WEBVTT 00:00:00.000 --> 00:00:04.880 All of these cases have a big problem. Can you tell what it is? 00:00:04.880 --> 00:00:08.080 If you said the intake fans are starving for airflow, 00:00:08.080 --> 00:00:13.320 give yourself a gold star. This was a widespread problem in case design for years, 00:00:13.320 --> 00:00:17.240 but new airflow edition cases have flooded the market, 00:00:17.240 --> 00:00:21.440 completely solving it, or have they? See, there's nothing that Lee and Lee 00:00:21.440 --> 00:00:26.540 or Fractal can do about users, shoving their cases right up against the wall 00:00:26.540 --> 00:00:31.080 or on top of a shag carpet. You could be totally killing your gains, bruh. 00:00:31.080 --> 00:00:32.080 Killing your gains. 00:00:34.080 --> 00:00:38.040 But by how much? How much space does a fan need 00:00:38.040 --> 00:00:42.800 before it gets starved for fresh air? To find out, we sent Adam all the way 00:00:42.800 --> 00:00:46.200 to NASA's Langley Research Center in Virginia. 00:00:46.200 --> 00:00:50.760 Oh, sh! To work with some of the top scientists in America 00:00:50.760 --> 00:00:53.820 to figure out just how close is too close 00:00:53.820 --> 00:00:59.120 for good PC cooling. And how close is too close? 00:00:59.120 --> 00:01:01.800 For me to segue, to our sponsor. 00:01:10.920 --> 00:01:16.200 I know how I would try to answer our question, but where would NASA start with a problem like this? 00:01:17.120 --> 00:01:20.160 Great question. We're at the Hypersonic Test Complex 00:01:20.160 --> 00:01:24.920 at the NASA Langley Research Center, and we're gonna do some super scientific stuff, 00:01:24.920 --> 00:01:28.960 and that means we're using this tape and this string. I'm not even joking. 00:01:28.960 --> 00:01:32.080 This is called tufting, and it is decidedly low-tech, 00:01:32.080 --> 00:01:36.120 but it remains one of the most important forms of aerodynamic testing that can be done, 00:01:36.120 --> 00:01:39.400 and it's basically where NASA starts every single time. 00:01:39.400 --> 00:01:42.000 You see, air is difficult to study because it looks like this. 00:01:44.360 --> 00:01:47.420 So we rely on the way it affects other things 00:01:47.420 --> 00:01:51.440 to understand its behavior. Wow, I wasn't that far off. 00:01:51.440 --> 00:01:55.160 What you're looking at is a Noctua NF-A12X25, 00:01:55.160 --> 00:01:59.160 with some acrylic acting as our airflow restrictor or our panel. 00:01:59.160 --> 00:02:02.880 With tufts or little pieces of string on the backside of the fan, 00:02:02.880 --> 00:02:07.160 we can see how adjusting the distance of our panel from three and a half centimeters 00:02:07.160 --> 00:02:10.320 down to just 0.5 centimeters affects our airflow. 00:02:10.320 --> 00:02:13.480 We'll put the specifics up on the screen, and you can pause and read them. 00:02:13.480 --> 00:02:17.520 Actually, we changed our minds. It's an article on the lab's website. It's in the description. 00:02:21.680 --> 00:02:24.760 So what did we find out? Well, a few things. 00:02:24.760 --> 00:02:29.200 As you'd expect, at an ample distance from our panel, our fan performs admirably, 00:02:29.200 --> 00:02:34.080 sucking air in on one side and blowing it out the other. What I didn't expect, though, 00:02:34.080 --> 00:02:38.320 was that the tufts near the center only started to get a little floppy 00:02:38.320 --> 00:02:42.760 when the front panel got really close, just one and a half to two centimeters 00:02:42.760 --> 00:02:47.400 from the face of the fan. Another thing I didn't expect was that getting even closer 00:02:47.400 --> 00:02:50.560 caused the fan to not only blow in effectively, 00:02:50.560 --> 00:02:54.400 but even start sucking the tufts back into the blades, 00:02:54.400 --> 00:02:59.040 indicating reversed airflow. But it's a little hard to see like this. 00:02:59.040 --> 00:03:04.000 Oh yeah, I forgot. These are strings they glow in the dark. We took this big-ass NASA grade ultraviolet lamp 00:03:04.000 --> 00:03:08.680 to get maximum glow from our tufts while we recorded on this beast, 00:03:08.680 --> 00:03:12.520 the Kronos 4K12 high-speed camera. 