Honda Insight Forum banner

1 - 20 of 271 Posts

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #1 · (Edited)
@retepsnikrep and I are considering designing a simple battery management system that supports Honda's 3rd generation Insight lithium battery packs. We've discovered that these packs mechanically fit in the G1's OEM IMA battery bay, which makes them an IDEAL candidate for simple lithium conversions. These batteries are currently for sale here at the bargain price of $400, plus ~$40 shipping (CONUS):

These batteries were initially posted by @mmdepace here:

This thread is a continuation, since it's a separate topic. Peter and I are mid-discussion at the time of this writing... I'm going to answer the questions he proposed in said thread here... eventually Peter will probably move those comments here (since our discussion is an offshoot topic).

...

Right now Peter and I are trying to determine how much existing technology we can cobble together to produce this BMS with minimal effort. At this point we HAVE NOT COMMITTED TO BUILDING ANY PRODUCT, so do not purchase batteries if a $440 investment will keep you from paying rent. Right now I'm envisioning this project being sold to verified members that know what they're doing. Eventually we could open it up to more people, but I'm not looking to hold anyone's hand right now.

Full disclosure: I previously began the Linsight project way back in 2015... to date I have not shipped it. Linsight is an incredibly complicated project from a firmware perspective, because it replaces both existing IMA computers. As such, Linsight must handle everything the OEM system handled, plus all the bells and whistles that feature crept into it (e.g. support with Pegasus, manual controllers, car charger support, etc). For the past five years I haven't been able to set aside the time needed to finish Linsight.

This BMS discussion is a much simpler project... at present we're considering leaving the OEM computers in place, and then hacking together our existing designs into a compatible product. This is maybe 5% of the engineering effort compared to Linsight.

...

So with that introduction out of the way, the following posts pick up where Peter and I left off.
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #2 · (Edited by Moderator)
[Peter, please insert posts from existing thread here]

RetepSnikrep said:
John.

Can I suggest your BMS offering supports upto the full 3 x 18S modules as well as perhaps other combinations of packs thereof. (Note my 3x12S comments in later post)

So as you have them in the case have two socket footprints on the BMS board at each space.

Both sockets don't need to be physically present as they will probably be through hole people can/could solder in the socket combination they need.

So a footprint for a 12S and 18S socket in each of the three gaps.
May need some jumpers for configuring it at each socket.

Just an idea so we are not tied into only one layout and one module arrangement/combination.

That will also make better use of the available modules in multiple sets as well.

Yes I managed to sort shipping using a firm I have used before.
Fingers crossed.. ;)

I'll have to look and see how much complexity 54S adds. There are many concerns that would require additional thought:

-Usability. The 12S and 18S pinouts are different, which would require two separate connectors per "bay" (in the OEM enclosure).

-Safety. The non-used male header (either 12S or 18S) would be energized once the used header was connected to the battery. Customers touching this header could kill themselves.

-Cost. Each LTC6804 circuit costs about $25 in parts... the IC itself is $21.40 EACH. Obviously we can depopulate the PCB, but that adds multiple build options. Adding this all up, a PCB that supports 48S is going to be ~$125 in parts, plus the cost of the PCB (~$20 in low volume).

-End use. I believe the primary benefit is for customers to have a simple lithium solution. Most of those customers probably aren't going to want to swap out their OEM DCDC converter. For those of us tinkerers that want to push the voltage, I'll probably end up recommending people manually wire a 5th LTC6804 (e.g. using LT's demo PCB, which would require manual wiring).

-PCB real estate. This PCB is quite restrained by the OEM pack's small top space (with the OEM computers installed). Particularly the folded steel piece that surrounds those computers is all kinds of in the way. Therefore, I probably won't have enough room for a 5th isolated BMS circuit.

I think I'll go into this project with these configurations in mind:
QTY-12S QTY-18S Result Vnom Vmax
3 0 36S 133 150
1 2 48S 178 200
0 2 36S 133 150
After I've roughed it out, if I can make more configurations work for free (or little effort) I'll look into that. I agree 54S would be great, but I don't believe it will make the initial roadmap.
That's actually more effort than it sounds... I plan on using the 328p, which only has two hardware serial ports. I'll bit-bang a third serial port if I have to, but I'd rather not waste the CPU time, since I plan to use that to rapidly query SPI battery voltage and make realtime decisions. 25 MHz only goes so far ;).

