My Raven fairly consistently reports significantly higher altitude from the accelerometer than from the barometer. The last five flights are as follows:
Motor.....Accelerometer.....Barometer
H151.....2685.....2271
H170.....2677.....2458
H159.....3996.....3600
H195.....3401.....3048
I205.....5110.....4562
Any thoughts? Is this amount of variation typical? Any keys to identifying the more accurate number?
The baro and accel sensors are complimentary, and each is more accurate than the other in different parts of the flight. At apogee, the baro sensor is far more accurate than the accel-based altitude. The rocket has been coasting quietly along, going slower and slower, which provides very good conditions for accurate baro measurements.
Small errors in the accel calibration or linearity accumulate over the duration of the flight, and so the accel-based altitude accuracy tends to degrade with time of flight. Others have mentioned that the flight path can affect the results, since the Raven can't tell which way the rocket is pointing, so it assumes that the gravity measured when it was resting on the pad is acting against the measured acceleration throughout the flight. This can lead to significant errors when the rocket is far from vertical. The accelerometer in the Raven is also only moderately accurate to begin with, since it's an analog sensor that is used to make measurements well beyond the +/- 1G range used for calibration.
However, during the boost phase of a typical rocket flight, the accel-based altitude can be more accurate than the baro-based altitude. The baro sensor measurements are affected by any transients changes to the av-bay pressure, which can be common during boost, particularly in high-G flights, when the parachute shifts backwards. In a high-G flight, even a well-sealed av-bay can have significant pressure transients due to the mass of the air in the bay causing increased pressure at the bottom. In transonic and supersonic flights, there can be pressure transients when the shock wave moves over the vent port. The accel is immune to those error sources. One thing that can be fun to look at is to zoom in on the initial rocket motion and measure how long the rail or tower was. You can do this by displaying the accel-based altitude along with the lateral acceleration, and look for the accel-based altitude at which the last impact that shows up in the lateral accel data.
The accel data is also useful for judging the strength of deployment charges. Zoom in to the data around a deployment, and you can tell whether or not there was a big jerk when the rocket pieces take out the slack in the recovery harness. Let's say the rocket has a typical configuration with the main chute above the av-bay, and the first separation charge below the av-bay. The apogee charge will show up as a big positive impulse, and the shock cord jerk will show up as a negative. You can even measure the average deployment speed by dividing the time between the +and - impulses by the length of the shock cord. If the charge is wimpy, you'll know it because there isn't any clear negative impulse. You can still measure the initial velocity in that case by looking at the initial change in velocity when the deployment charge goes off, and adjust according to the portion of the rocket mass holding the altimeter compared to the mass of the ejected part.
In summary, the Raven's accelerometers can provide lots of interesting or useful information, but for the apogee altitude, stick to the baro sensor.