00:03:12.520 --> 00:03:16.360 We're prepping our 1000 FPS 4K camera. 00:03:16.360 --> 00:03:19.560 That's about 11 gigabits per second of data, 00:03:19.560 --> 00:03:24.480 which means that recording on this for a minute is going to be bigger than your call-of-duty install. 00:03:24.480 --> 00:03:27.840 And in slow motion, we can get a much closer look. 00:03:27.840 --> 00:03:32.400 Note the increased tuft movement in our close-up test condition, indicating more turbulent flow. 00:03:32.400 --> 00:03:35.680 And you can even see some of the tufts dragging into the middle of the fan 00:03:35.680 --> 00:03:41.880 where there's areas of low pressure. Pretty cool, huh? Super cool, Adam, but we have Kronos camera at home. 00:03:41.880 --> 00:03:45.720 Why did we need NASA for this? We didn't, but for what we're going to do next, 00:03:45.720 --> 00:03:49.360 we sure did. And besides, if NASA invites you out, 00:03:49.360 --> 00:03:52.720 are you really going to say no? They were even nice enough to give us a whole tour 00:03:52.720 --> 00:03:56.560 that included an incredible new makerspace that they use for rapid prototyping. 00:03:56.560 --> 00:04:00.120 Our supporters on Floatplane get exclusive access to that and a ton of other content, 00:04:00.120 --> 00:04:04.280 so if you want to check that out, head over to LMG.GG slash Floatplane. 00:04:04.280 --> 00:04:07.640 For now, I'll just give you a TLDR. The Langley Research Center 00:04:07.640 --> 00:04:11.920 is kind of the OG NASA facility. In fact, it predates NASA itself, 00:04:11.920 --> 00:04:15.880 founded in 1917 as part of the NACA, which is the aeronautics organization 00:04:15.880 --> 00:04:19.120 that would eventually become NASA. The Langley Research Facility 00:04:19.120 --> 00:04:22.640 was keyed in numerous discoveries and improvements to early flying machines, 00:04:22.640 --> 00:04:25.720 creating the first hypersonic jets and paving the way for space travel 00:04:25.720 --> 00:04:30.680 and would go on to play a key role in the Apollo missions. Currently, one of the focuses of the research center 00:04:30.680 --> 00:04:35.280 is assisting the Artemis campaign to get to the moon as a means to get to Mars, pretty cool. 00:04:35.280 --> 00:04:39.840 A little bit out of the scope that we're going to be working on today. So we're going to steal a couple of the 3,500 employees 00:04:39.840 --> 00:04:43.360 that come here every day to do some more down-to-earth testing. 00:04:43.360 --> 00:04:46.800 We're inside of the Hypersonic Test Complex at the NASA Langley Research Center 00:04:46.800 --> 00:04:50.360 to answer that... Wait a second, this is the same lab? 00:04:50.360 --> 00:04:53.840 Yep, same lab. I thought we were supposed to go somewhere that was like more high-tech. 00:04:53.840 --> 00:04:55.600 Oh, it's more high-tech over there. 00:04:56.880 --> 00:05:00.400 Oh, right, and this is Dr. Lewis Edelman. He's a researcher here at NASA, 00:05:00.400 --> 00:05:03.400 and he's helped design all of these experiments that we're going to be running today. 00:05:03.400 --> 00:05:06.720 Hello. So what the heck are we doing? So we've moved on from tufting 00:05:06.720 --> 00:05:09.920 to particle image velocimetry, or PIV. 00:05:09.920 --> 00:05:13.760 Oh. Flashing lights in fancy cameras. Sounds like my kind of deal. 00:05:13.760 --> 00:05:23.440 First, we shine a bright light focused 00:05:23.440 --> 00:05:29.000 into a thin vertical sheet. Then we fill the air with, you guessed it, particles 00:05:29.000 --> 00:05:33.560 and start rapidly taking pictures. Now, that might sound like video, 00:05:33.560 --> 00:05:38.520 but it's different in one important way. Instead of taking images at a constant frame rate, 00:05:38.520 --> 00:05:42.360 we instead take two samples just microseconds apart, 00:05:42.360 --> 00:05:45.440 followed by a short gap, then another two samples, 00:05:45.440 --> 00:05:48.640 and so on and so forth. We can then use really clever math 00:05:48.640 --> 00:05:54.800 to determine the velocity of the particles. Note that this is not speed, but velocity, 00:05:54.800 --> 00:05:59.120 because we're talking about how fast they're going and also their direction. 