The Importance of Prominence Ratio in Acoustics

In the field of acoustics, prominence ratio is a crucial concept that helps us understand how annoying a given sound can be to our ears. The prominence ratio looks at the critical band, which is the range of frequencies that are perceived by the human ear as distinct from one another. By analyzing the critical bands and their averages, we can determine if a sound is going to be noticed or not.

Prominence Ratio vs. Tone Noise

There are two methods that use prominence ratio to assess sound annoyance: prominenc ratio and tone noise. Both methods have their individual applications, but they share similar goals. Prominence ratio looks at the tone and then analyzes the critical band to determine if a sound is going to be annoying. Tone noise, on the other hand, focuses solely on the tone of the sound. While both methods can detect annoyance, prominence ratio may be more effective in certain situations.

The Role of Critical Band in Acoustics

Critical band is a fundamental concept in acoustics that refers to the range of frequencies that are perceived by the human ear as distinct from one another. When we analyze critical bands, we look at their averages and compare them to determine if a sound is going to be noticed or not. The prominence ratio method takes this analysis a step further by looking at the average of the critical band and comparing it to the average of the critical band next to it.

The Similarities Between Aero Acoustics and Thermal Radiation

Aero acoustics, which studies the interaction between sound waves and solid objects, has several similarities with thermal radiation. Both fields use similar methods to analyze light propagation and acoustic propagation, despite being distinct disciplines. The development of new computational methods has made it possible to simulate complex systems using ray tracing techniques, which have been adapted from thermal radiation simulations.

The Impact of Computational Power on Aero Acoustics

The advent of powerful GPUs has revolutionized the field of aero acoustics by enabling faster and more accurate simulations. This has opened up new possibilities for researchers and engineers to analyze complex systems and optimize their performance. The use of GPU acceleration has made it possible to simulate intricate designs, such as fan blades and heat sinks, with unprecedented accuracy.

The Role of Acoustic Simulation in Product Design

Acoustic simulation plays a crucial role in product design, particularly when it comes to designing fans. Fans are designed to be optimized for airflow, but the interaction between each blade and the fins can significantly affect their performance. By using acoustic simulation, engineers can analyze the interaction between these components and optimize the design to minimize noise and improve efficiency.

The Significance of Ray Tracing in Aero Acoustics

Ray tracing, which is traditionally used in computer graphics, has been adapted for use in aero acoustics. The technique involves simulating light propagation and acoustic propagation using complex algorithms. This allows researchers to analyze intricate designs and optimize their performance. The use of ray tracing in aero acoustics has far-reaching implications for product design and engineering.

The Future of Aero Acoustics

As technology continues to advance, the field of aero acoustics is likely to play an increasingly important role in product design and engineering. The development of new computational methods, such as GPU acceleration, will enable researchers to simulate complex systems with unprecedented accuracy. This will open up new possibilities for optimizing fan performance and minimizing noise.

Nvidia's Role in Aero Acoustics

Nvidia has been at the forefront of developing new technologies that apply ray tracing techniques to aero acoustics. The company's expertise in GPU acceleration has enabled researchers to simulate complex systems with greater accuracy than ever before. This has significant implications for product design and engineering, particularly when it comes to designing fans.

The Interview with Nocto

Nocto is a company known for its engineering focus, and the interview with them provides valuable insights into their approach to product design. The company's experience in optimizing fan performance using acoustic simulation highlights the importance of this technique in reducing noise and improving efficiency.

Revolutionizing Fan Design

The development of new technologies and methods has revolutionized the field of fan design. What was once a simple device has become a complex system that requires careful optimization to minimize noise and improve efficiency. The use of acoustic simulation and ray tracing techniques has enabled researchers to analyze intricate designs and optimize their performance.

Ray Tracing in Marketing

The advent of ray tracing technology is set to revolutionize the way we market products, particularly those related to acoustics. Ray tracing will enable companies to accurately simulate how a product performs under real-world conditions, making it easier to demonstrate its benefits to customers. This has significant implications for the marketing landscape and will likely change the way we approach product promotion in the future.

Conclusion

Acoustic simulation is a powerful tool that can help us understand how annoying a given sound can be to our ears. The prominence ratio method, which analyzes critical bands, provides valuable insights into sound annoyance. Aero acoustics has several similarities with thermal radiation, and the development of new computational methods has made it possible to simulate complex systems using ray tracing techniques. As technology continues to advance, the field of aero acoustics is likely to play an increasingly important role in product design and engineering.

