How Noctua measures fan noise
Noise measurements are an extremely complex topic. Imagine you have ten different sound sources. The noise from each will radiate through the space in different ways, and measuring ‘which is the noisiest’ and coming up with a clear verdict on that is a really hard nut to crack. To give you a better idea, we’ve interviewed Noctua. Different measurement techniques yield potentially different and sometimes contradictory conclusions.
This is the second interview with a Noctua representative in a relatively short period of time, but an important one, both thematically and in terms of timing. The release of the NF-A14x25 G2 PWM fan is imminent, and different reviewers will measure its noise level differently and come up with different results. By these we mean different relative to other fans as well. That’s because everyone will be sensing the noise from a slightly different angle, in a different way. We talked more about this topic with Jakob (Dellinger), who is always happy to answer any technical details that concern Noctua products.
HWCooling: Hello, Jakob, the opening question will be quite short, but the answer will probably be more dense. How does Noctua measure fan noise for internal purposes and why exactly in that way? I’m referring to the position of the fan and the noise meter relative to each other.
Noctua: Measuring fan noise is a science of its own. There are many challenges with it, but two of the most important ones are that firstly, fan noise has dipole characteristics, which means that it does not spread in a uniform, spherical way, and secondly, that the radiation patterns can vary depending on frequency range, distance, reflections and interactions, especially in real-world environments where you typically have reflective surfaces as well as other sources of noise.
In an ideal world, you would tackle these issues by measuring from an infinite number of points in space that are evenly spread around the fan in all directions and at several distances, then combine these measurements with additional measurements in any possible real-world scenario. For obvious reasons, this is impossible. What we can do is find a reasonable balance regarding the number of measurement points so that is practical to implement in the desired time frame on the one hand and but still gives us sufficient resolution on the three-dimensional radiation patterns on the other hand.
However, doing this full-scale spatial analysis is not always possible or even necessary. The reason for this is that due to the dipole characteristics, the highest noise intensity is in axial direction, so you can capture the most critical aspect by measuring from the inlet side of the fan in axial direction. Measuring from the outlet side is generally not recommended because flow turbulence will be dominant. Other directions such as from the side, perpendicular to the axis of the fan, are problematic as well because these measurements will fail to capture the maximum noise intensity and are generally more prone to variations due to frequency-dependence, reflections and interactions as well as other distance specific phenomena.
HWCooling: Thank you for a detailed and at the same time very clear introduction to the issue.
Isn’t there some influence of the airflow on the measured noise values even if the noise meter is placed in the axial direction in front of the fan in the way you prefer? After highlighting the air currents (e.g. by smoke) we know that they are also in front of the fan, albeit with a smaller vector. I assume my concerns are unfounded, as the noise meter will be sufficiently distant when measurements are taken that there can be no distortion at this level?
Noctua: Yes, this is correct, the distance has to be sufficient for intake turbulence not to be a problem, but unless you are dealing with extremely high airflow performance fans or high ambient noise levels, this shouldn’t be an issue.
HWCooling: So if frontal sensing is chosen to capture as much noise as possible, doesn’t the resolution decrease by requiring a greater distance than would be sufficient when sensing from the side? For example, a very faint motor noise could already be below the resolution from a meter away when measured frontally, whereas from the side, from some 20 centimeters away, the noise meter would still pick it up.
Noctua: For this, it’s important to note that different noise components can have different directivity patterns, so whereas the noise caused by blade passing has dipole characteristics, other components of fan noise such as potential motor or bearing noises often have different radiation characteristics. This depends on where and how these noises occur, their frequency ranges as well as how they are transmitted or dampened by the structures of the fan.
Generally, it is not uncommon for these side noises to be easier to pick up from the side or even from the outlet side of the fan, so when analysing specific phenomena like these, it can indeed make sense to diverge from measuring from the front as long as you keep in mind that this is a means to zoom into specific phenomena rather than trying to get the big picture where blade passing noise is usually the key component. Once you know the frequency range of the phenomenon you’re analysing, another way of approaching this is to isolate this range in measurements taken from the front.
HWCooling: Okay, let’s stay with the aerodynamic noise measurement and take the specific situation of your two fans – models NF-A14 PWM and NF-A14x25(r) G2 PWM. Is it possible to outline what ratios, in terms of noise measurements, we could arrive at with different measurement techniques? What I mean now is from a sufficient distance (ensuring that the measurements are not distorted by airflow) from the front, then from the side, then for example after shortening the distance (typically we always talk about 1 m, do you also maintain such a distance in your internal analyses?). In short, we would like to make it easier for our readers to understand how different approaches to noise measurement can affect the resulting values (of noise levels).