00:05:59.120 --> 00:06:03.000 Sounds simple, right? I mean, not really, but for starters, 00:06:03.000 --> 00:06:08.480 the particles we're tracking are not air, which means that they won't move exactly like air. 00:06:08.480 --> 00:06:12.480 Second, to do this right, we need the velocity of many particles. 00:06:12.480 --> 00:06:16.400 That means thousands of samples across thousands of images. 00:06:16.400 --> 00:06:19.520 And third, we need some pretty special cameras. 00:06:19.520 --> 00:06:23.520 The Levision Flowmaster is a super high precision machine 00:06:23.520 --> 00:06:28.480 that uses a double frame buffer to take photos just nanoseconds apart. 00:06:28.480 --> 00:06:31.840 All right, it's finally time to test. 00:06:31.840 --> 00:06:37.680 Hit it. So this is our first test result. 00:06:41.920 --> 00:06:45.200 I mean, there's no obstruction on this fan. No obstruction, pure control. 00:06:45.200 --> 00:06:48.240 So we're doing the analysis now on one image pair 00:06:48.240 --> 00:06:53.360 as a test for the processing stream, and we go, okay, that's not very clean. 00:06:53.360 --> 00:06:56.720 Once we take an average of the 200 or 158 images 00:06:56.720 --> 00:07:00.800 we just took, this will fill out to look like a fairly nice picture. 00:07:00.800 --> 00:07:04.720 Importantly, that this is kind of the hub, like we're really only looking at the top half 00:07:04.720 --> 00:07:08.920 of the fan right now. We're looking at the top half of the fan. Would it be fair to expect that there would be, 00:07:08.920 --> 00:07:13.680 I mean, it's a circle, so there'd be symmetry on the bottom? I would expect radial symmetry in all things 00:07:13.680 --> 00:07:18.120 you wanna try and measure in as much detail as possible. So if we were fully zoomed out 00:07:18.120 --> 00:07:22.280 and looking at the whole fan, we'd be wasting resolution effectively. 00:07:22.280 --> 00:07:27.440 So this is it, applying the scaling from pixel space to physical space 00:07:27.440 --> 00:07:31.360 from our calibration plate. Now it's starting to do the PIV, 00:07:31.360 --> 00:07:35.000 and we can see for our 158 frames, 00:07:35.000 --> 00:07:38.960 that's probably gonna take about 20 minutes. What is this computer running on? 00:07:38.960 --> 00:07:41.080 This is a 00:07:42.280 --> 00:07:47.280 i9-14900K with 192 gigs of RAM. 00:07:48.880 --> 00:07:53.240 So this just takes a long time, is what you're telling me. Yes. After considerable number crunching, 00:07:53.240 --> 00:07:57.440 we've got our results, and what you're looking at is a narrow slice of air 00:07:57.480 --> 00:08:02.000 that is flowing out of the fan. The colors indicate the stream-wise velocity of the air, 00:08:02.000 --> 00:08:05.200 so how fast it's going away from our fan blades, 00:08:05.200 --> 00:08:10.200 and then the arrows are kinda just an easier way to visually process that same information. 00:08:10.200 --> 00:08:16.640 Then to make it even simpler, we added these dots. Without a front panel, flow is smooth and fast moving. 00:08:16.640 --> 00:08:20.400 There's a small section where you can see there's no flow, but if you look closely, 00:08:20.400 --> 00:08:23.840 that's right behind the fan hub where there are no blades. 00:08:23.840 --> 00:08:28.720 Now, it is important to remember that this is just a thin slice of the overall airflow 00:08:28.720 --> 00:08:32.000 in what is a 3D space. Our test isn't gonna capture 00:08:32.000 --> 00:08:37.000 the spiraling 3D vortex of the air, but the overall direction away from the fan 00:08:37.000 --> 00:08:40.720 is what's most important for cooling, so this is good enough for our purposes. 