"WEBVTTKind: captionsLanguage: enyou might remember this guy Malcolm Gutenberg a thermal engineer and if you don't he's the one who cut the RTX 490 cooler in half with a water jet for us previously stand by Malcolm is a thermal engineer at Nvidia and he's joining us again for another of our educational engineering discussion video series and this time he brought lasers so uh this is a laser infometer he also brought brought this highly specialized acoustic testing senior technician today's content focuses on learning about the crossroads of thermal engineering and fan design and psycho Acoustics as the three all converge in this Confluence of science when building a thermal solution and just like all our other engineering discussions we asked Malcolm not to dumb it down and he definitely didn't optimize it for the human year and use something called critical band but it can be thought of as analogous to an overall one3 octave band centrifugal fan I was talking about the fan Affinity laws previously so p and Q are probably the most overused variables in thermal engineering but so basically what happens is this is your maximum coefficient of lift although not everything requires such technical terms which might sound good if you're into EDM but uh as as far as the graphics card does not what we want at all right so although maybe a marketing opportunity and as always my objective in hosting these discussions is the same as yours for watching them I'm here to learn from experts earlier you you asked a question of uh do you do any prominence ratio something or other after that and I was like I don't know what any of those words mean so I I don't think so as a great followup to our previous case fan engineering Deep dive with Jacob Dinger from noctua let's get started with this fascinating insight to Acoustics thermals and product engineering Acoustics thermal performance all kind of really tied together um so here we are in the Acoustics lab uh we got a lot of cool stuff Hemi an aoic chamber and uh we we really care a lot about the Acoustics of our product not just the overall SPL as we'll we'll get into the uh the overall listening experience how annoying something is to listen about or to listen to um so the psycho acoustic aspects as well yeah so the the video today is going to be very focused on the engineering side of things Y which will make a lot of fun so uh we've been talking for hours now with the whole team and the various topics that have come up I psycho Acoustics measurement methodologies things like coil wine all this type of stuff uh so we'll we'll pick a few of those and get into them today this caught my eye and I did want to ask what is it so uh this is a laser infometer and basically what it does is uh you can see here it's shooting at a PCB and uh we'll have this card running a certain stress test any workload and we'll have a mesh of different points there so basically we'll we'll have a set of different points along the board that we think may be causing any component noise coil wine and this machine here will shoot a laser at the board and then measure how much it reflects back and then through that you can tell what vibration is going on and uh typically the the total amount of display here we're talking about is in the picer scale right it's incredibly small so it's super accurate machine usually we have just a a water block we can expose the rest of the PCB yeah it looks at how much is reflected back and it can calculate what degree of deflection there is based on that so it's a pretty fascinating piece of equipment takes usually uh it can take a while to run the full Suite of the test but the results are very interesting on this case in particular uh the you can see that the inductors are usually causing the noise but the noise might and the peak vibration might not be corresponding to exactly where the inductor is okay cuz the whole PCB is kind of flexing so depending on where that is you can see sometimes it's the capacitors moving um so it's a pretty fast and this is where where we get to earlier uh your colleague Noah noting that coil wine is can be a misnomer I guess because it might not be literally the coil exactly yeah sometimes it's the whole PCB flexing and moving that causes the noise right I don't know if we might come back out here later but this is all hooked up as measurement equipment for what we're going to walk into which the chamber exactly we got a few mics going uh in the chamber right now a mix of half inch and full inch mics we'll get into the difference and why we might use one or the other um but you can see here this is split up into 1/3 octave bands we'll get a better idea of why that is later um how it impacts the overall user experience and uh what what it means to how the ear kind of perceives it perceives noise exactly yeah okay let's check out the chamber so we're gonna how's it how's it going guys good to see you do you normally um you normally have this many people I have never SE more than five people at once think you know it's a it's a big it's a big moment um as you can see right when you walk in it's it's a incredible difference door has a large impact as well but we've already lost a lot of that Echo so exactly I mean especially with 15 people outside the door might be a bigger impact today than other days there are a lot of similarities in certain aspects of this chamber it's the one we just had built exactly so the wedges are very familiar to me cuz this is same type of wedge we're using where it's a perforated metal correct and you've got I guess is it a fiberglass installation one of the topics from viewers we've had is but why right like okay we get it you can get a really low noise floor nobody lives in an a Hemi anoqui chamber you why do you use a chamber in testing yeah it's a good point and uh you know most people won't be subject to this level of of quiet in their in their daily life but for True engineering evaluation of any product you really need to have a very low Baseline to different differentiate between what's background noise and what we can improve on the product um if your background noise is too high then you're not going to be able to pick out specific frequencies or specific traits that could impact some users more than others right right so this is really the ideal case yeah and if you're in a a large campus with a lot of moving you know people and parts then yeah you're not going to be able to tell everyone at the door to please be quiet today we're doing some acoustic testing on the second floor blade design um yes we previously we talked about it primarily from a thermal perspective you kind of briefly mentioned you know there's acoustic considerations too I know you've got a plan here for what you want to talk about I'll let you lead the discussion so this is basically an evolution from our first aial design here which is 20 series to 30 series to 40 series and the first thing you kind of notice is lower total blade count so we went from 13 to 9 to seven here and there's a very good reason for that it really has to do with those frequencies again right as you lower the overall blade passing frequency typically we attenuate those frequencies more in our ear so you can't hear those frequencies as well so you want to shift that frequency as far down as possible okay so reducing the blade count is a great way to do that and also as you go to a larger diameter I was talking about the fan Affinity laws previously if you want to maintain ISO Acoustics you then run at a lower over overall speed you still gain airf flow and you run at a lower speed and you have lower uh blade count so it's just benefits on top of benefits so if it's if it's known that um in theory you might want a lower blade count for for acoustic reasons you know what because there's a lot of considerations that if you know okay perfect world I want lower blade count how do you end up with the higher blade count still I what considerations are there in that design process you know that that make you not just jump to the okay this is the best fan we can make in the world we're done well through time uh I think we've improved our overall understanding of how the how the overall uh F structure fan structure works um one of the key breakthroughs that we had was if we look at these fans here uh typically most of the noise is created on the the leading and trailing edges okay the actual middle part of the blade doesn't create much noise right much aerodynamic noise so if we remove blades you would think there would probably be an impact on either pressure or flow rate but in reality pressure pressure and flow rate don't have an exact correlation there's some empirical correlations so for static pressure typically what you want uh static pressure is proportional to the number of blades times the blade length right so so this uh this length here right so you would think okay reducing blades so you're you're not talking about this you're talk just yeah the width actually I guess yeah sure length um so you would think that if you reduce the number of blades then therefore you reduce performance but the key is to extend that to fill in the extra missing blade right so you actually don't lose much pressure or flow rate by going to a lower blade count um so is it that is it primarily the the thickness or the length I guess of the blade that helps to make up for the reduction in total blade count I mean is there an angle of attack that comes into play in the design yeah angle attack is really important so uh one of the goals when you're designing a fan is for acoustic reasons and overall thermal reasons is you want the velocity to be the velocity of the air to be about cons constant As you move across the fan right but the problem is the tip speed here is three times that of it at the Hub right and as you get closer to the the center of the Hub it's zero right yeah so it's very difficult to create uh