Stack Fuel Gauge (2S–6S) with Cross-Cell Sync

Why 2S+ Needs a Dedicated Stack Fuel Gauge

A single-cell fuel gauge and a multi-cell (2S–6S) stack fuel gauge solve two different problems. For series packs you must measure each cell, keep the cells in sync, and present one pack-level result to the BMS host. Otherwise the upper controller has to chase several voltage/temperature registers sampled at different moments.

This chapter targets small-batch, multi-brand, sometimes unstable supply projects — tools, light EVs, in-vehicle auxiliaries, telecom/industrial backup — not phones or wearables that run on a single Li-ion cell.

Typical 2S–6S scenarios

• Power tools / light mobility (2S–4S) that need to see which cell sags first.
• In-vehicle auxiliary / small 12 V-to-Li packs where ambient temperature varies a lot.
• Small telecom / industrial backup packs (2S–3S) that require stable SOC reporting.
• Prototype / retrofit BMS builds where the intended brand is not available every week.

Why single-cell logic breaks on 2S+

For series packs, three issues appear immediately:

1. Cell-to-cell mismatch: one cell ages faster or has higher ESR — total pack voltage hides it.
2. Uneven temperature rise: cells closer to the enclosure or PCB run hotter and need per-cell T capture.
3. Unequal charge/discharge moment: charger “looks OK” at pack level, but one cell is already limiting.

Because of this, the fuel-gauging IC must speak from pack perspective — give the BMS host a single, coherent SOC/power prediction that is already derived from all cells. The host does not want to maintain different register maps when you swap brands.

Out of scope for this chapter: cell pairing / pack manufacturing process, choosing cell chemistry/brands, and charging current/JETA-style curves — those live in other pages.

System Architecture & Signal Path

The stack fuel gauge sits between the series-connected cells and the BMS/vehicle controller. It is parallel to the multi-cell charger, not a sub-module of it. Cells report electrical state to the gauge; the gauge frames and aggregates it; the BMS host uses it to take system-level decisions.

Two typical deployments

1) Standalone stack gauge + MCU

• IC does per-cell V/T capture and builds a synchronous frame.
• MCU/host consumes the frame and runs pack-level prediction or reports upstream (CAN/LIN/other).
• Good for small-batch builds where the gauge IC may change brand.

2) BMS AFE with stack measurement + host-side gauging

• AFE measures all series cells and exposes them to the MCU.
• MCU performs SOC/SOH logic, possibly with vendor-provided tables.
• Better when you already have a strong host MCU and want to keep one bill of materials.

We do not cover balancing FET drive, high-current busbar design, or large daisy-chain AFE networks here — those belong to other BMS pages.

For higher-voltage packs, long harnesses, or automotive EMI environments, isolation or differential sensing around the stack gauge may be required. Keep this in the BOM notes so purchasing cannot silently downgrade to a non-isolated monitor.

Series-Stack V/T Capture Requirements

For 2S–6S battery stacks, the fuel-gauging function must capture per-cell voltage and per-cell or per-pack temperature, and it must do so inside a synchronous frame. Pack-voltage-only monitors are not sufficient, because they hide cell-to-cell mismatch and make cross-brand replacement risky.

This section breaks the requirement into two clear parts — voltage capture and temperature capture — and then ties them together with sampling / refresh / sync so that the BMS host can trust the data at pack level.

1. Per-cell voltage capture

The stack gauge must read every cell in the series string, not only the total pack. A practical target is:

• Cell count: 2S–6S (must match charger / multi-cell controller capability).
• Per-cell voltage range: 0–4.20 V or 0–4.35 V per cell (Li-ion stack typical).
• Accuracy: ≤ 1% or “must be aligned with charger CV accuracy”, so the charger and the gauge do not disagree.
• Common-mode tolerance: high enough for the full stack voltage (up to 6S worst case).

Write this into the BOM so purchasing cannot replace it with a “pack-voltage-only” monitor.

2. Temperature capture (per-cell or per-group)

Voltage alone is not enough for reliable SOC/SOH estimation on series packs. The gauge should support:

• Single NTC per pack for simpler builds.
• Multiple NTC / AUX temperature inputs when cells are placed in different thermal zones.
• Note for cross-brand alternatives: temperature channels, pinout and scaling are not always compatible, so re-validate T capture when swapping ICs.
• Board trace vs. cell surface: if you measure at the board, document the offset; if you measure on the cell, document the probe placement.

3. Sampling & refresh policy

How often the gauge samples all cells determines how often the BMS can safely make a pack-level decision.

• Full-frame sampling: all cells (V/T) in one shot — best for cross-cell comparison.
• Polled / rotating sampling: cells are sampled over several cycles — host must reconstruct a full view.
• Pack-level refresh: typical 50–500 ms depending on system load and logging rate.

The key requirement is: per-cell values that belong to the same frame must represent the same moment. Otherwise, a transient on one cell will be misread as “this cell is weak”.

