Our laboratory conducts measurements all day long on every product you can imagine, from GPUs and CPUs to power supplies and fans. But let’s start by answering some plain questions to get you slowly into today’s hot subject.
What is a measurement?
A measurement is a reading or property of something. For instance, the efficiency of a PSU at some specific conditions is a measurement, as is the 12V rail’s voltage level acquired through a multimeter. You need to use an instrument (e.g., a multimeter, a thermometer, etc.) to get a measurement.
So, a measurement consists of two parts: a number and the unit that describes the measurements. For example, we can have 28 degrees Celsius in temperature, and for power, we say 120W.
Measurement Uncertainty
There is an existing doubt about the result of any measurement. How can I know that the X laboratory provides accurate measurements? How can I be sure that the thermometer I use is trustworthy and provides accurate readings? For any measurement, even those conducted in NASA’s labs, there is always a margin of doubt that it is not 100% accurate. This is called uncertainty.
How can we express the Measurement Uncertainty?
As I already mentioned, there is always a doubt about any measurement, so we have to consider the uncertainty margin—how large is the doubt I have about the provided measurement(s)? Two numbers can show this. These two numbers can quantify uncertainty.
- Interval: Width of margin.
- Confidence level means how sure we are that the actual value is within the interval.
To make it clear, let’s provide an example:
- You have a PSU whose efficiency at 50% load is 90% plus or minus 0.1% at the 95% confidence level. This can be written:
- 90% ±0.1%, at a level of confidence of 95%
This means we are 95% sure that PSU’s efficiency is between 89.9% and 90.1%.
Uncertainty Types
There are two ways to estimate the uncertainty of a measurement:
- Type A uses statistics.
- Type B uses any information.
Have you participated in any inter-laboratory comparison (ILC) or proficiency testing to validate the accuracy of your methods? If not, do you have plans to?
Now your laboratory is accredited, and you have an ISO/IEC 17025 certificate? You might be the first hardware reviewer who is legitimate.
Our lab works under ISO 17025 for quite some time now. We have our system implemented several years ago, based on our knowledge and through some external consultants we hired.
We will have the first inspection in the following months, and we also applied for ISO 17065 which will allow Cybenetics to enter Energy Star. We would obtain 17025 years ago, but we moved the lab 2x times in the last two years. Now that we bought the damn building, we go full speed to it!
To get ISO 17025, you must operate for at least six months under this ISO’s guidelines. We operate for many years under that since this is the way.
This kind of ISO I am afraid is NOT for reviewers because it is too complex, too expensive and requires a crazy tight system. If there wasn’t Cybenetics I would never be able to even think of it. Basically Cybenetics gets it. HWbusters just uses Cybenetics data 🙂
Into this year we will get the cert, we DONT have it yet. We filled the application in December 2024 and we are waiting for the first pre-inspection in the next month (hopefully). My goal is to have both ISOs within 2025 (17025 and 17065)
Wow! Impressive