Inaccurate HR during weight training and HIIT

[samulrich]

I upgraded from a Change 5 to the Sense 2 and the short-term HR accuracy is horrible. Am I doing something wrong? Any tricks to improve?

The Charge 5 wasn't perfect but still measured my HR averaging around 125 (which is actually low for my weight training  - more precise measurements put it in low 130s) which was fine with me. The sense 2 on the exact same schedule at the same time will show an average of 90. 

It works great for long-term HR tracking like running, biking, elliptical but short-term is just so bad. 

[Dopovo]

Unfortunately there's nothing to be done, Sense 2 HR (like Sense 1) is actually unfortunately very bad. For more information you can watch this intesting analysis

[samulrich]

Annoying, the way it behaves leads me to believe it's a software issue. I'm guessing they run a moving average to smooth (what is probably) erratic HR readings from the sensor. An exponential moving average on a smaller set of steps would fix this (hint hint if any fitbit devs see this).

Regardless, I wonder if the Pixel Watch has the same issues

[JerryHam]

I'm amazed that you find it OK for running or biking. Many of us find it to be junk for those activities too. Just today I had it off by 100% during a warmup for a distance jog. The iFit band showed 77 (again just the warmup) and the **ahem** Sense 2 showed 153. Seeing as how I wasn't breathing hard at all - just walking at that point - 153 was completely bogus. It often shows the AVERAGE for a run as higher than the actual MAX was for that run (when measured by a real HR monitor).

[samulrich]

I've only used it for a couple days so my sample set is very small. What's weird is mine consistently under estimates. I haven't seen any number HR past 120 and 100+ is a very rare occurrence. 

[Sparhawk69]

If you watched the video that was sent by @Dopovo above, you will find that the Sense 2 is always reporting lower HR.

Hopefully Fitbit will release an update to rectify it.

[t.parker]

As @JerryHam said, cycling and running seem to struggle a lot, too. Running sometimes gets better (that's just random) but cycling... well... I'd like to see at least one tolerable HR readings for outdoor cycling (not indoor spinning). Here's an example of my recent bike rides with Sense 2:

Orange is Sense 2, Blue is a PolarH10 chest strap. Since original Sense, I have never had remotely accurate HR readings during cycling. I think it's even worse than weight training (although those two activities are hardly comparable). If you need accurate HR, Sense 1, Sense 2, Versa 3 and Versa 4 are not for you.

[samulrich]

If anyone on the software team at Fitbit sees this, I will fix your embedded HR code for free if you give access to the codebase. Seriously, respond to this post or email me and I will fix it for you. 

As for the rest of this, how can you have a worse HR sensor system in your premium line???

@LizzyFitbitAny thoughts on this problem???

[JerryHam]

They never fixed it on the Sense 1 even after a couple of years and teardowns have suggested this is the same sensor as Sense 1. Likely they are unable to do anything about it due to the hardware of the sensor. Either that or they just threw all their development into Premium (which is funny since the HR data drives most of Premium and garbage in, garbage out). Reminds me to go cancel my free trial of Premium again as the data is useless.

[samulrich]

I don't know for certain but the behavior appears to be more software like. At bare minimum better software could fix most erratic sensor issues. They might not be able to put that software directly on the device but they could run it on the app or in the cloud and update the device accordingly. 

If the deal was that immediate on-device readings are 50% accurate during workouts but within 20s your corrected value were shown in historical data I think most people would be fine with that. If Google can artificially insert "natural light" into a dark image (night sight) and remove people from photos by generating photo-realistic backgrounds to replace them (magic eraser) they can definitely use a neural net to repair any inaccurate HR sensor data.  

Again, dev team at Fitbit feel free to take the hint. 

[SunsetRunner]

For this device weight training and HIIT is hard for it to track, Quantified Scientist did a review on it as well as other devices. From my own research it is great for everyday monitoring and walking. In some cases running but anything else is mediocre at best. I think this device's sweet spot is for the basic user and not the user who does anything too complex for it. 

Then again, I came from Apple and see they have their own issues as I have seen that heart rate unreliable at times. After all, these are all mass made and use LEDs to measure heart rate and have to stay under a certain price point. They also are not calibrated against medical devices. So, i would use these as the benchmark. For example, when I am on the treadmill sometimes it is spot on and other times its high. When it is high, it seems to be off by around 10 beats per minute. I use an oximeter for verification which still is not medical grade but seems to have been accurate in the past. 

[samulrich]

So I definitely understand that these are mass market and should not be counted as medical devices. With that said, 10 bpm is a reasonable error rate. Most of my issues are 40 to 60 bpm delta.

Additionally, my issue is simply that this can be fixed with software and is fundamental functionality for their offering. It's like a car manufacturer claiming that a non-functioning engine is acceptable because they have to make a lot of them. If it's core functionality that all other aspects of your product depend on, it needs to be within a reasonable margin.

