Lot testing will not fix problems that are the result of design decisions.
The specifications indicate that there are missing codes and it isn't monotonic.
I have never heard these assertions made against the microcontroller I'm using for the Raven, and I can't find anything like that in the spec or the errata for the part (
https://www.zilog.com/index.php?opt...ne=1&parent_id=2&familyId=6&productId=Z8F0880) Moreover, those assertions are not supported by any recorded flight data I have ever seen. So either David has gotten confused (hey everyone has a bad day) or he's talking about a different microcontroller. Now that said, the nonlinearity error of +-5 LSB isn't a great spec, considering how it's only practical to calibrate at +-1 G and we're talking about a 70G or 250G range. I've looked into changing microcontroller families, and the trade just hasn't come out in favor of making the switch away from Zilog so far. Apparently Zilog stopped developing new ones some time ago, so at some point I'll need to switch, (probably to a 32-bit ARM) but not any time very soon.
But that doesn't address the problem of running a 5V +/-5% accelerometer at 3.6V. A simple fix here is to switch from the ADXL78 to the ADXL001. Same package and an almost compatible pinout but it is single axis and costs more.
First of all, the accelerometer operating range isn't 5V +/- 5%, it's specified to operate from 3.5V to 6V.
https://www.analog.com/static/imported-files/data_sheets/ADXL278.pdf See page 5. I know from experience it really operates down to about 3.3V. Operating the Raven at 3.6V puts allows the Raven to run from a single Lipo cell, which is key for packaging it into a tiny av-bay. That's one of the main reasons why it was chosen by Tripoli for all of its barometric-based records, and why I couldn't fly any other altimeter in most of my rocket designs even if I wanted to. I can install 2 complete av-bays for about the same mass as a single 9V battery.
I would also contend that an altimeter with a dual-axis accel and 2%-10% accelerometer error has more useful and informative information than one with a single-axis accel and 1%-2% error. It's especially useful when diagnosing a problem flight, when data is needed most. After a shred, did the rocket go aerodynamically unstable as it hit Mach 1.8, or did it throw a fin? How high up was the rocket when it stopped fishtailing? What was the sideways force on an eyebolt that got bent? You've got a good chance of figuring that out with a dual-axis accelerometer, and almost no chance if you don't. Cost aside, I would rather fly a a dual-axis accel that gives me data that's probably good to a few percent, maybe 10% error if I'm unlucky, rather than a single-axis accel with better linearity. Not everyone will think the same way, but anyone is welcome to design their own Raven-beater and see how that fares in the marketplace.
For every decision that went into the Raven's design, I have done my best to make choices that balance the needs and desires of the largest number of users, while maximizing the Raven's compatibility with record-setting rocket performance. Up to about 110,000 feet, the Raven is pretty hard to beat by those criteria. It did successful deployment duty and recorded excellent-accuracy barometric data on the Carmack prize-winning flight to 105,000 feet. It has been used to break altitude records for H, I, J, and L impulse motors. Its potential in multi-stage flight has only begun to be tapped. Is it the best choice for a 130,000 foot shot where you would like to have a really accurate accelerometer-based apogee detection, preferably with a Kalman Filter? Probably not. For just about any other flight, there's no altimeter I would rather fly (even if I didn't own the company).
May your rocket flights be safe, fun, and informative.