WEBVTT

00:00:00.000 --> 00:00:04.920
How do you fix a problem you can't even see?

00:00:04.920 --> 00:00:11.040
In smaller systems, it's not that hard. Air leak on your tire, just throw some soapy water on it

00:00:11.040 --> 00:00:15.480
and find the bubbles. Easy, but in a larger system,

00:00:15.480 --> 00:00:22.640
like our workshop's compressed air hoses, it could take a fluke to find a tiny leak somewhere.

00:00:22.640 --> 00:00:27.080
A fluke II915 acoustic imager to be exact.

00:00:27.120 --> 00:00:32.120
This $25,000 handheld camera does not work

00:00:32.560 --> 00:00:37.320
like a normal camera, you know, taking in light and transforming it into internet cloud.

00:00:37.320 --> 00:00:42.160
Instead, it takes in sound and uses it to find leaks

00:00:42.160 --> 00:00:45.200
and equipment malfunctions, even detecting noises

00:00:45.200 --> 00:00:50.520
beyond the range of human hearing. It then represents that audio like a heat map.

00:00:50.520 --> 00:00:55.440
So let's see if we can use it to find some really useful stuff like the leak in our system

00:00:55.560 --> 00:00:59.600
and maybe even a segue to our sponsor.

00:01:09.600 --> 00:01:15.000
Like so many of our videos, this one was prompted by an actual problem.

00:01:17.160 --> 00:01:21.160
That, obviously it helps a little that we shoot most of our workshop videos

00:01:21.160 --> 00:01:25.800
on the other side of this wall, but it's still been a major source of disruptions

00:01:25.800 --> 00:01:30.440
over the last few years. Now, an obvious solution would be to just

00:01:30.440 --> 00:01:34.440
not have a compressed air system, but that's not really an option.

00:01:34.440 --> 00:01:42.400
We use it for everything from cleaning things off to cutting and polishing to hacking things apart.

00:01:42.440 --> 00:01:45.600
Excuse me. And I've done a fair bit of sandblasting with it too,

00:01:45.600 --> 00:01:50.440
mostly for my personal motorbike project. So, okay, we need an air compressor.

00:01:50.440 --> 00:01:54.160
Deal with it, right? Well, see, it wouldn't be that big of a problem,

00:01:54.160 --> 00:01:57.880
except that somewhere in the system is a leak.

00:01:57.880 --> 00:02:01.000
So the compressor goes off way more often,

00:02:01.000 --> 00:02:04.240
disrupting us way more often than it needs to.

00:02:04.240 --> 00:02:09.280
That's where Fluke's acoustic imager comes in. Theoretically with this, in a matter of minutes,

00:02:09.280 --> 00:02:13.280
I should be able to track down the leak, saving us both time and money.

00:02:13.280 --> 00:02:17.640
Theoretically, I'm gonna put that to the test. I've never actually used this before,

00:02:17.640 --> 00:02:22.000
so I don't know how I'm gonna tell the difference between RHVAC system and air leaking out of a hose,

00:02:22.000 --> 00:02:23.960
but presumably it's pretty easy.

00:02:25.040 --> 00:02:26.840
Oh, there's something in this wall.

00:02:28.040 --> 00:02:31.240
Oh, there's something bouncing off this wall, I think.

00:02:31.240 --> 00:02:34.600
You'll get a lot of sounds reflecting from that sometimes.

00:02:34.600 --> 00:02:37.840
That makes sense. Keep an eye on the right side,

00:02:37.840 --> 00:02:41.840
and you'll see the range of frequency band that it's looking for.

00:02:41.840 --> 00:02:45.440
Okay, I feel like I could move pretty quickly at this point.

00:02:45.440 --> 00:02:48.680
Oh, hello, Captain, I think we've found something.

00:02:50.880 --> 00:02:54.360
Right there. Okay, but would I even be able to hear this?

00:02:54.360 --> 00:02:58.720
Oh, I can. I would have to be within like 10 inches of it, though.

00:02:58.720 --> 00:03:01.500
One to go? And that took like two minutes.

00:03:03.920 --> 00:03:07.000
Oh, geez. Okay, once you know what you're looking for,

00:03:07.000 --> 00:03:08.840
it stands out like a beacon.

00:03:10.240 --> 00:03:13.640
Oh yeah, this one's a lot more obvious.

00:03:13.680 --> 00:03:17.600
So between these two, this one was a pretty substantial one.