Finally, this is fixable via software. I can't stress that point enough. Google basically practices magic with a ton of their software offerings. Correcting this data via a cloud based AI model is not only feasible, it is directly within Google's skill set. Now will they do that, I don't know and tend to think not but I'm still annoyed by it. If they intended to have a low performing product, they should mark it as such and disclaimer with "HR measurements during exercise as inaccurate." They got this to work better on the tracker lines so I won't give them an easy pass on the Sense 2.

P.S.
Why did "40 - 60" without the space trigger the "not permitted" message? Is there some slang I'm unaware of???

[sgbednar]

You know, I am a very "basic" user, and I'm not happy.  I am 73 years old, track steps, monitor sleep, and work on very basic goals, like increasing my weight-bearing exercise, sleeping better, and preserving my cardiovascular health.  Still, when activities that earned me 151 zone minutes on my Inspire 2 suddenly earn me 1 zone minute on my brand new Sense 2, I am concerned, I think justifiably.  How do I trust this device to monitor my health?  Perhaps it was my old device that was off base, and I am more sedentary than I thought.  Even so, it alarms me.  Both devices are Fitbit products, allegedly performing the same basic functions.  How to give a little old lady heart failure?  This may be it!   I decided to stick with Fitbit because of the detailed sleep data (I have a sleep disorder).  So I'm pretty "basic" here.  If HR monitoring is so dramatically different between the products, what do I trust?  If the thing tells me there is something wrong, do I even take it seriously?  Do I run to my BP cuff or my doctor every time the charts show an irregularity?  A certain degree of accuracy is necessary, even for the most basic of users!  

[JerryHam]

I don't know about it being good during walking and everyday use. I have many examples from walking on the treadmill (granted power walking) where it is off by 100% (showing 145 when reality is 72) for long swaths of time. The other day, I was sitting on the couch, reclined, and it said 146 for two minutes. Then it said 51. The 51 was close, since my resting HR is 46. That 146? Just dead wrong. At 146 I am in my Zone 4 and I am sweating and breathing fairly hard. As far as software, if the sensor can't lock on and returns readings that are just way off - and you don't have a way to know which direction they are off, you aren't going to fix it.

[sgbednar]

I am waaaaay older and less fit than you, but it still matters to me.  It seems to me it never acknowledges anything in my fat burn zone at all.  Then when I do start to exert myself a bit it jumps up to cardio zone, with not a whole lot of transition.  At may age, I don't do anything that suddenly!  😁  If my heart is suddenly behaving like I do, then I'm concerned!  I don't have to be a marathon runner or a hot shot programmer to recognize something is not right here.  Oh, by the way, the only zone minute I got yesterday was for brushing my teeth.  Brushing my teeth!  I went out and mucked out horse stalls (shoveling, pushing loaded wheelbarrows, carrying water, etc.)  Nothing.  Yawn .  If the thing were consistently off in a particular pattern, I could compensate.  But erratic like this is rough.

[t.parker]

@sgbednargeneral rule is not to blindly trust those devices. One reason is that the wrist HR sensor is probably the worst when it comes to placement. The wrist keeps bending, flexing, and moving a lot. Additionally, the shape of the forearm often makes it impossible to keep the watch at recommended position (further from the wrist). Sometimes they will be more accurate, other times less, and yet another time not accurate at all. To trust something, you need to be able to verify whether the data is valid against bullet-proof sources or against multiple sources (if two agree and the third isn't then it's (almost) obvious which one is correct and which isn't). Basically, always consider readings inaccurate unless proven otherwise. I pasted the above charts from bike rides where HR couldn't be possibly worse. The remedy for it is simple - using different devices when accuracy counts. It is what it is and won't get any better for Sense 2 (as it didn't for Sense). Anyway, most of the time if the device doesn't do what you expect it doing it means that it's time to change the device.

[sgbednar]

Unfortunately, I did change my device.  I got this.  So I will just need to adapt until I can afford to do something better.  Probably not another Fitbit/Google device, but we'll see.  I've already adjusted my goals downwards and set up a custom zone, so at least I have data that is meaningful to me.  I hope some of the more ambitious and athletic users here can find something that works for them.  I guess I'll follow along and see what I can learn.

[t.parker]

@sgbednar there's nothing else you can really do 😕 Fitbit is designed for rather casual use (it isn't watch for athletes, too simple, lack of many features athletes use) but somehow it struggles to get basics right.

[RonaldM]

@t.parker I finally understand your referrals to Garmin. I got an early Christmas gift from my wife two days ago and lets say I'm a bit overwhelmed 😅

[Jerseycitysteve]

I feel that I've wasted my money on the Sense 2. The heart rate monitoring is inaccurate. For example, on a brisk walk, my pulse rate reads 90bpm when my old Charge 5 would have me at least at 115. As a result, I get very few active minutes. 

I can't believe that Fitbit is unaware of this problem.