Honda Insight Forum banner

41 - 60 of 69 Posts

·
Registered
Joined
·
921 Posts
The dimensions of this socket are quite similar to the 20 pin plug:

The drawing above references part numbers for the other components. I only looked briefly for the plug at Digikey and TE.com but not very whole-heartedly.

Over at TE.com this appears to be from the TH/.025 connector system.

I was getting a headache trying to find the 24 pin socket/plug at the TE site. Here's where I last gave up (doing this in the morning might produce better results):

Seems that Honda has had a long relationship with TE Connectivity/AMP.
 

·
Registered
Joined
·
921 Posts
Actually there was a catalog marked TH/.025 which had a picture of every connector, but none of the connectors matched either connector here or the socket found above.

Other potential TE Connectivity series: 0.64/2.3. If those are larger power pins on either end of the double row of pins on the 24 pin connector, this might be the series they are from. Bad news, though - a number of them are marked "restricted" and no part information is available:

 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #43 (Edited)
Peter, for your over-current hacking pleasure, how much current do you want the battery sensor to be able to measure:
-During Assist?
-During Regen?

The higher the current we have to measure, the lower the current resolution. For this reason, I don't want to add too much headroom versus the stock values (which I believe are 50 A regen, 100 A assist). Basically what's your highest stable overcurrent hack?

Related: You had emailed me a couple years ago asking about converting this bipolar current sensor to single ended. I never tested out the circuit design I sent you to do this, but I just built it and it is very linear... IIRC you had had non-linearity problems during regen (as the current sensor drives Rsense more negative). I'm not sure if you ever figured out why, but this is due to the current sensor's headroom, which appears to be ~1.75 volts from either rail; note that we'll never hit the positive rail because it's much further away. Anywho, the solution here is to reduce the current-to-voltage converter ratio (i.e. decrease precision power resistor resistance). I'll choose this value based on your answers above.

Here's the current sensor circuit I mentioned a couple years back. Note that the final Rsense resistor value (81.1 Ohms in picture) will change (decrease) based on your answers above:
 

·
Registered
Joined
·
921 Posts
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.
I was wrong about the "small FETs". They are diodes, perhaps zeners, and there are 20 of them. The silkscreen direction-indicating arrow looked like a third pin.
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #45
The dimensions of this socket are quite similar to the 20 pin plug:
Thanks for joining the search. Two reasons this connector won't work:
-This is an in-line connector, whereas we're looking for a PCB-mount connector. I had previously restricted my search to PCB-mount connectors, so I didn't happen across this guy, which indeed has the correct pitch. Obviously if we can find who makes the part, then we can explore that series, so thanks for increasing the connector scope.
-The tab portion of the Honda battery's 12S connector is too wide for 1473750-1:


Over at TE.com this appears to be from the TH/.025 connector system.
I'll look more into it. Thanks for the lead.

Seems that Honda has had a long relationship with TE Connectivity/AMP.
My very first thought when I saw the connector was "This is a TE Connector". Unfortunately there's ZERO stamping/stickering/inking anywhere on the connector.
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #46
I was wrong about the "small FETs". They are diodes, perhaps zeners, and there are 20 of them. The silkscreen direction-indicating arrow looked like a third pin.
Zeners make sense. LTC pushes them hard on large series stacks, in an effort to prevent IC blowouts. Unfortunately, the only way to prevent zener blowout is to add series resistance between each cell and the LTC IC... which adds voltage measurement error, particularly when the discharge FET is active on any given channel.
 

·
Administrator
Joined
·
12,616 Posts
Re current hack. (+50%)
The BCM sensor should cope with 150A Assist and 75A Regen.

So +/- 150A is a good round number.

We physically hack the BCM current sensor at present using a 220R resistor in parallel with the 100R OEM loop resistors inside the BCM. Many people already have these installed in their BCM's.

How would we hack yours? Physically or in the firmware?
How or will you adjust the current counting for the SOC if a hack is fitted?


For people following along on this thread maybe playing catch up..

The Orion2 BMS is a very good (IIRC) LTC chip? based documented BMS design.
The website, user manual and technical resources are great general reading. (y)

 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #48 (Edited)
If you're ok with 150A Assist & 75A regen, I'll use those values as maxima (versus 150 A both ways).
If you previously physically hacked the BCM current sensor, my recommendation would be to remove the hack. However, we could correct for the modification in firmware, too... it'll just decrease the resolution on those BCMs with the hack installed. I've made a note to document how to change this value in the firmware.
Correction (brain fart): My plan right now is to remove the BCM entirely... so no more hack at all.

I do plan on current counting with this sensor. The microcontroller's built-in ADC has 10b resolution, which lets me resolve current to within 250mA. The circuit I posted previously configures the ADC's voltage reference to the 5 volt rail, which is great because the current sensor's output voltage is ratiometricly scaled to the 5 volt rail... this effectively removes a ton of noise from the ADC circuit. For example, if the 5 volt rail droops to 4.8 volts (for some reason), the current sensor output will drop proportionally, which will cause the digitized value to remain constant (assuming the current has not changed).
 

·
Administrator
Joined
·
12,616 Posts
The BCM Current sensor hack was usually done by soldering the extra resistor across the sense ones inside the BCM, so this hack will disappear along with the OEM BCM.
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #50 (Edited)
Peter, my previous post was a major brain fart: I'm planning to remove the BCM entirely, hence no need to worry about it at all... your hack will go away entirely (as will the entire BCM). I've corrected my post.