00:03:17.600 --> 00:03:21.400
We found both of them within just a couple minutes. This is a lot like the video Adam Savage

00:03:21.400 --> 00:03:25.160
did with this thing, hey? Two cool people can have the same cool idea

00:03:25.160 --> 00:03:29.120
at the same time. Adam Savage and ElectroBoom.

00:03:29.120 --> 00:03:33.240
And also me, just slower. I wanna show you guys,

00:03:33.240 --> 00:03:36.800
separate from the video mode, you can put it in leak Q mode.

00:03:36.800 --> 00:03:40.400
This allows you to take a picture of a leak,

00:03:40.400 --> 00:03:45.120
then according to your settings. So gas price, compression energy cost,

00:03:45.120 --> 00:03:48.200
as well as your currency, it will tell you how much that leak

00:03:48.200 --> 00:03:52.920
is gonna cost you a year. I had no idea that just a little air leak

00:03:52.920 --> 00:03:57.560
could cost $45 a year. So replacing the fitting is a no-brainer.

00:03:57.560 --> 00:04:00.560
That assumes it's running all year. I mean, it is running all year.

00:04:00.560 --> 00:04:04.560
We never turn it off. Like that's the whole thing with a leak. It's like a faucet, right?

00:04:04.560 --> 00:04:09.000
Like every once in a while, drip will waste like more gallons in a month

00:04:09.000 --> 00:04:13.040
than you could possibly fathom. Now, obviously, if the only gases you're working with

00:04:13.040 --> 00:04:16.360
are compressed air, it would take a little while for an investment like this

00:04:16.360 --> 00:04:22.400
to pay for itself. But in certain industries, this could be worth its weight in gold in cost savings.

00:04:22.400 --> 00:04:26.840
Now, while the workshop team gets those fittings sorted out, I'm gonna go see what else this thing can do.

00:04:26.840 --> 00:04:31.320
The magic of this device is really in its algorithm and precise calibration.

00:04:31.320 --> 00:04:35.920
The hardware itself is pretty simple. This array of 64 microphones

00:04:35.920 --> 00:04:39.200
is arranged in a logarithmic spiral to capture sound,

00:04:39.200 --> 00:04:42.600
and it uses beamforming to determine the source of the sound.

00:04:42.600 --> 00:04:48.480
Then it overlays that on the camera image as what Fluke calls a sound map.

00:04:48.480 --> 00:04:52.200
In the default mode, a sound map looks an awful lot like a heat map.

00:04:52.200 --> 00:04:58.200
All of this is possible because the microphones are all in known locations relative to both each other

00:04:58.200 --> 00:05:04.240
as well as to the camera right in the center. So sound hits the array and with a little bit of maths,

00:05:04.240 --> 00:05:07.480
the source can be determined by analyzing the reception delays.

00:05:07.480 --> 00:05:12.020
There are some limitations though. It's microphones pick up sound between two kilohertz

00:05:12.020 --> 00:05:15.960
and 100 kilohertz, which is well above the range of human hearing.

00:05:15.960 --> 00:05:19.560
So if you wanted to find something like a small liquid leak,

00:05:19.560 --> 00:05:25.200
say in our water cooling loop here, the frequency of a drip drip drip

00:05:25.200 --> 00:05:28.440
might actually be too low for the Fluke to pick up.

00:05:28.440 --> 00:05:34.360
Fortunately, we have an idea for that. Why not turn our liquid leak into

00:05:36.840 --> 00:05:40.200
an air leak? All we need to do is pressurize our loop.

00:05:42.400 --> 00:05:45.760
Well, that was easy. Oh, you put a hole in the tube.

00:05:45.760 --> 00:05:49.520
Usually the leaks are a lot smaller than that in these water cooling systems.

00:05:49.520 --> 00:05:54.760
I kind of wanna know if it can handle a much smaller leak like an O-ring issue or something.

00:05:54.760 --> 00:05:59.480
Every once in a while, especially on hard line fittings, these O-rings will go missing

00:05:59.480 --> 00:06:06.200
and the system can look like it's all sealed up, but when that drip gets on the back of your GPU,

00:06:06.200 --> 00:06:10.960
you're gonna have a pretty bad time. Put this up in the blue zone.

00:06:10.960 --> 00:06:16.240
You can see, even though there's no O-ring in there, it's sealed right now,

00:06:16.240 --> 00:06:19.800
but let's see if our Fluke can pick it up. Oh, dude.

00:06:21.520 --> 00:06:27.540
Easy. And that is a much, much smaller leak.

