WEBVTT 00:00:00.000 --> 00:00:08.000 Like the thirsty cast members of The Bachelorette, Samsung Aura and, if 00:00:05.480 --> 00:00:13.000 rumors are correct, [music] Apple all want to put a ring on it. And why not? 00:00:11.400 --> 00:00:18.920 The promise of a smart ring is undeniably sexy. A screen-free titanium 00:00:16.240 --> 00:00:24.080 hoop that tracks your vitals 24/7. Who wouldn't want it? But, like a reality TV 00:00:21.720 --> 00:00:28.720 marriage, buying a smart ring is sometimes followed by regret. If you've 00:00:26.640 --> 00:00:32.160 ever wondered why your smart ring thinks you were napping while you were working 00:00:30.320 --> 00:00:36.520 and yet misses your PR at the gym, that's because, turns out, shrinking an 00:00:34.560 --> 00:00:40.760 entire lab's worth of equipment down to the size of a piece of jewelry requires 00:00:38.480 --> 00:00:45.000 some serious engineering compromises. Compromises that could be fat-fingering 00:00:43.000 --> 00:00:48.560 your biometric data. Let's break these devices down, starting with the main 00:00:46.880 --> 00:00:53.360 sensor. [music] PPG, or photoplethysmography, is a fancy name 00:00:50.800 --> 00:00:57.320 for shining light into your skin and measuring how much bounces back. In a 00:00:55.320 --> 00:01:01.800 hospital, they'll clip a sensor onto your fingertip, yield transmissive PPG. 00:01:00.080 --> 00:01:06.600 It's incredibly accurate because the signal is unobstructed. Smartwatches, on 00:01:04.160 --> 00:01:10.840 the other hand, use reflective PPG, where the sensor and the photodiode are 00:01:08.440 --> 00:01:14.800 on the same side. Fun fact, by the way, they use green light because your blood 00:01:12.840 --> 00:01:20.000 absorbs it like a sponge, providing a high-contrast signal. Depending on, you 00:01:17.320 --> 00:01:23.200 know, how far apart your photodiode is, if it's, say, on the curvature of a 00:01:21.400 --> 00:01:27.920 ring, you may actually kind of be between these modalities, where you're 00:01:25.000 --> 00:01:31.240 somewhat like in a transflective approach. You may see behavior of both 00:01:29.760 --> 00:01:36.200 like a reflective and a transmissive PPG. Smart rings can occupy a weird 00:01:33.800 --> 00:01:39.440 middle ground. Because the sensors wrap around your finger, light travels 00:01:37.800 --> 00:01:44.680 further through your skin than it would from a smart watch. This gives the ring 00:01:41.560 --> 00:01:46.880 a slight signal advantage, but it's one 00:01:44.680 --> 00:01:49.440 that unfortunately gets dragged down by the constraints that arise from its size 00:01:49.386 --> 00:01:55.480 >> [music] >> and from its battery. >> The amount of light you need in order to 00:01:53.440 --> 00:01:58.400 measure an appreciable pulse is proportional also to the active area of 00:01:57.200 --> 00:02:02.720 the photodiode that you're using to sense it with. So, a bigger photodiode, 00:02:00.480 --> 00:02:07.440 a bigger sensor, means you can use less light, less current, in order to perform 00:02:05.280 --> 00:02:11.480 the same relative measurement. If you only just shrink the photodiode itself, 00:02:09.960 --> 00:02:15.120 you're going to be decreasing your your signal-to-noise ratio 00:02:13.360 --> 00:02:18.240 and have to, you know, boost your LED light in order to accommodate for that. 00:02:17.040 --> 00:02:22.320 You can think of it kind of like listening to a conversation in a crowded 00:02:20.040 --> 00:02:25.360 bar. A smartwatch stands next to a person with a large, directional 00:02:23.800 --> 00:02:28.720 microphone. A smart ring, on the other hand, is trying to hear them from across 00:02:27.320 --> 00:02:32.520 the room using a tiny little spy mic taped to their collar, 00:02:31.040 --> 00:02:37.856 which brings us to the battery. The Samsung Galaxy Watch Ultra packs 590 00:02:36.520 --> 00:02:43.840 mA hours. >> [music] >> The Galaxy Ring, 00:02:39.760 --> 00:02:46.720 17 to 22 and 1/2. And yet, the ring 00:02:43.840 --> 00:02:50.880 claims it lasts for up to 7 days. How? Well, 00:02:48.840 --> 00:02:54.120 because it doesn't. Both devices use duty cycling to extend their usable 00:02:52.600 --> 00:02:57.960 charge time, but because the watch has a larger battery, one intense session 00:02:56.160 --> 00:03:02.600 isn't going to drain it. The ring, on the other hand, has about 00:02:59.440 --> 00:03:04.320 the battery of a Cheerio. [music] 00:03:02.600 --> 00:03:07.200 They have also large dormant periods, where the ring simply is not measuring 00:03:06.120 --> 00:03:11.262 at all. And and so you can kind of think of it as like interlacing on top of 00:03:09.760 --> 00:03:15.520 interlacing, you know, in order to [music] make it so maybe you only