Instinct - 9.20 Beta Release Candidate

trux said:

Garmin uses a simplified formula without temperature compensation. It is mentioned for example in the document Barometric Altimeter Accuracy: 

"The barometric altimeter is not temperature compensated."

Nope. You are misreading the document and Gamin's statement. The full statement (made for Edge devices) is:

"The barometric altimeter is not temperature compensated. Temperature changes in the measuring device will affect the barometric pressure sensor and altimeter readings."

Therefore, it is simply saying that the sensor is not temperature compensated (something to be expected), which has nothing to do with the temperature being or not being used in the algorithm used to convert the sensor-measured pressure to mean-sea-level pressure. Different ballgames, altogether.

But let us assume, for the sake of argument, that after some more work, or after asking someone with inside knowledge, we can determine that, in fact, Garmin does use a simplified algorithm that does not use the temperature to map the temperature to sea level. That still does not change anything. Remember "altitude"? You will certainly not try to argue that altitude is not used in the algorithm; therefore, a change in the perceived altitude (by calibrating the altimeter, for example) may result in a change in MSL (possibly triggering a storm alert ?) even if the measured ambient pressure is constant. Therefore, it is now my turn to repeat: These two pressures may have different change patterns, and are thus not "directly proportional," as you put it (unless some malevolent behavior is at play from Garmin's part, specifically designed to prevent it :)).

Bottom line: You have to work harder in your effort to try to show that my point is a moot point. Maybe you will want to make experiments with changing temperatures and altimeter calibrations while recording both pressures, graph them, and see if the change patterns are the same in both pressures with no exceptions whatsoever? It is a lot of work, for which I have no time, but you seem sufficiently motivated to disprove me, and thus might be willing to do it.

All I ask from you, for the sake of efficiency (time is costly), is that you keep your reasoning clear, objective, and to the point, as you did this last time.

No, the sentence "The barometric altimeter is not temperature compensated" means exactly what I stated - the temperature is not used for the compensation in any way. And it also cannot be used, because the device does not know it. At Instinct even less than at Edge, since it is much more influenced by the body temperature. And the temperature from the weather widget is not used either - if it relied on the weather station data (which is often not available anyway), it would read the pressure from there too.

As for the altitude - you have to compensate for it in one way or another anyway. Normalizing the local pressure to the MSLP (without the temp) is the simplest and the most common way of doing it, but you can do it in another way too, if you wish. Whatever method you choose, you still get the same results in the final, since the formulas are perfectly reversibe in both ways.

trux said:

"The barometric altimeter is not temperature compensated" means exactly what I stated

As someone who has professionally designed many temperature-compensated circuits (and many more that were not temperature compensated, of course), the phrase in question seems crystal clear to me. In fact, I myself wrote many identical ones. But I fully respect your right to read anything different from it, probably as a result of a different background and/or a different life experience.

trux said:

And it also cannot be used, because the device does not know it. At Instinct even less than at Edge, since it is much more influenced by the body temperature. And the temperature from the weather widget is not used either - if it relied on the weather station data (which is often not available anyway), it would read the pressure from there too.

This notion that the watch cannot use temperature to perform the conversion to MSL because the two temperatures available to it are unreliable or only approximate, is deeply flawed (even if we discard that last baseless IF-THEN statement). It reminded me of the mantra of an old professor of mine who would always receive his new Estimation Theory students with the phrase "If you know what you are doing, using an approximate estimate is always statistically better than using no estimate at all." The watch will always perform statistically better extrapolations to MSL by using an approximate value of the local temperature than by using no estimate at all (and thus, a fixed value, be it summer, winter, night, or day).

Anyway. I feel that the arguments in this conversation are not being productive anymore, We are approaching the point where things become dogmatic ("it is so because I think and wish it to be so"), and that is the moment when these conversations must end. From here on, the utility is gone, and only the time loss is real.

Therefore, if you agree, and from my part, I consider this exchange terminated. Be happy, and continue enjoying this great watch.

PMO said:

As someone who has professionally designed many temperature-compensated circuits

I am on the same boat. And additionally, in case of Instinct I verified it experimentally, so I do know it does not adjust the MSLP with changing temperature.