###  This is excerpted from the Prius_Technical_Stuff yahoogroup,
###  starting at around message # 16003 in the archives.


Date: Tue, 28 Mar 2006 03:07:51 +0000 (GMT)
To: prius_technical_stuff@yahoogroups.com
From: hobbit@avian.org (*Hobbit*)
Subject: dTPMS preliminary

Ahh, spring!  Time to start tearing the car apart again!

While catching up on some of the early archives of P_T_S, I noticed
a flurry of discussion about indirect tire pressure monitoring, i.e.
a system that does differential analysis on the wheel-speed sensor
inputs into the ABS controller.  This led to major debate back and
forth between the "radialists", who held that a soft tire would sink
down and decrease the effective rolling radius and thus speed up,
and the "circumferentialists" who countered that since the same
linear amount of rubber had to pass underneath, rotation speed would
be unchanged and the tire would simply squirm more [and heat up] to
accomodate the larger contact patch.  I don't know if that debate
was ever resolved, and by now there are huge NPRM documents issued
by the NHTSA and other agencies on the required specifications of
TPMSes in general regardless of what method is used.  Well, your
answer lies ahead.

It's interesting to read about how even the people drafting those
documents still aren't sure about the indirect approach, even though
it's cheaper to implement than outfitting each wheel with a direct
pressure sensor and RF link and all.  There are a number of different
approaches to indirect speed analysis, that try to group diagonal pairs
of wheels to compensate for different-side speeds during turns, or that
try to do hairy FFTs on the frequency variations of the speed-sensors
to sense tire "resonance" which changes with pressure.  Reliability of
these seems to be questioned at every turn.  And out on the fringes are
a couple of other approaches to the problem, one of which I'm guessing
is using ultrasonic distance sensors to measure the height of a
suspension member above the ground.

All of this has been interesting me intensely since I picked up a screw
last winter and the only saving grace was that it was a warm day, I had
the windows open, and happened to hear the "tik, tik tik" as the screw-head
kept hitting the pavement.  I got lucky that day and was able to simply
have the tire patched.  How far would I have driven, with the damn thing
slowly leaking down, if I hadn't heard and suspected something and pulled
over?  With the news emerging that the '06es would come with factory TPMS,
I wanted the equivalent in my '04.  I didn't want to keep living with the
risk of having a tire slowly go soft, such as during the LONG TRIP I was
about to go on at the time, with possibly disastrous results on a highway.
Why let a tire get totally shredded and destroyed when early detection
of a slower leak would allow safe stopping and probably saving it?  As
almost everyone has noticed by now, TPMSes make a lot of sense.

There seemed to be three possibilities open to me:

	Spring for a full-blown TPMS at the same time as getting new tires.

	Crock together something involving clamps and fiberglass drag rods.

	Investigate differential wheel-speed analysis.

The advantage of a full-blown aftermarket TPMS of the right type would be
direct pressure readings from each tire, hopefully including the spare,
with appropriate instrumentation "cool factor", albeit with a significant
outlay and PITA factor of getting the sensors installed.  Not to mention
the extra RF blather flying around the vicinity of the car.

I began trying to find parts to construct the moral equivalent of "curb
feelers" but which would clamp to suspension members right next to each
wheel and extend downward to just above the road.  The idea is that
a soft tire would cause the rod to start dragging on the pavement,
hopefully in a way that would be hearable inside the cabin.  I haven't
gotten very far with this, since I'd be looking for durable hardware
[stainless, etc] and little or no modification to suspension arms.  The
pimp-your-caddy feelers in the bling section of Auto Zone do not come
close to qualifying in this regard.

So I read up on speed sensors again.  They've even got little waveforms
in the repair manual, and I figured that bringing those four signals
into somewhere and conditioning them to logic level wouldn't be too
tough.  But then what?  Frequency to voltage converters?  Count the
pulses and have an MCU run some algorithms to sniff out anomalies?
What's the right way to determine the "average" of these four, anyways?
How much delta would I get when turning?  What sort of indicator would
be appropriate?  If all the wheels have the same number of reluctor
teeth [42, in fact], they should all be at almost the same frequency
on a straight road.  As I thought further about it, something else occurred
to me.  On my gauge panel are three "leftover" red LEDs that used to form
sort of a left arrow for a directional indicator, e.g.