measure the pulsations, measure the 00:03:13.960 --> 00:03:19.520 heart rate, pulse ox, you know, for a minute, you know, every 10, 15, 20 00:03:18.040 --> 00:03:23.760 minutes, for example, in order to to keep conserving uh power. [music] To put 00:03:21.760 --> 00:03:28.160 this in perspective, when the Franklin Research Lab built their own DIY smart 00:03:25.920 --> 00:03:32.960 ring, the battery gave out after about 4 to 6 hours of continuous use. So, if we 00:03:30.720 --> 00:03:37.960 take that as a baseline for true continuous tracking, to survive a full 00:03:35.320 --> 00:03:42.880 168-hour week of nonstop data collection, you'd need to wear 34 rings 00:03:40.480 --> 00:03:46.800 and swap them out like ammunition magazines. By Wednesday, you'd look more 00:03:44.960 --> 00:03:49.320 like the Divination instructor at Hogwarts than like a fitness enthusiast. 00:03:49.252 --> 00:03:54.492 >> [music] >> The reason that smart ring bands can claim a week while Dr. Franklin gets 00:03:53.400 --> 00:03:59.560 only 5 hours >> [music] >> is the sampling rate. When a researcher 00:03:56.880 --> 00:04:04.320 says continuous, they usually mean high-frequency sampling. When a consumer 00:04:01.800 --> 00:04:08.640 wearable says continuous, they often mean, "Ah, we check in every now and 00:04:06.720 --> 00:04:13.360 then, and then we interpolate a bit so the graph looks like a solid line." And 00:04:11.200 --> 00:04:16.560 if you have a darker skin tone, the LEDs actually have to work harder to get a 00:04:14.840 --> 00:04:19.320 readable signal, which can cut that already slim battery life nearly in 00:04:18.720 --> 00:04:25.240 half. [music] Now, that isn't to say that the form factor has no benefits. There is a 00:04:23.000 --> 00:04:28.680 reason that hospitals clip sensors onto your fingers. They are densely packed 00:04:27.160 --> 00:04:32.240 with arteries that are right close to the surface of the skin. But for smart 00:04:30.520 --> 00:04:36.520 rings, that's kind of a double-edged sword. Besides being so darn cute and, 00:04:34.448 --> 00:04:40.400 [music] you know, fingery, fingers are your body's temperature control meat 00:04:38.200 --> 00:04:44.680 sticks. When you have vasoconstriction or dilation 00:04:42.520 --> 00:04:48.800 uh due to being cold or due to, say, being in a sauna, that will, again, 00:04:47.040 --> 00:04:52.400 change the distribution of blood within the skin and impact your 00:04:50.200 --> 00:04:56.200 photoplethysmography signal. Typically, vasoconstriction 00:04:54.040 --> 00:04:59.280 or having that blood redistributed away from the surface of 00:04:57.760 --> 00:05:03.360 the skin can have a detrimental effect on the PPG. So, it'll decrease your 00:05:01.560 --> 00:05:06.680 pulses, it will decrease your signal-to-noise ratio, and then the 00:05:05.120 --> 00:05:11.480 opposite is true when you have perfusion. Cold hands equals a weaker 00:05:09.480 --> 00:05:15.440 signal, which equals your ring struggling to find your pulse. Lock 00:05:13.280 --> 00:05:19.840 outside in the winter or sit next to an aggressive air conditioner, and your 00:05:16.960 --> 00:05:25.400 ring huh might think you're a spooky vampire with no pulse. Or Sam Altman. 00:05:22.600 --> 00:05:28.360 Unlike the finger, the wrist sits closer to the body's core, [music] staying 00:05:26.560 --> 00:05:32.400 warmer and maintaining more consistent blood flow, giving the watch a natural 00:05:30.320 --> 00:05:35.320 edge in colder environments. But hey, I mean, if sitting still in a freezing 00:05:33.600 --> 00:05:39.880 room isn't your only hobby, what about if you actually exercise? 00:05:37.760 --> 00:05:43.760 Even a simple jog is going to slosh your blood around enough to give the sensors 00:05:41.480 --> 00:05:48.040 a headache. But for hand heavy lifting, the ring becomes basically decorative. 00:05:46.280 --> 00:05:53.440 See, the problem isn't just motion, it's grip. Odoo, your business doesn't need 00:05:51.240 --> 00:05:57.720 different apps for signing agreements, running a POS, doing inventory 00:05:55.440 --> 00:06:01.720 management, or running email marketing. All you need is Odoo. It's a business 00:06:00.040 --> 00:06:05.720 management software with just about anything you could imagine needing 00:06:03.160 --> 00:06:10.160 integrated into a single platform. It's customizable and built to be easy to 00:06:07.880 --> 00:06:15.920 navigate since you pick and choose only what you need, no extra fluff. And if 00:06:13.320 --> 00:06:20.600 you only need one app, it's free. Use our link for a free 15-day trial with no 00:06:18.800 --> 00:06:24.320 credit card required. When someone tightly grips a steering 00:06:22.440 --> 00:06:27.800 wheel or a