a uniform velocity gradient without twisting the blade so you can see here on 20 series there's a bit of a Twist but we've really gone to when you talk about twist where this uh don't try this with your fans at home by way don't take yours apart this way yeah um you could see here how the blade starts off at a shallow angle and then ends at a much sharper angle right so the blade angle changes as you go uh as you go closer to the hub and the reason for that is to maintain the constant velocity across the fan blade and uh and the blade angle the real key to the blade angle is avoiding stall like I was talking about before so basically very similar to an airplane wing uh you have stall issues where you have flow separation behind the behind the Leading Edge of the of the blade in this case and stall is really bad for Acoustics because it's a lot of turbulence it's really noisy uh and terrible for flow rate as well so you really want to avoid that stall range so if you look at this fan here it's fascinating it kind of looks like we had no idea what we were doing with the with the Hub right I mean you know all this wasted space you think um man these guys in Nvidia have you know nothing going on you know but there's actually a reason for this so okay uh obviously with the FDB we can go to a much smaller Hub size if we want yeah and uh and that might be a good idea initially but if you extend the same blade angle to that new Hub size so effectively you're reducing the total uh like external diameter of the Hub exactly exactly and you would think like as a blade passes through the air you know you have more overall net volume yeah you have more hole to push the air through exactly you think this is this is fantastic but the issue is you know if you look at this fan here you're pushing against a certain back pressure right so there is a back pressure there and you'll actually start to see recirculation through the center okay so you'll see negative velocity where that Hub used to be so you think okay well let's just increase the overall blade angle at that point so we can reduce that recirculation but then you start to get into stall issues and aerodynamic uh uh flow separation between the blade and the the air so this is actually the optimized Hub diameter okay which is really interesting I think that was fascinating yeah cuz I guess purely from a physical standpoint you have the space to shrink it down increase the blade length right I have some other questions for the chamber but um the uh specifically talking about where stall region appears I think on that curve might be helpful for people so looking at any fan really the first thing you need to look at is what's called a pressure versus flow rate curve so p and Q are probably the most overused variables in thermal engineering but in this case it means pressure and volumetric flow rate and for a given axial fan you'll typically see something like this and this hump is very important but to kind of understand what this means is this point here if we just take this point this is if the fan is just operating in free air with uh with no blockage behind it so no static pressure it's just a fan nothing this if we take the common sort of acoustic testing setup is this when you have a fan suspended and exactly c yeah yeah yeah yeah yeah so this is typically the point where you see the acoustic value for a case fan this is where you see the volumetric flow rate you'll get the highest numbers exactly yeah it looks amazing right but the issue is this is not where you're going to be operating at there's in in any actual thermal engineering case there's always going to be some pressure that you need to push through and the pressure part is very interesting here I'll use red if you were designing something like um uh like a wastewater treatment plan or something and you just need to push water or push air through water then the pressure you would need would just be the hydrostatic pressure of the water and as you get to a laminer system you'll have more of a straight line but since ours is a turbulent system we have a quadratic line that this is our impedance curve and this line right here so we can call this q1 and this pressure here is actually where we're going to be operating at and this point is really important because if we look at the overall efficiency of the fan we need to make sure that this point aligns with the overall efficiency because typically the overall efficiency aligns within plusus 5% typically with the lowest noise level and you're thinking that doesn't make any sense because you know you're increasing the overall volumetric flow rate you would expect more Aero acoustic noise more Broadband noise you know so if you were to add a right Axis or something plot noise on that so good thing we have three markers here I guess so here if we just do sp right uh we would expect something like this but in reality it looks something like this okay and the reason for that is this area here is all stall so you don't want to be in this region you really don't want to be in this region so basically what happens is this is your maximum coefficient of lift similar to how an aircraft works or so and after this point the coefficient of lift drops right off and this basically if it wasn't for the fact that it's spinning around you would get no pressure no flow it would be terrible but uh since we're still spinning it around it just basically basically becomes a mixed flow fan at this part so if we could see the air uh and you're looking at a fan and its blades does that look different you know than when the air when you're say at the operating point oh for sure yeah so you'll definitely see flow Separation on the back side on the trailing edge of the blade um you'll see a lot more turbulence turbulence is pressure fluctuation pressure fluctuation is noise so it's certainly not something that you really want to uh be there so one of the Beautiful Things is you know the issue is you could say okay well let's just move the operating point to the right right you get more flow no problem but the issue is you also want fins right for thermal surface area so you want to add fins but you also want to decrease pressure and so reason yeah and let's let's expand on that a little bit just for background so adding fins uh I'll take that part you take the pressure part for it no problem yeah adding fins the obvious I guess being surface area exactly you want more surface area to help deal with the heat uh how about the pressure side of things so uh the pressure side I mean you want more flow right you want more flow and you want more more fins as well so it seems like a bit of a you yeah exactly so the beautiful thing is you know even if you just plot our East fan and our West fan on 4090 blow through makes a huge difference so basically instead of being at this line you could be operating somewhere here and then you can add more fins and you can gain fin surface area and still be operating at that perfect like lowest noise highest maximum you know mechanical efficiency it's like the The Best of Both Worlds it it's one of the big benefits of blowr yeah and um so chamers side I guess if we go yeah sure back in with the fans can I ask about this setup first Butcher Block Top's pretty self-explanatory but how about what's underneath it yeah so this is a really interesting thing that we've just been kind of working with more recently so for a lot of our testing um we want to be able to test while the GPU is running uh we don't necessarily if we just want to test fans then this setup is totally not needed right we can externally power those but a lot of the times we want to look at electrical noise we want to look at how everything kind of comes together as a final product test fan curves all that stuff so for that you do need a system and as good as passive systems get we also want it to be as quiet as possible and support a lot of power you know so that is the idea that you put the host system in here is that exactly exactly and then what you run a riser up through a hole or something exactly yeah that'll and and the issue is with you know longer pcie like you can have those extenders right but you sort of lose out on a lot of stuff but yeah some have some insertion loss there is this something we can show what's going on here if you come around here vital wait let me get a shot yeah so this is how we dampen some of the noise that's created inside of the uh inside of the box there we have a serpentine so is there there's the case fan there okay and then the airf flow will come up through here so is this this is I guess it's pushing in is it is it effectively getting its air from up here through this hole Yeah so there's a bit of pressure drop that's a long path yeah it's a bit of a long path hopefully we don't have to support too much power in there right um but so idea being that uh you're trying to get some Cooling in for whatever's under there but also I guess minimize external noise is that exactly exactly yeah so we try and keep everything as quiet as possible but um do see do you have um do you have an exhaust on the other side or something yeah this one might be oh yeah okay I see similar Style just uh yeah exactly and then through the bottom we try and get the air flow you know exhausting out right below and then of course we have a lot of microphones here as well um the main reason why we have so many is a lot of our products uh are very directional when it comes to sound so uh depending on where you measure you can have different overall sound pressure level values different tones different uh annoying sounds right yeah so we try and measure in as many locations as possible to simulate however users might use them whether you know it's in a chassis normally or you know this way anything anything that could in theory be used and we have an assortment here of half inch and uh one inch microphones can you walk