Honda Insight Forum banner

Jim's LTO HCH1 Conversion

5.1K views 24 replies 5 participants last post by  HCH03MT  
#1 ·
Per Peter's request I'm starting a new thread. Originally I posted this at:
OBDIIC&C LTO BMS CAN INTERFACE

Having retired and moved from Chicago to St Pete, Florida 21 months ago, my ‘03 HCH1 MT is definitely needing to retire its original pack. I’ve not frequented Insight Central for about 6 years, but grid charging for years I have reached diminishing returns. I came back to see what others were doing with their vehicles. I discovered the FIT LTO and quickly concluded a retrofit would allow me to happily keep my HCH1. I’ve done a lot of reading here to get reasonably up to speed and have recently purchased and received 3 packs of Fit LTOs. I’ll probably do 72s, but might play with 60s.

It may be a while before I can find time to retrofit my HCH1 (It seems I'm busier in retirement than when I was working full time!), but I want to bench test the packs under heavy load to identify any bad cells before the Greentec 3 month warranty ends. Quick unloaded voltage readings of half packs show I have one bad half pack (~0V).

Rather than tear them open before I do any warranty returns, I’d like to use the OBDIIC&C to read all cells through the BMS boards.

So cutting to the chase, I built my own OBDIIC&C on a plug in breadboard back in 2013 posted here:
OBDIIC&C HCH1 Gen 1 Civic Hybrid

pic of breadboard that Peter might remember:
hi res OBDIIC&C breadboard.jpg

I have a new PIC 18F2680 to replace the 16F886 and an MCP2551 CANBUS board will arrive next week.
So I’d like to purchase your latest firmware to read the LTO cells. I also will contact minor4326 to see if he has any harness connectors left. I read that you abandoned HCH1 development for the OBDIIC&C, but that is a separate matter that hopefully we can discuss at some point.
Thanks,
Jim

Peter also recommended several threads I need to digest:

I assume you read my HCH1 Lithium conversion thread from years ago?
There are also several other relevant threads.

www.insightcentral.net

HCH1 A123 60ah Pack Install
Started on this now. ;) Couple of videos showing modification of the Civic switch board to accomodate the MCM fooler resistor matrix, 3kw mains charger and four battery connection points. HCH1 Honda Civic A123 60ah Pack Install Video 1 - YouTube HCH1 Honda Civic A123 60ah Pack Install Video...
www.insightcentral.net
www.insightcentral.net

www.insightcentral.net

HCH1 Inside the IPU and more power.
Having been delving around this a bit recently and studying the manual I think we might have another option for more power with a Lithium upgrade. The HCH1 IMA current as we know can be increased tweaking a number of current sensors for an 18.5% increase. Matt says this works well on his car...
www.insightcentral.net
www.insightcentral.net
 
#2 ·
The Insight G1 OBDIIC&C CAN LTO firmware is $25 per unit.
Of course this does not work on the HCH1. :(

We have a few options.

1) Merge the HCH1 OBDIIC&C and CAN code so we have an HCH1 OBDIIC&C CAN.
This will take time effort and $$.
I also don't have an HCH1 and haven't looked at the code for 5 years so testing will be problematic.

2) Use an OBDIIC&C CAN to just talk with the LTO 60/72 cell blocks and get the voltages and have alarms/ warnings/charger control etc . This will also take some time and $ but is more doable.

3) If you just want a standalone OBDIIC&C CAN to talk with a single LTO block (12 cells) at the time to read the ID and voltages then the attached will do it.

It also squirts out the cell voltages in a serial 9600,8,N,1 datastream to a plx-daq excel spreadsheet once a second.

(Only one block must be plugged in at a time)

Rename it to .HEX it's not a pdf!
 

Attachments

#3 ·
Thanks for the quick response. Number 1 would be ideal in an ideal world and I'm tempted to go for it down the road. Number 2 would be a reasonable shorter term goal, but even that isn't likely to fit my calendar commitments for some time. Our "new" (1956 mid-century modern) retirement home has taken up most of my time since moving here. I designed and built the kitchen and have plans to do a major addition. Also I still have some off and on immediate commitments to my previous colleagues in Chicago.

For the moment Number 3 fills the bill and will at least protect my investment in the LTO packs (warranty wise).

I'll keep you up to date with results.

Thanks, again.
 
#5 ·
You can't get "loaded" measurements, but you can simply pop the covers on each half-pack and manually measure each individual cell voltage with a good voltmeter.