As much as I'd love to port my existing code over to another microcontroller, my existing code works on the 328p... technically it also works on a much larger microcontroller, too, but that's more expensive. The code I kludged together doesn't have a simple HAL that I can easily rewrite... in short, using another microcontroller is more effort than I'm willing to commit to this project.

Is there a way we can tie my PCB into your existing BCM fooler? I can even replicate your (PIC?) circuit on my PCB, and then load your code.

People will scratch their heads, thinking "why does this PCB have both Atmel and PIC uCs?"

Re the BMS PCB Random thoughts in no particular order..

1) I was thinking put your PCB/s at the end of the pack where the fan would be.
A thin? PCB the width of the pack, or three small interconnected pcb's that fit into the openings, one per block. Connectors facing inwards accepting those plugs.
Maybe use the fan mount bolts holes or studs on the end of the packs to secure it/them, whatever.....

The OEM fan could mount over it as normal and provide airflow for the bleed resistors when balancing.

(I'll do a YT vid with some thoughts.)


2) If you want to replace the BCM you have to do a lot of stuff to keep the MCM happy. :eek:

a) Measure/count current in out and have a running or faked SOC that is stored between drives.

b) Have a serial data stream on BATTSCI containing current, voltage, temperature, Soc, flags etc.

c) Accept serial data from the MCM on the METSCI bus..

d) Other stuff I have probably forgotten.


3) Yes it would be possible to add a BCM Interceptor to your physical PCB.

The current version of the BCM Interceptor PCB fits inside the MCM and has two simple 5V logic level high/low inputs to control (enable/disable assist/regen).


I think we should keep the OEM BCM and let it do it's thing. That also avoids a lot of hardware development and harness chopping modifying. We get to put the 4x OEM temperature sensors in the new packs and use the OEM current sensor, SOC counting etc all as normal.

Fake the ptc strips with a 30R resistor as we do now for most setups.

Feed the BCM taps with a simple BCM Fooler resistor matrix.
That's well proven and reliable and is easily configurable for higher voltages than OEM with a pre-resistor. (This could be on your pcb)

Use a BCM interceptor (fitted inside the MCM or on your pcb) to interface between your BMS and the car. Again that's well proven and reliable.

Your BMS cpu just has to output a 5V logic high/low on two discrete lines to enable disable assist/regen.

The cell voltages could be output via CAN or more likely serial data on the HLine and be picked up by an OBDIIC&C like the LTO etc.

Most people doing mods like this will have an OBDIIC&C plugged into the car, might as well use it ;)

You could have your CPU listen on the serial HLine and when it gets a send me the cell voltages request packet from the OBDIC&C/Android app/etc etc it just spits them out onto the HLIne/Bluetooth for the OBDIIC&C/Phone to display.

That would be fairly simple to implement.


So if we kept the OEM BCM then we would need....

1) Your simple BMS pcb with two 5V logic level outputs to enable/disable assist/regen via the onboard BCM interceptor.
A 5V HLine serial 9600 baud I/O port to exchange cell voltage data or commands with the OBDIIC&C on request.

2) A BCM tap fooler resistor matrix. (You could put that on your pcb and use your HV connections etc)
The builder/converter would simply cut off the tap harness from their orange end board and solder it your pcb.

3) A simple BCM Interceptor inside the lightly modified MCM or on your PCB.

4) An optional OBDIIC&C to display stuff from your BMS and/or send/receive commands like adjust voltage/balance setpoints etc.

5) Your BMS would do all the balancing stuff autonomously and just toggle the logic level protection request outputs when things (cell voltages) get out of hand!

Peter, I'll have to dust off my Linsight IMA schematic tomorrow and see if there's a simpler method. I'll read your reply in more detail then (it's 4AM here right now).

...

