Pulse ox accuracy

How do you know? Did you compare with a medical device whose accuracy has been certified?

I have an average of 96% over the last 7 days.  I have a pulse ox meter on my Samsung phone that uses a finger to measure and every time I have checked it I get the exact same reading that my 945 shows, which always seems to be at 96 or 97.  So I feel pretty confident that it is accurate.

I have compared to a high-end Nonin Pulse Ox and the 945 is at typically 4-5 points lower. I wish there was a way to calibrate this sensor. 

My sleeping pulse ox seems to dip into high 80's/low 90's at certain times during the night...i can almost certainly confirm through my wife that i do not have sleep apnea.  i think the dips happen when you are sleeping on the side of the body the watch is on, possibly interfering with the circulation of blood to your arm.  i just turned on all day pulse ox as an experiment, but it seems to register 96 to 97 percent consistently for me when i take a manual reading, which is the same as my finger pulse ox that i used before getting this watch.

Same here.

My FR945 reads about 3 pts lower than my fingertip meter and my Fenix 5x plus.  I have compared the F5x+ reading with a reading in the doctor's office several times and they were within 1 point each time, so it's reasonable to conclude the FR945 is reading consistently low.  It's interesting that the F5x+ is accurate and the FR is not.

I can easily and often compare it to the medical pulse ox on the Corpus C3.

If I were the watch tight and relax the arm, all measurements are the same. (in my case mostly 96%)

The Garmin reads 4-6 % below actual.  I have checked it against a Nonin medical grade device at the same time on a finger of the same hand so the difference is not positional related.  I’ve also repeated the test with doctors equipment.  It is ALWAYS 4-6 % lower.  Your O2 level does decline on average 4% during sleep.  That is normal.  But the measurement itself is inaccurate.