CRITICAL. If you remove the covers, be absolutely certain to remove the small "pucks" at each end terminal BEFORE you remove the plastic cover. Otherwise there is considerable risk of the puck moving around and shorting to a fusable link, which will destroy that link.

I did a bunch of loaded testing on all my packs using the Revolectric.com PL8 charger. One charger will test up to 40A discharge rate and two chargers used together will test to 80A. It is complicated and expensive, but it will get you to the charge/discharge data and it also does cell balancing. A bit time consuming to hook up since it can only do 8 cell at a time.

Discussion thread here:

 
#6 ·
I've finally got time to get back to testing my LTO blocks and have successfully pulled no load CANBUS data from one block (or brick, do we have a preference what to call them yet?) using Peter's above attached firmware as well as opening 2 blocks and directly measuring each cell with a lab grade DMM (Keithley 196). The Keithley has a 6 1/2 digit display giving 1uV resolution on the 3V range. It was last calibrated in 2000, so I can't vouch for absolute accuracy other than it's good enough for now. I just set up a picaxe so was able to download to excel.

I noted a couple of things worth mentioning.

1) The first block ($0206) DMM measurements were unbelievably well matched. All except 2 were within +/- 0.1mV of 12 cell average, and the other two were at +/- 0.2mV. Simply astounding. I haven't done CANBUS download on this one yet.

2) The second block ($0202) was problematic in that the overall voltage before opening it was at ~2V and very noisy which had me quite concerned. I opened it up and found visible corrosion on the aluminum spacers at the end terminals. The end terminals were only finger tight as received so I suspect Greentec's actions allowed this to occur. I will post photos but not sure if this should go in another thread (Peter - what do you recommend?) This will require gentle removal of the corrosion and use of a noalox type anti-corrosion compound.

DMM data was not as well matched, but still reasonable (worst case +1.06 / -1.24mV re. 12 cell average).

3) After getting the OBDIIC&C CANBUS up and running I tested the second block and found the data to be quite noisy and all over the map compared to DMM data. I then got the picaxe running and downloaded data into excel (242 data sets). I did histograms for each cell and found what looks like a noise problem in the data.
Image

Since I haven’t run any other blocks yet I don’t know if this is particular to this block or a typical occurrence or a noise artifact in my OBDIIC$C setup. I did an average of all 242 data sets and compared to the DMM data as follows:
87603


Main takeaway is an average offset of 1.8 mV between CANBUS and DMM data. Not a big deal, But note cell 12 is an outlier, 2mV above CANBUS average while it was -0.9mV re. DMM average, pointing to individual cell errors in the CMU board. It will be interesting to see how the other blocks come out. This is a lot of work, and may not result in anything useful, but after seeing the super matching of block ($0206) and the obvious noise in CANBUS data I wanted to reconcile that if possible. My main goal for this exercise is to get an accurate measurement of cell impedance by comparing open circuit voltage to a loaded voltage (maybe as high as 100A). So to do this I want as noise free data as I can get. The average of 200 or so data sets gets rid of most of the noise so that may be my route. The fixed error sources in the CMU board should cancel out of this equation.

I notice that my pasted excel data is barely readable (uploaded as a .png), so I may have to upload as tables in the future.
 
#7 ·
Just a quick heads up. I will be posting a lot tomorrow in this thread as I've finished testing my packs (at least for now).

I have done lots of reading here and I have a few questions that I hope to resolve by consensus from those who've been working with the pack longer than I have. I also had a nice discussion with Art at Greentec a few weeks ago and have received a warranty replacement for one of my packs as well as bought a spare pack (one of the last) as he warned me they were about to sell out.

So first topic. I have a dented cell on the bottom (pics tomorrow) and it doesn't affect the measurements I've done. Jime said he'd like to do an autopsy on a cell he punctured,


but so far I don't think he or anyone else has torn a cell apart. The dent is shallow, ~1mm and I'm hoping there is some internal space or compressible protection between the cell guts and the interior of the aluminum cell case.

Lots more tomorrow...
 