     *
   *
     *

and there's room to drill in a fourth, making an obvious diamond:

     *
   *   *
     *

and suddenly, the approach to the display was clear to me.  Each of
those LEDs could indicate anomalous speed differences between any two
adjacent tires, and the most simplistic way to detect such would be
to mix the wheel-speed frequencies and let them "beat" together.  So
on a straight road the LEDs should remain fairly stable in various
degrees of brightness, and vary slowly as minor differences happened.
A soft tire would in theory make two of the LEDs cycle between bright
and dark more often, giving a very simple and clear indication of where
a problem might be.

So I needed to test if this was even remotely feasible.  It looks like
the speed sensors produce in the neighborhood of 10 hz for every 1 mph,
with amplitude falling off down near zero.  Their shielded two-wire
outputs all arrive at the brake ECU, which I didn't feel like digging
into quite yet.  But the two rear sensors pass through connectors at
the right and left kick panels, making them much more accessible.  It
wasn't very productive to run the sensor into the listener-widget, but I
*could* hear a very soft note once up to a reasonable speed so I knew
I had the right wires.  So later today I scoped things out [literally]
and found that the RL+ output is the one that really varies its output
voltage [car jacked up at RL, spin wheel by hand] while RL- remains at
or near ground level.  So it evidently isn't a truly differential input,
even though all the sensors have two wires AND independent shields, meaning
that tapping the + side of each sensor should suffice to determine rotation
speed.  Oh, and that + signal sits at a bias of about 1.2V when the car is
powered up, with signal amplitude about 1.5Vpp at meaningful speed.  The
scope traces in the manual are a little misleadingly drawn.

Anyway, here was a quick opportunity to find out if the whole idea has
any merit at all.  I already had the RL sensor paper-clipped under my
feet.  Under the right-hand kick panel I found the connector holding
the RR sensor, tapped that one too, and brought both into the 2-channel
scope.  Fired up the AC inverter in the back to power the scope, and
went for a ride.  Both rear tires were around 43 psi to start off.

Right away it was clear that turning makes a BIG difference between right
and left of the car.  But once I got to a place where I could go dead-
straight, I switched the scope channels to "add A+B" mode.  This made
a single waveform from both sensors, which would be double-height
when they were in phase and close to zero when they canceled.  So the
varying amplitude of this would be my beat frequency.  And guess what?
While going straight, the beat was *very* slow.  I could cause it to
vary a little more with *very* gentle drifts within my lane, but for the
most part it reflected the fact that the tires are identical.

And now, if you've read this far, you're waiting for the BIG TEST.  I
stopped and let one rear tire down to the !Toyota-recommeded! 35 psi
or so, and took off again.  And guess what?  The radialists WIN!  At
40 mph, I could see a definite beat of at least one Hz, maybe a little
faster.  The tire didn't really *look* any softer than the rest, because
radials are sort of hard to judge that way -- maybe a little longer
contact patch when viewed from right at pavement level.  But that
difference between 43 and 35 psi, still quite tolerable from a safety
standpoint, created a significantly measureable delta in wheel speed.
The variance from taking curves was still in there, of course, and I
found it entertaining that a gentle turn toward the left would bring
the beat back to zero [the RL tire was the soft one].  I stopped and
pumped it back up to match the other, re-verified zero beat on the
straight, and declared this preliminary test run a success.

So I think the way to go is to simply condition the sensor outputs to
a uniform level, and then for each LED in the diamond [e.g. each pair
of tires], simply borrow a page from PLL design and XOR the adjacent
signals together to turn their phase into a duty-cycle.  A nominal
low-pass and driver stage can then go straight to the LED.  So the little
diamond shape can sit there in front of me and beat slowly on the straight,
beat faster at top and bottom on turns, and my brain will just handle the
rest of the analysis and notice any pair blinking faster than they should.
This seems minimalist and strange enough for my purposes, and yet solves
the original problem without big capital outlay or drilling into A-frames.
It will involve uprooting all the skid-ECU harness again, but I sort
of anticipated that at some point anyways.  This should also help indicate
how to *trim* tire pressures and figure out what the pressure difference
between front and rear *really* should be for identical wheel speed.  In
fact, it may be prudent to pre-test front vs. rear the same way before
committing to a particular design.

Given this strong evidence that the principle of indirect TPMS works and
knowledge of the caveats about turning, it seems like some relatively
simple algorithms based on periodic pulse count could allow an MCU to
detect a soft tire without much of the uncertainty that seems to surround
the statements in the NPRMs and patents.  One side's pair of sensors
returning a uniformly lower count for a while would clearly indicate a
turn, which could be ignored, and the code should be able to determine
variation on ONE sensor over a fairly short time.  There are proposals
to simply have ABS ECUs run a little more code to do this, and have a
"tire pressure" idiot light, but the way they're going at it seems unable
to determine if any *two* tires are going.  When a modern ECU also has
access to the steering position output and the yaw sensor, that seems
like a totally solvable problem.

