Sleep Tracking

wakarimasen

@dimitrios-kanellopoulos Yeah…

I remember Alex Hutchinson (of Endure fame) mentioning some research stating people actually sleep worse and get anxiety about their sleep, because gadgets tell them how bad it was. You know… You had 20 sleep disturbances and have been awake for 90 minutes during your sleep… Your deep sleep was 16 minutes. This kind of stuff.

Yup - that’s me

Dimitrios Kanellopoulos

@wakarimasen and me sometimes.

Oh shit , this night the guys came and I drunk 3 beers and look at my heart rate now. Trying to sleep and it’s f6373 like 80bpm. Oh man I am going to have a bad sleep again and how will I train tomorrow. I wished I had just watched a series.

Me a couple of years ago :

yeah what a great night, I’ll sleep like a bird don’t even need to watch tv. (And it was like that)

wakarimasen

@dimitrios-kanellopoulos
WHY IS MY WATCH JUDGING ME…

Dimitrios Kanellopoulos

@wakarimasen because you ask it to “guide” you.

I agree that it it’s guiding is fine, but ain’t then a point when you need no guidance ?

That said we are out of topic. The guys need help with accuracy in sleep tracking.

@dimitrios-kanellopoulos I beg to differ! We are helping!

Why track accurately something that shouldn’t be tracked at all?

Dimitrios Kanellopoulos

@nickk of course you differ. And actually that’s great. Thank you for that. And mostly for the debate.

wakarimasen

@dimitrios-kanellopoulos Only joking @Dimitrios-Kanellopoulos
I find sleep tracking pretty helpful, and especially like to understand sleep phases. I’m assuming it’s also used to ‘inform’ recovery and resource levels.
Unfortunately, I find my S9B not as accurate for me, compared my old Vantage V. I understand this can be personal (and may be based on the sensor used) but am not worried too much, as I don’t want to stress too much about it.
I do wonder if the S9P would be more accurate though…

Patrick Löffler

I am currently trying out the Suunto 7. I like the sleep tracking very much! I know also already discussed. But again: The hardware of the S9P has at least the same capabilities. So why not improve the software to the same level? Please do not tell me it is a licensing question. If so, just get it licensed.

@patrick-löffler Just get it licensed… And who’s gonna pay for that? It was actually quite nice for Suunto to add all the FB pieces like daily stress, body resources, VO2 Max to existing S9 a whole year after the series went out. I don’t know another manufacturer, Garmin including, who licenses new third party bits for existing units and does so free of charge.

Anyways, it’s not just licensing. S7 being WearOS is a completely different animal (read: codebase) than other Sx watches. I’d imagine S9P and S9 are pretty much the same code. So, even without FB licenses involved, pulling features from one line and into another, let alone specifically S9P… I think it’s a tough ask.

Not impossible, but certainly not a question of removing comments around a code block…

DMytro

@patrick-löffler I think it was mentioned somewhere, that broader sleep feature analysis might come to s9p eventually, as well as 1s daily HR (and spO2?) tracking. Dunno whether the same holds for s9B and whether firstbeat/suunto algos will be used.

@dmytro I don’t see how we can have 1 sec continuous HR and keep current 7 day battery life on S9P. There will have to be an awful lot of optimizing happening elsewhere for this to happen.

Also, I hope Suunto doesn’t go and try develop their own sleep analysis.

Even Garmin with their practically unlimited resources had to abandon a home grown machine learning cloud solution and switch to FB algos running on the watch. And accuracy of everyone else, who had a stab at it, be it Polar, Oura, or even WHOOP, leaves a lot to be desired.

I’d rather Suunto focus on other stuff… You know, multiple sensors per type, full Stryd integration like COROS, structured workouts, multiple S+ fields that get persisted between activities. Maybe fixing those zombie notifications and making notifications actionable, i.e. being able to dismiss them on the watch or act on them when connected to Android phone. Weather, like Polar, would be nice. As will be sunrise/sunset calendar like Garmin or COROS. Multiple alarms? Multiple timers? Maybe music controls for the phone…

Super-basic map outline for loaded routes, around waypoints or turns, so you know which trail to take…

A man can dream, and his dreams aren’t about sleep tracking

DMytro

@nickk s9p already has LEDs always on for HR monitoring, but the data gets acquired only every 12min as part of legacy code from s9b (source: chasethesummit review on s9b, commentaries from dcrainmaker).
Suunto already uses its algos for sleep tracking on s5/s9b/s9p, works spot on for me. So, if they continue development - great, of they get 3d party algos - great as well. Wherever they lead, I’ll follow.

DMytro

@nickk this turns outline you mention might be coming sooner than you think:)
There was a survey on that matter quite recently and they don’t make surveys for nothing I don’t think.

@dmytro said in Sleep Tracking:

@nickk s9p already has LEDs always on for HR monitoring, but the data gets acquired only every 12min as part of legacy code from s9b (source: chasethesummit review on s9b, commentaries from dcrainmaker).

Just checked on mine. It doesn’t. It’s off most of the time, and it takes upward of 8-10 sec to acquire a lock on HR if done on demand.

Suunto already uses its algos for sleep tracking on s5/s9b/s9p, works spot on for me.

Except the only thing provided right now is awake/sleep time and “deep” sleep, with the exact definition of deep sleep being up for grabs. The awake time isn’t terribly accurate. Unless you get up and walk in the middle of the night, it might miss quite a few minutes when you aren’t sleeping and actually know that.

Now, if you were to go full sleep analysis, you’d have to break out REM (exceedingly hard I’m hearing) and your deep sleep must be N2/N3 slow wave sleep and not just lack of body movement. So far only Fitbit scores okeyish in that department.

