How is deep sleep differentiated from light sleep?

You really can't trust garmins sleep tracking. Last night i was sound asleep and my stress levels where in the high 50's almost all night. And I sleep well.

It changes with each update it seems, one update ago stress was working ok, but on the same update, steps were overcounting by almost a 100%.

So don't listen to your watch, it will stress you out.... it is not ready for the market, its still a beta device...

Thank you! That makes me feel better. 

Yeah just FYI, none of these sleep trackers actually "know" what sleep phase you're in, it's a "best guess" based on whatever algorithm they came up with to guess what sleep phase you're in based on what the various sensors are reporting. Every company seems to have a different algorithm, some better and some worse. Garmin isn't great. Fitbit seems to be the best I've tried.

I've struggled with sleep issues my entire life so I've read into this quite a bit and listened to podcasts with various experts - the only way to know what sleep phase you're in is a polysomnograph - basically a sleep study in a lab.

Fortunately, it sounds like Garmin is going to farm out the sleep analysis to FirstBeat, at least for the Fenix series, hopefully it'll trickle down to the Vivoactive series too.

Just to maybe help with your own interpretation of your data, the following is what I posted in another thread regarding my own data and use of a Garmin watch over the past 6 months:

I have researched the limited information online about how Garmin and others, like Oura, track and report on sleep.  I have also had two conversations with Garmin about the sleep tracking on my watch.  The whole wearable sleep tracking world is really entertainment rather than true sleep analysis. Sleep tracking and reporting on the world of wearables is somewhat inaccurate and highly variable person to person.  What I have come to conclude about my sleep is this:

I feel I sleep well, although I do move around and reposition quite a bit. I have no trouble breathing and do not snore.  At times during the night, almost nothing will wake me up.  At other times I remember that I was sleeping lightly.  And at still other times I recall dreaming vividly.  My watch consistently reports zero to 10 minutes of deep sleep per night. I have no way to be 100% sure, but I believe that is likely inaccurate.  Someone further down in these comments reports over 4 hours of deep sleep.  That is also likely inaccurate.  Garmin analyzes heart rate and movement to determine sleep stages.  I move myself and my wrist quite a bit during sleep so the watch cannot really detect when I am in deep sleep.  Some people may move less so Garmin thinks they are in deep sleep even when they are not.

I feel like the reported REM sleep is more accurate for some reason. My watch consistently reports about 2 to 3 hours of REM, spread across several periods throughout the night. On the occasion that I wake up from a dream and notice the time, the next morning Garmin does report that sleep time as having been REM.

The light sleep reporting for me is probably too high because of the difficulty the watch has determining when I am in deep sleep.

Also, because the watch depends so heavily on the accelerator to determine sleep time and sleep stages, it is often inaccurate knowing exactly when I fall asleep.  If I am reading in bed before going to sleep the watch often reports that time as sleeping.  So because I do fall asleep pretty quickly, I am in the habit of looking at the time when I turn out the light so that I can adjust the sleep start time in the morning. In the morning, if I wake and then lay still in bed for several minutes before getting up – the watch shows me awaking, but then thinks I have gone back to sleep for those few still minutes before I get out of bed.

Most mornings I look at what the watch has reported and adjust the sleep and wake times accordingly.  There is nothing I can do to adjust or more accurately know my actual deep and light sleep times.

Maybe Garmin will become better with all this with future updates to software?  But likely it will take not only new software, but new hardware before at home sleep analysis becomes accurate.  Hopefully some real effort will be put into improving this feature.  Garmin recently make a big deal about bringing an on device sleep widget to some watches. They are now using the company Firstbeat for sleep tracking and analysis. Firstbeat adds a bit more detail to the analysis of the date and also adds a "coaching" aspect.  But as the actual hardware/software detection component has not changed significantly, all this really does is being the inaccurate information on Connect to the watch. It then adds some analysis and recommendations based on that inaccurate data but does nothing to improve the accuracy of the detection process itself.

Thanks!

Thank you WTAFT!

And just to add a couple more thing I have noticed.  The Garmin sleep tool also often gets awake time wrong.  It sometimes shows me awake for long periods during the night when I was clearly not awake.  Of course this destroys the accuracy of knowing total sleep time unless you add that time back in mentally.  You can not correct it in the app though as the only adjustments Garmin allows are to edit the go to sleep and awake in the morning times.

And on the other hand, it also often shows me as in deep sleep when I am awake. For example last night it showed me as in deep sleep from about 6:30 am to 6:50 am.  In reality I had awoken and decided to go back to sleep and was in bed, very still. After about 15 or 20 minutes I fell back to sleep until 8:40 am. The app got the REM and light sleep from 6:50 to 8:40 correct.  But the awake time was registered as deep sleep.

I think these errors are because the Garmin algorithm depends so much on movement and heart rate to guess at sleep.  So if you lay very still, and breathe very smoothly the app thinks you are in deep sleep.  If you move at all it thinks you either awake or in light sleep.  Overall still a long ways from being very useful in monitoring sleep.

At 6 am it showed 3 full hours of DEEP + 3 full hours of LIGHT, starting from 12pm, nothing ELSE.

But at 9am, it showed 25 minutes of deep, 1h of rem, 6u43m of light, 2u12m awake, starting from 11pm.

So it would seem they are not comparing to any already recorded data from weeks and months, but just the last couple of a few hours.

Unbelievable results all over, even about breathing, and stress (supposedly their extracted HRV).

I see this frequently on the first sync of the morning. Deep sleep at 2 or 3 hours. Then after subsequent syncs it is more like 50 minutes or an hour, even when the 2nd sync is within minutes of the first. So it is hard to understand what is happening there. Garmin-AmberD any thoughts here?

@Maryamk1e The best short answer most probably being: AT RANDOM..

Just read along a bit more to see.