Oh, and of course none of this is a substitute for CHECKING your tire
pressure on a regular basis, as all the documents are quick to say.

_H*


Date: Tue, 28 Mar 2006 06:01:19 +0000 (GMT)
To: prius_technical_stuff@yahoogroups.com
From: hobbit@avian.org
Subject: oop..

the wheel-speed rotors have 48 teeth, not 42.

A little more quantitative frequency determination:

      revs      teeth      mi         secs
  855 ---- * 48 ----- * 40 --- / 3600 -----  = 456 Hz  [teeth / sec]
       mi        rev       hr          hr

[855 r/m taken from TireRack figures for goodyear integrity 185/65R15]

Assuming a soft tire produces 457 Hz instead, reversing the mess above
gives 856.875 revs / mile.

855 revs/mile --> effective radius = 11.794 in.  @ 43 psi

856.875 revs/mile --> effective radius = 11.768 in.  @ 35 psi

Nice testimony to the consistency of tire manufacture, if nothing else...

_H*


Date: Tue, 28 Mar 2006 14:03:19 -0500
To: Prius_Technical_Stuff@yahoogroups.com
From: "Gary Novosielski" 
Subject: Re: dTPMS preliminary

Sorry, but the Department of Transportation has already considered
(and rejected) indirect tire pressure monitoring based upon ABS
sensors.  They found that while it could detect severe loss of
pressure, it wasn't able to detect slight losses as effectively as
they wanted.

So the mandated TPMS systems are direct pressure measurement types.

=Gary

On 3/28/06, JerryW  wrote:
> One of the most interesting posts I've read in a while. It seems to me that
> you have the basis of a fine monitoring system here; I strongly advise you
> do delete the post immediately, and write instead to Toyota. Remember me,
> when you're a millionaire ;-)


Date: Tue, 28 Mar 2006 21:16:19 +0000 (GMT)
To: Prius_Technical_Stuff@yahoogroups.com
From: hobbit@avian.org
Subject: Re: dTPMS preliminary

The effectiveness of indirect TPMS probably has *everything* to do with
sidewall flex, and isn't guaranteed to work with all tires.  Low-
profiles especially seem to render the approach largely useless.  But
I still think it's worth pursuing for the typical Prius tires, and at
some point I'll probably get to repeat the observations for a set of
Hydroedges or something.

There's an interesting little chart on page 8 or so of a set of slides at
  http://www.unece.org/trans/doc/2003/wp29grrf/TRANS-WP29-GRRF-53-20ebis.pdf
illustrating the relative radius dropoff by TIRE BRAND AND MODEL.

The NHTSA tests may also have been done at "placard pressure" and down,
which in this case I'll argue is already too soft.  Were they looking
for the ability to sense a 20% delta away from *there*?  I wouldn't be
surprised if the sensitivity curve starts to drop off once you get that
far away from the sidewall-rated 44 psi in the first place.  [It sounds
like eventually having my typical "raw display" may be useful in quickly
validating any other thought experiments, but said display is a long way
from built yet.]

I did find various mention of tire manufacturers thinking about ways to
help indirect approaches work, by putting some sort of optically-readable
indicators on sidewalls or whatever other wacky schemes.  So the whole
subject may get revisited based on support from tire makers themselves,
or else everyone will give up and just start building RF transponders
straight into rims.  Wouldn't it be a hoot if you could unscrew a little
plug on the inside of the rim to replace a battery contained in a sealed
compartment that wouldn't also let the air out...

Didn't the original Hummers have some sort of air-hose feed through the
hub centers that could *re-inflate* tires on the fly?

_H*


Date: Wed, 29 Mar 2006 06:49:15 -0000
To: Prius_Technical_Stuff@yahoogroups.com
From: "Cor van de Water" 
Subject: Re: dTPMS preliminary

Hobbit,

Truly remarkable and exactly the test that needed to be done.

I am not worried about all 4 tires going soft at the same time.
I am worried about a nail in one tire and shredding it at the
Freeway, at night. Actually damaging 2 tires, so a tire swap
with the spare is out of order.

With a simple (counting) monitor and some SW the tires can
very well be monitored, just using the 4 sensor signals.
Sounds like a new after-market system...