Out of scope here: bringing the measured temperature into JEITA / fast-charge curves (charging topic), or full pack thermal modeling (pack/vehicle topic).

Cross-Cell Sync Mechanism

Even if every cell is measured, a stack gauge is only useful when all cells are measured in the same time window. Otherwise one cell might be read during discharge and another during idle — and the BMS will believe the first one is weak. That is why cross-cell synchronization sits at the top of the validation list when you swap brands.

1. Why sync matters

• Comparable cell voltages must originate from the same sampling instant.
• Pack-level SOC/power prediction must be based on a complete and coherent frame.
• When charger and gauge disagree, the first thing to check is: “are cell readings synchronized?”

2. Common sync methods

A stack fuel gauge typically supports one of these:

1. Built-in frame sync: the IC captures all cells and exposes a “frame ready” / “frame valid” flag — easiest to integrate.
2. External trigger / timer-aligned: an external event or timer aligns several acquisition channels (or ICs) to the same instant.
3. MCU poll + timestamp: the host polls cells in sequence but tags each read with a timestamp and later re-aligns them — a fallback when sourcing forces a simpler device.

3. What desync looks like

• One particular cell always appears lower than others.
• Pack-level SOC or available power jitters at every refresh.
• Charger says “terminate/float”, while gauge still says “charging/low”.

4. Sync validation experiments

When you drop in a second-brand or substitute IC, run at least these two tests:

• Load step on all cells: apply/remove a common load and check the reported voltages all change in the same frame.
• Slow temperature ramp: heat up the pack or put it in a chamber and verify all channels progress smoothly without lag between cells.

Out of scope here: BMS CAN message formats, large daisy-chain AFE networks and multi-board addressing — those belong to higher-level system communication pages.

Pack Capacity / Power Prediction

A series-stack fuel gauge only becomes useful when all per-cell measurements are consolidated into one pack-level decision: SOC%, available power, and (optionally) SOH hooks. For 2S–6S systems, this consolidation must follow a “weakest-cell-first” rule: if one cell is aging faster or runs hotter, the pack prediction must respect that cell, not the average.

Also note a procurement-critical point: if the selected IC does not provide pack-level prediction, the MCU must do a secondary composition step. This must be written into the BOM to prevent buyers from replacing the stack gauge with a cheaper “voltage-only monitor”.

5.1 Inputs from the series stack

Minimum input set for pack-level prediction:

• Per-cell voltage frame (2S–6S): already time-aligned from Chapter 4.
• Per-cell / per-pack temperature: at least 1 NTC, preferably multiple for uneven thermal zones.
• Pack / branch current: from shunt, AFE, or integrated CSA in the same BMS path.
• Chemistry / OCV table: so the pack-level algorithm can map V/T/I to SOC.

5.2 Aggregation strategies

A practical way to explain the consolidation is to show three steps — voltage coherence, weakest-cell gating, and temperature-aware derating:

1. Voltage coherence: use the synchronous frame (V1..Vn, T1..Tn) to make sure every cell is from the same instant.
2. Lowest-cell gating: identify the cell with the lowest usable energy (often the lowest V or the highest T) and make it the reference for the whole pack.
3. Temperature-derated capacity: apply high/low temperature factors to reduce available capacity and power if any cell is outside the preferred window.
4. Aging / cycle hooks: if the IC tracks cycles/impedance, use it to gradually tighten the derating.

If the fuel-gauging IC only outputs per-cell values and not pack-level SOC, the MCU must re-create this 3-step logic. This is the “host composite” pattern you should document for small-batch projects.

5.3 Pack-level outputs & consumers

The final deliverables of the stack fuel gauge should be easy for the rest of the system to consume:

• Pack SOC % — single number, already gated by the weakest cell.
• Available / deliverable power — pack voltage × allowable current, derated by cell limits.
• SOH / RUL hooks — if the IC exposes them, forward to BMS/vehicle controller.
• Recipients: BMS MCU / vehicle controller (active use), charger / power-path manager (read-only).

5.4 Error sources & procurement risks

• Different OCV tables: TI, ST, NXP, Renesas may use different OCV grids; pin the chemistry in the BOM.
• “Capacity-only” parts: some ICs output only remaining capacity %, not remaining power/time → state this explicitly.
• Lack of sync: replacing a synchronous part with a non-synchronous monitor will make pack-level SOC jump.
• Temperature channel mismatch: if a brand has fewer NTC inputs, derating will be less accurate.

5.5 Example PNs for pack-level or host-composite designs

Brand / PN Mapping Strategy (2S–6S Stack Gauging)

Different vendors place the multi-cell measurement function in different places — sometimes in a dedicated stack fuel gauge, sometimes inside a BMS AFE, and sometimes as part of an MCU+AFE combo. This chapter shows where to look in each of the seven brands and how to phrase your BOM so procurement cannot buy a total-voltage-only monitor.