I've been trying to find more information about these batteries. Here are my findings thus far:

The best rabbit hole I've explored so far is from the serial number label that was attached inside the box... it contains the text "UF121285H". That leads to this website:

Findings from that website:
From Honda immd hybrid system, which is used in both 2018-2020 Accord Hybrid & 2019+ Insight
Chemistry: LiNiCoMnO2 (ternary lithium battery)
"Panasonic used after localization, model UF121285H". I think he means that the battery was originally designed by Blue Energy (which itself is 49% owned by Honda)... and now Panasonic is making the pack instead. Remember: the box that these modules shipped in has Panasonic written all over it, as do the various labels.
"At present, the detailed specifications of this battery of Panasonic are not available online, but because Honda directly replaces it, we have reason to believe that the specifications of the two batteries (Panasonic vs Blue Energy) are compatible."
"UF 12 12 85" is based on cell dimensions (120 x12 x 85 mm).

Cell information:
Model: EHW5
Ah: 5 (when charged to 4.2 volts)
Vnom: 3.6
Operating range: -30 degC to 55 degC (131 degF)
Cycle life: 50,000 (@10-85% SoC), continuous 40A charge/discharge!
Time Life: 10 years
Recommended charge/assist current: 200 A
Maximum charge/assist current: 300 A (60C)
Power density: 4.9 kW/kg

The reason the part number on the box doesn't pull up on Honda's website is that the pack itself is a non-user-purchasable subcomponent of the following parts:
1D070-6C2-305 (2018-2020) Accord Hybrid). OEM cost: $2762
1D070-6L2-A00 (2019 Insight Hybrid). OEM cost: $3126

So yeah, this $440 battery is ~$2800 from Honda.
A rambling video is worth 1000 words!!


Peter,
-I like your idea to use the fan shroud to electrically insulate the open connection.

-I think we'd need to move/add the/a fan to the right side of the pack... that's the way the air slots flow through the lithium cells. The "unused" side you mentioned (covered by the plastic cover) is where the fans would likely need to go. However, I think we'd leave the OEM fans installed, too, as they would cool the BMS discharge resistors. We can figure that out later.

-The master/slave board idea is good, too. The interconnect would be 2-wire isoSPI.

-I agree we can't replace the BCM with the amount of effort I'm willing to focus on this project. Replacing the BCM very quickly approaches the effort of Linsight.

-I like your earlier idea of cutting the OEM tap connector off and screwing it into the BMS PCB.

-9600 baud H-line is easy... 328p has a dedicated serial hardware protocol.

-Remind me: does the OEM BCM mind if the battery voltage is 200 volts?

-I'll think about this in more detail tomorrow.

With an appropriate value pre-resistor for the 10x10k 0.1% BCM Fooler matrix and identical VPIN resistor to bring the working voltages down into the OEM range, the car can tolerate actual battery voltages up to ~220V (DC-DC Shutdown)
+ 5/12V Power/Ground presumably? Or will each slave board have a regulator/separate pwr/gnd??

Very easy to implement this with the OBDIIC&C as well,
it already talks to the IMAC&C P&P inside the MCM on the HLine using exactly this method.
(HLine 5V 9600,8,N,1) Idle line high.

We just have to give your device a unique ID on the H BUS.
I used a 5 byte packet including header $BB & simple summing checksum.
So ' BB 05 00 80 C0' for example is a packet I send to the IMAC&C P&P every 100ms or so..

I suggest your BMS could listen on $CC, I think that is free on the HLine

It can then send out the cell voltages with the header $CD when asked.
One large packet (max length 256 bytes including header and checksum)

So $CD + (48 cells x 16 bits per reading) 96 bytes + Checksum) = 98 bytes total or about 100ms
Or we can pack each cell voltage into 12 or even 8 bits bits depending on the resolution required.

It would be good if your cpu has enough free cycles to do the conversion from the weird LTC data into a real voltage say 2/3 decimal places it then sends out.

If you want to do total pack voltage, highest cell, lowest cell, average cell, cell deviation calculations and send those in the packet as well it means I don't have to do it in the OBDIIC&C. ;)

We probably only need/want to poll your CPU/BMS for all this stuff once a second or so.

The voltages/configurations in red will be too low IMHO for the IMA without boosting it at the VPIN and the Voltage taps.