barbell, we typically call this motion artifacts. At the uppermost 00:06:26.280 --> 00:06:32.000 level of your skin, you have a capillary bed, and when you apply pressure to this, that's the first layer that 00:06:30.520 --> 00:06:36.680 actually gets compressed. And so you end up with a a large spike in your photoplethysmography signal. That's not 00:06:35.120 --> 00:06:41.680 attributed to your pulsations or your heart rate, which is the thing we want to measure. These motion artifacts and 00:06:40.000 --> 00:06:45.960 pressure-induced [music] signal loss hit simultaneously, and they compound fast. 00:06:43.840 --> 00:06:48.920 White-knuckling a barbell squeezes blood out of the very capillaries the ring is 00:06:47.320 --> 00:06:52.680 trying to read. So, by the time your heart rate has climbed to 160 beats per 00:06:51.280 --> 00:06:56.080 minute mid-set, the ring might be reporting a breezy 80. A quick 00:06:54.440 --> 00:07:00.240 TechQuickie safety tip, by the way, wearing a metal ring during heavy lifts 00:06:57.880 --> 00:07:03.760 is not a good idea. Beyond scratching the finish of your ring, you're actually 00:07:01.760 --> 00:07:07.160 risking a degloving accident. Seriously though, do not Google that. The standard 00:07:05.360 --> 00:07:11.720 way to deal with motion artifacts is to cross-reference the accelerometer data. 00:07:09.560 --> 00:07:16.600 If motion exceeds a specific threshold, then the PPG data just gets discarded 00:07:13.640 --> 00:07:22.000 entirely because a gap is better than a wrong number. A 2024 study even found 00:07:19.000 --> 00:07:23.920 that you might need to discard up to 70% 00:07:22.000 --> 00:07:28.160 of heart rate variability readings due to poor signal quality. To combat this 00:07:26.160 --> 00:07:33.320 data trashing, researchers are working on transformer-based AI models to 00:07:30.480 --> 00:07:37.360 reconstruct corrupted PPG signals. So, the sensor's tiny, the battery prevents 00:07:35.440 --> 00:07:41.120 continuous tracking, cold hands kill the signal, gripping blinds it, and most of 00:07:39.520 --> 00:07:46.040 these rings aren't even FDA-regulated medical devices, making them pretty much 00:07:43.120 --> 00:07:49.320 high-end accessories with an asterisk. So, are they an overpriced piece of 00:07:48.040 --> 00:07:53.160 e-waste? Actually, no. There is one time of day 00:07:51.680 --> 00:07:57.800 where almost all of these problems disappear, >> [music] >> and that's during sleep. Unless you're 00:07:56.080 --> 00:08:01.240 David Goggins, you're probably aren't gripping a barbell in your sleep, so 00:07:59.400 --> 00:08:06.440 your hands are relaxed and your arteries are right at the surface. Smart rings 00:08:03.400 --> 00:08:07.880 can get incredibly accurate resting data 00:08:06.440 --> 00:08:13.800 while being far more comfortable to sleep in than this darn thing. Aura even 00:08:11.120 --> 00:08:17.960 boasts about their .999 correlation with medical-grade ECGs for 00:08:16.000 --> 00:08:22.560 resting heart rate, which checks out. Even with a few misbeats, basic average 00:08:20.080 --> 00:08:25.760 is easy. Their .984 HRV correlation claim, however, has a massive catch. 00:08:24.400 --> 00:08:31.520 Heart rate variability measures millisecond differences between individual beats, and it needs [music] 00:08:29.360 --> 00:08:35.000 beat-to-beat continuity. So, a gap in the data doesn't just reduce accuracy, 00:08:33.440 --> 00:08:38.719 it invalidates the entire measurement window, which is why smart rings filter 00:08:37.000 --> 00:08:43.320 out low-confidence data before reporting. To legally be able to claim 00:08:40.919 --> 00:08:48.400 this HRV accuracy, many validation studies use sedentary participants. In 00:08:45.960 --> 00:08:51.200 one, no subject's heart rate broke 90 beats [music] per minute, so it's kind 00:08:49.640 --> 00:08:55.720 of easy to be accurate when your user's just vibing on the couch. As for the 00:08:53.760 --> 00:09:00.000 high-level scores, like sleep score, readiness, and stress, those are 00:08:57.680 --> 00:09:04.400 proprietary algorithms that arguably don't mean much. It's kind of a way of 00:09:02.280 --> 00:09:08.840 gamifying your sleep to see who [music] can have the most readiness, I suppose, 00:09:07.200 --> 00:09:13.000 and to help you justify your $5.99 monthly subscription for the rest of your life. If you want to see the 00:09:11.240 --> 00:09:16.400 testing facilities that smart wearables, like rings, are designed in, check out 00:09:14.720 --> 00:09:21.400 this video where I visited one in China. It's actually, for all of the dunking on these devices 00:09:19.480 --> 00:09:24.840 that we just did, a lot more elaborate and way cooler than you might think.