I'd said
>> .... then perform a small electronic offset or trim to give the desired
>> amount of under-correction. Inherently runs in full servo, doesn't require
>> feedback loops >to the actuators, etc.<
and Kit said -- correctly without knowing my context (which wasn't clear)
> It doesn't? How do the actuators know which direction to go and how far?
What I meant was without feedback from the actuators to give relative
position (e.g. via our beloved LVDTs). The adder (or functional equivalent)
that gives the offset will give the right sum 'automatically' based upon
which 'side' of the bob is actuating. The 'feedback loop' is entirely
related to gravity and the instantaneous local orientation of the bob (as it
tilts in response to the active mechanism); logically there would be a small
'dead zone' on center, or some form of time-delay or buffering, to prevent
oscillations similar to 'porpoising' in aircraft control.
(By the way, I LOVE the story about the cephalic negative-cant-deficiency
sensor suite! That's what I call 'using your head for more than a
hatrack'...)
Now, of course, I was being somewhat rhetorical in describing the
'simplicity' of a tilt system controlled that way. In practice there would
be all sorts of other systems, both to improve response and to attempt to
provide fail-safe operation. For example, if each side of the 'bob' has a
strain gage that works in both tension and compression, the opposite side
can act as a cross-check or redundant system to determine the magnitude and
direction of tilt.
Naturally, a full practical system would use some sort of linear encoder to
determine the tilt mechanism's absolute excursion, rate of change, etc. and
this would be used to control and modulate the active tilt modus. I believe
digital encoders can be made better and cheaper than those with LVDTs at
present (there's a lot more processing and tinkering involved, but the
materials and cost are very small and, once fabricated, are much easier to
adjust and maintain). I would expect the Pendolino's system to use some
form of optical device for this -- my systems for suspension and valve gear
use a coded strip similar to the material used in CDs, with Gray-coded
information that is 'scanned' relative to a jittered laser pickup when the
active element moves. You can easily code some 'absolute' position markers
into this, and assign them any type of value 'in software', as well as get
the sort of pure differential information that LVDTs provide.
Gerry -- you'd be correct about the added cost and expense of the gyro if it
were a mechanical unit that was sourced uniquely for an advanced passenger
train. I had been assuming that 'sensitive gyro' was either a
second-sourced commercial navigation unit, for example from an obsolescent
aircraft, or was something easy to manufacture like a modern fiber ring
gyro. Note that regardless of the number of axes of 'tilt' you choose to
monitor with a ring gyro, most of the physics, and the fab cost, remains
constant. (Of course, these gyros are essentially immune to curve-entry
lateral acceleration per se, as their axes are vertical; they'll be mounted
with a top or bottom 'hinge' or gimbal in the appropriate plane, which will
cause them to tip on lateral acceleration and thereby induce the error
signal sent to the tilt mechanism for correction.) I'd be reluctant to use
only a lateral-plane accelerometer (e.g., cantilever-beam piezoelectric) for
this, as there are going to be moments in the suspension, of various kinds,
that I'd think would make 'clean' sensor information extremely rigorous to
acquire...
Correct me if I've missed something -- but my understanding about the balise
system was that it provided the appropriate 'prediction' of curve entry to
allow the system to be in full effective compensation without inherent lags.
It would be possible to incorporate enough information into circuitry in the
balises to be able to predict what the tilt system would do based on known
curve geometry and observed train speed -- but this would involve much more
complexity in the trackside systems and balise-to-train communications. So
I would suspect it's just a means of indicating something like 'distance to
curve entry' and 'permission to tilt' rather than a factor in intelligent
negative cant deficiency management (insert your chosen acronym or buzzwords
here, please ;-})
