Re: flight stability

On Tue, 4 May 2004 [email protected] wrote:
> > if its size is
> > well-known and there's a bit of advance warning, so you can feed in the
> > correction in advance.  
>
> An advanced correction results in some degree of advanced anticipatory 
> spacecraft corrective movement which is not good.

Only if the correction starts too early.  The idea is to match it to the
disturbance, essentially eliminating the lurch caused by having to wait
until the disturbance starts to correct it.  (This is why it's necessary
to know the size of the disturbance as well as when it will arrive.)

Even if the timing or magnitude of the correction isn't quite right,
typically the resulting motion will be rather smaller than if no attempt
is made to anticipate the disturbance. 

>   A rather better idea is to have "rate of change"  output, as well as the 
> "direction of change" output, from the stable platform system. This
> information is used to vary the force of the corrective motion in real 
> time, concurrent with, and equal to, the disruptive force.

Unfortunately, there is no such thing as "real time" when there are
significant time lags in both the detection and the correction... as there
inevitably are.  (E.g., gyro data is sampled at a finite rate rather than
perfectly continuously, and engine gimbaling takes time.)  That's where the 
problem *comes from*.

Not attempting "feedforward" advance compensation means you have to live
with the lurch resulting from correction time lags, and that can have
repercussions elsewhere in the design.  For example, the Saturn V had its
first-stage engines splayed out slightly, rather than pointing straight
rearward, through much of its burn.  This cost a bit of performance, but
put the thrust lines of the individual engines closer to the center of
mass, reducing the strength of the lurch from an engine failure... which
would otherwise have been violent enough to overstress parts of the Apollo
spacecraft.  Since the onboard computer almost always had advance warning
of engine problems -- the major reason for an engine out was the computer
deciding to shut it down! -- this could have been handled with feedforward
instead. 

                                                          Henry Spencer
                                                       [email protected]