[time-nuts] matlab, python, etc.
Bob Camp
lists at rtty.us
Mon Jan 7 17:25:23 UTC 2013
Hi
It being lunch time, I got out my stopwatch. On my machine here, Matlab
takes a bit over 30 seconds to load. FDATool takes another 30 seconds to
load. Loading a saved session on a fairly simple FIR into FDATool adds a
bit over a minute. If I tell it I want to load a fairly complex FIR that
bumps it up to three minutes. None of that actually involves any "work",
that's just bringing up the session. Yes, it does a bunch of math to do the
display, and that's why it takes a while.
The machine has it's own license, so it's not waiting on FlexLM to do some
mumbo jumbo. Nothing else was running on the machine.
This is not something I would want to try on a single core running at 500
MHz...
Bob
-----Original Message-----
From: time-nuts-bounces at febo.com [mailto:time-nuts-bounces at febo.com] On
Behalf Of Jim Lux
Sent: Monday, January 07, 2013 9:40 AM
To: Discussion of precise time and frequency measurement
Subject: [time-nuts] matlab, python, etc.
On 1/7/13 4:30 AM, Bob Camp wrote:> HI
>
> Well if you are getting it done in seconds on Matlab, then you likely
don't need Matlab very badly. Around here a typical Matlab setup is
indeed CPU bound for a *lot* longer than that during a normal work day.
Two or three hours a day is not at all unusual.
I do tons of stuff on Matlab that runs in seconds, because it's a easy
and facile development environment. Here's a time-nutsy sort of example..
I've been working on systems for time and frequency distribution among
orbiters and landers at a planet. Say you're flying one of those
newfangled Deep Space Atomic Clocks on an orbiter, and you want to
propagate that to other orbiters or landers. The questions that come up
are things like "how much RF power do we need to radiate", "what
performance is achievable at the lander, for an orbiter with orbit
parameters X and Y"..
So you've got to run an orbit propagator, then take that data and run it
through some sort of ad-hoc models for the time/frequency transfer
performance, and then take that and come up with some meaningful plots
to put on a slide or in a report. And then, someone asks, ok, that was
for a 400km circular orbit around Titan, but what if we make it
elliptical with a 2000km apoapsis, oh, and by the way, what's the
maximum time we can NOT run the transmitter during periapsis, and still
get decent nav performance.
sure, I could load up the orbits in something like STK or SOAP, generate
a file of positions or link analysis, then ingest that into something
that does the time analysis, etc.
Or, I can just use an existing Matlab toolbox that does the orbit
propagator, run it through the existing Matlab code that does the link
and time modeling, and use Matlab to generate the pretty plots. And
when the weird question comes up, I can just change a few numbers,
re-run it, see that it's going to be a disaster, performance wise, go
back to them and say, "are you sure you want to do that, why not a 1200m
apoapsis?" or something like that.
There's a whole lot of stuff that time-nuts do in terms of data analysis
that is pretty quick and easy in Matlab (or Octave), especially for
"fooling around". I'm not wild about Matlab's data acquisition
capabilities, but then, I'm less wild about LabView (because under it
all, I'm a "edit the text file, compile and run" kind of guy).
And if you're doing something like filters and control loops, Matlab
makes it pretty easy, and has a lot of library stuff. Granted, all this
also exists for most other languages, e.g. Python.
A couple years ago, I was trying to figure out a way to measure the gain
and noise behavior of a GPS receiver with just the raw bits from the
sampler. Matlab makes it easy to generate data and do the analysis.
Ultimately, Matlab (and related products) saves developer time which is
expensive. It's faster (wall clock wise) to just call a toolbox routine
to, say, do some curve fit or digital filter than it is to code it up in
something else. This is especially true when the developers are more
domain knowledge specialists rather than software engineers: they like
using Matlab, they're familiar with it.
_______________________________________________
time-nuts mailing list -- time-nuts at febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
More information about the Time-nuts_lists.febo.com
mailing list