<Perhaps inane comment>

RE:  ... If you're willing to hack the firmware ...

It's *so* cool that the hardware/firmware/software is open source!

</Perhaps inane comment>


On Fri, Nov 15, 2013 at 7:27 PM, Keith Packard <[email protected]> wrote:

> Casey Barker <[email protected]> writes:
>
> > What's the expected sample rate during the ascent? Storage limits aside,
> > what could the hardware do?
>
> We're running everything at 100 samples/second, and the GPS at 1
> sample/second. If you're willing to hack the firmware, you could
> increase the rates significantly for some of the sensors. Just to remind
> you, here are the sensors that we're using:
>
> sensor                  model                   resolution      max rate
>
> Z axis accelerometer    Freescale MMA6555       12 bits         500
> IMU (3 axis accel/gyro) Invensense MPU6000      16 bits         1000
> Mag sensor (3 axis)     Honeywell HMC5883L      12 bits         160
> Baro sensor             Measspec MS5607         24 bits         100
> GPS                     u-Blox MAX 7Q                           10
>
> As you can see, the baro sensor is the slowest at a 100 samples/sec max
> to get the full 24 bits. Reduce the resolution of that sensor and you
> can make it go faster, but that means more noise in the height data.
>
> We've got 8MB of flash space, and the logging format stores 64 bytes per
> sample, so enough space for 128K samples.
>
> I've got profiling support for the hardware, and at the current sampling
> rate we're using less than 10% of the CPU, so it's pretty clear we've
> got lots of head room for faster sampling.
>
> I haven't measured SPI or I2C bus utilization. The hardware is designed
> to minimize contention by avoiding long delays with the bus busy
> (thanks, MS5607), but it's possible that you'll just run out of steam --
> the IMU, Z accel and baro sensor are all on a single SPI bus as the
> other SPI bus is used for the radio, which uses the bus for extended
> periods and would cause a bunch of sample rate jitter if we stuck a
> sensor on that.
>
> From a firmware perspective, we've always used a 100Hz clock
> tick. There are a bunch of pieces of the system which have that value
> baked-into the algorithms, from the Kalman filters for the vertical
> motion to the quaternion integration of the gyroscopes. Note also that
> the flight logging and telemetry uses 100Hz timestamps on everything, so
> changing that rate would mean hacking all of the ground software too.
>
> Another trick here is that you'd be running the z-axis and baro sensors
> at different rates, and that would significantly complicate the Kalman
> filter implementation, which currently assumes that the measurements are
> perfectly synchronized.
>
> Overall, I think this would be quite do-able, but it looks like a fair
> bit of work.
>
> > Our major point of interest is the first
> > ~second after ignition, down to the effects of buttons leaving the rail.
> We
> > really want to understand that phase. Beyond that, staging and motor burn
> > are more interesting than coast. I know on TeleMetrum, descent has a
> slower
> > sample rate, so I wonder if we could further refine the rate over the
> > flight profile.
>
> Yeah, that'd be an easy hack in the firmware; there's a flight-state
> dependent logging rate which would be easy to extend.
>
> > We'd still like to include a 70-cm 100-mW Beeline APRS, mostly because
> > we've got all the gear and familiarity tracking these.
>
> Backups are good. Note that TeleMega can also send APRS beacons,
> although as they consume 1 second of bandwidth per position fix, it does
> mean a significant reduction in telemetry transmission.
>
> > I know the Beeline would interfere with rx on the TeleMega, but do you
> > foresee any issues with the TeleMega in pad/flight transmit-only mode?
> > Could they be in the same bay?
>
> There shouldn't be any problem; TeleMega doesn't care about other RF
> transmitters as long as they're of modest power. A 5W transmitter right
> next to the board might cause issues. Of course, ground testing is your
> friend here; just make sure you light up *all* of the electronics when
> checking for interference.
>
> > Whatever the answer, we'll bench test it. Just looking for a swag, and
> > a sense of whether it might kill the rx front end. Plan B is to move the
> > Beeline up to the nose.
>
> I suspect your range for RX on the TeleMega board itself would be
> dramatically reduced with another transmitter in the same bay, but that
> would only affect the 'idle' mode command link.
>
> It shouldn't have any effect on the TeleDongle (or TeleBT) receivers, as
> long as you're well away from them in frequency.
>
> --
> [email protected]
>
> _______________________________________________
> altusmetrum mailing list
> [email protected]
> http://lists.gag.com/mailman/listinfo/altusmetrum
>
>
_______________________________________________
altusmetrum mailing list
[email protected]
http://lists.gag.com/mailman/listinfo/altusmetrum

Reply via email to