Honda Insight Forum banner

21 - 40 of 69 Posts

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #21
2) If you did cut the case internal webbing out would the three packs go in rotated 90 degrees.

So side to side rather than front to back?
Peter, I don't believe there's any other orientation where these packs will fit. That they fit without modification is amazing... but no, even with the webbing removed there's just no other way they'd all fit in there nicely.
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #22
Happy to adapt my BMS measurement cards (lt6804) if you guys want to use them for this. probably saves you guys a little work. I would recommend having separate boards in at least 24s increments. safer for the end user to limit the max shock you can get from touching a board.
Tell us more about your boards. Do they have an isoSPI header?
One major concern with separate boards is blowing out the LTC6804 when the pack is disconnected... also, with 18S packs, there's no "clean" way to split the remaining six cell taps over to a separate PCB.
Also, we're gonna end up needing to add our own circuitry to either fool the existing BCM, or replace it entirely.
One final thought is that designing and laying out this PCB isn't going to take more than 40 hours of work, and in the end we'll have a single-board solution, instead of a rats nest of different PCBs cobbled together.
I'd definitely like to look at your schematic/layout if you have it on github or something.

Also, thanks for commenting!
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #23
After reviewing the BCM's role in the IMA system (it's been a few years), I've decided to just remove the BCM entirely... that will be easier. I'll probably just place a second microcontroller on the PCB specifically for this purpose. Spoofing the BCM is entirely trivial:
-We need to read METSCI data and then send it over BATTSCI to MCM. I believe Peter and I have previously reverse engineered this data stream entirely.
-We need to measure bipolar battery current sensor (IPWR+/-). No problem there, just an analog front end to center the output between 0 & 5 volts (i.e. 2.5 volts is 0A).
-Probably add a battery temperature sensor or two (e.g. intake and exhaust).
-we'll need entire stack voltage (from other microcontroller).

That seems pretty simple, and now we don't need to kludge our way around BCM. Also, I believe the ONLY other required change becomes dropping the voltage slightly on the connector that goes to the MCM. Everything else goes away. Please correct me if I'm wrong... I haven't actually reviewed the schematic, but I know it pretty well in my head.
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #24
I reached out to @black-cat & @Puggie over in their BimSight thread:

After deciding to remove the BCM entirely, our two projects are essentially identical... the main difference is that these Honda packs don't have an integral BMS... since it looks like the BimSight project is already functional, that would mean our custom PCB could just reuse their BCM replacement firmware. We'd only need to replicate the CAN bus architecture on our BMS, which is pretty awesome.
 

·
Super Moderator
Joined
·
8,774 Posts
This is beginning to look like a real drop-in, without much owner difficulty. I'm sure there is a huge market, since there are no complex fabrication issues. Someone could take this on as a business sideline. With slight tinkering of the operating voltage range, looks like it will deliver about same energy as the 3.9 effective output of the OEM NiMH, be more reliable, and have a backup source in salvage 3rd gen Insights.

Like Natalia, I don't expect to follow suit, being pretty commited to LTO, but the possibilities excite me:D
 

·
Registered
insight 2006
Joined
·
16 Posts
I am new on the forum, I read a lot. Maybe I miss something. On recent CRZ and other Honda car there is lithium battery at 144V.
Can we use an original Honda BMS made for lithium and make it communicate with G1 insight with just a protocol adapter board ?
 

·
Registered
Joined
·
921 Posts
I'm going to bridge a couple of threads and suggest that perhaps the LTC6804s should be on a separate board near the battery communicating back to the controller via CAN bus.

The CR-Z/HCH3 blocks with their attached LTC6803s and Renesas R8Cs already output CAN voltage messages AFAIK. If someone has a running CR-Z or HCH3 Lithium car, and can publish a capture of the CAN bus messages between the battery and its controller from an overnight cold start through a drive and shutdown, then we may be able to find balancing messages.

If this is the case, and that protocol is reused for the G3 Insight packs, then one is not restricted to the G3 Insight packs (which might not last long at the current price/condition) but would also be able to use the CR-Z/HCH3 packs as a direct plug-in replacement, and maybe even the Fit EV LTO packs opening widely the range of packs available.

The firmware for the board for the G3 Insight blocks that would have to be developed can then be open-sourced and reviewed by the community and tweaked by people who want to do things like have their version of that board do temperature sensing or have a safety interlock or crush detection and report it via new CAN messages, rather than run more wires from the block.
 

·
Administrator
Joined
·
12,616 Posts
No one yet ASFAIK has sniffed the CAN balancing messages between the Honda Lithium BMS boards and the BCM controller due to lack of a suitable vehicle/pack time/resources/$$$.