me through so at a top level broadly speaking um what types of considerations are there when choosing the microphone for the chamber because like you wouldn't want to take this lab you know and use it so low noise floor is kind of the obvious one to me uh for the mic but what what other considerations are there so fundamentally when you look at the microphone you know it's just very the top capsule there and there's a diaphragm and the diaphragm has a charge and then there's a back plate with a negative charge and you can think of it as kind of a capacitor so as the diaphragm flexes it changes the capacitance and then you read that as overall pressure and then you can take the roots some squared and get a uh like an actual value of what total pressure you would be able to see so the bigger the diaphragm the easier it is to kind of flex right so if you go to a 1-in microphone typically you have lower overall background noise that you can measure right so you can measure down to uh well below zero DB um and the issue with I want to come back to that in a second too but I'll let you continue on sure sure and uh the the smaller microphones the half inch microphones typically they're better for uh overall frequency range so the human hearing range like I said before goes from about 20 hzz to 20 khz so you want to be able to capture that full dynamic range so you mentioned below Z DB which I I think this is probably worth talking about sure sure uh so scale doesn't stop at zero here yeah yeah uh is this something you can walk us through yeah yeah so the whole DB scale is very interesting like it's fascinating it has nothing I don't think it was ever uh made for Acoustics I think it was made for telecommunications I think Alexander Graham Bell or not I'm not an expert on that but uh it's just all all all of what the decibel is is just comparing a reference pressure value to whatever pressure value you're measuring so let's say you measure a pressure value you divide it by 2 * 105 pascals is the is the number and then you take the log of that and multiply by 20 so instead of us saying like oh this card this card is 0.0000 three pascals it might be a little more you know confusing for the end person so the decel scale makes it a lot easier so basically if you're below zero that just means you're below the reference pressure uh and then talking on the chamber side because that that kind of reminds me of uh Microsoft I think has one of the famous Chambers for if I remember correctly I think it's like underground or something it's like it's at least at one point had the quietest room in you know in the world Ward but um there are different types of Chambers too so there's fully an aoic Hemi anaco you've got a hard surface yeah I think Springs underneath correct yep so uh the hard surface is more representative of a typical user case um typically there's always kind of something below so instead of using a full anaco chamber we use a Hemi anico chamber as it's more representative you'll see the same things in like the Auto industry because there's always a road beneath right so it gives a overall better representation of uh kind of the the typical user use case for the product yeah yeah I guess testing a a car suspended in the air in a fully anticor CH yeah and and the reason it it's kind of large right I mean the the GPU here makes it look a little uh overk Overkill um but there there is a reason um you know in case we obviously we make a lot more products than just like dgx or something exactly exactly so if we want to test let's say the full sound sound power of a given rack a server rack then we can put it in here and hopefully have you know that's why the ceiling is so high have the space required in order to test that um also as you go to a larger chamber typically you can measure uh lower overall frequencies just because of at the peak of human hearing like the wavelength is incredibly small like 17 m I think okay uh but at 20 HZ it's 17 M long right so typically if you want lower overall frequency you have to go to a larger larger chamber right and my understanding too is that the wedge depth also affects the frequency cut off exactly how about uh do you do you spend personally when you're doing engineering work do you spend a lot of time on the fan curve and V bio like figuring out the fan response oh for sure yeah it's a it's a very important part um you know looking at how different parts of the fan curve CU obviously that PQ curve I showed is just for one RPM and then you can use the Affinity laws to kind of scale it based on RPM right so um does it hurt you when someone manually adjusts their fan RPM uh you can do whatever you want but personally I think I think this is the the best one you can have um yeah I'll I'll I'll stay out of that one um but yeah yeah I I think cuz we separate these two frequencies of the blade pass at least right by 2 25 Herz in order to try and give the best listening experience so maybe you can think oh well why do they even separate them let me just put them at the same level but uh you might get a a worse listening experience overall yeah depending on where they land on it yeah um so to really get into some of the reasons we do the design changes that we do you really can't just look at overall all sound pressure level so I know typically a lot of case fans and product reviews you see just one level SPL this is what it is and and then the story kind of stops there but you can have two different products one with uh slightly higher overall SPL and one with slightly lower overall SPL but depending on which frequencies are causing that uh or are prevalent in that uh tone or in that sample you could pick the lower SPL version as sounding worse one of the main things that we see when we look at the Acoustics of our card is What's called the blade passing frequency so every time you spin this fan you have blades passing a certain spot and those create pressure fluctuations in the air and pressure fluctuations are what we hear so um both of these fans are seven bladed so they both if operate at the same RPM will have the same overall um blade passing frequency and if you look at a chart of the frequency versus SPL you'll see that um you'll see these tones very obviously in the overall frequency spectrum so what we want to do is eliminate how annoying those sounds are right because if you start to have tones stacked on top of each other you really have uh a very unpleasant user experience to really get into that we need to think about uh how the human ear hears different tones so uh we like to break up the uh frequency spectrum that we can hear from 20 HZ to 20 khz into bands that we call critical bands and I think most people are might be familiar with the octave term yeah so an octave is a doubling of frequency so if you go up an octave you double the frequency so if you break that up further into onethird octave bands like I was showing on the the screen out there you get to a place that's more representative of how humans here so uh if a tone is in the same octave one/ third octave band as another you typically can't really hear the difference between the two tones so what we do is we further optimize it for the human ear and use something called critical band but it can be thought of as analogous to an overall one3 octave band and what we want to do is we want to make sure those two tones for a given fan don't overlap on each other but we also don't want them in the same or too close to each other because if the tones are in the same critical band but separated by only let's say 4 Hertz or uh less than 25 Hertz we start to hear what's called modulation which is this w w which might sound good if you're into EDM but uh as as far as the graphics card goes not what we want at all right so although maybe a marketing opportunity that's true so we we like to separate them by about 25 Hertz uh past 25 to about 125 is what call roughness okay and roughness is much more desirable than uh pure modulation much less noticeable uh and better overall listening experience and so this this kind of bridges into the topic of you know talking about psycho Acoustics where when I was first starting to we we only really got into past the DB meter in the past couple months right so total novice to this on my side um but we started speaking with Mike chin from Silent PC review some other people AR Harbor Busters and um one of the common themes I've gotten from people more experienced is I'll ask a question and approach it from sort of the scientific standpoint and the answer I get back is like well this is kind of psycho Acoustics it's a special area yeah because you're talking about it's not just the levels it's the perception of the levels exactly exactly it's very uh you know very human and it's looking at how we listen to different sounds what sounds are typically louder and more an annoying and it's really um it's really how the inner ear is structured there's two parts the ear in general that really determine what sounds are annoying what we hear louder and if you look at the overall uh loudness Spectrum you can see that around 2 to 5 khz are what we perceive as the loudest sounds right okay so those sounds you know if we had ISO SPL in those frequencies we would hear it the loudest and that's because of how our inner ear or ear canal amp ifies those different frequencies and then the frequency range that we hear is based on how the CIA which is part of the inner ear perceives different frequencies so it's this logarithmic spiral shape which is very interesting and you can pinpoint exactly what parts of that are have different resonant frequencies okay so as you get closer to the center you get to uh the lower frequency so you have 20 Herz to 20 khz on average right and uh and if you we're just roll it out uh I'm not a biologist but let's say in theory you could roll it out uh you have more linear space for the lower frequencies