00:06:28.720 --> 00:06:33.200
What a trouble saver that would be. Now it is worth noting that our brilliant strategy

00:06:33.200 --> 00:06:36.480
of just emptying the loop and pumping it up

00:06:36.480 --> 00:06:39.920
isn't gonna work in every situation. Clean rooms, for instance,

00:06:39.920 --> 00:06:42.920
need to maintain strict control of airborne contaminants.

00:06:42.920 --> 00:06:46.280
And one of the ways that they do this is by building them to be fairly airtight

00:06:46.280 --> 00:06:49.520
and to maintain positive pressure inside, not a ton of pressure,

00:06:49.520 --> 00:06:52.720
cause like humans need to work in there, but some pressure.

00:06:52.720 --> 00:06:56.320
So how would they track down then a low pressure leak?

00:06:56.320 --> 00:07:02.580
Turns out the answer is Fluke's Beacon. This little device emits sound at 40,000 Hertz,

00:07:02.580 --> 00:07:07.740
well beyond what would bother any human. You just pop it into your system, seal it up,

00:07:07.740 --> 00:07:12.300
tune it to 40,000 Hertz, and theoretically we should be able to find our leak.

00:07:12.300 --> 00:07:16.940
I can definitely detect it through the acrylic now, but our system may be too complex

00:07:16.940 --> 00:07:21.740
for such a short wavelength to travel through like that.

00:07:21.740 --> 00:07:26.460
It's probably just because this knows there's sound coming from here or somewhere off the side,

00:07:26.460 --> 00:07:29.980
and so it's not gonna pick up the same frequency somewhere else. Hold it up over there then.

00:07:29.980 --> 00:07:33.420
Oh, dude, bingo. You know what, this would be a great opportunity

00:07:33.420 --> 00:07:37.740
to check the multiple source feature here. Ooh, yeah.

00:07:37.740 --> 00:07:40.980
So in acoustics, we can click multiple sources,

00:07:40.980 --> 00:07:44.540
and then, oh, dude, that's so cool.

00:07:44.540 --> 00:07:49.700
It'll show the different levels, however, those still totally overpower it.

00:07:49.700 --> 00:07:53.380
Really? Yeah, it really doesn't care about this leak unless it's the only one on screen.

00:07:53.380 --> 00:07:56.600
That's so funny. Very interesting.

00:07:56.600 --> 00:08:00.760
Wonder what else we could use this for? A really cool application for acoustic imagers

00:08:00.760 --> 00:08:06.760
is identifying problems with transmission towers at a distance, say, from the ground.

00:08:06.760 --> 00:08:10.560
It apparently works at up to 120 meters, or 393 feet,

00:08:10.560 --> 00:08:14.280
and can be used to detect partial discharge or arcing,

00:08:14.280 --> 00:08:18.680
the sort of thing that could lead to a forest fire. Now, we don't have anything like that to test,

00:08:18.680 --> 00:08:23.520
but we do have some whiny electronics like this GPU

00:08:23.520 --> 00:08:28.060
that has some coil wine. Now, coil wine isn't dangerous like those transmission tower problems.

00:08:28.060 --> 00:08:31.060
It's just vibrations in board components,

00:08:31.060 --> 00:08:36.380
specifically the coils, as electricity passes through. It's just, it can be pretty flipping annoying.

00:08:36.380 --> 00:08:41.260
Now, you can see that I'm not picking anything up right now. That's because this GPU isn't under load.

00:08:41.260 --> 00:08:45.700
Let's just plug in our keyboard using a handy-dandy TrueSpec cable from lttstore.com

00:08:45.700 --> 00:08:49.180
and fire up a GPU-intensive application.

00:08:49.180 --> 00:08:52.500
There it is, but where is it coming from specifically?

00:08:52.500 --> 00:08:56.520
There's a hotspot right where the 12 volt two by six

00:08:56.560 --> 00:09:00.240
comes into the back of the board. Yep, it's right across the back of the card

00:09:00.240 --> 00:09:04.520
where all the VRMs would be. Doesn't really help us solve the problem, but now we know.

00:09:04.520 --> 00:09:09.400
Actually, I had another situation very recently where coil wine was driving me absolutely nuts.

00:09:09.400 --> 00:09:12.400
I used these USB fiber optic docks

00:09:12.400 --> 00:09:17.420
in order to have my system downstairs, but my monitor and all of my peripherals upstairs.

00:09:17.420 --> 00:09:22.480
I was hearing this like high-pitched, almost like a CRT TV in the next room.

