Lambda, deriving from the Greek letter “λ”. The same Greek letter represents wavelength, which is a fundamental property of sound waves.

The same conditions with efficiency testing also apply to our noise measurements. 
                                                                                       
Noise Levels (115V/230V Input) Noise Requirements (1)
A++ <15 dB(A)
A+ ≥15 dB(A) & <20 dB(A)
A ≥20 dB(A) & <25 dB(A)
A- ≥25 dB(A) & <30 dB(A)
Standard ++ ≥30 dB(A) & <35 dB(A)
Standard + ≥35 dB(A) & <40 dB(A)
Standard ≥40 dB(A) & <45 dB(A)

(1) In case the DUT is close to the lower limits, with a maximum margin of 0.25 dB(A), we will provide a second chance to the manufacturer, who will have to provide us with another sample. In case the second sample registers a significant overall difference compared to the first one [>2 dB(A)], we will ask for a third one, and the average noise measurements of all three samples will deliver the final result. 

Beyond the average: load ranges and the High-Load Index

LAMBDA is one number. The evaluation page of each desktop PSU also shows where that noise comes from.

The LAMBDA figure is the energy average of the noise over all the measured load combinations of the certification sweep. It is what the grade is based on and it does not change. A single average can hide how a power supply behaves at different loads, so each desktop PSU evaluation page also shows, where its measured data allows and for information only:

  • Load ranges. Five ranges of the rated output (0-<20, 20-<40, 40-<60, 60-<80 and 80-100 %). For each range: the average noise, calculated the same way as LAMBDA, the loudest point, and the share of load points at which the fan is stopped (semi-passive operation). The efficiency and the power factor are shown for the same ranges.
  • High-Load Index. The noise of all measured load combinations at 70 % of the rated output and above, averaged the same way. It tells how loud the unit is when it is really worked.

These figures are calculated from the measured load combinations of the certification sweep, and a figure is shown only when its five ranges add up to the average of the whole sweep. They are informational: they never change the ETA or LAMBDA grade, the score or any published average.

Redundant server PSUs have the efficiency and power factor ranges on their evaluation pages too; for them the ranges start at 2 % of the rated output, as in the Redundant Server PSU Test Protocol (Rev 1.3), section "Load-range results (informational)".

Example: almost the same average, different behaviour under load

Two certified desktop units at 230 V input (names withheld). Their average noise is within 0.5 dB(A) of each other, so LAMBDA sees them alike, yet their High-Load Index differs by 4.6 dB(A).

At 230 V inputUnit AUnit B
Rated output650 W1000 W
Average noise (LAMBDA), dB(A)22.222.7
High-Load Index (70-100 % load), dB(A)22.627.2
Loudest point at 70-100 % load, dB(A)29.031.0
Fan stopped, share of load points0 %17 %

The full load-range results of these two units, with efficiency and power factor, are on their evaluation pages (section "Noise by load range"): Unit A · Unit B

COMPARISON
Clear All