On Tue, Nov 20, 2007 at 04:01:24AM -0800, Andrew Greenberg wrote:
> Hi Mark,
> > did you find out how fast you could get away with sampling the various
> > devices?  I have some robot applications where I found that if I sample
> > the data too fast I get different result than if I give things a chance
> > to settle.  Its a function of the device you are measuring.
> We *very* carefully chose the sample rates for all of our inertial
> sensors on that IMU. You can argue whether or not they're optimal, but
> they're definitely not too fast. They've got solid anti-aliasing filters
> on them, and the whole system was designed off the noise floor of the
> sensors.
> Also, we very carefully examed the raw data coming off the IMU (in terms
> of CAN packets) before flight, and it was just what we expected. So I
> don't think there's anything wrong with the IMU.
> > have done the actual sampling at a very regular interval. It was just a
> > micro-controller, and had low latencies in its timer interrupt handler.
> > I believe Andrew tested that with a good oscilloscope.
> Yep, I did, and it was dead on.

So what was the dead on sample rate of the IMU packets?

> >> I'm sure it did, at one point, but once everything was integrated and
> >> additional data was put on the CAN bus I'm thinking things where
> >> different.
> CAN packets are transmitted based on their message ID, so they can get
> all out of order. That's half of the "fun" of CAN :/

Hmm, that could be something to look at in the post processing.

> >> So before lift off the data seems to come once a jiffie, after launch I
> >> get 25ish samples per jiffie (most times) with spikes of 70 and lows of
> >> 1.  How do I infer a uniform time per sample?
> Huh. So, I want to look at that. I really thought we solved this problem
> before flight, so I think we should all sit down and make sure the
> post-processing isn't missing something. I swear the ADEOS patch fixed
> all this junk.
> Anyway: happy processing! See you all at the next meeting.

When will the next meeting be?  This is Thanksgiving week you know.


Attachment: signature.asc
Description: Digital signature

psas-software mailing list

Reply via email to