The Samsung Gear S gets the smartwatchcity treatment…
We conducted a number of tests for which we devised the methods ourselves as stated. We do not claim these to be clinically accurate nor statistically significant (several thousands of tests under a variety of laboratory conditions would be necessary to give those sorts of results), however we feel they give a broad representation of the integrity of the units.
Smartwatches which are tethered via Bluetooth to smartphones do occasionally lose the connection. This is unfortunately an unavoidable fact of life. This can be due to the connection being disrupted, or because the watch has strayed too far from the phone. What is crucial, however, is how quickly that connection is then re-established.
Test: to test this, we deliberately broke the connection between watch and phone by progressively increasing the distance between them (in a straight, unobstructed line-of-sight) until the watch indicated that it had disconnected from the phone. We noted this distance, and then brought the watch back into proximity to the phone and measured the length of time it took to re-connect to the watch automatically:
Disconnection distance: >15m
Time to automatically reconnect: 30 secs
Battery & Charging
300mAh battery, micro-usb cradle with its own built-in battery
Time to charge: 25% in 0.5 hour, 50% in 1 hour, 75% in 1.5 hours, 93% in 2 hours and 100% in 2.5 hours
Time to charge from battery dock: a maximum of 52% in 1 hour 15 minutes (starting with fully charged dock)
Battery life on minimum use: 144 hours (turned on but unconnected to phone and in standby mode, checked every 12 hours for < 3 secs just to check battery level)
Battery life on average normal use: 40 hours (about 50 notifications per day plus steps, heart rate, searches, notes – averaged over 5 work-days usage)
Test: time taken to complete action via voice command, tested 1) in an absolutely quiet room then 2) in busy London rush-hour traffic (roadside, Marble Arch, 5 pm weekday) – averaged 5 attempts each test. N.B, where the unit failed to correctly recognise the command, a penalty time of 20 secs was entered, and it is noted how many failures occurred out of the 5 attempts:
“Show me my steps” – Quiet: 2.2 seconds, Noisy: 9 Seconds (1 fail)
“Show me pictures of monkeys” (always fun) – Quiet: 1.8 seconds, Noisy: 11 Seconds (1 fail)
“Note to self; drink more water” – Quiet: 2.4 seconds, Noisy: 14 Seconds (3 fails)
“Set alarm for half-an-hour” – Quiet: 1.8 seconds, Noisy: 14.4 Seconds (3 fails)
“What is the population of Lithuania?” – Quiet: 6.4 seconds, Noisy: 16.2 Seconds (3 fails)
An overall average time of Quiet: 2.92 seconds (no fails), Noisy: 12.92 Seconds (total 11 fails)
Test: Conducted by walking 1000 paces in a straight line along a high street (at a quiet time) in a brisk, striding manner, arms swinging, in order to maximise step detection:
Steps average variation: 7 (a variability of 0.70%)
In normal activity testing, where uniform strides were not used, the accuracy was very poor with a variability of over 50%. This was found with all of the watches we tested.
Heart Rate Monitor Accuracy
Test 1: Time to take heart rate reading. We took 5 consecutive readings, timing each one:
Seconds – 8,9,9,9,8 – average 8.6 seconds
Test 2: Compare heart rate to electrode-based system as found on treadmills. Since these watches presumably average the rate over their recording period, while the electrode-based monitor takes continual spot-readings, the procedure we used was as follows: we observed the fluctuating readings on the electrode monitor throughout the period over which the watch took its reading. If the electrode monitor readings fluctuated more that 5 bpm over that period (from highest figure to lowest) that particular reading was discarded (this did made for a very long test). We then took the average of 5 of the accepted median readings. We also undertook mild exercise to compare a range of heart rate readings (all in beats per minute).
1st reading: electrode median: 86, Watch: 79 – variation 8.48%
2nd reading: electrode median: 86, Watch: 84 – variation 2.35%
3rd reading: electrode median: 101, Watch: 93 – variation 8.25%
4th reading: electrode median: 89, Watch: 84 – variation 5.78%
5th reading: electrode median: 98, Watch: 98 – variation 0%
Overall variation: 4.97%