There are probably not many measuring devices based on a physical phenomenon as unstable as barometric altimeters. And yet, they are still widely used, for want of anything better. Technological progress even makes it possible to go even further in the accuracy of this technique, which seems to defy common sense. Welcome to the YoctoAltimeterV2!
A barometric altimeter measures the atmospheric pressure to determine the altitude. As the barometric pressure at any given point doesn't depend only on the altitude but also on the state of the surrounding air mass, that is of the current weather, a barometric altimeter is usually used in differential mode: you give it a known barometric pressure at a given nearby location and the altimeters computes the altitude variation, based on the pressure variation.
Theoretically, it seems simple and Cartesian. When you look in more details, you discover many issues, of more or less importance:
 The formula to compute the altitude is an approximation, because the physical model is actually complex. There are several formulas: one used by pilots, one by meteorologists, and they vary depending on the country...
 Weather conditions at the reference location and at the measuring location are never exactly the same, which distorts the measure
 Weather conditions change continually, which requires continuous updating of the reference pressure, and not a unique update
 Digital pressure sensors based on the MEMS technology have a non negligible error on the absolute value of the pressure measure
 The absolute error of MEMS pressure sensors often tends to drift with time
Despite all these difficulties, if you define a scenario with care, you can work small miracles with a barometric altimeter. And as we discovered a little while ago a new barometric pressure sensor, which is very stable and has an incredible sensitivity, we couldn't resist the temptation to offer you a new YoctoAltimeterV2.
YoctoAltimeterV2 vs. YoctoAltimeter
As usual, the new version fits in any location where the YoctoAltimeter was already used, without even changing the software. But the new version if both much more accurate (in absolute value) and much more sensitive to variations.
YoctoAltimeter  YoctoAltimeterV2 


The absolute accuracy of the YoctoAltimeterV2 can even be improved by one order of magnitude with a simple one point calibration, as this new sensor has an impressive stability and repeatability of 0.01 mbar on any variation of 10 mbar between 950 and 1050 mbar.
An astonishing aspect of this new sensor is the low noise, the sensitivity to very small pressure variations, and the strong correlation between measures provided by two distinct sensors. To demonstrate this, we implemented a small experiment consisting in measuring the relative altitude between two YoctoAltimeterV2 which travel in a miniature elevator along a wall. Each elevator car is a small metallic box containing a YoctoAltimeterV2, a YoctoAltimeter for comparison purposes, and a YoctoHubWirelessg to transmit the measures in real time:
An experimental car
At the top of the wall, a motorized reel moves the elevator cars up and down alternately. The reel is made of a simple DC gearbox engine driven by a YoctoMaxiIOV2, which also allows you to read limit switches.
The installed miniature elevator
Below, you can find in green the estimated altitude measure of one of the two cars, on the left scale, with the other car as reference altitude. This graph was drawn on purpose when the temperature  in red  was significantly decreasing. Thanks to the use of altimeters in differential mode, compensated every five seconds, we keep a very stable measure. The sensitivity of the height measure, in the order of a few centimeters, seems to be a miracle...
The results of the YoctoAltimeterV2
In comparison, the first generation YoctoAltimeter drifts a lot faster compared to the reference sensor:
The results of the YoctoAltimeterV1
Beware, don't rejoice too fast: hard reality is ready to catch up with us fast. The experiment above happened in almost ideal conditions, where both cars moved in the same air and stayed at the same temperature. If you now stop the cars for several hours at different heights, and if you let one of the metallic boxes heat up in the sun, we note that a heat difference of 5°C of one box compared to the other makes it take 0.5m of altitude, while it was naturally stationary:
Impact of the temperature difference on the altitude measure
Seeing the very direct correlation of the temperature difference between the boxes and the altitude error, it seems that a small linear correction should be sufficient to solve the problem for this particular application. Maybe... but in any case, remember to be wary of barometric altimeter drifts!