The CANary project

[turbo2ltr]

We use Eagle extensively. Very good program and if you compare it with some other packages, it's very reasonably priced.

Don't bother with the autorouter. Not that I've tried it, but typically you want more intelligence when routing.

 

[chris1howell]

One question. What's the motivation to move the SMDs to one side? Using both sides gives me a lot more real estate and it's easier to route the wires.

Using one side for SMDs will allow small batch production at home. If you are using a skillet or toaster oven reflow you can only do 1 side. Hand soldering SMD is very time consuming and tedious. It is okay for a prototype but when you make 5 - 10 - 25+ using a solder paste stencil and reflow saves a ton of time.

With OpenEVSE I have all the SMD on one side and through hole on the other, I can efficently use board space and still build SMD quickly at home. It did take careful placement and a lot of work to route the traces.

 

[TickTock]

We use Eagle extensively. Very good program and if you compare it with some other packages, it's very reasonably priced.

Don't bother with the autorouter. Not that I've tried it, but typically you want more intelligence when routing.

I am using Eagle, too. VERY reasonably priced I am using the free version since I only need 2-sided and can easily fit in the max board size the free version allows. The UI is very retro, though. I feel like I've gone back 15 years. I tried the auto router a couple of times. Agree it isn't very useful.

 

[garygid]

I did not think that the around $600 for a single
user for the standard version was "inexpensive".

However, if the free version does the job, then
there is just the learning curve, so I will try it.
However, it is for non-profit use only.

Eagle Light (same limits as the free version) is just $69
and can be used for commercial purposes.

So if you start with the Free version, then decide to
sell some boards, you should upgrade to Eagle Light.

I found a nice Eagle Viewer for Android for under
$2, with a free demo version that seems to show a
good, high resolution, zoom-in rendering of the board
layout and the schematic files.

 

[turbo2ltr]

EAGLE Standard
(Limitations: 99 schematic sheets, 6 signal layers, 160x100mm routing area)
Layout+Schematic $575

Considering a single seat of orCAD or PADs will cost you $5-15k, it's a bargain!

I have the professional, so I didn't realize the cheaper ones have board size limitations. 160x100 isn't exactly a huge board...

 

[turbo2ltr]

Anyone interested in a group buy on the display? Shipping is a bit pricey if you just get a couple. ($25 for up to 5kg)

I'm placing an order either way. If we get a few people we also get slightly cheaper displays.

 

[TickTock]

No boards yet. :-( but I did get to work on the s/w some. Thanks to Greg's unearthing of the cellpair readout, I was able to code up this display which shows all the cellpair voltages, computes the CVLI judgement value, and indicates the max and min cellpair as well as any that fall below the judgement value. Looks like all of mine are within tolerances.

 

[garygid]

Really splendid work: all the cell-pair voltages, and the 4 temperatures.

What is the formula for CVLI?

I was thinking (but probably just do not remember correctly) :

CVLI = AVE - 1.5 (MAX - AVE)

You seem to be using 2.5 instead of 1.5, right?

 

[garygid]

The most frequent value seems to be 3724.

A histogram (frequency of occurrence of values)
as a vertical bar graph would be really nice.

And, looking at two sets of voltages, one from low SOC,
and the other from a high SOC, would really be good.
Take each CP's high-soc value minus its low-soc value = v-delta.
Then the CPs with the larger v-delta would presumably
be the lower-capacity cells.

Cells that are at the highest voltage at high-soc data
will be the cells that cause charging to stop.

Cells that are at the lowest voltage in the low-soc data
will be the cells that limit the range.

 

[TickTock]

Good catch! You are quite right. The SM has two formulas - one using the 2.5 scalar (but applied to the total pack volts) and the other using 1.5. I got the two mixed up. With the proper formula, my inspection judgement value should have been 3693 so cell 32 should have been highlighted red. I just fixed the code, I'll look again tonight when I get back down into the proper testing range (car is fully charged right now)

Really splendid work: all the cell-pair voltages, and the 4 temperatures.

What is the formula for CVLI?

I was thinking (but probably just do not remember correctly) :

CVLI = AVE - 1.5 (MAX - AVE)

You seem to be using 2.5 instead of 1.5, right?

 

[garygid]

You just did it to see if I was paying attention, right?

I would like to see the voltages after a fill-up (and a chance
to equalize several times). The voltages are probably the
best after the car has a chance to "sit" (neither charging
nor being used or driven) for around a half-hour.

 

[garygid]

These Request and Response sequences are BEST done
when parked, for safety. Most likely Nissan has debugged
then for use while "in service", but might not have done a lot
of debugging of these "intrusive" functions while driving.

So, I intend to implement these functions so that they will
be executed once on user command, and not have them
execute every N seconds while driving.

While stationary (and filling the tires with air), it might be useful
to have the tire pressures updated once a second.

One might even want to log the CP voltages once every N seconds
during charging or equaling, but normally (except in rare circumstances)
would it be desired by "normal" drivers, to monitor the CP voltages
while driving, right?