#8 · (Edited)
Here is the worst dent (marked as 1/8” deep on pack $020C below) and is the only one to have punctured the translucent bottom cover and obviously telegraphed all the way into the aluminum case of cell #9 about 1mm deep.
88305

88306


I don't know what the grey / beige located in the honeycomb holes is but it is hard and well stuck and looks similar to the glue used to hold the pack together. Just doen't make sense where it is.
88309


Profile of dent.
88310


In addition, $020C has a strange blue-green deposit (below) next to and under the negative busbar of cell#1, and bridging over to the adjacent BMS tap screw from the positive end of cell #2, possibly providing a leakage path between those two cells.
88311

Those two issues on the same block caused my warranty request to Greentec. The balance performance of this block was the second worst of all blocks, -6 / +4.9 mV re. the average of all cells in that block, prior to heavy load discharge testing, but probably still acceptable. I’ll show summary balance graphs of all blocks later (prior and post load testing).

I peeled back the honeycomb foam (some say thermal pad?) to check all my blocks.

Here is the next worse looking bottom cover. $0206 (see upper left arrow in below pic).
88312


The two photos below show a barely visible very slight dent into the aluminum case of cell #2 near the edge adjacent to cell #3.
88313
88314


It's hard to quantify without a surface profiler, so I guess I need to ignore that to keep my OCD in check. Besides, this is also one of the best balanced blocks along with paired $0205, so my OCD hates to give that pair up!

Cover and pad took a beating ($0201):
88315

but cell looks OK underneath (not shown).

One of the two final packs recently received (duplicate ID-#2_$020B – no pic) has a shallow dent that falls into the OCD range and also has quite good matching.

These dents were obviously caused in handling and probably due to stacking the blocks on top of each other many deep without sufficient protection. It is what it is now, but I’m curious if others ran across these and lost any sleep over them.

To be continued shortly…
 
#12 ·
The post was flagged by the "akismet" anti-span system which V-Bulletin apparently has running on the site. I approved the post, as I have done once before, since they were of obvious non-spam content.

I will raise the matter with the site admin folks at HQ. I have no tools to troubleshoot the problem otherwise. I will also make a note to other moderators via the moderator channel to approve your posts for the time being. Hopefully non will be deleted.

(BTW, I inadvertently "likeed" the post - not that I dislike it ;))
 
#11 ·
No idea why it sometimes selects posts for moderation. Anyway it's approved now.

IMO The dent is a complete non issue.

However the funny green deposit need investigating.

Carefully remove the BMS board from that block and investigate, remove and or neutralise the corrosion.
 
#13 ·
@jime
Thanks Jime,
By the way, did you ever autopsy your punctured cell? Peter's comment about the worst dent does calm me, but I'm still curious about the innards and interior spacing. I was somewhat expecting that every available cubic mm would be stuffed with active structure to maximize capacity.
 
#14 · (Edited)
On to test methodology and results. I’ll post the test system pics last.

Steps in the process for each block:

1) Measure each cell voltage as received before any discharge with Keithley 6 ½ digit DMM. On 3 volt range I can get precise 0.1mv resolution with 0.01mv noisy resolution. Lowest digit is discarded. This becomes the reference for all other measurements

2) Collect one minute (~ 1 per second) of unloaded BMS cell data using PICAXE into EXCEL and average the results over time. This reduces the noise in the BMS measurements and would allow getting better than 1mv resolution if the BMS errors were truly random (but they aren’t random based on what I see).

3) Brief sequential test application (a few seconds) to verify each of 4 available 0.96 ohm (forced air cooled) nichrome wire loads giving design goal of 30 amp at 28.8 volts, manually switched by circuit breakers. Follow by several minute no-load rest. Current is monitored by reading the voltage (with the Keithley) across a 1 milliohm nichrome current sense resistor.

4) Apply ~1 minute load in increasing increments; 30, 60, 90, 120 amp to monitor for misbehavior in case there’s a bad cell (to minimize any pyrotechnic surprise), with no-load rests in between each for voltage to recover to ~ steady state. Each block was tested starting from the as received charge condition. Discharge was stopped when ~2.0 V/cell was reached. For blocks starting closer to full charge, I simply repeated the ~120 amp (and falling) pulse until 2.0V was reached.

5) Repeat manual DMM measurements of each cell after final recovery to ~ steady state. For the last four blocks this was about a 15 minute recovery. For the first six blocks, I didn’t think to do the post DMM at the time of other measurements. About two weeks later I did the post DMM and repeated the post no-load BMS for those blocks.