IIRC If the unloaded pack voltage falls below ~144V the car will not IMA start etc..

It will also be trying to regen a lot of the time to get the voltage up into a more normal range.
It won't stop charging naturally until the detected IMA voltage gets well above the 150V, upto around ~175 in fact.
A BMS/control failure could be very risky with a low voltage poly pack and a lot of regen.

I found it much easier to reduce higher pack voltages with cheap resistors,
than boost low ones with op amps and inline dc-dc converters etc.

We can choose the BCM Fooler pre resistor and VPIN resistor to get the pack down into a nice optimal OEM comfort zone. The car will then naturally cut regen and assist as it's detected voltages reach the normal IMA limits adding some free backup/fallback inherent safety.

I think 48 cells should be the standard for most people.

All just my two cents of course.. ;)
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #3 ·
First things first, Peter made some comments about the DCDC's supported Vin range. I sat down tonight with some power supplies and a programmable load to measure these parameters:


Here are my most important findings:
The DCDC converters I tested shut down at 221.7 & 221.8 volts. This is a digital controller shutdown - based on a 0.02% precision reference - which means we can push right up to 220 volts without worrying about tolerances across parts.

Given this number, that means we can in fact use a 54S pack (e.g. QTY3 18S modules). In this configuration, we cannot charge the cells above 4.07 volts (if we do, the DCDC will turn off).

So yes, Peter, I think we should support the 54S configuration (that we previously proposed).
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #4 ·
Peter mentioned 36S wasn't going to work well, due to voltage being too low. I'm not certain I agree with this, but Peter is probably the resident expert on the MCM's behavior. However, I'd like to discuss the following data I've gathered:

I put my worst OEM NiMH pack into the test mule tonight and drove it around. With 50 A assist, the pack voltage dropped to 117 volts. With 50 A regen, the pack voltage was 178 volts. Therefore, this pack has an ESR somewhere around 610 mOhm. These voltages are as measured by Pegasus, which are retrieved from HLine. I believe the pack voltage is measured directly by the MCM (via the 2-cable connector that plugs into the rear side). Therefore, the ESR I calculated above should include only the pack voltage. Peter let me know if I'm wrong here.

In comparison, a 36S lithium pack has 43 mOhm pack resistance, which is 14x less than the worn out NiMH pack I tested above. This means that given the above test conditions, a 36S pack resting at 133 volts (3.7x36) will drop to 130.8 volts at 50 A assist, and rise to 135.2 volts at 50 A regen. Given that most of the energy within a lithium pack is consumed while the pack is ~3.7 volts, I propose the MCM won't care that a 36S lithium pack is installed.

With a 36S configuration, we won't be able to empty the pack to 2.75 volts, because that results in a 36S stack of 99 volts. However, I believe we can bring the IMA battery voltage down to 110 volts or so without upsetting the BCM. I can certainly gather this data directly from the car with the above-mentioned test equipment, but hopefully Peter just knows what the minimum supported voltage is.

Obviously there will positive/negative recal quirks, but the system should work without issue. We as a community have years worth of experience understanding how the MCM/BCM behave with lithium packs.

Peter, I trust that you are correct regarding 36S being too low, but please let me know where I went wrong.
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #5 ·
Regarding PCB footprint/topology.

We should probably have just a single PCB. The primary driving force here is that the LTC6804 supports up to 12 cells in series. Given that these Honda packs are 18S, we need to be absolutely certain that the full pack voltage is never applied across a single LTC6804... if that happens, the LTC6804 will blow up. Since we can't easily divide these packs into even 12S, we therefore need to be able to put 12 cells from one 18S module's connector to one LTC6804, then the remaining 6 cells into another LTC6804... we'd then place the first 6 cells from the next 18S module to the remaining six inputs on the 2nd LTC6804. The remaining 12 cells in the second 18S module would connect to a third LTC6804.

Don't worry about the specifics in the above paragraph. The conclusion is this: If we made separate PCBs for each pack, we'd need QTY2 LTC6804 modules on each PCB. Given that each LTC6804 costs over $20, this is not a practical solution (i.e. the LTC6804s alone would be $120 for this project with separate PCBs). The only other alternative would be to add six additional conductors to the connection between each PCB, but this gets clunky fast, would also blow up the LTC6804 if any cable came unplugged.

