It is good to note that sleep tracking is one of those things that is kind of a hand-wavy thing right now And it has actually shown to increase anxiety when using sleep trackers when people get less than great sleep, and can make things worse when the algorithm incorrectly estates quality (which all if those proprietary apps do instead of just tracking sleep state).

There is also the issue of comparing with eeg which the head unit that the qualified scientist is also not completely accurate, it is just the best that is readily available.

Sorry that I can't be helpful, but I dived into it a while ago and what came out of my search was pretty much that it is not reliable and people pay way too much attention to it when they should pay attention to how they feel.

https://intrasleep.com/blog/article/top-open-source-sleep-tracking-tools-for-better-dream-insights

Garmin Fenix 6 (or equivalent) and above work fine for sleep tracking with gadgetbridge and endurain, and are pretty cheap secondhand. Battery life and build are stellar (and seeing as you charge it at most weekly unless hammering GPS and Music, 500 charge cycles is a decade, batteries are replaceable too, screws, but I'd be leery of diving with it afterwards) and you get all the smartwatch / sportswatch stuff. It's sleep, as long as it's +- 5 min, it's good enough to work with, doesn't need to be the best on the market.

I've been interested in tracking my own sleep too. I wonder how useful these commercial products are in producing data that is correct. They come from technology companies who primarily want to collect data, as much data as possible. They're not at all interested in providing their users with an honest interpretation of anything. A lot of them probably don't even have the expertise to interpret whatever they've gathered.

I've been involved in sleep studies in the past and none of them involved using a smart-watch or other commercial wearable. They used electrodes (?) that were stuck to my body and fed into a computer. What data was being collected and how to properly interpret that data is beyond me.

Making a DIY device is probably feasible, but again, I wouldn't have a clue on how to interpret the data gathered. I don't think it'd be too complicated though.

Tools built specifically to monitor one signal are probably better than a generalized thing, such as a smartwatch, from a company that produces surveillance equipment for marketing purposes.

Pebble: https://repebble.com/watch#pt2

The quantified scientist would be the first to tell you that he's making these determinations on a sample size of one (sometimes he gets a friend to help, and then it's a sample size of two).

So maybe what ever tech you're thinking about will work for him and not for you or vice versa.

Unfortunately, it's kinda like the wild west out there. The tech changes so fast that no one has time to do a big study and figure out what actually works.

I have heard that a dedicated sleep tracking device will work better than the watches, but I don't have any direct experience there.

Maybe you can find one that can export the data over a USB cable