which means like I was saying earlier with the one one3 octave band the critical bands you can TP typically differentiate between lower frequencies better but in terms of overall hearing based on how the the ear canal is uh amplifies different sounds you can't hear those as loud so it kind of all plays into that yeah but yeah psycho Acoustics like you're saying very subjective there are some numbers that they put to it right what prominence ratio can be maybe we'll get into what that means actually but um earlier you you asked a question of uh do you do any prominence ratio something or other after that and I was like I don't know what any of those words mean so I I don't think so uh so those I would say those three items uh are some of the most important in terms of overall uh Gathering how annoying a given sound is so prominence ratio looks at the critical band so you take that critical band and you take the average of that and then compare to the average of the critical bands next to it okay and then if it's above a certain level it's very noticeable to the ear and tonen noise does a very similar thing I mean tonen noise is very welln named I think it just looks at the tone and then it looks at the critical band so it looks at the noise um both have their individual application if you have multiple tones in the same critical band uh prominence ratio can be better at determining if that's going to be aning or not anything else in here that you want to cover or externally no I mean um the only other the only other really I think fascinating thing when it comes to Acoustics that's kind of really On The Rise right now because you know when you look at a fan I don't know when the first fan was invented but probably like ancient Egypt you know yeah a long it's been a very well-developed industry and you may think okay there's not a lot to improve nowadays and if you just look at the fan on its own you know free air then it's true there isn't too much you know it's it's a very optimized fan but as you start looking at it in terms of overall product and you look at how each blade interacts with the fins you know the heat sink fins right you can get some really interesting data and and we use Aero Acoustics for this so basically you take this card and then you simulate a box around it right but the issue is at 20 khz your frequency is or your wavelength is like 17 mm right and you want to have somewhere in the range of 100 points along that wavelength so let's say you had this box that you wanted to simulate and now you have to break it up into little cubes that are no longer than I guess that would be .17 mm long right so it's incredibly computationally expensive but the beautiful thing is light propagation and acoustic propagation are very similar and they're different but they're they're relatively similar so a lot of methods have been have been produced that use some of the ray tracing methods oh interesting it's s very similar to Ray tracing so that's why a thermal radiation and acoustic like uh Aero acoustic simulation have like skyrocketed now because of how in terms of computational time like reduced yeah reduc so the abilities are amazing now when you look at how GPU accelerated some of these are so I think it'll be revolutionary going forward I mean we've already learned a lot like even for example when you look at the blade shape and how it interacts with the fins and and everything it's like fascinating yeah it kind of comes back to what we talked about last time which was uh Nvidia in a unique position to use its own GPU Hardware to simulate its own next gen GPU yeah yeah a lot of the data center things that we've been doing just feed right back into G4 it's fascinating to look at and um it's it's like revolutionary for Acoustics and uh like you're saying back to Total product level it's also why you can have a I'll I think nocto is fairly well known for the engineering Focus you can have a company like them invent what is right now to them the best fan they've ever made but then as they tune it for different products you might have to tweak the design which is something we have an interview with them about just because at a you know product level once you put a fin stack behind it Like Everything Changes Everything Changes and how you optimize the fins change too right and uh yeah it's just fascinating um to look at kind of the similarities to R tracing you know yeah yeah so uh next Nvidia product will have Ray traced air R TRC yeah R traced Acoustics that'll be on the marketing box with the actual R tracing performance Malcolm thank you again for the walkthrough always a pleasure always a pleasure I yeah I love talking about this um yeah it's it's uh fascinating stuff we really I mean we really care about Acoustics the overall lising experience so and what better place to do it than here and check out our last video with Malcolm I'll link it below if this interested you thank you all for watching we'll see you all next timeyou might remember this guy Malcolm Gutenberg a thermal engineer and if you don't he's the one who cut the RTX 490 cooler in half with a water jet for us previously stand by Malcolm is a thermal engineer at Nvidia and he's joining us again for another of our educational engineering discussion video series and this time he brought lasers so uh this is a laser infometer he also brought brought this highly specialized acoustic testing senior technician today's content focuses on learning about the crossroads of thermal engineering and fan design and psycho Acoustics as the three all converge in this Confluence of science when building a thermal solution and just like all our other engineering discussions we asked Malcolm not to dumb it down and he definitely didn't optimize it for the human year and use something called critical band but it can be thought of as analogous to an overall one3 octave band centrifugal fan I was talking about the fan Affinity laws previously so p and Q are probably the most overused variables in thermal engineering but so basically what happens is this is your maximum coefficient of lift although not everything requires such technical terms which might sound good if you're into EDM but uh as as far as the graphics card does not what we want at all right so although maybe a marketing opportunity and as always my objective in hosting these discussions is the same as yours for watching them I'm here to learn from experts earlier you you asked a question of uh do you do any prominence ratio something or other after that and I was like I don't know what any of those words mean so I I don't think so as a great followup to our previous case fan engineering Deep dive with Jacob Dinger from noctua let's get started with this fascinating insight to Acoustics thermals and product engineering Acoustics thermal performance all kind of really tied together um so here we are in the Acoustics lab uh we got a lot of cool stuff Hemi an aoic chamber and uh we we really care a lot about the Acoustics of our product not just the overall SPL as we'll we'll get into the uh the overall listening experience how annoying something is to listen about or to listen to um so the psycho acoustic aspects as well yeah so the the video today is going to be very focused on the engineering side of things Y which will make a lot of fun so uh we've been talking for hours now with the whole team and the various topics that have come up I psycho Acoustics measurement methodologies things like coil wine all this type of stuff uh so we'll we'll pick a few of those and get into them today this caught my eye and I did want to ask what is it so uh this is a laser infometer and basically what it does is uh you can see here it's shooting at a PCB and uh we'll have this card running a certain stress test any workload and we'll have a mesh of different points there so basically we'll we'll have a set of different points along the board that we think may be causing any component noise coil wine and this machine here will shoot a laser at the board and then measure how much it reflects back and then through that you can tell what vibration is going on and uh typically the the total amount of display here we're talking about is in the picer scale right it's incredibly small so it's super accurate machine usually we have just a a water block we can expose the rest of the PCB yeah it looks at how much is reflected back and it can calculate what degree of deflection there is based on that so it's a pretty fascinating piece of equipment takes usually uh it can take a while to run the full Suite of the test but the results are very interesting on this case in particular uh the you can see that the inductors are usually causing the noise but the noise might and the peak vibration might not be corresponding to exactly where the inductor is okay cuz the whole PCB is kind of flexing so depending on where that is you can see sometimes it's the capacitors moving um so it's a pretty fast and this is where where we get to earlier uh your colleague Noah noting that coil wine is can be a misnomer I guess because it might not be literally the coil exactly yeah sometimes it's the whole PCB flexing and moving that causes the noise right I don't know if we might come back out here later but this is all hooked up as measurement equipment for what we're going to walk into which the chamber exactly we got a few mics going uh in the chamber right now a mix of half inch and full inch mics we'll get into the difference and why we might use one or the other um but you can see here this is split up into 1/3 octave bands we'll get a better idea of why that is later um how it impacts the overall user experience and uh what what it means to how the ear kind of perceives it perceives noise exactly yeah okay let's check out the chamber so we're gonna how's it how's it going guys good to see you do you normally um you normally have this many people I have never SE