So my conclusion is that we need a single PCB, and that it needs to support 48S, might support 36S, and would be nice to support 54S. Given Peter's previous thoughts, and my tinkering with the OEM components tonight, I believe a 355x70 mm PCB bolted to the air intake side of the OEM battery bay is the most obvious solution. The narrow width allows airflow to route through the battery using the OEM fan.

Peter: I was wrong when I previously stated that adding a fan to the right side would allow us to blow air through the channels between the cells... since the modules are sideways, there is no way to use these channels (they point up/down as installed). As previously discussed, though, I don't believe we need airflow between these cells... the OEM fan doesn't have much room to suck air through the lithium-loaded IMA enclosure, but it'll certainly be able to pull some air around the outside of each case. Given that I don't believe cooling is required at all (see previous discussion), I'm not worried about the limited airflow.
 

·
Administrator
Joined
·
12,790 Posts
Good idea new thread..

The DCDC converters I tested shut down at 221.7 & 221.8 volts. This is a digital controller shutdown - based on a 0.02% precision reference - which means we can push right up to 220 volts without worrying about tolerances across parts.
Where is this reference out of interest and can it be changed?

Onto the BMS and a possible 36S x 3.7V nominal = 133V.

It all depends on what voltages the MCM and BCM see.

If we fake these low voltages into the right (Nimh) areas/zone then it will work with caveats, but is less safe from a potential overcharge scenario than a higher voltage pack.

My earlier comments about faking voltages down being a lot easier than faking them up still apply.

The Nimh pack is 144V nominal and is basically pretty much empty at that resting voltage.
The car will simply not allow IMA start if the detected pack resting voltage is much less than this +/- a volt or two.
I haven't checked this for a while, so you might want to confirm the exact value.
The 36S 133V Nominal is way below this no IMA start threshold. :(

The MCM will also aggressively cut back assist current if the detected voltage under load drops below ~120V. (1V per Nimh cell)

The cutback is progressive and it will cut right back to virtually nothing in a determined effort to prevent the pack dropping much below ~120V. There are likely some variations between cars due to component tolerances and even firmware revisions.

Because the pack voltage is so low the system will also naturally try to charge it a lot more.

My comments about building in some intrinsic safety with these cells suggests to me we should operate at a voltage where the car will naturally cut back regen of its own accord with a high pack voltage...

Which is more dangerous for these new cells? Over or undercharge?

Of course all that being said we can make/force it to work with lower voltages.

You might remember I did a 48 Cell LTO ~96-120V setup (faking the voltages up) that worked quite well, but they don't go on fire if overcharged.


I just remembered. We did have that weird ~2000 rpm related assist dead zone with the LTO low pack voltages so that's another negative. :(

Lower pack voltages = more current/losses for the same Kw power output.

Uncorrected lower pack voltages will stress a standard 100A fuse more as the car will tend to push more current.

With a high voltage pack and standard 100A IMA fuse 200V x 80A = 16kw out of the box without current hacking! Instant +50% power upgrade..

Low voltage 133V x 90A = 10.6kw max :cry:
 

·
Administrator
Joined
·
12,790 Posts
With the OEM fan operating I'm not bothered about extra cooling either.
As you say they have a low ESR so self heating is unlikely to be an issue.

I saw there was a few spare mm around the packs when inserted into the case.
Can they be centralised in each void/slot so air can flow past on all sides?
Perhaps very thin self adhesive beading strips or something... :unsure:
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #8 ·
Minimizing unusable "leftover" 12S modules:

One goal with this project is to consume entirely the QTY2 12S and QTY2 18S modules included with each $400 purchase. Let's explore the various options assuming we purchase QTY4 complete sets (i.e. QTY8 12S & QTY8 18S modules):

If we support 36S, then we can always consume any remaining 12S modules, regardless of the number of 48S/54S conversions. For example, given the above input modules, we could build the following cars:
18S+18S+18S = 54S
18S+18S+12S = 48S
18S+18S+12S = 48S
18S+12S+12S = 42S
12S+12S+12S = 36S
Leftover 18S modules: QTY0 (there will always be QTY0 leftover 18S modules).
Leftover 12S modules: QTY1
Total battery connectors required: QTY6 (because we need to support both 12S & 18S in each bay).

