BMS Wiring - complexity at connector vs in harness

methods

1 GW
Joined
Aug 8, 2008
Messages
5,555
Location
Santa Cruz CA
I am struggling with some decisions around wiring a BMS Master.

Sorry for a long post with no picture breaks.

For cost and IP I have settled on a single IP66 35way single connector for my Master
It is to be mounted in a box with a gasket.

Slave units get built into 4S-12S modules and connect via a single twisted pair (isoSPI)

I think I already know the answer... but I need to talk it out.
There are two paths forward - as I have the pins to support it:

Peripheral Components, such as the Primary Contactor, Precharge Relay, Throttle cut Relay, DC-DC... they all need between 2 and 7 wires. They all need different voltages based on their independent requirements. Example:

We must support contactors with coil voltages ranging from 12V to 116V

We can centralize... meaning the BMS would have: (in the case of the Coil)

* Coil Ground... which I control ON/OFF with an N-Channel (or possibly PWM). This is good for 12V - 130V, up to 7A.
* Coil Positive... which is an always on Connection... which can come directly from the B+, directly from 12V, or as a pass-thru from me
* Coil Source..... Which would be a pass-through from me in the case of a BMS centralized wiring.

SO:

1) External Components can get their Positive excitation via direct pickoff from the harness (with fusing of course)
2) External Components can get their Positive excitation from the BMS as a centralized point

Option 2 minimizes and standardizes harness complexity and makes for adjustments that can be made with jumper pins at the BMS allowing us to keep harness complexity down. With this approach we can also include self-resetting fuses in the BMS on the PCB... but... it makes for a lot of extra crimps and a lot of wasted I/O on the primary connector.

Option 1 greatly reduced pin count and complexity at the BMS. It forces fusing to happen out in the harness... with actual fuses... it adds a lot of ring terminals.. It could still work with a pretty standard harness but I would have to have some "dead ends" built into the harness from birth.


Since I am the BMS component designer... of course I think the world revolves around me... but in reality the world probably revolves around other components - like the controller or primary ECU. In this case I am sort of the primary ECU... but we want to get away from that.

I am thinking that it would be best to keep the BMS simple. In the case of driving all those contactors... I am considering exposing ONLY the low side switch... so for a few examples:

* Primary Contactor gets its excitation from either the B+ terminal, the 12V buss, or some other source.
* I provide primary contactor low side coil switching only... and I have 4 ports for this (Primary, Precharge, Throttle, and DC-DC control - all interchangeable and all rated for full spec)

I may be able to downsize my 35 pin (HUGE) connector to a 24 pin if I am smart. The 24 pin is also widely available in a 90 degree PCB mount variant which is MUCH more desirable from a packaging perspective (picture supporting the PCB in an off the shelf housing... no custom molds will be built). With a 90 degree connector I may be able to leverage an ebike-type extruded aluminum housing that the PCB slides into, where we have some heat sinking options, where the end caps can be fitted with a gasket, where only the end caps need to get machined for the connector hole. Alternative is a standard plastic housing with either the box end or lid end getting CNC'ed in a batch - and that soft plastic could allow me to have them machined by a hobby grade machine.

I think my answer is that I need to provide 5-10 "reference diagrams" for how customers will handle their specific use case.
I need to build a lot of custom fuses that (instead of being ceramic or glass...) are modules that plug in and leverage self-resetting restive fuses.
Customers can wire to suit... and instead of the BMS trying to be the end-all "we have this for you" solution... it is aimed more as a support component.

Folks will need to understand fundamentals... like how low-side switching architectures work.

Speaking of low side switching... I am currently debating whether I need to protect from overcurrent or just put that responsibility on the user.
There are plenty of N-Channel mosfet drivers which include inputs for a shunt resistor, have timing inputs for over-current protection, and provide feedback to the MCU... these are expensive and require a lot of parts... so I am leaning toward simpler drivers that go down easy in common footprints ... and just trust that the user wont do anything epic stupid.

The Sevcon is a pretty good example of a totally protected self contained unit... but... I am not at that level.
If I can just provide a robust and reliable switch, that when run inside of spec, will not fail under any circumstances ... I think that's enough.

V1 will NOT do any PWM on the contactor drives. Sevcon implemented this and I think it actually lowers reliability.
Low side contactor drive needs to be one of the most reliable parts of the system... and with enough TVS diodes and margin... I can converge on 99.999% reliability... but if I start trying to PWM - all that goes out the door.

So... IF we wanted to PWM the coils in the circuit... we could centralize and provide that service to components... by knocking down pack voltage to coil voltage... but this only scales up to maybe... 120V. Over that it is not really a reasonable option.. so... yea.. drop that. Furthermore... to ensure reliability... both the top side and bottom side would need to be Mosfet controlled, to address the issue of low side failure and over current.

BAH - BOM COST, MANUFACTURED COST, TIMELINE, CUSTOMER REQUIREMENTS... Stay on point methods... this aint Disneyland we are trying to develop... its V1 of a Park Ranger hut... KISS.

