Tracking the 12v Battery with OBD2

I'm not sure whether to post this here or in the "12V battery" thread... I was actually
reading some of the "flat battery" threads over at speakev. Anyway, I went to revisit
the Jeju PID for "aux battery current" which didn't seem to work for me as given.

The file has "Aux Current,Aux Batt CURR,2102,(SIGNED(W)*256+V)/1000,-100,100,A,7E2"
or the like. It's mostly right for the Kona, at least the ID and PID and bytes are okay,
but the math is apparently wrong. Looking at the raw 16-bit value from W and V seems
to center around 32768, with deviations above when the car is in ON and below when
it comes to RUN state. I hooked up the ol' current-clamp so I could correlate values
to the reality of what's going through the wires.

Semi-long story short, it seems that this works well: (32768-((W*256)+V))/100
showing discharge as positive, same as with the traction pack. Adapt as needed
to your own PID definitions. Note that it's just current in/out of the 12V battery
itself, and doesn't represent running 12V loads of the car once the DC/DC comes
on. It stayed fairly negative on a drive this evening but slowly worked its way
back toward 0, holding at about -2 amps by the time I got home. That's not
enough to run headlights, butt-burner, gauges, etc ... but that's what the DC/DC
was floating it at, still around 14.9 volts. I guess it wasn't long enough to reach
that state where it drops to 13.6V during driving.

The original equation must have been assuming milliamps centered around 0
or something. This doesn't seem to work that way. It makes sense, as most
current sensors power from 5V and swing output either side of 2.5. The question
now is, where *is* the 12V current sensor, or is it a derived value from other
hardware?? I dont see any hint of one in the schematics.

_H*

 

So that little doohickey *is* a current shunt after all! I suppose I should have
guessed, given the different metals little 2-wire connector coming off it. It seems
to be pretty accurate, too, against my clamp-probe. *Absolutely* no hint of it
on the schematics; the bit on the positive terminal is undoubtedly a remote
voltage sense. [Prius has one too, and has to run the length of the car]

_H*

 

I started thinking that maybe it's supposed to be "+X" instead, since it comes
after W and a lot of other 16-bit values from the same ECU / id are presented
as high,,low in the alpha order. It almost certainly isn't (V,,W) as W
tracks rough current deltas reasonably well and V jumps all over the place.

I could easily see different engineering groups for HV vs. 12V systems
having their own preferred endian-ness for reporting and submitting their
arbitrary parameter specs to whatever collective finally put it all into the
EPCU code. We may have a "what were they thinking" reverse-engineering
type of problem here...

OBDlink is not your friendliest CAN debugging tool, but it was easy
enough to make three simple "gauges" to just read V, W, X as raw (decimal)
values, and play with various loads [lights, fan, butt-blasters]. I got
these example combinations with higher and higher currents drawn, all
in ON-mode discharge as it's easier to control:

W V value
124 156 31900
123 212 31700
122 168 31400
122 68 31300
121 224 31200
121 124 31100
120 180 30900
119 236 30700

X just sat there showing "69" the entire time, so it's evidently not in
the picture despite my thinking that the value might be (W,,X) instead.
But Jeju has that as 12V SOC [directly in decimal percentage, yet], so
I wouldn't expect it to move much.

It's clear that the combined value is stair-stepping by decimal hundreds when
taken as a 16-bit value. The V figure sat quite steadily at the observed
values, and didn't "bobble" around much at all. I would expect it to be far
more variable, with minor changes such as a different seat-heater level or
fan speed. Nope. Rigidly granular.

My "/100" interpretation would also only deliver a maximum range of
+/- 32.7 amps, and we know the 12V system is capable of much more. But
reading this way *was* matching my amp-clamp quite closely in realtime,
so while it's useful to have as is I'd rather figure out WTF is really
going on here. I should repeat this table with the amp-clamp in, but
it's dark and cold out at the mo'.

Given that the SOC reports in decimal, maybe this is all about W times
some factor centered around 128, only reporting in 8 bits, and the V byte
is something else. A simplified (128-W) or the like wouldn't match my
amp-clamp very well; maybe this too is a percentage of some known
max/min value that the shunt can measure?