Also, the 81.1 Ohm resistor I randomly selected in the above schematic was a lucky guess: it's linear from 68 A regen all the way to 160 A assist:


I'll decrease the final Rsense value to 75 Ohms and that will give us ~74 A to 175 A (pending me gathering actual data). Resolution with 10b ADC is 244 mA per count... I don't see a reason to be more accurate than that.
 

·
Administrator
Joined
·
12,616 Posts
No problem. Just to be clear....

So the BCM correction for those with the hack will now be done in your firmware or on your hardware?

The (+40% power) current hack (most common and reliable) in its entirety involves fooling all five current sensors in the car with resistors.

(4 sensors in/via the MCM using 8 x 1k 0.1% resistors) and (one in the now defunct BCM using the 240R part)

We probably need to chat about this
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #52
It probably makes sense to just add some extra pads in parallel on my PCB, and then have those with current hack populate resistors to match whatever they've fooled elsewhere. Of course, we can modify the value in firmware, too, but I worry that's more complicated to most people (versus just soldering parts onto empty pads).
 

·
Linsight Designer
Joined
·
2,068 Posts
Discussion Starter #53
Nightly update: I spent a few hours reviewing and thinking this project over today... at this point I've mentally mapped out the project to work with just a single Arduino Uno (328p) Microcontroller. This gives us the following capabilities:
-Completely replaces BCM... OEM BCM removed entirely. All OEM functionality preserved.
-Support for Peter's over-current hack (via PCB population and/or firmware value change).
-Native support for QTY3 Honda G3 Insight lithium modules, in either 42S or 48S configuration. External support for up to 192S (via additional 3rd party LTC demo boards).
-USB Communication allows cell voltage/temp/current dumping onto a computer. No speed restrictions.
-USB Communication allows simple firmware upgrades.
-HLine Communication allows cell voltage/temp/current dumping to OBDIIC&C. We might need to be careful about how much data OBDIIC&C requests in one block... I'm port expanding BATTSCI (9600b), METSCI (9600b) and HLine (9600b) onto the 328p's single unused I2C port. The I2C-to-UART chip I'm using can only queue 8 commands at once. More details to follow once we see how much CPU time the uC has... it will work, just not sure how fast. Honestly a minor detail right now
-Balances cells while car is off (draws minuscule amount of power from 12 volt battery... maybe 2 mA).
-Prevents overcharge/discharge by over-riding QBATT PWM signal (man-in-the-middle).
-Contrary to what I wrote earlier, I do not plan to support CAN bus... it's just not in the cards with the effort I'm willing to put into this project. CAN requires quite a bit more work for several reasons... any future external communications to this board will need to occur over USB/HLine/isoSPI/UART... but not CAN.

I'm gonna throw together a pencil-on-paper schematic tomorrow.

Note: I'm working on two other projects right now. This project is not my primary focus.
 

·
Registered
Joined
·
921 Posts
Looking very good!
just a single Arduino Uno (328p) Microcontroller
Is the atmega2560 (Arduino Mega) an option with an expansion header to enable experimental builds with custom hardware, say, an MCP2515 for CAN, an SD card interface for logging, or an extra serial interface Bluetooth via an HC-05 or HM-18 for a status display on a phone, and enough RAM/ROM for the associated code? Or are you considering a command/logging interface, perhaps via Serial or even USB, so that this could be done on a separate nearby board (another Arduino with a USB host chip, a Pi, or similar?)

The CAN interface in particular would become popular if the cheap source of G3 Insight packs runs out, and used CR-Z/HCH3 packs become more available, and the CAN protocol to those is found to contain balancing commands. Or, perhaps your thinking is more, "let's limit V1 to this specific use case, and learn from it, and use that to inform these considerations on a V2"?
 

·
Registered
Joined
·
40 Posts
Not wishing to step on any toes but I would suggest something a little more modern than the ATmega. The Teensy 3.6 Puggie and I are using for our BCM has 2 CAN, lots of serial and a very good ADC.
 

·
Registered
Joined
·
921 Posts
You might need to set the ALF/RLF flags in the BATTSCI data stream as well as control QBATT to completely prevent over/under charging IIRC.
I would imagine that the BCM replacement firmware would act similarly to the Insight, in that assist fades gracefully but still is somewhat available with increasing throttle position?

When I'm driving with a passenger, monitoring the car is often forgotten, and if I am in an environment where the pack voltage has been getting lower and lower (hills or long uphill) and I need to pass and get no feedback in the foot that the pack is low and full assist suddenly goes to zero, this is a show-stopper for me.
 

·
Registered
Joined
·
921 Posts
Not wishing to step on any toes but I would suggest something a little more modern than the ATmega. The Teensy 3.6 Puggie and I are using for our BCM has 2 CAN, lots of serial and a very good ADC.
I was thinking the same thing (I have an M0 in the form of an STM32F103C8T0 sitting in front of me) but stopped short since these are 3.3V devices in a 5V car. Where did you have to do level conversion in your BCM?
 

·
Registered
Joined
·
40 Posts
There are lots of level converters out there. For the CAN and serial busses it isn't an issue as you have to use an interface chip anyway.
 

·
Registered
Joined
·
921 Posts
<long rant reduced to>

Seems now might be a good time to open a GitHub repo, wiki, and start getting used to using the issues feature for wishes, requirements, and later bugs.
 
41 - 60 of 69 Posts
Top