Base 7 equal noise levels…
To write that we have something mapped out to the last detail is perhaps too bold, but after proper preparation, few pieces of hardware are as easy to evaluate as fans. Of course, this had to be preceded by long preparations, developing a methodology, but you already know the story. What you don’t know yet is the first fruit, or rather the results of Akasa, SilentiumPC, SilverStone, Xigmatek or more exotic Reeven fans.
Base 7 equal noise levels
There are several options by which to normalize the test modes for fans. In the previous chapter, we wrote that perhaps the least appropriate option is equal speed.
Settings according to the same static pressure or flow are for consideration, but we find it most sensible in the long term to normalize the measurement modes according to the same noise levels. Firstly because decibels are a logarithmic unit and all others scale linearly, but mainly because you can orientate fastest by the same noise levels. The easiest way to compare the efficiency of fans is just by how they perform at the same sound pressure level. Of all the options, this is the one that most people can best imagine and bounce off of when considering other variables.
The individual noise level modes are adjusted from low levels continuously to higher levels. All users will find their results in the tests, regardless of whether they prefer very quiet operation at the limit of audibility or whether high performance is paramount.
The quietest mode corresponds to 31.5 dBA, followed by 33 dBA, and for each additional mode we add 3 dBA, which always doubles the noise level (36, 39, 42 and 45 dBA). Finally, we measure the fans at maximum power. Here, each one already has a slightly different noise level, which we also report. If there are missing measurements between the results for any of the fans, this means that it was not possible to set the target noise level. Either because its minimum speed exceeds the quietest mode of 31.5 dBA or vice versa because the fan is quieter than 45 dBA at maximum power.
It is important to add that our noise level measurements are incomparable to the values quoted by the fan manufacturers in their specifications. One of the reasons is because we use a parabola-shaped collar around the sensor of the noise meter, which increases sensitivity. This is important in order to distinguish and set to the same noise level even modes at very low speeds, especially 31.5 dBA.
The noise meter next to the fan is quite close for sufficient resolution. The distance between the frame and the sensor is 15 centimeters. The sensor is positioned in such a way that there is no distortion or that the noise level measurements are not affected by airflow. Therefore, the noise meter is centered perpendicularly to the frame that defines the depth of the fan. Everything is always at the same angle and at the same distance. We use an inclinometer and markers to set the distances precisely and always the same.
We use a Reed R8080 noise meter to measure noise levels. This allows real-time averaging of samples, which is important for fine-tuning individual modes. We tune the fans until the specified noise level is reached to two decimal places, for example 31.50 dBA. The noise meter is the only instrument we calibrate inside our testlab. The other instruments have been calibrated by the relevant technical institutes. However, in the case of the noise meter, calibration is required before each test and we therefore have our own calibrator. This is already calibrated externally according to the standard.
- Contents
- Cooler Master MasterFan SF120M in detail
- The basis of the methodology, the wind tunnel
- Mounting and vibration measurement
- Initial warm-up and speed recording
- Base 7 equal noise levels…
- .. and sound color (frequency characteristic)
- Static pressure measurement…
- … and airflow
- Everything changes with obstacles
- How we measure power draw and motor power
- Measuring the intensity (and power draw) of lighting
- Results: Speed
- Results: Airlow w/o obstacles
- Results: Airflow through a nylon filter
- Results: Airflow through a plastic filter
- Results: Airflow through a hexagonal grille
- Results: Airflow through a thinner radiator
- Results: Airflow through a thicker radiator
- Results: Static pressure w/o obstacles
- Results: Static pressure through a nylon filter
- Results: Static pressure through a plastic filter
- Results: Static pressure through a hexagonal grille
- Results: Static pressure through a thinner radiator
- Results: Static pressure through a thicker radiator
- Results: Static pressure, efficiency by orientation
- Reality vs. specifications
- Results: Frequency response of sound w/o obstacles
- Results: Frequency response of sound with a dust filter
- Results: Frequency response of sound with a hexagonal grille
- Results: Frequency response of sound with a radiator
- Results: Vibration, in total (3D vector length)
- Results: Vibration, X-axis
- Results: Vibration, Y-axis
- Results: Vibration, Z-axis
- Results: Power draw (and motor power)
- Results: Cooling performance per watt, airflow
- Results: Cooling performance per watt, static pressure
- Airflow per euro
- Static pressure per euro
- Results: Lighting – LED luminance and power draw
- Results: LED to motor power draw ratio
- Evaluation
It looks like the frequency response charts do not show.
Thanks for the heads up, it’s now corrected.
In some of the older articles in English, imperfect processing techniques combined with a lack of focus resulted in incorrect paths to the spectrograms being left in the source code. If you should come across something like that again, the original language version (there are graphs with EN descriptions anyway, you can get to it by clicking the flag in the upper right corner of the page) is always correct. Alternatively, the spectrograms of the individual fans can be accessed by entering the URL according to the format “name-of-the-fan-g***”, where the *** a number from the interval 233 to 244. Of course, this is only a temporary solution until we fix it (after you reporting it). But otherwise it must always work without such complications. 🙂
Thanks, I will keep it in mind if I encounter such an issue in the future and report it so that everyone can have it fixed.
I wanted to see the frequencies because of Noctua NF-A12x25 which has a highly disturbing peak right before 400 Hz. The frequency response of SF120M looks so much better at “33dB” in comparison. That’s why it would be great to have at least 1 sound sample per fan at 33dB, to judge the noise according to personal preference.
This is true, but the reason for this is that the SF120M has an overall “dropped” aerodynamic noise spectrum in this mode due to the higher tonal peaks of the motor and bearings. This is also why it achieves a lower airflow. If the test modes were aligned for equal airflow (i.e., the SF120M would have higher RPM), I expect that the NF-A12x25 might not be noisier even at 380 Hz.