BMS to Controller Integration: Contactor Sequencing, Charge Enable and Derate Handshakes

When a lithium conversion misbehaves, the battery is rarely the problem. The pack, the controller and the charger are usually all working exactly as designed. What is missing is an agreement between them about who is in charge of what, and what each one does when another goes quiet.

That agreement is the integration work, and it is almost never written down before the parts are bolted in.

Decide who owns the contactor. Once. In writing.

The single most common architectural fault. The BMS has a contactor output because it must be able to disconnect the pack. The controller has contactor control because it must sequence precharge. If both are wired to influence the same contactor without a defined hierarchy, you get a machine that drops out under load for no reason anyone can reproduce, or — far worse — one that closes into a fault because each device assumed the other was inhibiting.

Pick one owner. The defensible choice in almost every case is that the BMS has final authority to open, because it is the only device that can see cell-level conditions, and the controller requests closure and performs the precharge sequence. Then write down:

  • Which physical output drives the coil, and through what.
  • What the other device does with its contactor capability — ideally nothing, explicitly disabled and documented.
  • Whether the BMS can open under load, and what the controller does when the bus voltage disappears mid-torque.

That last one deserves bench time. A pack opening at full current into an inductive load is a real transient, and “we will never do that” is not a design.

Precharge, and why it is not optional

Closing a contactor directly onto a controller’s DC link capacitance is an inrush event that pits contacts and, repeated enough, welds them closed. A welded main contactor on a lithium pack is the failure you least want, because the pack can no longer disconnect itself.

The sequence is unglamorous and must be respected: precharge resistor path closed, DC link allowed to rise to near pack voltage, link voltage verified rather than assumed by timer alone, main contactor closed, precharge path opened. If your controller provides this, use its implementation rather than rolling your own. If you are sequencing it yourself, verify against measured link voltage and keep the timeout short enough that a shorted link is caught instead of cooking the resistor.

Charge enable is a separate conversation from discharge enable

These are two different permissions with different conditions, and conflating them causes a specific and seasonal complaint.

Typically limited by What it must reach
Discharge enable Low cell voltage, high temperature, low temperature, current limit, fault state The traction controller, as an inhibit and as a current limit
Charge enable High cell voltage, cell imbalance, and above all low cell temperature The charger, and anything else that can push current into the pack — including regenerative braking

Regen is charging. A pack too cold to accept charge current is also too cold to accept regen. If your charge-enable logic only reaches the charger and not the controller, a cold machine going downhill will push current into a pack that has told you it cannot take it. Lithium plating from cold charging is cumulative and invisible until capacity is gone.

So charge enable has to arrive at the controller as a regen inhibit or a regen current limit, not only at the charger. On older lead-acid-era controllers, that path frequently does not exist at all — which is the real reason some otherwise sensible controllers are a poor choice behind a lithium pack.

Derate: the limits have to actually do something

A good BMS publishes live discharge and charge current limits, pack state of charge, and cell temperature extremes. Publishing them is not integration. Integration is the controller consuming them and reducing torque accordingly.

What to define explicitly:

  • How a limit becomes torque. Does the controller take a current limit directly, or do you have to scale it into a torque or current parameter? On a CANopen drive this is a mapped object; on a simpler controller it may be an analogue input or nothing.
  • How fast a derate is allowed to apply. Stepping instantly from full torque to 30% is a lurch that an operator feels and may misread as a fault. Ramp it.
  • Whether the operator is told. A machine that has quietly halved its torque because the pack is at 2 °C will be reported as broken unless the display says why. This is a display requirement that comes out of the BMS integration, and it is routinely missed.
  • What happens at the floor. There should be a defined minimum capability — enough to limp off a ramp — rather than torque simply reaching zero wherever the curve lands.

Timeouts: design for silence, not just for faults

Every fault condition you can enumerate is the easy half. The hard half is a device that stops talking. A BMS that has crashed is not reporting a fault; it is reporting nothing, and nothing looks a lot like “no problems” to naive code.

For every cyclic message your system depends on, write down three things: the expected period, the timeout, and the action. Then test it by physically unplugging the device mid-operation.

  • BMS goes silent. The controller must assume the worst, not the last known good value. Derate to a safe limp capability or inhibit, and tell the operator.
  • Controller goes silent. The BMS and display both need a defined response. Does the pack stay closed? For how long?
  • Display goes silent. Covered in the display integration article — and the answer must never be that the operator loses the ability to stop the machine.
  • Charger goes silent mid-charge. Does the contactor open? Does the pack sit at full voltage with a charger attached that may resume unsupervised?

The anti-pattern to look for in your own code is a variable holding “last received current limit” with no age attached to it. Stale data must be distinguishable from fresh data at the point of use, not just at the point of reception.

Five failures we actually see

  1. Welded main contactor from a missing or timer-only precharge. Pack can no longer disconnect. This is the one that matters most.
  2. Cold-weather complaints that are correct derate behaviour, invisible to the operator, and therefore reported as a fault. Fixed on the display, not in the pack.
  3. Regen into a charge-inhibited pack, because charge enable never reached the controller.
  4. Charger that will not start because it is waiting for a BMS permission that is being published in a format it does not read. Two correct devices, no agreement.
  5. Both devices inhibiting, each assuming the other would close. Machine simply will not wake up, and nothing is faulted.

Note that four of the five are agreement failures, not component failures. Replacing parts does not fix them.

The document to write before you wire anything

One page, and it prevents most of the above:

  • Every device on the system, with its protocol and bitrate.
  • Who owns the contactor, and what every other device’s contactor capability is set to.
  • The precharge sequence, with its verification condition and timeout.
  • Each permission — discharge enable, charge enable, regen enable — its source, its destination, and its format.
  • Each derate input, how it maps to torque, its ramp rate, and its floor.
  • Every cyclic message, with period, timeout and timeout action.
  • What the operator sees for each degraded state.

If you cannot fill that in, the integration is not designed yet, however complete the bill of materials looks.

We do this work

BMS-to-controller integration, contactor and precharge sequencing, derate strategy, and the display work that makes degraded states legible — see custom electronics and control system development. We also stock the lithium packs, controllers, contactors and chargers the design calls for — and on charger selection, the charge algorithm guide explains why a lead-acid profile on a lithium pack is not a small mistake.

Email matt@electricdrivesolutions.com or call (717) 421-9988.

This article describes general integration practice and is not a substitute for the BMS, controller and charger documentation for your specific parts, or for the applicable standards for your vessel or machine. Verify every permission and limit against the manufacturers’ own documents and test the degraded states on the bench before the machine carries a person or a load.

Back to blog