Ford TPMS

[Greg Troxel‏]

I've been digging into the Ford TPMS decoder a bit. Here's some notes,
sort of note to self shared with those of you that might be interested.

All of these observations are in the US.

This decoder is said to be for more or less all Ford passenger cars and
maybe light trucks. So far I've heard it from a ~10 year old passenger
car and from what I think are Transit delivery vans. The van pressures
are very high, but that matches their data from Ford's website.

I am hearing these on 315 MHz, -- and not at all on 433.92 MHz as the
comments say. I suspect that's a regional difference, and have the
impression 315 MHz isn't allowed for this sort of thing in EU.

The temperatures seem about right; tires in the garage, not having been
driven, are close to that from a BME680 also in the garage.

The pressures are a bit mysterious. I have 8 datapoints from winter
tires after taking them off yesterday (1 each), around 34.5-35 psi, and
then after letting air out to about 32 psi to prompt them to transmit so
I could verify which wheel has which code. I am using a Topeak bicycle
gauge that seems non-junky, and of course don't know about that.

Overall, the rtl_433 reported pressures are high, and multiplying them
by 0.949 makes them on average about right.

I then found the formula in the source, which has two regions, and each
with an offset and per-code pressure increment. This seems kind of
strange to me; I would think one would just pick a scaling factor, as
low pressures are of interest and accuracy isn't that big a deal as this
is a warning device. I did a regression on my data and got a negative
offset and a higher psi/codestep rate. I don't believe it so far.

I can sort of see how it might be "psi = 7 + 0.2 * bits", and the
designers thinking that a pressure of 7 or below is effectively
deflated. That's pretty close to the code right now which is:
pressure_psi = 6.8f + psibits * 0.2122727273;

I would certainly expect a roundish number to convert bits to either psi
or kPa. It's hard to imagine an engineer using some calculated value
for a range that fits and not tweaking it to sort of round. But that's
just me having thoughts and reality is of course out there.

I just did an experiment with my data, backing out the psibits from the
computed pressure, and then applying a formula of
pressure_psi = psibits * 0.25f
in a spreadsheet, and comparing to my gauge. The results are that the
max residual is 0.5 psi, and the sum of the 8 residuals is 0.2 psi.

Therefore, my current working theory is that 0.25f * psibits is the
right conversion. That fits the data, and is something that I think an
engineer would choose. I am curious what others think.

Finally, there was discussion earlier about temperature correction of
pressure. When the car returns after driving a decent distance, both
pressure and temperature are elevated, and the data fits the theory that
the pressure is the measured pressure, not adjusted for temp, and the
data really does not fit the theory that the pressure has been adjusted
to 20C or some other reference temp.


If anyone has a recommendation for a gauge that's moderately priced (say
under $50) that you actually believe is accurate to 0.5 psi, please let
me know.

[Greg Troxel‏]

Here is my histogram of temperatures. The 8-28 are plausible garage and
returning temps. The high ones look 128 too high; could be bit errors.


10 8.0
12 11.0
4 12.0
35 13.0
61 14.0
52 15.0
48 16.0
20 17.0
22 18.0
8 19.0
4 22.0
8 23.0
2 26.0
2 27.0
11 28.0
12 142.0
1 146.0
2 156.0

[Greg Troxel‏]

Earlier I wrote about the Ford TPMS decoder. I've been working on
improvements, and have just submitted a pull request:

https://github.com/merbanan/rtl_433/pull/2071

Note that I am in the US. My only hesitation is that there is a slight
possibility that Ford is using similar but not quite the same encoders
in US and EU. I think though, that the previous notes are consistent
with the same underlying behavior, and just are an intermediate step of
analysis. It would be great if someone could run my PR code in EU and
look at Ford decodes as a sanity check. Or really anyone, anywhere.

I have added a lot of comments, but the highlights are:

pressure is in 0.25 psi units
I confirmed the temperature decoding as at least being within a degree
or two C (and didn't change it).
high-bit of temperature byte means the byte is something else, not
temp, actual meaning TBD. This means some reports don't have
temperature.
decoding of learn mode bit
decoding of moving bit

I am not 100% sure everything is right, but I feel that it's a big
improvement. I've been running it and looking at both a 10-year old
Ford sedan as well as neighbor's cars and guest/delivery vehicles.

Comments etc. welcome.

Greg