Results comments:

The following charts were generated in EXCEL and nicely show the immediate voltage change when load is applied or disconnected from which we can get the DC component of internal resistance. We also see the time dependent sag or recovery. I plotted min, max and average values. The X axis was the test number, roughly the same as time except for removal of some portion of the rest periods. I also plotted the max-min difference on a secondary vertical axis which shows the noise in the BMS measurements. It’s interesting that during high current load, the max-min noise increases significantly and drops back immediately when the load is disconnected. Also, there seems to be some drift in max-min over time, and some blocks are worse than others.

Overall, all blocks look to have roughly equivalent internal impedance relatively independent of state of charge (in the range tested) so no bad news here!

From the DMM data, I also plotted the max-min for each cell in each block normalized to the average max-min of each block, with cell # for the X-axis. I did this twice, using prior load and post load data. This is an easy way to spot outlier cells and it pointed out one cell, #4 in one block #2_$0202. That cell is at -17mV re the average cell in that block for prior load. Notably, for blocks with larger max-min data, prior measurements were larger than post measurements.

This might be explained if the BMS did balancing near the bottom end of the state of charge cycle (bottom balancing).

What this doesn’t answer is how that cell got that far off, and it makes me question whether that cell will rapidly get out of balance.

I made separate plots of the difference between BMS no-load data and the DMM measurements as reference, both prior and post load tests. This shows an obvious error “tilt” with larger positive errors for cells near the positive end of the block. Overall worst case across cells was -5mV to +9mv. This suggests that we can’t use the BMS CANBUS data for any kind of precision, but it also suggests maybe we don’t need to. The FIT EV worked, but it also had working balancing. So +/- 3mV matching criteria can only be had with accurate DMM measurement. At some point (down the road) I’ll try to analyze the BMS to see if I can explain the error and maybe reduce it if it's an analog error before the A/D.

Your thoughts on any / all of this are greatly appreciated.

I’ll post the charts starting in the next post. I feel a nap coming on. I will spread them over several to avoid the forum robo-cop.
 
#16 · (Edited)
@HCH03MT. Awesome. A real baseline for the LTO blocks. Can't wait for the data.

Keithley 6 ½ digit DMM
Nice.

I am VERY interested to see how your LTO cells respond when the different loads are applied. I hope that you are logging the CAN voltage reports at full data rate, even if it is a lot of data. Thanks for doing these tests.

Are you using Arduino/Pi for any of this?
 
#17 ·
Everyone - thanks for the complements:)

@retepsnikrep I will remove the BMS that is hiding the green stuff but probably not right away.

@*sean* No Arduino or Pi yet. I'm an old analog guy with some DSP for audio. The data rate is set by Peter's OBDIIC&C which is close to one second for one full block of 12 cells. I say close, because it occasionally skips a data point per the PICAXE time stamp.

Finally the EXCEL charts:
Nit-picky disclaimer: The test methods evolved over time and are presented below in the order done. I wish the timing was more consistent but I didn't go back and re-do anything. There are few glitches due to being distracted and or missing time alarms, e.g. see #2_$0202.

I'll post the Cell Balance charts later today.



$0201-Good
88331


$0202-Good
88332


$0205-Good
88333


$0206-Very shallow dent on cell #2 - but otherwise Good
88334


$020B-Note Larger than typical max-min offset but probably Good
also has largest BMS Board error.
88335


$020C-Dented cell #9 and Blue-Green deposit that may provide a leakage path around cell #s 1 and 2.
88336


#2_$0201-Note larger than Typical max-min offset
88337


#2_$0202-Large Drifting max-min mostly due to Cell #4
88338


#2_$020B-Very shallow dent on cell #9 but otherwise Good
88339


#2_$020C-Good
88340
 
#23 · (Edited)
I found the cell in my garage trash. I will be doing an autopsy and providing some pictures under a separate new thread. But for now, The active part of the battery is spaced from the aluminum case by a piece of plastic along the ends of the cells and for about 20mm along the bottom. This leaves the main "roll" of active battery elements spaced about .025" away from the bottom of each cell. Dents raise different kinds of issues, depending on where they occur and how deep they are.