DMytro

@nickk Ok, I don’t have an s9p so won’t argue, but most watches are able to provide 24/7 1s HR with reasonable battery life, don’t think s9p would be an exception.
Yes, I agree that sleep is not as detailed as competition but at least start/end times look reasonable for me. As for awake time - can’t judge as I sleep through the night and rarely get any awake time.

@dmytro said in Sleep Tracking:

@nickk Ok, I don’t have an s9p so won’t argue, but most watches are able to provide 24/7 1s HR with reasonable battery life, don’t think s9p would be an exception.

Yeah, about that. Not most, no… Only Fitbit and Garmin and recently Polar, after they released Precision Prime sensor as part of their Vantage series. Before that Polar was pretty much every 5-10 min on all devices that supported OHR, like their fitness bands and M430. Poor M600 never got daily HR at all, even though it had enough battery me thinks. In fact, while LEDs are always on for recent Polar watches, I’m not sure their sampling rate is as fine grained as Fitbit and Garmin. COROS remains every 5-10 minutes, even on super-bulky Vertix with its big battery. I think Apple Watch is not continuous either, at least prior to the current gen. Neither are any of the WearOS watches.

As for awake time - can’t judge as I sleep through the night and rarely get any awake time.

I don’t always wake up in the middle of the night, but when I do… I can be awake for 10-15 minutes, can check my watch, and in the morning it would still report no awake time or just a minute or two.

I don’t blame Suunto. Sleep analysis is hard, even for such basic things as being awake. Which is why I hope Suunto will get algos from an established source like FB or work out a deal with somebody like Oura. Maybe Google decides to turn Fitbit into Firstbeat. That would actually be pretty cool.

Patrick Löffler

Thank you very much for all these insights. To me already the discussion is valuable. I have tried several different devices. Including Oura Ring, Whoop 3.0, Fitbit Sense, Polar Vantage V2, Garmin Forerunner 945 LTE. I know it is not scientific at all and I know it’s a lot of different devices. But Suunto actually is one of the few of those devices that can really differentiate between being in bed reading and sleeping. So Suunto definitely has some great working tech already. And at least for me tracking my sleep has lead to some changes in behavior already. Good ones. I go to bed on a much more regular schedule and I can see when I had a late dinner or sleeping at higher altitudes. And I find that interesting. I still can enjoy a good dinner or a glass of wine… and I’m not obsessed. Or not too much at least. So sleep tracking for me is a super helpful tool.

Jamie BG

@nickk
And even then despite 1s continous tracking with HR no one graphs it like that. S7 is graphed at 10min average?, Fitbit is a 10min? (or 5min can’t remember, but pretty sure it was 10min) average, Garmin is a 2min average.
Apple is dynamic - means at points its continous, and other times its periodic. They also don’t mention if they use actual points to graph or an average and if so the average duration.

You are incorrect about wear os - Suunto 7 is continous - 10min average graph. Also Fossil has cardiogram app so can adjust to continous (though not advisable as their battery sucks without 24/7 monitoring so continous wipes their battery). Also can used 3rd party apps like Cardiogram/Trace etc to get continous tracking, though again due to battery life it isn’t advisable. That will all change with Wear 3.0 as it will include using sensors on the low power coprocesser like the S7 does, and considering Tizen and Fitbit already do that, it will definitely be included for all.

Tizen have a continous measure option, but again not sure of their graphing method.

@jamie-bg said in Sleep Tracking:

@nickk
And even then despite 1s continous tracking with HR no one graphs it like that.

Maybe because it’s not easy to graph 50,000 data points in space that’s 1,000 pixel wide? Google Maps doesn’t show you streets and buildings in continent view either. Both Garmin and Fitbit let you zoom in on heart rate details and go down to 2 min intervals (I think 1 min for Fitbit). As does WHOOP. There’s, however, data to show when you zoom in.

You are incorrect about wear os - Suunto 7 is continous - 10min average graph.

Are you sure? Because I have S7, and I don’t remember it being continuous. The whole of last year when it was syncing to Google Fit there were huge breaks in the HR chart discussed in this forum. Precisely because the measurement wasn’t continuous, but was happening every now and then.

Well, let me grab my S7 (see below, good I didn’t bet $20 on last year behavior)

Also Fossil has cardiogram app so can adjust to continuous (though not advisable as their battery sucks without 24/7 monitoring so continuous wipes their battery).

Right, it can be continuous but it’s not advisable. Like jumping without a parachute from a plane can most certainly be accomplished. But for some reason, not advisable either.

How about we consider normal device functioning that gives us advertised battery life?

That will all change with Wear 3.0 as it will include using sensors on the low power coprocesser like the S7 does, and considering Tizen and Fitbit already do that, it will definitely be included for all.

The future is bright and marvelous, sure. I can imagine each of us having Suunto 99 Deep Space with hydrogen fusion cell and daily DNA mutation analysis. But what about now? The today’s WearOS watches? You know, the ones that will be left behind because nobody will be updating them to anything?

Can you show me a WearOS wearable that has a truly continuous tracking and lasts more than 12-14 hours here and now?

UPDATE: S7 indeed is tracking continuously, with LEDs always on (or at least, have been for the past 10-15 minutes the watch’s been on my hand). The HR acquisition on its tile is also instantaneous. So yeah… There’s one wearable that does it here and now. I humbly stand corrected.

Dimitrios Kanellopoulos

@nickk Suunto 7 has continious HR sampling 1s (actually faster as its HRV)