Anyone know how to contact the mysterious Jeju??

_H*

 

I got that bit about 327xx/100 limited to 32 amps wrong; it would be more like
+/- 327 amps, wouldn't it. I did a little more measuring; the shunt seems to
develop about 1 millivolt / 10 amps, so it's pretty wide range.

Changing the constant in the equation to 32700 seems to still match reality,
and then gives integer amp values for which we can knock off any decimal
places. For this we don't really need any more granularity than 1 amp,
especially when you can't use OBD2 to measure currents when the car
is powered off.

_H*

 

12V battery current aside, I went after a slightly different aspect today.
With the current shunt right on tne battery, that's not an indication of
how much overall 12V current is being sent to tne car, once we come
up into full RUN mode. Aux battery current eventually gets down to an
amp or less of charge. What I wanted was the proper PID for low-voltage
DC converter current, feeding everything. There are a couple of
possibly-theoretical PIDs for the Niro or Soul, but they seemed to
be wrong.

I took another look at Jeju's research spreadsheet. In the "VCU" tab
under PID 2102, we see "Aux battery curr" filled in with some sampled
values. Given the foregoing discussion, the resultant value shown indicates
that the battery is discharging, i.e. the car is in ON mode and the LDC
isn't running at all. There are nearby pairs of bytes that are 0. I made
the assumption that LDC current would never be negative, and could
thus be zero-based rather than 327xx midpoint-based. I also made the
assumption that since this all talks about many parameters of the
VCU or PCU in general, the group concerned with the LDC might all
be contained in this small block.

I fiddled my "raw bytes" PIDs to watch P and Q, confirmed that they're
00 00 in ON mode, and went to RUN. The two parameters gratifyingly
changed as the 12V rail voltage went up. Q was a modest positive
value, albeit somewhat variable; P was all over the place. Aha: probably
high,,low ...

Long afternoon short: I think I have a working PID for LDC current.
ID 7E2, mode/pid 2102, math ((Q*256)+P)/100 ... same little-endian
methodology, similar reduction from 16-bit. I managed to get my small
current probe around the big LDC wire near the relay box. Problem is,
current draw is highly variable -- I watched it on a scope, it has somewhat
periodic waveforms varying by several amps sometimes. The scope's
averaging function helped a little, but I still wasn't sure.

Can someone with a bigger current clamp get it around the main LDC
wire coming off the PCU, it's on the passenger side and the output fitting
is black plastic with a red "+" on it? Bang this PID into your favored OBD2
tool and run different loads ... rear defrost, fans/lights, butt-burners, the
non-driver power windows are the big ones. See if I'm actually in the
right ballpark WRT measured current here. Different refresh rates might
give varying results. RUN with nothing else turned on seemed to sit around
9 -10 amps; with bloody everything in play I could drive it to upwards
of 100, and those wires were slightly warm.

I didn't look at ab, ac, or anything up from there yet.

_H*

 

So here's the present state of my "dashboard", based on all of the
foregoing halfassery ...



Keeping it running takes a lot of phone battery, because it's polling
for a lot of stuff. But I think I've got the important diagnostic basics,
which are definitely worth checking in on here and there. Now it's fun
to correlate some of this to what I see in the "electricity use" screen,
and things match up pretty well, math-wise.

This is just after charging to 90% this morning. Ambient temp around
30F, but the pack has clearly retained a bit of heat from when charging
ended sometime in the wee hours.

_H*

 

At the time of plunking in battery temps I was more interested in heat, so
watching the BMS "max temp" seemed like the best idea. I threw in temp
3 just to see a single sensor too.

Obviously I'm not using the Jeju CSV names and labels, I'm entering my own
into "obdlink".

I need to watch all the LDC stuff more on a "real drive", as I never went to the
13.x volt state while just sitting in the driveway. For voltage I just use the one
that the OBD2 widget directly reports at the DLC.

_H*

 

I've added one more picture to my "auxbat" page, to wit:



Scoping LDC current, showing a lot of variation at ~100 Hz or more.
So our OBD readings are going to get messed up by a simple matter
of sample rate.

Remembering those old buzzer-style voltage regulators...

_H*