Key takeaway: With a 36S option, we can always consume all 12S modules. This requires QTY6 battery connectors (yuck, but ok).

..........................

If we only support 48S, then we get:
18S+18S+12S = 48S
18S+18S+12S = 48S
18S+18S+12S = 48S
18S+18S+12S = 48S
Leftover 12S modules: QTY4
Total battery connectors required: QTY3 (ideal)

Key takeaway: We'll always have QTY1 unused 12S module for each $400 set purchased. This seems wasteful.
..........................

If we only support 48S & 54S, this gets even worse:
18S+18S+18S = 54S
18S+18S+18S = 54S
18S+18S+12S = 48S
Leftover 12S modules: QTY6
Total battery connectors required: QTY4 (one bay has 12S & 18S, the other two are 18S only)

Key takeaway: We CANNOT offer 54S if the only other option is 48S... this leaves the most unused 12S modules.

..........................

If we support 42S, 48S, and 54S:
18S+18S+18S = 54S
18S+18S+12S = 48S
18S+12S+12S = 42S
18S+12S+12S = 42S
18S+12S+12S = 42S
Leftover 12S modules: QTY1
Total battery connectors required: QTY5 (two bays have 12S & 18S, one has 18S only)

Key takeaway: Theoretically we can reduce the number of leftover 12S modules. HOWEVER, I don't expect too many customers would choose the 48S option, given that the cost difference is trivial between 12S & 18S modules. Certainly cash-strapped customers will purchase the 42S option, but everyone else would want the 54S option... I just don't see any market for 48S packs with this offering.

..........................

Given the above concern, in the real world the above case reduces to 42S & 54S (only):
18S+18S+18S = 54S
18S+18S+18S = 54S
18S+12S+12S = 42S
18S+12S+12S = 42S
Leftover 12S modules: QTY4
Total battery connectors required: QTY5 (two bays have 12S & 18S, one has 18S only)

Key takeaway: We still have QTY1 leftover module per $400 set (same as 48S only condition above)... seems wasteful. Also we've managed to require QTY5 connectors, whereas 48S (only) condition only requires QTY3 connectors.

..........................

If we only 42S & 48S:
18S+18S+12S = 48S
18S+18S+12S = 48S
18S+18S+12S = 48S
18S+12S+12S = 42S
18S+12S+12S = 42S
Leftover 12S modules: QTY1
Total battery connectors required: QTY4 (one bay has 12S & 18S, one has 18S only, one has 12S only)

Key takeaway: We've reduced leftover modules considerably. This only required QTY4 connectors, which is really nice.

.................................................................

Given the above, I think we should support ONLY 42S & 48S.
I know the die-hard 54S crew will be let down, but 54S adds the following additional burdens:
-If 36S isn't technically possible (see previous post), then we end up "wasting" a bunch of 12S packs.
-However, adding 36S support requires two separate connectors for each battery bay. This DRASTICALLY complicates things... really anything more than four connectors makes the routing tree (i.e. where each tap voltage goes) considerably more difficult.
-Adding the 54S option requires an additional LTC6804, which adds $25 to the raw PCB cost.
-54S Vcell(max) is different (because DCDC shuts down at 220 volts). Therefore we have a different cutoff voltage for 54S configuration (4.07 volts per cell). I don't want to have to deal with that.
-Customer confusion... five separate ways for you to hook something up wrong and kill yourself, or blow out the PCB and then gripe to me that "it must have been broken when I got it."

I am certainly open to feedback on my thoughts, but I am not excited about the drawbacks that a 54S system adds. Please convince me otherwise if you find a better solution.
 

·
Administrator
Joined
·
12,790 Posts
I agree 48S & 42S looks best.

My preferred choices in order..