00:09:22.480 --> 00:09:27.060
Why? Now I was like, can anyone else hear that? Where is that coming from? Everyone's looking at me like I'm crazy.

00:09:29.060 --> 00:09:35.100
I'm having trouble getting it to go lower. Change the decibels from automatic to manual.

00:09:35.100 --> 00:09:39.820
It's trouble-loat when you get into the low frequencies because if you talk, you can actually see it.

00:09:39.820 --> 00:09:43.660
Exactly, when it's something that you can hear, like coil wine, it's always gonna be

00:09:43.660 --> 00:09:46.820
in the frequency that we can hear, so it can be a little bit of trouble.

00:09:46.820 --> 00:09:50.260
Right at like 10,000 Hertz.

00:09:50.260 --> 00:09:54.580
That's not even that high-pitched. Why couldn't anybody else hear this? It was driving me freaking nuts.

00:09:55.100 --> 00:09:59.420
Everyone can hear 10,000 Hertz, right? Some with hearing loss or significant age-related

00:09:59.420 --> 00:10:02.420
hearing decline might struggle to detect it. So you shouldn't have had any problems.

00:10:02.420 --> 00:10:06.940
Listen, listen. It's really cool how adjusting the frequency band,

00:10:06.940 --> 00:10:10.100
as well as the sensitivity, can really help you dial it in

00:10:10.100 --> 00:10:13.860
to detect particular things. Of course, your next question is probably,

00:10:13.860 --> 00:10:17.700
but what can I do about coil wine Linus? Some people have reported success

00:10:17.700 --> 00:10:24.380
putting schmoo over these vibrating components, but your mileage, not to mention your warranty,

00:10:25.340 --> 00:10:29.620
may vary. Is this kind of analysis a practical application for such an expensive device?

00:10:29.620 --> 00:10:33.660
Yeah, probably not, but there are more affordable options

00:10:33.660 --> 00:10:37.980
for this sort of thing. The Fluke II 500 is about eight and a half thousand dollars,

00:10:37.980 --> 00:10:41.020
but is just limited to the leak Q functionality

00:10:41.020 --> 00:10:44.380
that we showed you guys earlier with a more limited frequency band.

00:10:44.380 --> 00:10:48.100
Our II 915 has some added functions like mech Q,

00:10:48.100 --> 00:10:53.460
which detects things like a faulty bearing in an assembly line to help identify and fix problems

00:10:53.460 --> 00:10:57.260
during planned downtime, rather than when they just come up unexpectedly.

00:10:57.260 --> 00:11:01.180
But if you want to go even cheaper, you totally can. Other options are out there,

00:11:01.180 --> 00:11:07.780
and there are even open source acoustic imagers. The Cessenta, a modular system, looks like a standout.

00:11:07.780 --> 00:11:12.060
Of course, that sort of thing means that the calibration is entirely up to the user.

00:11:12.060 --> 00:11:15.500
If you're a business fan, you might not want to deal with that. All right, do you think they're done

00:11:15.500 --> 00:11:20.300
with the fixes in the shop? How'd all the fixes go? We're thinking the regulator was leaking.

00:11:20.300 --> 00:11:24.340
It may have been the hose, but we swapped both out. I wanted to replace the hoses anyway.

00:11:24.340 --> 00:11:28.900
The connection before was questionable at best. Tore the regulator down, rebuilt it, and no leaks at all.

00:11:28.900 --> 00:11:31.460
Sometimes you just have to take it apart and put it back together.

00:11:32.380 --> 00:11:35.900
This one was a little bit different. I knew that there was a leak in here because this wasn't tight enough,

00:11:35.900 --> 00:11:40.100
but I discovered that there was also an O-ring missing in her fasteners.

00:11:40.100 --> 00:11:43.820
Just completely missing? Yeah, I was actually quite surprised

00:11:43.820 --> 00:11:47.820
that it wasn't leaking worse than it was, but we ended up getting it put back in

00:11:47.820 --> 00:11:51.300
and now it's good and tight. So that was never in there?

00:11:51.300 --> 00:11:54.740
No. The whole time. Yeah.

00:11:56.500 --> 00:12:01.980
So it was only the two leaks? That was it? Well then, what is this that I'm picking up right now?

00:12:03.340 --> 00:12:07.740
Oh, I knew it. It's another segue to our sponsor.

00:12:07.740 --> 00:12:12.460
If you guys enjoyed this video, why not check out the one we did on a Caterpillar branded smartphone

00:12:12.460 --> 00:12:15.940
that has a FLIR thermal camera built into it? That was pretty cool.