00:08:40.720 --> 00:08:44.960 Fun observation, by the way. Dr. Edelman noted that the way that Noctua's fans 00:08:44.960 --> 00:08:49.600 throw momentum inward more than a typical fan contributes to reducing their overall noise. 00:08:49.600 --> 00:08:53.760 Good job, Noctua. When we move the plate closer, we don't see much change. 00:08:53.760 --> 00:08:57.200 That is, until, just like in our Tuft test, 00:08:57.200 --> 00:09:02.920 we get as close as about 15 millimeters. Take a look at how large our dead zone has become now. 00:09:02.920 --> 00:09:06.960 We also noticed that the flow of air is starting to curl outward 00:09:06.960 --> 00:09:12.680 rather than coming straight out of the fan. What that means is lower stream-wise momentum 00:09:12.680 --> 00:09:18.000 to blow air across your components or to pass through a restrictive heat sink or radiator. 00:09:18.000 --> 00:09:21.440 But why? Well, it's due to the difference in radial pressure. 00:09:21.480 --> 00:09:27.800 The air that passes through the tip of the fan blades is lower pressure and almost bounces off the stagnant 00:09:27.800 --> 00:09:31.240 and thus higher pressure air that's right by the fan hub. 00:09:31.240 --> 00:09:35.040 This isn't optimal, but our overall airflow 00:09:35.040 --> 00:09:38.040 is still pretty good, so for a case fan, 00:09:38.040 --> 00:09:43.320 it's probably still fine to have this much restriction. We will check on this test condition again later 00:09:43.320 --> 00:09:47.640 once we add a radiator. For now, let's look at our worst case scenario in open air. 00:09:47.640 --> 00:09:53.720 Imagine your case is right up against a wall or your power supply intake is on the floor on a carpet. 00:09:53.720 --> 00:09:57.960 This is what you're doing to your poor, poor PC. 00:09:57.960 --> 00:10:01.880 Not only is the fan barely pulling any air in at the edges, 00:10:01.880 --> 00:10:05.520 it is so starved for air that there's a reverse flow 00:10:05.520 --> 00:10:09.800 that causes the air to curl into a vortex that isn't gonna move heat anywhere 00:10:09.800 --> 00:10:14.760 and that's your best case scenario. What if we were trying to cool something directly 00:10:14.760 --> 00:10:19.080 with our starved fan? To find out, we whipped out a water cooling radiator 00:10:19.080 --> 00:10:24.520 to create a scenario with much higher back pressure. You can think of back pressure as the friction 00:10:24.520 --> 00:10:29.240 that a fluid experiences in movement, and as you can see, adding a ton of friction 00:10:29.240 --> 00:10:34.960 to an already restricted fan results in a two-word review of my debut rap album, Zero Flow. 00:10:34.960 --> 00:10:39.760 Now naturally, backing this off to a 15 millimeter gap yields much better results, 00:10:39.760 --> 00:10:44.720 but it's still worth noting that this is a massive drop in performance compared to our free air test, 00:10:44.720 --> 00:10:47.960 especially with respect to the size of our dead zone over the fan hub. 00:10:47.960 --> 00:10:53.280 Practically speaking, we're only getting flow on about the outer 50% of the fan blades. 00:10:53.280 --> 00:10:56.760 Interestingly though, the radiator completely straightens out the flow 00:10:56.760 --> 00:11:01.120 and we don't see that same outward curling, but the speed is cut about in half. 00:11:01.120 --> 00:11:05.840 So what does all of this mean? Well, we can't draw overly broad conclusions. 00:11:05.840 --> 00:11:13.120 We only tested the NFA 1225 at full speed, but it seems like you can get as close as about 15 millimeters 00:11:13.120 --> 00:11:17.360 or a little over half an inch from your fan with reasonable performance. 00:11:17.360 --> 00:11:21.080 Any closer and you are seriously harming its cooling ability. 