I am looking for a CR-Z Lithium pack in the UK, but who knows when one might pop up and if I can afford it when it does.

The CR-Z/CIVIC Lithium BMS boards support ten cells each.
You are limited to the stock 4 boards on a CAN bus as the fixed ID's cannot be changed.
So 40 cells max unless you have more than one CAN BUS.
 

·
Registered
Joined
·
921 Posts
You are limited to the stock 4 boards on a CAN bus as the fixed ID's cannot be changed.
With enough effort the CAN IDs can be changed. However it would be easier to add a second CAN bus interface. For the STM32F103 family for instance you could use the built-in CAN interface and add a second interface using the MCP2515 via SPI. Of course the latter can be left unpopulated for someone with no intention of using the CR-Z/HCH3 packs, but having this as an option for the future would probably be desirable.

Again the long pole here is not the hardware, but learning the balancing commands from the CAN protocol. Given that there are a bunch of these cars out there, this is probably just a matter of time.

Incidentally, there are eight DC1894B boards on Ebay right now at $100 each / $95 for two or make offer.
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #32
What about the LTC6813 to handle the 18s packs?
On paper, the LTC6813 is a great solution for 18S modules. However:
-It is completely backordered for at least 12 weeks everywhere I look. Let me know if you can find it.

-The LQFP-64 footprint is a chore on my pick and place machine... Pretty much guaranteed rework on every IC. We just got a new machine with a better vision system, but it's still an entry level unit. We use an LQFP-48 component on a production board; at this point my techs don't even populate it before running through the oven... they prefer placing it by hand. Note: I'm aware the LTC6804's leads are the same pitch; the difference is how the solder wicks the pads down with leads on only two sides (6804), versus four sides (6813).

-I already have a partial reel of LTC6804 chips ($6400) "leftover" from the Linsight project.

-18S support is only useful if we support a single battery configuration (e.g. 48S = 18S + 18S + 12S). Our present plan is to support both 48S & 42S ( = 18S + 12S + 12S). I guess maybe we could have QTY2 LTC6813 modules (for bays 2 & 3), and also QTY2 LTC6804 modules (for bays 1 & 2). That certainly solves the

Interestingly, the LTC6813 is substantially cheaper per channel ($0.70/channel) than the LTC6804 ($1.79/channel). Of course, this is moot (to me) because I already have a reel of LTC6804s.

...

None of the above shortcomings preclude us from using it. Let me think about it some more. Thanks for posting!
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #33
I'm going to bridge a couple of threads and suggest that perhaps the LTC6804s should be on a separate board near the battery communicating back to the controller via CAN bus.
Sean, this is a great idea. I agree we should separate the BMS proper (i.e. LTC6804s) from whatever else we dream up. We can place a simple serial interface between the two. In fact, I'll probably just throw headers on it that accept an Arduino Uno for the controller. This will make the PCB much more useful (even on other projects). Also, it means I can start that PCB today... while we work out everything else. The Uno's USB port is an added plus... also, I have about QTY1000 Arduino Unos left over from years of manufacturing; I'm not ashamed to say I still use the Uno in shipping products (running highly optimized C). Our main shipping Uno product has 2 bytes free memory ;).
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #34
If anybody wants to help out right now:
I still haven't found the connector. The pitch is quite non-standard:
12S connector:
-10 columns, 2 rows
-column C2C: 2.2 mm
-row C2C: 3 mm

Note that I haven't even found a non-shrouded male header that has the above pitch... 2.2 mm is very uncommon. Right now the only way I can make a mating connector is to use QTY10 2-pin headers with 3 mm pitch. Annoying, but it is a solution.

So if anyone can find the 12S connector, that would be a better solution. If we can find the 12S connector, then we can move onto the 18S connector, which is even more non-standard. Here are pictures of both connectors:

 

·
Registered
Joined
·
921 Posts
Does Honda balance lithium cells with a variable load - more load for cells with greater imbalance - and do they make the balancing decision prior to a cold engine start?

I was looking at pictures of the HCH3 board and noticed that it uses two LTC6803s to balance 10 cells. Reading the datasheet for the 6804 and the demo circuit it looks like the 6804 is capable only of turning the FET on off and can't vary the amount using PWM or so forth. It looks like there are at least three layers on the board so I did not attempt to guess at the routing. But there are two sets of what are likely to be FETs - one large near the top on one side and another set of smaller devices on the other side. Nope - not a great photo, the small FETs appear to be zeners according to the silkscreen, and there are 20 of them. This tells me that maybe Honda is able to apply up to three levels of balancing load (four if you count "no load").