150V IR mosfet
Proper Linear Technology Driver
1.5KW strapping TVS diodes
Standard strapping diodes that are tuned tight and can sink current or short out
Proper bypassing with caps
Heavy heat sinking on a surface mount

I think if I spec 12V to 150V at up to 7A I will be safe.
At 7A it may start to overheat... but really... contactors need only a handful of amps to start and an amp or two to maintain... so thats proper margin.

To provide 4 identical ports - I am looking at a footprint about 130% the size of 4pcs D2 pack.
Thrift says I should run a much smaller fet for driving the Throttle Cut relay - but I want them to be interchangeable in case someone pops a channel.
I cant see any instance where someone would want to drive a coil at over 120V... and the only reason I support that is to support the full range of Zero legacy gear... where through the years... Zero ran 12V, 24V, 48V, 86V, 96V, and 116V contactors (bastards).

I want to keep unique parts count down and duplicate as much as possible so I can actually afford reels and get some volume pricing.

-methods
 
As for grounding....

My only connection to the pack is IsoSPI - transformer isolated
My hall sensor is non-contact
My contactors are totally isolated
Most DC-DC's are somewhat isolated - but can pass spikes

In theory I could run isolated from Battery Negative.

I am not going to do this. I am going to follow Sevcon's lead and hard central point ground right on the battery negative.
I am going to ride the noise... with a heavy duty thick single point.

I totally get that in theory we are isolated and that in theory isolated is better.
In practice... we are not actually isolated... we are inductively coupled all over the place.
We are subject to capacitive coupling and all sorts of EMI
I want a hard connection to true ground.

As I get older... I think of Ground in less of an absolute way... less of a static way.
I think of ground in relative terms. It is not fixed in space with noise bouncing above it.
Ground is relative to all the positive in the system. They bounce together.

I can try to hold all that down... or isolate myself from all of that... but in the end... if something is bouncing like hell... and you need to keep a hold of it... its better to hard-strap yourself with it and accept the ride. Trying to stand next to it, keeping your arms on it, with a pair of wet oven mitts... is just fail.

Every major aerospace design I have seen was common ground eventually... out at the carrier.
Independent components - especially between active components and telemetry... practiced strict grounding rules... but at the system connector - where the rubber actually hits the road... all things are one.

Chassis?
Wont tie it to battery negative - except perhaps with a RLC of some sort to sink noise

Battery Ground?
Thats every-bodies ground
We work hard to have not a single loop anywhere... we ensure that battery ground is fastened tight (as if it comes loose... those electrons WILL find a path.

-methods
 
Correction:

I need to read Switched Pack + off the outside post of the contactor...
so unless I wanted to do all sorts of isolated voodoo... I am on the hook to reference ground at least there.

I will also need to perform isolation tests - those will require ground.

I will also need to accept DC-DC's that are not isolated... that requires ground.

I may also need to interface with Chargers... those likely reference ground.

Isolation is a fleeting idea. I am letting it go.

-methods
 
Take a look at the JAE MX23 series if you are after sealed right angle connectors. Way less bulky than (I assume) the ampseal you were thinking about. Termnals are cheaper, and housings cheaper aswell. JAE likes to frock people on the crimping tool though, so its kinda the high buy in and low running cost option.

For the inputs and outputs that have high current potential, PTC all the things. Ill dig up the part numbers for some that work rather well.
 
I support your grounding scheme... I always called it a star point ground. Also applaud the idea that ground is never absolute.

" I am currently debating whether I need to protect from overcurrent or just put that responsibility on the user.
There are plenty of N-Channel mosfet drivers which include inputs for a shunt resistor, have timing inputs for over-current protection, and provide feedback to the MCU... these are expensive and require a lot of parts... so I am leaning toward simpler drivers that go down easy in common footprints ... and just trust that the user wont do anything epic stupid."

... now if you were production, with total control of all aspects I would say this is applicable. When provisioning this BMS to folks, one off, there is a huge potential for "oopses" Now whether suitable protection can fit the perceived cost model is a separate story, but folks are going to blow stuff by mis wiring, dropping washers onto live components, etc. Doing that will "blow" the simple interfaces, resulting in a dead box. Now can they cost effectively repair the box? Do you get service calls? Do they berate the design on message boards? Will they think they won't blow stuff, and the price for a protected design is too high?

All part of the system design space as you are doing. Technically I would say in this "one off" to a single user type hardware, the protection is technically needed. I am not sure the buyer will perceive the value in it though. Everyone thinks they won't make a mistake.

farfle, the MX23 look sealed, but not gasketed. Did I miss something? As you said $2400 for a hand crimp tool from Digikey... WOW :shock:
 
They are not gasketed (IMHO, we call sealed boxes "Aquariums" as they tend to pump humidity out of the air into the box) but they are designed with open face potting the container in mind. they are also somewhat freindly to fence-and-fill coatings.
 
Back
Top