00:11:21.080 --> 00:11:24.220 An obvious question would be, why go to all this trouble? 00:11:24.220 --> 00:11:29.280 Couldn't you have just sent a piece of acrylic and a fan to Cybernetics to put in front of their fan tester? 00:11:29.280 --> 00:11:33.840 Well, yes, but also no. While a fan tester would tell us the performance 00:11:33.840 --> 00:11:38.720 of the fan under various conditions, we were more interested in measuring the behavior 00:11:38.720 --> 00:11:41.960 of the air, which tells us not only the answer, 00:11:41.960 --> 00:11:45.440 but it also lifts the veil on why the answer is what it is. 00:11:45.440 --> 00:11:50.000 It also gave us an excuse to check out another really cool piece of kit down in Virginia. 00:11:50.000 --> 00:11:53.240 If you thought air was complex, we'll just wait till human perception gets involved. 00:11:53.240 --> 00:11:56.740 Yeah, we're gonna talk about sound. Now, we have been hard at work improving 00:11:56.740 --> 00:11:59.960 our audio testing at LTT Labs. I mean, we even just built a home theater room 00:11:59.960 --> 00:12:06.160 to test more speakers now. But what we don't have is a NASA grade audio chamber. 00:12:06.160 --> 00:12:09.880 This is the shack and it's a little old place where we can do some testing. 00:12:09.880 --> 00:12:16.840 The small hover anechoic chamber was constructed in the late 80s and it gets as quiet as 18 dB, 00:12:16.840 --> 00:12:20.360 which is disconcertingly quiet. 00:12:20.360 --> 00:12:24.440 For our testing today, we're using two different arrays, a linear one up there and a spiral one. 00:12:24.440 --> 00:12:28.040 Why are they different shapes? Well, they have two different jobs. The linear array is a directivity array 00:12:28.040 --> 00:12:31.320 that gives us a broad idea of where sound ends up in the chamber. 00:12:31.320 --> 00:12:36.160 The spiral array, also called a phase array, is a collection of 40 beamforming MEMS microphones 00:12:36.160 --> 00:12:40.520 that allows us to get super detailed information about the source of a sound. 00:12:40.520 --> 00:12:44.440 These microphones are so precise that we can map the location of the sound 00:12:44.440 --> 00:12:47.920 onto footage from the camera that sits in the middle of the array. 00:12:47.920 --> 00:12:52.760 Essentially, you can define a region in here and it will perform an integration. 00:12:52.760 --> 00:12:56.760 So right, it's a little hard to see, but right now it's just region one, two, and three. 00:12:56.760 --> 00:13:00.960 And then you can basically go ahead and have it process it and give you a spectrum like this 00:13:00.960 --> 00:13:04.560 with the different contributions of those regions to the total sound field, 00:13:04.560 --> 00:13:09.320 or at least what their array picked up. Why would folks at NASA need to know this kind of stuff? 00:13:09.320 --> 00:13:13.520 Well, when they're testing the acoustic properties of something, they want to know where that sound's going to be coming from. 00:13:13.520 --> 00:13:18.160 Like this, for example, they might want to understand if the sound is coming from the tips of the blades 00:13:18.160 --> 00:13:22.240 or if it's coming from the rotor itself. There's only one way to find that out, 00:13:22.240 --> 00:13:25.960 and that's to use the phase array. I mean, there's probably another way to find out, 00:13:25.960 --> 00:13:29.040 but the way we're going to do is the phase array. So stop asking questions. 00:13:29.040 --> 00:13:32.880 Well, you can ask one more question. If you noticed, our fan is looking a little pink. 00:13:32.880 --> 00:13:35.880 Why would folks at... 00:13:35.880 --> 00:13:40.160 That's because this apparatus was designed to test rotors for things like drones. 