1) 48S = 132-201V
Best overcharge voltage safety headroom.
Highest IMA power as standard with OEM 100A fuse.
Requires only simple 2 x 10 cent resistor voltage down faking. (Maybe -10/15V or so)

2) 42S = 115-176V
Less overcharge voltage safety headroom.
Slightly lower IMA power.
May not require any resistor faking (Apart from the BCM Fooler).

I would not be worried about spare extra packs.
They can be picked over by the scavengers and hackers on here LOL.
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #11 · (Edited)
The voltage reference is built into the DCDC converter. All the components inside the chopper are rated to at least 400 VDC, so certainly the OEM DCDC converter could be modified to allow higher voltages. However, while simply increasing the voltage reference would allow higher input voltages, it would also modify the output voltage, too... the output voltage is regulated by IC201, so you'd need to proportionally lower the output there.

Increasing the OEM DCDC's allowed input voltage is certainly possible, but I'd need to twiddle with it a bit to fully figure it out. I am certain, though, that is could be made to work up to 250 VDC... above that and the main chopper's drain current starts to drift outside the maximum safe operating area (MSOA); even though input current decreases with increased input voltage, the MSOA derating is at a higher slope... you might be able to hit 275 VDC, but it would reduce the maximum allowed output current (e.g. 50A max output).

I'm going to table this discussion for now.
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #12 ·
Peter, my BMS will disable regen when any cell is full, and assist when any cell is empty.
There are multiple ways to achieve this... we'll need to decide which one to pick (e.g. QBATT hacking, BATTSCI spoofing, etc). This is a hard requirement.

If you're comfortable with 42S & 48S, I think that's a clean solution... it greatly simplifies this project.
 

·
Administrator
Joined
·
12,790 Posts
All sounds good.

A couple of final thoughts for now before I must start work on another project... ;)

1) The/a 12S block should always go in the slot nearest/behind the switchboard.
Its shorter length leaves more room for routing high current cables out of the back of the switchboard.

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?

I'm thinking high current connectors all right behind switchboard.

BMS connectors all at other side, and your/our BMS could then go in that secure space with the opaque cover well away from any high current wiring. .

Unobstructed pack airflow vents in correct orientation for OEM fan flow.
(The pcb is not in the way)
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #14 ·
More thoughts on mechanicals: I'd love to use the existing Fan Mount Threaded inserts to mount the PCB, but the batteries fit better with the PCB mounted to the air inlet side. The reason is the OEM enclosure's inlet side has tapered ribs, which allow the PCB to sit further forward, which makes cable management easier. Mounting the PCB will require drilling a few holes into the enclosure's plastic... whatever.

I want to use the threaded inserts (on the exhause side) for the plate that mechanically secures the batteries to the enclosure... I don't want these batteries coming loose in a wreck. Since the 12S modules are shorter, there will be a "mounting plate extension" required for each 12S module. This could be as simple as a long threaded spacer... TBD, minor detail... the end goal is that the modules are coplanar on the intake side (with the cell tap connector)... otherwise it gets hard to plug the connectors into the PCB.

///////////////////////////////////////

Required hacks with 42S+48S design:

The PCB I proposed previously has tons of room on it... we'll have room for whatever "hacks" we come up with to make the BCM+MCM happy. However, I'd like to minimize the number of "extra" hacks beyond the PCB (e.g. the BCM interceptor); ideally everything is contained on this one PCB.

If that is not possible with the BCM in the picture, I could study the thought of generating the BATTSCI data stream (which I know requires METSCI feedback, too). I looked over my notes last night and it seems like you and I have collectively deciphered this data stream. My prior complaint that we don't have enough hardware serial ports becomes less valid if complexity increases elsewhere... If nothing else, I could just use a separate microcontroller (they're cheap) to handle BATTSCI+METSCI, so that I don't have to waste time software bit-banging while also reading each cell voltage. My rough calculation is that reading 48S cells is going to take about 100 ms... if I have to bit-bang that I'll quickly run out of time.