What do you think?

 

[TickTock]

I am not doing any auto poll. While it may be nice to try that feature during charging for qualitative reasons (would make for an interesting plot - especially if we managed to catch balancing in action), I agree that for accurate measurement the battery should be at rest as you say (neither discharging nor charging). My thought was this would be a diagnostic screen available for occasional use but not the display you want to have up during normal use. Presently, when the cellpair screen is up you have to tap the "send request" button to send a single request after which the transmitter is put back into snoop mode. While driving, I plan to have the tx_enable jumpers pulled as a secondary precaution.

These Request and Response sequences are BEST done
when parked, for safety. Most likely Nissan has debugged
then for use while "in service", but might not have done a lot
of debugging of these "intrusive" functions while driving.

So, I intend to implement these functions so that they will
be executed once on user command, and not have them
execute every N seconds while driving.

While stationary (and filling the tires with air), it might be useful
to have the tire pressures updated once a second.

One might even want to log the CP voltages once every N seconds
during charging or equaling, but normally (except in rare circumstances)
would it be desired by "normal" drivers, to monitor the CP voltages
while driving, right?

What do you think?

 

[garygid]

Sounds good.

With my electric motorcycle, I instrumented it to be able to
read each cell voltage, even while driving or charging, log
the data onto a netbook (small PC), and display a time-scrolling
graph in real time.

It has only 21 cells, instead of 96, but seeing the cell voltages rise
(and fall) during charging (after driving), was very educational.

As CAN-Do logs the Requests and Responses for cell voltages,
I hope to strip out the CP voltages, and either display the data
in graphic form, or send the data to a modified version of my
cell-voltage logging/graphing program. As I recall, it can handle
48 cells now, but I could expand it to handle 96, I think.

I will put that on the project list for "after taxes".

 

[GregH]

I'm not auto-polling either.. but I have repeatedly polled group 4 (temps) while driving to catch temperature transitions.

 

[Ingineer]

Guys: Don't worry about overloading the CAN bus, as this is why there is a priority mechanism. The Diag request frames have a high value for this reason. (And the safety-critical frames are low values!)

Worry more about your electrical connections. A CAN bus is supposed to be just that; a BUS. Not a star topology. When you connect the the DLC3 connector, you are adding a tap onto the bus and creating a standing wave. While the CAN bus is designed to resist electrical disturbances like this, it's still the biggest risk. Cables should always be twisted-pair, and shielded if possible. Most of the cables I have seen used are neither shielded or twisted!

I *highly* recommend you install CAN noise chokes close to your transceiver, and it wouldn't hurt to install one in the DLC3 connector if possible. If you open any Nissan device on the CAN bus you will find these chokes. (Every device has them!)

A easily possible scenario is: you add your "permanent" non-shielded non-twisted-pair bridge tap to the CAN bus, then one day you get a cell phone call and your LEAF shuts down on the highway inexplicably!

-Phil

 

[Ingineer]

Also, if you notice, most professional scantools have either only very short cables from the DLC3 or no cables at all (dongle style). This preserves the bus topology.

(Hopefully I'm not the "Canary" in the coal mine here!

-Phil

 

[TickTock]

Thanks for the tip! Timely - I will definitely get chokes incorporated in the next PCB rev. FWIW, I am using CAT5 cables for my connection using the diff pairs for the signalling. I, too, noticed that the pre-fab OBD connectors did not maintain the diff-pair integrity so I made my own low profile connector with CAT5 (encased in epoxy after taking this pic).

 

[TickTock]

Ingineer,
Did you see how many uH's the LEAF's systems were using to protect themselves?

 

[Turbo3]

Great work TickTock!

I ordered several of the displays and according to DHL tracking they just left Germany. So I should have them by next week. The mbed is also in the mail.

I am planning on building my own circuit board (hand wired for now). Your circuit board hole centers appear to be 4” by 2.5” and the board 4 3/16” (107mm?) by 2 11/16 (69mm?). Holes are 3/32” from the board edges. Is that about right?

I will use the WattsLeft cable I make up as it has both the EV and CAR CAN buses in it. It is 8 pins and the first four match yours. The next two are for my rotary control (+V and pot center) so I could use an A2D input to determine position and control functions. The last two are the always active +12v and ground. The cable is CAT-5 (so twisted on CAN buses) like you are using.

I have found that you can bring the CAT-5 up the center console without any need to modify the plastic edge so the console fits back in flush even with the cable. I use stranded CAT-5 not solid. That is very important.

For the power supply I will use the DC to DC switcher from WattsLeft to keep idle power down. I assume the mbed can be put into sleep mode and wake on CAN activity.

Also, I prefer using Bluetooth for data logging as it avoids cables and most devices support Bluetooth serial. Is is just a 4 pin module 1.4" x 0.63".

How would I go about getting the case made? Will there be a group buy? Put me in for one if there is.

Thanks again for all your hard work on this project.

Jim