00:13:40.160 --> 00:13:43.720 So our NF-A12X25 that we were using before, 00:13:43.720 --> 00:13:47.560 it was a bit too quiet, and we swapped it out for this industrial version 00:13:47.560 --> 00:13:53.360 that runs considerably faster and considerably louder. The paint that's on it is this funky pressure sensitive paint 00:13:53.360 --> 00:13:56.360 that we were going to use for another test, but we ran out of time 00:13:56.360 --> 00:13:59.720 because NASA has a lot of important work to do. Anyway, I bring up the paint 00:13:59.720 --> 00:14:04.000 because it might make the fan perform a bit worse than what Noctua would ship from the factory, 00:14:04.000 --> 00:14:08.120 but we aren't comparing it to other fans. We're only comparing the intake clearance, 00:14:08.120 --> 00:14:13.800 so we're not worried about that. Again, because of time constraints, we decided to just test the fan without the front plate 00:14:13.800 --> 00:14:18.760 and at the 15 millimeter point. Now, intuitively, you might think that covering a fan 00:14:18.760 --> 00:14:22.560 would decrease the amount of noise that it makes, right? 00:14:22.560 --> 00:14:27.920 But if you've ever tried placing your hand in front of your PC fan, you might have noticed that it often gets louder 00:14:27.920 --> 00:14:32.880 up until the point where it becomes completely started there. And in our test conditions, that's true. 00:14:32.880 --> 00:14:37.240 You can see a broad spectrum increase in noise when the front panel's present. 00:14:37.240 --> 00:14:40.400 Why? Well, referring back to our PIV results, 00:14:40.400 --> 00:14:45.840 remember the stalled flow in the middle? That causes the overall airflow to be more unsteady, 00:14:45.840 --> 00:14:49.400 which makes it louder. Think of it like roaring rapids 00:14:49.400 --> 00:14:52.800 versus the smooth flowing water in the Mississippi Delta. 00:14:52.800 --> 00:14:56.840 And this increase in noise also shows up in our phase array results. 00:14:56.840 --> 00:15:01.000 Yet another reason to not let your fans get too close to obstructions. 00:15:01.000 --> 00:15:05.000 Even if you have cutouts in the side panels, those can still cause annoying resonances. 00:15:05.000 --> 00:15:09.320 Fractal terra owners will know this very well. In summary then, for performance, 00:15:09.320 --> 00:15:15.080 keep your fans more than 15 millimeters away from any surfaces and 20 millimeters or more 00:15:15.080 --> 00:15:18.960 if you're gonna be contending with other obstructions like a heat sink or a radiator. 00:15:18.960 --> 00:15:23.440 As for noise, it seems like you can't have too much clearance from the intake, 00:15:23.440 --> 00:15:27.080 but you can definitely have too little. There are a hundred other questions 00:15:27.080 --> 00:15:31.560 we would have loved to answer, but frankly, NASA was already extremely generous 00:15:31.560 --> 00:15:35.280 with their time and they are hard at work answering much larger questions. 00:15:35.280 --> 00:15:38.520 This may not have been the be all and end all answer that you were looking for, 00:15:38.520 --> 00:15:42.440 but that's just how science is. It's the accumulation of many tiny discoveries, 00:15:42.440 --> 00:15:47.080 solving many tiny mysteries that build up to create a knowledge base that allows us to 00:15:47.080 --> 00:15:50.080 have a greater depth of understanding of the world that we live in. 00:15:50.080 --> 00:15:53.880 Before we go, I wanna thank all the folks that Massa, who helped made this possible, 00:15:53.880 --> 00:15:58.000 Lewis, Brittany, Nick, Jordan, so many more folks that I haven't named 00:15:58.000 --> 00:16:03.120 helped make this trip possible. Thank you so much. Thank you for watching and of course, 00:16:03.120 --> 00:16:06.200 thanks for the segue to our sponsor. Thanks for watching. 00:16:06.200 --> 00:16:09.280 If you like this video, why don't you watch another one of our videos where we tour some cool place, 00:16:09.280 --> 00:16:12.880 like how about there's an internet exchange in Toronto 00:16:12.880 --> 00:16:17.000 that was pretty dang neat. Thanks again to NASA. Thanks for watching. 00:16:17.000 --> 00:16:17.840 Bye.