One reason for this might be that the only time we can get an accurate indication of imbalance is when the pack is quiescent, namely, after the cells have stopped recovering from last use and the temperature is even across the pack. The only time this would occur in the car is when the key is turned in the ignition before starting a cold engine.

The controller may very well decide at that moment to measure the resting voltage. Rather than come up with some duty cycle algorithm it could simply apply a load proportional to the amount of balancing each cell needs. Then it could simply let it go for a certain amount of time and stop it after that time has passed. If the car is turned off early, at least progress has been made evenly in the right direction.

Subsequent cell monitoring by the controller might look for gross imbalance due to some failure, but temperature variations within the pack, recovery rate differences, or whatever, probably might make cell voltages an unreliable predictor of imbalance, except just before engine start after a cold soak.

It might be worth the engineering effort to understand the HCH/CR-Z behavior in case it provides some valuable insight that might influence the balancing module design.
 

·
Registered
Joined
·
700 Posts
Looks like a double row version of an AT Computer PSU

Would need 4x of the things and a saws all to mod two of them
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #38
Does Honda balance lithium cells with a variable load - more load for cells with greater imbalance - and do they make the balancing decision prior to a cold engine start?
Many controllers allow PWM loading. This is only necessary to prevent overheating the PCB proper... with enough ground plane to spread the heat there's no reason to PWM the passive balancing load. Our PCB will have plenty of room for cooling, plus forced air cooling from the OEM fan, plus onboard temperature sensors on each bank (to turn off the balancing if the board gets too hot).

I was looking at pictures of the HCH3 board and noticed that it uses two LTC6803s to balance 10 cells. Reading the datasheet for the 6804 and the demo circuit it looks like the 6804 is capable only of turning the FET on off and can't vary the amount using PWM or so forth. It looks like there are at least three layers on the board so I did not attempt to guess at the routing. But there are two sets of what are likely to be FETs - one large near the top on one side and another set of smaller devices on the other side. This tells me that maybe Honda is able to apply up to three levels of balancing load (four if you count "no load").
Most BMS boards are going to have at minimum four layer (three layers is hard to manufacture). Most commercial BMS systems have redundant systems, so that one system can take over if the other fails. In most "real" BMS' I've looked at, one IC can balance the cells (and read voltage), whereas the other has very high series resistance, such that it can only measure voltage. This prevents IC damage (to the voltage-only IC) if the pack opens somewhere.

One reason for this might be that the only time we can get an accurate indication of imbalance is when the pack is quiescent, namely, after the cells have stopped recovering from last use and the temperature is even across the pack. The only time this would occur in the car is when the key is turned in the ignition before starting a cold engine.
Yes, balancing typically does not happen while the cells are under load. If balancing does occur when the car is on, it's only using data from the last time the car was off.

The controller may very well decide at that moment to measure the resting voltage. Rather than come up with some duty cycle algorithm it could simply apply a load proportional to the amount of balancing each cell needs. Then it could simply let it go for a certain amount of time and stop it after that time has passed. If the car is turned off early, at least progress has been made evenly in the right direction.
I don't intend to use PWM, but the theory is otherwise correct.

Subsequent cell monitoring by the controller might look for gross imbalance due to some failure, but temperature variations within the pack, recovery rate differences, or whatever, probably might make cell voltages an unreliable predictor of imbalance, except just before engine start after a cold soak.
There shouldn't be appreciable temperature variation in the pack (at least if all cells are good). The ESR on these cells is more than an order of magnitude lower than the NiMH it's replacing. If a cell goes bad, we don't need to measure temperature to see it... our per-cell BMS will see the voltage drop (assist) or rise (regen). Temperature difference of even tens of degF won't appreciably affect cell voltage.

It might be worth the engineering effort to understand the HCH/CR-Z behavior in case it provides some valuable insight that might influence the balancing module design.
I always advocate studying existing designs. I've got a pretty firm grasp on what's required, but certainly if anyone has any input (like you're providing here) I'm all ears. I don't plan to have redundant voltage measurements, which is a "risk" per se, but instead I'll just disable the system entirely if anything comes back wrong (cell voltage wise).
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #39
Looks like a double row version of an AT Computer PSU

Would need 4x of the things and a saws all to mod two of them
The hard part is finding a connector with the correct pitch... 2.2 mm is very non-standard. The only connectors I've found with that pitch have an 180 degree offset between rows (i.e. the columns aren't lined up perpendicular to the rows).
 
21 - 40 of 69 Posts
Top