Noctua: For these two specific models, their sound pressure level is very similar measuring from the front, but if you look at the radiation patterns, the G2 model looks broadly elliptical whereas the G1’s pattern has a more figure of 8 like shape with notches of lower noise emission right at the sides. This means that when you measure from the side, the G1 model may appear 1-3dB quieter because it does not have these distinct notches. To put this into perspective though, the G2 model is still 4-5dB(A) quieter measured from the side than from the front, so as I’ve explained, this difference is of limited relevance for the total noise emission of the fan.
HWCooling: Different directivity patterns of the noise… Is it possible to make some comparison with your fans? Between which models of the same format is there the biggest “difference”, if it is possible to discuss in this way, simply?
Noctua: Variations of 1-3dB between different models of the same size are not unusual, especially if there is a significant jump in performance like between NF-A14 and NF-A14x25 G2. This is due to the fact that the higher pressure fluctuations caused by higher performance blade geometries typically lead to increased noise radiation towards the sides. This is hard to avoid when transitioning towards blade geometries that can run an increased total aerodynamic load. By contrast, radiation towards the inlet and outlet side is somewhat easier to control if you manage to reduce flow separation and turbulence phenomena.
HWCooling: Are the numerical differences in noise that you provide for the NF-A14 and NF-A14x25 G2 fans based on measurements from a distance of one meter? It would seem that shortening the distance might reduce these differences. From the side (on a radiator, we don’t have results without obstacles), but from a significantly shorter distance, we have measured virtually comparable noise levels for the NF-A14 and NF-A14x25r G2.
If it weren’t for the fact that we use a parabolic collar around the sensor to increase resolution, the measured values would almost be blending together. Have you also dealt internally with measurements of aerodynamic noise from shorter distances or does this not make too much sense to you and you stick to the standards?
Noctua: Yes, these are from 1m. It’s good to know that you’ve measured similar noise levels from the side, this illustrates nicely how measurements from this direction can vary from setup to setup. Your parabolic collar may also play a role in this as it will let you capture a broader angular spectrum rather than straight perpendicular only, but as I’ve said in the beginning, we generally see more variation when measuring from the side due to factors such as frequency dependence, reflections or interactions. I suppose that one of these factors or a combination for them is why you’re seeing similar levels from the side.
As for your last question, we do measure from various distances not only due to different measurement requirements but also to make sure that we don’t fall prey to any of the factors mentioned above. While there are much, much less reflections when measuring in anechoic chambers, it’s never completely zero, and both interactions and frequency dependent radiation phenomena can occur, so it always makes sense to double check your data from different distances to be on the safe side. We also vary measuring distances depending on what we’re analysing, e.g. for some very low volume stuff or when tuning commutation acoustics, getting closer is vital to get a good picture.
HWCooling: Thank you for your comprehensive answer. I think we are coming to the end – the purpose of this interview is fulfilled. Readers should have gotten an idea of why when measurements are taken from different angles, there may be different results recorded to work with when evaluating fans. Perhaps you can still add something thematic that we didn’t ask about, but which might be important for completeness. If you have something like that, then here’s your chance. I think that’s about it from the HWCooling editorial team. It’s possible that some more questions will come up in the comments. I would then confront you with those and they would already be answered in the discussion below this article.
Noctua: We spoke a lot about what a complex phenomenon fan noise is and how something seemingly simple like establishing whether fan A or fan B is louder can sometimes be quite difficult due to directivity and different radiation patterns. However, there’s something we didn’t really go into yet and I think just briefly mentioning this could be a nice way of concluding our discussion: SPL or dB(A) measurements alone are a challenging endeavour, but this is really just the tip of the iceberg. While the A weighting has been created to better reflect the sensibilities of human hearing, even weighted sound pressure levels will never give us the full picture.
It’s only when we start analysing frequency spectrums, psychoacoustic criteria and modulation phenomena such as beat frequencies that we can get a bit closer to be able to determine how good or bad a fan or a cooler actually sounds. Ultimately, these dimensions multiply the complexities we’ve discussed because ideally, all these factors would have to be investigated for each measurement you take from various directions and distances. In short, we have only scratched the surface!
Read more: Noctua on NF-A14x25 G2 frame deformations (interview)
English translation and edit by Jozef Dudáš
- Contents
- How Noctua measures fan noise