In short, can you provide a list of those hacks (if any) that cannot be rolled into this PCB with a 42S & 48S configuration?
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #15 ·
The/a 12S block should always go in the slot nearest/behind the switchboard.
Its shorter length leaves more room for routing high current cables out of the back of the switchboard.
I absolutely agree. The order will be fixed:
-nearest to junction must be 12S
-middle must be 18S
-far is either 12S or 18S
 

·
Linsight Designer
Joined
·
2,238 Posts
Discussion Starter · #16 ·
Let's talk electrical:

Communication:
-The CPU will tie into the HLine... we'll need to tap into it. That'll be the only way to gather data from the BMS. Users will need something to read the data... the simplest solution is your OBDIIC&C. Given how good it is, I'm not going to take the time to build a cheaper (or free) product. Customers won't need the OBDIIC&C per se... BMS will still work 'headless'. "$CC" works for me... we can figure out the transfer protocol later... minor detail for now.

Note that converting the raw ADC output count to voltages is actually pretty time consuming on the CPU level. I'm sure I can make it work, but "decimals" and "microcontrollers" are always cumbersome. Minor detail for now.

I can certainly send pack voltage (adding is easy)(possibly as a 16 bit raw ADC sum). highest and lowest cell is easy, too, but average cell voltage isn't likely... that's a LOT of division. Division is only easy when it's 2^n (e.g. 32 cells or 64 cells)... in those cases we just right-shift the data n bits... in all other cases we actually have to do the math. Who knows, we'll see how much CPU time we have left at the end.

Certainly the 328p can handle decimals/division with a properly architected program... but I'm not wanting to dedicate the effort to develop a properly-developed system. It won't be spaghetti code, but it's not going to be an RTOS either.

Power:
BMS will need to tap into either the 12 volt or 5 volt lead, plus engine ground. I could generate 12 volts from the entire pack voltage, but that's a lot of work and makes bank isolation difficult. Much easier to just use 12 volt rail. I need to revisit the 5 volt rail generated inside the MCM (and routed to several sensors). My memory is that the phase current sensors are powered by them and that the rail can handle quite a bit of load. If that's a solution, then I won't need to generate 5 volts onboard (i.e. from the 12 volt rail). Minor detail, TBD later.

Safety:
The entire battery pack will be galvanically isolated from the control circuitry. There will be ZERO possibility that high voltage from the pack leaks into the digital control circuitry.

High current battery connections:
We MUST place the IMA power switch between the first (12S, closest to the junction board), and second (middle) module! Breaking the pack anywhere else will damage the LTC6804 ICs... because the full pack voltage will develop across one LTC6804. Customers that do not follow this guidance will damage the BMS PCB. There is no way to avoid this, unless we add an additional (fifth) IC... which adds $25 to the PCB COGs.

Cost:
The parts alone for this PCB are already more than $140, so we need to do whatever we can to keep the price down. As with everything insight, I'm not doing this to make money, but I won't sell the end product for a loss. Right now this BMS won't cost less than $300... that's pretty much guaranteed.

isoSPI (you had asked a question about this, although it's not relevant now that this is a single PCB solution. I still want to answer your question, though: isoSPI doesn't require external power or ground... just a differential serial bus (two wires). The power for each LTC6804 is generated from the 12 cells (or 6 cells) powering it.
 

·
Administrator
Joined
·
12,790 Posts
It's fair to point out IMA work is my living/income so there will be my costs on top.

Lets call it around $500 per BMS total approx for now.. John $300 + Peter $200.
 

·
Administrator
Joined
·
12,790 Posts
Communication: I can work with the raw ADC sum and highest lowest voltages.

Power: I would use the 5V MCM rail. I already use it for loads of gadgets.
It is well conditioned, can supply a fair chunk of power, and is very stable.
It also remains on after the car ign is turned off for ~ten seconds, allowing time for data saving etc

Safety: No obvious issues.

High current battery connections: Pack break is fine..

Cost: See my prev post.

isoSPI: I see.
 

·
Administrator
Joined
·
12,790 Posts
In short, can you provide a list of those hacks (if any) that cannot be rolled into this PCB with a 42S & 48S configuration?
Cannot be included.

(VPIN resistor)
This will have to be added into the harness orange VPIN wire from the IPU to the MCM or inside the MCM.
 

·
Registered
Joined
·
58 Posts
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.
 
1 - 20 of 271 Posts
Top