more than five people at once think you know it's a it's a big it's a big moment um as you can see right when you walk in it's it's a incredible difference door has a large impact as well but we've already lost a lot of that Echo so exactly I mean especially with 15 people outside the door might be a bigger impact today than other days there are a lot of similarities in certain aspects of this chamber it's the one we just had built exactly so the wedges are very familiar to me cuz this is same type of wedge we're using where it's a perforated metal correct and you've got I guess is it a fiberglass installation one of the topics from viewers we've had is but why right like okay we get it you can get a really low noise floor nobody lives in an a Hemi anoqui chamber you why do you use a chamber in testing yeah it's a good point and uh you know most people won't be subject to this level of of quiet in their in their daily life but for True engineering evaluation of any product you really need to have a very low Baseline to different differentiate between what's background noise and what we can improve on the product um if your background noise is too high then you're not going to be able to pick out specific frequencies or specific traits that could impact some users more than others right right so this is really the ideal case yeah and if you're in a a large campus with a lot of moving you know people and parts then yeah you're not going to be able to tell everyone at the door to please be quiet today we're doing some acoustic testing on the second floor blade design um yes we previously we talked about it primarily from a thermal perspective you kind of briefly mentioned you know there's acoustic considerations too I know you've got a plan here for what you want to talk about I'll let you lead the discussion so this is basically an evolution from our first aial design here which is 20 series to 30 series to 40 series and the first thing you kind of notice is lower total blade count so we went from 13 to 9 to seven here and there's a very good reason for that it really has to do with those frequencies again right as you lower the overall blade passing frequency typically we attenuate those frequencies more in our ear so you can't hear those frequencies as well so you want to shift that frequency as far down as possible okay so reducing the blade count is a great way to do that and also as you go to a larger diameter I was talking about the fan Affinity laws previously if you want to maintain ISO Acoustics you then run at a lower over overall speed you still gain airf flow and you run at a lower speed and you have lower uh blade count so it's just benefits on top of benefits so if it's if it's known that um in theory you might want a lower blade count for for acoustic reasons you know what because there's a lot of considerations that if you know okay perfect world I want lower blade count how do you end up with the higher blade count still I what considerations are there in that design process you know that that make you not just jump to the okay this is the best fan we can make in the world we're done well through time uh I think we've improved our overall understanding of how the how the overall uh F structure fan structure works um one of the key breakthroughs that we had was if we look at these fans here uh typically most of the noise is created on the the leading and trailing edges okay the actual middle part of the blade doesn't create much noise right much aerodynamic noise so if we remove blades you would think there would probably be an impact on either pressure or flow rate but in reality pressure pressure and flow rate don't have an exact correlation there's some empirical correlations so for static pressure typically what you want uh static pressure is proportional to the number of blades times the blade length right so so this uh this length here right so you would think okay reducing blades so you're you're not talking about this you're talk just yeah the width actually I guess yeah sure length um so you would think that if you reduce the number of blades then therefore you reduce performance but the key is to extend that to fill in the extra missing blade right so you actually don't lose much pressure or flow rate by going to a lower blade count um so is it that is it primarily the the thickness or the length I guess of the blade that helps to make up for the reduction in total blade count I mean is there an angle of attack that comes into play in the design yeah angle attack is really important so uh one of the goals when you're designing a fan is for acoustic reasons and overall thermal reasons is you want the velocity to be the velocity of the air to be about cons constant As you move across the fan right but the problem is the tip speed here is three times that of it at the Hub right and as you get closer to the the center of the Hub it's zero right yeah so it's very difficult to create uh a uniform velocity gradient without twisting the blade so you can see here on 20 series there's a bit of a Twist but we've really gone to when you talk about twist where this uh don't try this with your fans at home by way don't take yours apart this way yeah um you could see here how the blade starts off at a shallow angle and then ends at a much sharper angle right so the blade angle changes as you go uh as you go closer to the hub and the reason for that is to maintain the constant velocity across the fan blade and uh and the blade angle the real key to the blade angle is avoiding stall like I was talking about before so basically very similar to an airplane wing uh you have stall issues where you have flow separation behind the behind the Leading Edge of the of the blade in this case and stall is really bad for Acoustics because it's a lot of turbulence it's really noisy uh and terrible for flow rate as well so you really want to avoid that stall range so if you look at this fan here it's fascinating it kind of looks like we had no idea what we were doing with the with the Hub right I mean you know all this wasted space you think um man these guys in Nvidia have you know nothing going on you know but there's actually a reason for this so okay uh obviously with the FDB we can go to a much smaller Hub size if we want yeah and uh and that might be a good idea initially but if you extend the same blade angle to that new Hub size so effectively you're reducing the total uh like external diameter of the Hub exactly exactly and you would think like as a blade passes through the air you know you have more overall net volume yeah you have more hole to push the air through exactly you think this is this is fantastic but the issue is you know if you look at this fan here you're pushing against a certain back pressure right so there is a back pressure there and you'll actually start to see recirculation through the center okay so you'll see negative velocity where that Hub used to be so you think okay well let's just increase the overall blade angle at that point so we can reduce that recirculation but then you start to get into stall issues and aerodynamic uh uh flow separation between the blade and the the air so this is actually the optimized Hub diameter okay which is really interesting I think that was fascinating yeah cuz I guess purely from a physical standpoint you have the space to shrink it down increase the blade length right I have some other questions for the chamber but um the uh specifically talking about where stall region appears I think on that curve might be helpful for people so looking at any fan really the first thing you need to look at is what's called a pressure versus flow rate curve so p and Q are probably the most overused variables in thermal engineering but in this case it means pressure and volumetric flow rate and for a given axial fan you'll typically see something like this and this hump is very important but to kind of understand what this means is this point here if we just take this point this is if the fan is just operating in free air with uh with no blockage behind it so no static pressure it's just a fan nothing this if we take the common sort of acoustic testing setup is this when you have a fan suspended and exactly c yeah yeah yeah yeah yeah so this is typically the point where you see the acoustic value for a case fan this is where you see the volumetric flow rate you'll get the highest numbers exactly yeah it looks amazing right but the issue is this is not where you're going to be operating at there's in in any actual thermal engineering case there's always going to be some pressure that you need to push through and the pressure part is very interesting here I'll use red if you were designing something like um uh like a wastewater treatment plan or something and you just need to push water or push air through water then the pressure you would need would just be the hydrostatic pressure of the water and as you get to a laminer system you'll have more of a straight line but since ours is a turbulent system we have a quadratic line that this is our impedance curve and this line right here so we can call this q1 and this pressure here is actually where we're going to be operating at and this point is really important because if we look at the overall efficiency of the fan we need to make sure that this point aligns with the overall efficiency because typically the overall efficiency aligns within plusus 5% typically with the lowest noise level and you're thinking that doesn't make any sense because you know you're increasing the overall volumetric flow rate you would expect more Aero acoustic noise more Broadband noise you know so if you were to add a right Axis or something plot noise on that so good thing we have three markers here I guess so here if we just do sp right uh we would expect something like this but in reality it looks something like this okay and the reason for that is this area here is