Just an additional comment, IMO that one cell with the greenish deposit concerns me. I think it likely that that cell has ruptured the one time vent on the cell top. The vent is under a layer of plastic which runs down the center of the brick, so it isn't possible to view it directly, but I can't think of another explanation. Get close and give that cell a sniff. The electrolyte has a kinda sickening sweet smell, as I and others have described it. I'm sure the smell isn't safe for long exposure, but it hasn't hurt anyone who has smelled it - as yet.:(
 
#19 ·
This is right up Sean's street. Wow.. :eek:
No offence intended or hopefully taken we all have our areas of expertise..
That's what makes this forum interesting.. ;)
 
#20 · (Edited)
Here are the Cell Balance charts:
88357


Removing the three largest outlier block to provide a zoomed in view of the rest.
88358


Interesting that there's this much error in the BMS, and note the tilt from low to high cell number.
88359


Separating the prior and post measurements for a clearer view
88360


I would have expected the errors to not be dependent on prior vs post particularly because these are the result of averaging a lot of measurements for each.
88361


I would like to investigate what causes these errors but probably won't have time for a number of months. I suspect there could be digital noise on the BMS board that gets picked up by the analog lines on the board that feed the A/D converter. I haven't pulled a board yet but also suspect that there is an isolation network on the plate that holds the cell terminal screws. The purpose of such a network would be for current limiting if something (someone like us playing with it) shorted the top of the BMS board where all the cell taps enter the board in a very small area. Otherwise, major fireworks and damage could occur.

 
#21 · (Edited)
OK, something strange happened with the images in the last post. Clicking on the images for a full screen view puts a black band around the plot area that hides all the titles, axes and notes, rendering the full screen view useless.
Found the fix. The chart border area was transparent instead of white. Full screen view background is black. Makes sense now.o_O
 
#22 ·
The worst case imbalance is really quite good.

What is the small jump up between tests? Did you stop the measurements and wait for the pack to reach a steady state voltage?

What are the units on the x axis? Are they seconds? I would like to estimate how much capacity was removed during each test by measuring the width of each pulse. If you can provide the actual resistance for the different current loads that would be cool.

$0202 shows one cell down 25 mV from the rest at the start of the test, but nearly even with the others at the end of the test. If you charge this back up to 2.5V/cell (charging all in series) does that cell return to being 25 mV low? How long had $0202 been sitting at 2.5V before you did this test? Very curious about this as it might say something about how individual cells behave.

Also, IIRC the voltage read from the BMS is good to a few millivolts. I can't remember how noisy it is.
 
#24 ·
Thanks for the latest.
@*sean* Answers inserted inline in red:
The worst case imbalance is really quite good.

What is the small jump up between tests? Did you stop the measurements and wait for the pack to reach a steady state voltage? Yes indeed. I wasn't always consistent, but for the higher current bursts, I gave ~5 minute recovery of which I only captured the first minute. I then restarted capture for another minute before starting the next current burst.

What are the units on the x axis? Are they seconds? They are related to the row number in EXCEL of the data returned by the PICAXE, the first cell in each row is a timestamp and is nominally in 1 second increments. There are discontinuities in the timestamp since I stop/ started as mentioned above, as well as the occasional burp where the timestamp jumps 2 seconds. I would like to estimate how much capacity was removed during each test by measuring the width of each pulse. If you can provide the actual resistance for the different current loads that would be cool. The design value was 0.96 ohms per step. I simply parallel additional equal load resistances to achieve the nominal 30 amp per 0.96 ohm, at the nominal 28.8V I assumed.

I thought about doing what you proposed but simply didn't have the time. I'd be happy to have you do that if you want and will give you the complete .XLS file if you'd like. I also did a few series resistance measurements of the first block $0201 for the essentially "instantaneous" transient when switching the load. It was in the area of .6 to .8 milli-ohm for all current levels (columns T-U-V of tab $0202 full raw data.


$0202 (Do you mean #2$0202?- that is a different block as I have some duplicate IDs and is the one that has cell #4 at -17mV re average for the block, [with cell#1at +6mV], prior load, and -6mV post load based on the DMM measurements) shows one cell down 25 mV from the rest at the start of the test, but nearly even with the others at the end of the test. If you charge this back up to 2.5V/cell (charging all in series) does that cell return to being 25 mV low? How long had $0202 been sitting at 2.5V before you did this test? Very curious about this as it might say something about how individual cells behave. Yes I agree that is a useful test.

I haven't mentioned this but now is a good time. One of my immediate needs is to settle up with Greentec as far as return of a pack for the warranty claim I made based on test of the first three packs I bought. I thought I would be returning pair $020B / $020C. Based on the test of the next two, I think I want to split that pack and one of the latter two, returning $020C and #2_$0202. I realize this might be difficult, and that Greentec might not agree to it. In any case, I need to settle that soon, so I might not get to try your suggesting regarding cell#4 mentioned above.