I wish I was here back in the beginning (what a great project). As I jump in with my purchase of an APM 2.0, I am spending a lot of time reading to educate myself while waiting for the backorder to arrive. In the release notes for the APM 2.0, it said, "Three processors -- a triple-core autopilot!"
I am looking at the Eagle Files schematic. I find an ATMEGA1280-16AU and an ATMEGA8U2-MU. I assume that is two of the three cores. What/where is the third?
The release notes, under Main Features, also states that the MPU-6000 frees up more than half of the processing capacity of the ATMEGA2560. I don't see the 2560 on the schematic.
It's the DMP processor in the MPU-6000. The current code doesn't use it yet but it will (pending permission from Invensense to release our code as open source).
The board uses a ATMEGA2560, not a 1280. But they use the same footprint, which is why it shows up that way on the Eagle file.
So the ATmega8U2 appears to handle the USB connection, the MPU-6000 isn't used yet, therefore everything is on the 2560 which makes it a single core (waiting for the 6000 as the 2nd core). It the 8U2 doing something else that we can really give it any credit for?
I am not trying to knock it, I am just trying to understand. This is an awesome product.
It's not an ATmega8u2. It's an Atmega32u2, and it handles all the RC interface (PPM encoding), in addition to the USB interface. In other words, during flight it runs all the RC/failsafe side of the autopilot.
The fabled DMP huh. Is there any proof anywhere that it actually exists?
No code, no specs, nothing to bee seen anywhere on it.
Two paragraphs in a datasheet isn't really proof. The description is vague. In no way can I get any sort of "Motion Processing" if I had the chip and the datasheet.
I'm just asking for someone to say that it really exists, they've done some small bit of testing that proves the thing does anything at all. A little assurance that it's actually real.
I ask because I've seen writings on the interwebs that say it does not exist in a "real" sense or that it amounts to being able to tie an I2C mag line though the chip.
Hopefully someone can be the first to definitively say on the internet that it actually exists and works.
Jake: We have a version of the code that uses the DMP and is flying now. We just can't distribute it yet. It's a complete substitute for the DCM, except for the mag fusion, which we still do on the atmega2560.
Thanks for the info. Now we know it's actually something.