all stall so you don't want to be in this region you really don't want to be in this region so basically what happens is this is your maximum coefficient of lift similar to how an aircraft works or so and after this point the coefficient of lift drops right off and this basically if it wasn't for the fact that it's spinning around you would get no pressure no flow it would be terrible but uh since we're still spinning it around it just basically basically becomes a mixed flow fan at this part so if we could see the air uh and you're looking at a fan and its blades does that look different you know than when the air when you're say at the operating point oh for sure yeah so you'll definitely see flow Separation on the back side on the trailing edge of the blade um you'll see a lot more turbulence turbulence is pressure fluctuation pressure fluctuation is noise so it's certainly not something that you really want to uh be there so one of the Beautiful Things is you know the issue is you could say okay well let's just move the operating point to the right right you get more flow no problem but the issue is you also want fins right for thermal surface area so you want to add fins but you also want to decrease pressure and so reason yeah and let's let's expand on that a little bit just for background so adding fins uh I'll take that part you take the pressure part for it no problem yeah adding fins the obvious I guess being surface area exactly you want more surface area to help deal with the heat uh how about the pressure side of things so uh the pressure side I mean you want more flow right you want more flow and you want more more fins as well so it seems like a bit of a you yeah exactly so the beautiful thing is you know even if you just plot our East fan and our West fan on 4090 blow through makes a huge difference so basically instead of being at this line you could be operating somewhere here and then you can add more fins and you can gain fin surface area and still be operating at that perfect like lowest noise highest maximum you know mechanical efficiency it's like the The Best of Both Worlds it it's one of the big benefits of blowr yeah and um so chamers side I guess if we go yeah sure back in with the fans can I ask about this setup first Butcher Block Top's pretty self-explanatory but how about what's underneath it yeah so this is a really interesting thing that we've just been kind of working with more recently so for a lot of our testing um we want to be able to test while the GPU is running uh we don't necessarily if we just want to test fans then this setup is totally not needed right we can externally power those but a lot of the times we want to look at electrical noise we want to look at how everything kind of comes together as a final product test fan curves all that stuff so for that you do need a system and as good as passive systems get we also want it to be as quiet as possible and support a lot of power you know so that is the idea that you put the host system in here is that exactly exactly and then what you run a riser up through a hole or something exactly yeah that'll and and the issue is with you know longer pcie like you can have those extenders right but you sort of lose out on a lot of stuff but yeah some have some insertion loss there is this something we can show what's going on here if you come around here vital wait let me get a shot yeah so this is how we dampen some of the noise that's created inside of the uh inside of the box there we have a serpentine so is there there's the case fan there okay and then the airf flow will come up through here so is this this is I guess it's pushing in is it is it effectively getting its air from up here through this hole Yeah so there's a bit of pressure drop that's a long path yeah it's a bit of a long path hopefully we don't have to support too much power in there right um but so idea being that uh you're trying to get some Cooling in for whatever's under there but also I guess minimize external noise is that exactly exactly yeah so we try and keep everything as quiet as possible but um do see do you have um do you have an exhaust on the other side or something yeah this one might be oh yeah okay I see similar Style just uh yeah exactly and then through the bottom we try and get the air flow you know exhausting out right below and then of course we have a lot of microphones here as well um the main reason why we have so many is a lot of our products uh are very directional when it comes to sound so uh depending on where you measure you can have different overall sound pressure level values different tones different uh annoying sounds right yeah so we try and measure in as many locations as possible to simulate however users might use them whether you know it's in a chassis normally or you know this way anything anything that could in theory be used and we have an assortment here of half inch and uh one inch microphones can you walk me through so at a top level broadly speaking um what types of considerations are there when choosing the microphone for the chamber because like you wouldn't want to take this lab you know and use it so low noise floor is kind of the obvious one to me uh for the mic but what what other considerations are there so fundamentally when you look at the microphone you know it's just very the top capsule there and there's a diaphragm and the diaphragm has a charge and then there's a back plate with a negative charge and you can think of it as kind of a capacitor so as the diaphragm flexes it changes the capacitance and then you read that as overall pressure and then you can take the roots some squared and get a uh like an actual value of what total pressure you would be able to see so the bigger the diaphragm the easier it is to kind of flex right so if you go to a 1-in microphone typically you have lower overall background noise that you can measure right so you can measure down to uh well below zero DB um and the issue with I want to come back to that in a second too but I'll let you continue on sure sure and uh the the smaller microphones the half inch microphones typically they're better for uh overall frequency range so the human hearing range like I said before goes from about 20 hzz to 20 khz so you want to be able to capture that full dynamic range so you mentioned below Z DB which I I think this is probably worth talking about sure sure uh so scale doesn't stop at zero here yeah yeah uh is this something you can walk us through yeah yeah so the whole DB scale is very interesting like it's fascinating it has nothing I don't think it was ever uh made for Acoustics I think it was made for telecommunications I think Alexander Graham Bell or not I'm not an expert on that but uh it's just all all all of what the decibel is is just comparing a reference pressure value to whatever pressure value you're measuring so let's say you measure a pressure value you divide it by 2 * 105 pascals is the is the number and then you take the log of that and multiply by 20 so instead of us saying like oh this card this card is 0.0000 three pascals it might be a little more you know confusing for the end person so the decel scale makes it a lot easier so basically if you're below zero that just means you're below the reference pressure uh and then talking on the chamber side because that that kind of reminds me of uh Microsoft I think has one of the famous Chambers for if I remember correctly I think it's like underground or something it's like it's at least at one point had the quietest room in you know in the world Ward but um there are different types of Chambers too so there's fully an aoic Hemi anaco you've got a hard surface yeah I think Springs underneath correct yep so uh the hard surface is more representative of a typical user case um typically there's always kind of something below so instead of using a full anaco chamber we use a Hemi anico chamber as it's more representative you'll see the same things in like the Auto industry because there's always a road beneath right so it gives a overall better representation of uh kind of the the typical user use case for the product yeah yeah I guess testing a a car suspended in the air in a fully anticor CH yeah and and the reason it it's kind of large right I mean the the GPU here makes it look a little uh overk Overkill um but there there is a reason um you know in case we obviously we make a lot more products than just like dgx or something exactly exactly so if we want to test let's say the full sound sound power of a given rack a server rack then we can put it in here and hopefully have you know that's why the ceiling is so high have the space required in order to test that um also as you go to a larger chamber typically you can measure uh lower overall frequencies just because of at the peak of human hearing like the wavelength is incredibly small like 17 m I think okay uh but at 20 HZ it's 17 M long right so typically if you want lower overall frequency you have to go to a larger larger chamber right and my understanding too is that the wedge depth also affects the frequency cut off exactly how about uh do you do you spend personally when you're doing engineering work do you spend a lot of time on the fan curve and V bio like figuring out the fan response oh for sure yeah it's a it's a very important part um you know looking at how different parts of the fan curve CU obviously that PQ curve I showed is just for one RPM and then you can use the Affinity laws to kind of scale it based on RPM right so um does it hurt you when someone manually adjusts their fan RPM uh you can do whatever you want but personally I think I think this is the the best one you can have um yeah I'll I'll I'll stay out of that one um but yeah yeah I I think cuz we separate these two frequencies of the blade pass at least right by 2 25 Herz in order to try and give the best listening experience so maybe you can think oh well