> sidewall flex, and isn't guaranteed to work with all tires.  Low-
> profiles especially seem to render the approach largely useless.

All the tires in the overview showed between 05. and 1% change
in radius when pressure dropped.... Don't see what the problem is,
only where to "cut the line" of a tire pressure alarm...
Detecting a low pressure condition is NO question in my mind, with the
info from the sheet and your measurement validation.

Regards,
Cor.


Date: Wed, 29 Mar 2006 14:21:27 +0000 (GMT)
To: Prius_Technical_Stuff@yahoogroups.com
From: hobbit@avian.org
Subject: Re: dTPMS preliminary

Actually the right way to recharge a battery or ultracap or
whatever in a direct TPMS would be the equivalent of a magneto,
with a small coil embedded in the inner surface of a rim and
one or more magnets mounted just so to sweep past it.  Maybe
the same coil could be used for very short-range data delivery
on its way past, too, similar to a prox card.  But all that
could get into some unsprung-weight and balance issues, as
well as maybe compromising rim strength by having structural
discontinuities needed to mount all this stuff...

You'd probably want a small amount of local-to-tire storage
so that a low tire can be detected right at startup, instead
of waiting until reaching a healthy speed before being able
to sound the alarm..

Today will include some same-side-wheels testing on the
differential thing, along with a very early attempt at an
input-conditioning circuit.  It's basically turning audio into
a pulse train, not hard but I want to eventually minimize
parts count..

_H*


Date: Fri, 31 Mar 2006 05:04:43 +0000 (GMT)
To: prius_technical_stuff@yahoogroups.com
Subject: more proto-TPMS
From: hobbit@avian.org (*Hobbit*)

Well, I threw together a primitive XOR circuit to display a single LED
output, and moved one of the inputs to a front wheel.  So now the
right-hand wheels are instrumented, and I can glance down at the
proto-board in the footwell and see what it's doing.  This gives me
confidence that it's a nice way to represent wheel-speed phase.
For comparing tires on the same side of the car, I expected to see
very little speed delta.

Surprise.  Accelerating or decelerating, even modestly, has a rather
*profound* effect on speed difference.  Not so bad that the circuit
can't track it, but definitely able to vary the beat frequency between
zero and maybe 2 Hz or more just by pushing pedals.  Load is apparently
a large factor as well; we speak of the usual 2 psi difference to accomodate
the heavier front end... With about the same pressure in the two tires, I
was definitely seeing a running delta that I could bring down to close
to zero while *decelerating* on regen.  Obviously, on average, the front
wheel was turning faster.  [I could also see this by which way the two
scope traces drifted past each other.]  Highway speeds yielded a larger
delta than backroads, of course, since the front tires steadily pull harder
against air resistance.  Okay, so I worked on equalizing the speeds under
lower-speed "float" conditions, by putting more in the front...  and it
initially looks like that involves more like a *4* psi difference.

Not quite believing this, I measured tread depth at the center groove of
each tire and came up with .210" in both, averaged over a couple of
different measurements around.  This is fortuitous -- the two left-side
tires measure at .175" F and .215 R, and I have no idea why one of them is
more worn like that, especially considering that it was on the *rear* up
to late last summer.  No, I wasn't measuring at the wear bar.  Fortunately
the worn one was not involved in the both-rear-wheels tests.  But while I
think I might have two equal-circumference tires on the right side, I won't
truly know until I *exchange* them and run another round of testing.  Because
now I'm curious about the "proper" pressure difference with no load in the
back of the car [except a small UPS].

Tight turns appear to make the front wheel rotate faster, particularly left
ones -- this seems correct if I think about the paths each wheel takes.

Bumpy roads also have a profound effect on speed-sensor phase.  The phase
of where those 48 teeth are can change instantaneously, reflected by
very irregular flashing of the LED.

Bringing the front tire down to 35 and then 30 psi still shows a change,
though, and in fact gives the LED a rather "urgent" flash rate.  I think
the 4-LED display thing is still worth pursuing, but a meaningful reading
will need to be taken when straight-n-level on a smooth road, and any given
trip's baseline appearance will have to be sort of learned.  I think it
would take some *very* clever software, with additional inputs from steering
angle and driver-requested go/stop input, to make the same interpretations
that my brain just did in the foregoing.  And a computer has no way of
knowing that I just parked for 3 hours with the sun on one side of the
car, and that the shadowed ones are *not* going flat...

I wonder if it's worth trying to measure circumference with a flexible
cloth tape or something?  Maybe I'll try when I go to swap F and R.

_H*