why do they even separate them let me just put them at the same level but uh you might get a a worse listening experience overall yeah depending on where they land on it yeah um so to really get into some of the reasons we do the design changes that we do you really can't just look at overall all sound pressure level so I know typically a lot of case fans and product reviews you see just one level SPL this is what it is and and then the story kind of stops there but you can have two different products one with uh slightly higher overall SPL and one with slightly lower overall SPL but depending on which frequencies are causing that uh or are prevalent in that uh tone or in that sample you could pick the lower SPL version as sounding worse one of the main things that we see when we look at the Acoustics of our card is What's called the blade passing frequency so every time you spin this fan you have blades passing a certain spot and those create pressure fluctuations in the air and pressure fluctuations are what we hear so um both of these fans are seven bladed so they both if operate at the same RPM will have the same overall um blade passing frequency and if you look at a chart of the frequency versus SPL you'll see that um you'll see these tones very obviously in the overall frequency spectrum so what we want to do is eliminate how annoying those sounds are right because if you start to have tones stacked on top of each other you really have uh a very unpleasant user experience to really get into that we need to think about uh how the human ear hears different tones so uh we like to break up the uh frequency spectrum that we can hear from 20 HZ to 20 khz into bands that we call critical bands and I think most people are might be familiar with the octave term yeah so an octave is a doubling of frequency so if you go up an octave you double the frequency so if you break that up further into onethird octave bands like I was showing on the the screen out there you get to a place that's more representative of how humans here so uh if a tone is in the same octave one/ third octave band as another you typically can't really hear the difference between the two tones so what we do is we further optimize it for the human ear and use something called critical band but it can be thought of as analogous to an overall one3 octave band and what we want to do is we want to make sure those two tones for a given fan don't overlap on each other but we also don't want them in the same or too close to each other because if the tones are in the same critical band but separated by only let's say 4 Hertz or uh less than 25 Hertz we start to hear what's called modulation which is this w w which might sound good if you're into EDM but uh as as far as the graphics card goes not what we want at all right so although maybe a marketing opportunity that's true so we we like to separate them by about 25 Hertz uh past 25 to about 125 is what call roughness okay and roughness is much more desirable than uh pure modulation much less noticeable uh and better overall listening experience and so this this kind of bridges into the topic of you know talking about psycho Acoustics where when I was first starting to we we only really got into past the DB meter in the past couple months right so total novice to this on my side um but we started speaking with Mike chin from Silent PC review some other people AR Harbor Busters and um one of the common themes I've gotten from people more experienced is I'll ask a question and approach it from sort of the scientific standpoint and the answer I get back is like well this is kind of psycho Acoustics it's a special area yeah because you're talking about it's not just the levels it's the perception of the levels exactly exactly it's very uh you know very human and it's looking at how we listen to different sounds what sounds are typically louder and more an annoying and it's really um it's really how the inner ear is structured there's two parts the ear in general that really determine what sounds are annoying what we hear louder and if you look at the overall uh loudness Spectrum you can see that around 2 to 5 khz are what we perceive as the loudest sounds right okay so those sounds you know if we had ISO SPL in those frequencies we would hear it the loudest and that's because of how our inner ear or ear canal amp ifies those different frequencies and then the frequency range that we hear is based on how the CIA which is part of the inner ear perceives different frequencies so it's this logarithmic spiral shape which is very interesting and you can pinpoint exactly what parts of that are have different resonant frequencies okay so as you get closer to the center you get to uh the lower frequency so you have 20 Herz to 20 khz on average right and uh and if you we're just roll it out uh I'm not a biologist but let's say in theory you could roll it out uh you have more linear space for the lower frequencies which means like I was saying earlier with the one one3 octave band the critical bands you can TP typically differentiate between lower frequencies better but in terms of overall hearing based on how the the ear canal is uh amplifies different sounds you can't hear those as loud so it kind of all plays into that yeah but yeah psycho Acoustics like you're saying very subjective there are some numbers that they put to it right what prominence ratio can be maybe we'll get into what that means actually but um earlier you you asked a question of uh do you do any prominence ratio something or other after that and I was like I don't know what any of those words mean so I I don't think so uh so those I would say those three items uh are some of the most important in terms of overall uh Gathering how annoying a given sound is so prominence ratio looks at the critical band so you take that critical band and you take the average of that and then compare to the average of the critical bands next to it okay and then if it's above a certain level it's very noticeable to the ear and tonen noise does a very similar thing I mean tonen noise is very welln named I think it just looks at the tone and then it looks at the critical band so it looks at the noise um both have their individual application if you have multiple tones in the same critical band uh prominence ratio can be better at determining if that's going to be aning or not anything else in here that you want to cover or externally no I mean um the only other the only other really I think fascinating thing when it comes to Acoustics that's kind of really On The Rise right now because you know when you look at a fan I don't know when the first fan was invented but probably like ancient Egypt you know yeah a long it's been a very well-developed industry and you may think okay there's not a lot to improve nowadays and if you just look at the fan on its own you know free air then it's true there isn't too much you know it's it's a very optimized fan but as you start looking at it in terms of overall product and you look at how each blade interacts with the fins you know the heat sink fins right you can get some really interesting data and and we use Aero Acoustics for this so basically you take this card and then you simulate a box around it right but the issue is at 20 khz your frequency is or your wavelength is like 17 mm right and you want to have somewhere in the range of 100 points along that wavelength so let's say you had this box that you wanted to simulate and now you have to break it up into little cubes that are no longer than I guess that would be .17 mm long right so it's incredibly computationally expensive but the beautiful thing is light propagation and acoustic propagation are very similar and they're different but they're they're relatively similar so a lot of methods have been have been produced that use some of the ray tracing methods oh interesting it's s very similar to Ray tracing so that's why a thermal radiation and acoustic like uh Aero acoustic simulation have like skyrocketed now because of how in terms of computational time like reduced yeah reduc so the abilities are amazing now when you look at how GPU accelerated some of these are so I think it'll be revolutionary going forward I mean we've already learned a lot like even for example when you look at the blade shape and how it interacts with the fins and and everything it's like fascinating yeah it kind of comes back to what we talked about last time which was uh Nvidia in a unique position to use its own GPU Hardware to simulate its own next gen GPU yeah yeah a lot of the data center things that we've been doing just feed right back into G4 it's fascinating to look at and um it's it's like revolutionary for Acoustics and uh like you're saying back to Total product level it's also why you can have a I'll I think nocto is fairly well known for the engineering Focus you can have a company like them invent what is right now to them the best fan they've ever made but then as they tune it for different products you might have to tweak the design which is something we have an interview with them about just because at a you know product level once you put a fin stack behind it Like Everything Changes Everything Changes and how you optimize the fins change too right and uh yeah it's just fascinating um to look at kind of the similarities to R tracing you know yeah yeah so uh next Nvidia product will have Ray traced air R TRC yeah R traced Acoustics that'll be on the marketing box with the actual R tracing performance Malcolm thank you again for the walkthrough always a pleasure always a pleasure I yeah I love talking about this um yeah it's it's uh fascinating stuff we really I mean we really care about Acoustics the overall lising experience so and what better place to do it than here and check out our last video with Malcolm I'll link it below if this interested you thank you all for watching we'll see you all next time\n"