BOM remark (copy/paste): “Gauge IC must support per-cell V/T capture with synchronous frame for 2–6S stacks. Total-voltage-only or unsynchronized monitors are not acceptable.”

6.1 Mapping principles

• Prefer parts that already expose per-cell capture + frame-valid.
• If a part only outputs cell data, mark it as “MCU composite required”.
• Fix chemistry / OCV table in the spec so swap-ins do not silently change SOC behaviour.

6.2 Texas Instruments (TI)

Best coverage for 2S–6S packs, multiple ways to do pack-level prediction.

• BQ40Z80 — 2–6S gauge + protection, pack-level SOC/SOH ready.
• BQ34Z100-G1 — wide-range, use host to do pack synthesis.
• BQ78350-R1A + BQ76920/30/40 — AFE does stack sensing, host does gauging → perfect for small-batch.
• BOM: “use Impedance Track family only; monitors without pack-level prediction not allowed.”

6.3 STMicroelectronics

Good for 3–5S and automotive/industrial mixes; pack-level decision is usually on MCU.

• L9961 — multi-cell monitor/balancing/current sense; per-cell capture OK, do pack SOC in host.
• L9963E — higher channel count, same idea: host reconstructs SOC.
• Note: temperature inputs may differ by variant → write “multi-NTC required” in BOM.

6.4 NXP

AFE-first approach; always plan for an MCU that owns the SOC.

• MC33772C — 6-cell AFE, perfect for 2S–6S examples in this page.
• MC33771C — 14-cell, good to illustrate “frame-based reporting”.
• Spec: “requires host SOC / pack model on S32K or equivalent”.

6.5 Renesas

ISL94xxx line is a good fit for monitors that let the host do pack logic.

• ISL94212 — 3–8 cell monitor/protector.
• ISL94216 — up to 16 cells, chainable, host must do SOC.
• Note: specify “rev XX or later” to lock the feature set.

6.6 onsemi

Often protector + SOC indicator devices — you have to say “host composite required”.

• NCS35011 — 3–5S protector + SOC LED, not a full stack fuel gauge.
• LC709203F / LC709209F — 1-cell fuel gauge → if purchased by mistake, MCU must aggregate.
• BOM: “3–5S device acceptable only with MCU-side pack SOC.”

6.7 Microchip

Classic 6-cell monitor + Microchip MCU = flexible small-batch BMS.

• ATA6870N / ATA6871 — 6-cell measurement/balancing, proven for stacked packs.
• ATmega32HVB / PIC24FJ — run coulomb counting + pack model here.
• Note: when substituting ATA devices, re-validate “frame sync”.

6.8 Melexis

Use Melexis parts to make temperature / sensing channels consistent across brands.

• MLX90632 — non-contact, high-accuracy temperature for derating.
• MLX90316 — position / angle, mention only if pack has actuated elements.
• Clarify in BOM: “Melexis sensor is an assist channel, not the main fuel gauge.”

Small-Batch Procurement Strategy

For 2S–6S stack fuel gauges, small-batch projects should select by stack fit first, then by package, then by temperature/automotive grade, and only in the end by “nice-to-have” functions. This chapter gives a procurement-first view to prevent buyers from replacing a true stack gauge with a total-voltage-only monitor.

7.1 Selection order (what to check first)

1. Cell count (2S–6S): the IC must cover the intended series count; higher-cell parts may be used in partial-channel mode.
2. Package / solderability: QFN/QFP that your assembly house can handle for pilots or cut-tape sourcing.
3. Temperature / automotive grade: industrial-temp is acceptable for pilot; plan an AEC-Q100 or automotive-grade drop-in for mass build.
4. Interface / register map: SMBus/I²C alignment, frame-valid bit, cell-order indexing.
5. Price / channel: evaluated last, after technical fit is confirmed.

7.2 Channel priority for small lots

• 1) Original vendor / FAE sample: ensures you get the exact stack-gauging variant (with sync).
• 2) Authorized distributors: often stock only the industrial-temp version — use it for pilots.
• 3) Spot / cut-tape / e-commerce: okay for prototype, but BOM must say “retest sync on arrival”.

7.3 BOM remark templates (copy & paste)

7.4 Incoming acceptance / validation

Every time the brand or P/N changes, run this 4-step checklist:

• Power-up & read ID: confirm exact P/N and variant.
• Read one full cell frame: all used cells must appear in the same frame.
• Sync scenario switch: apply the same load step to all cells → verify all voltages move in the same time slot.
• Temperature channel presence: plug NTC(s) and confirm the IC reports them.

7.5 Recommended PNs by brand (pilot → mass)

Cross-Brand Alternative Paths

When the planned TI or ST stack gauge cannot be sourced, you can still keep your 2S–6S design moving by switching to NXP, Renesas, onsemi, or Microchip parts — but you must re-validate sync and temperature channels. This chapter shows three safe substitution routes.

8.1 Why cross-brand swaps can break gauging

Each brand may use a different OCV table, different frame-valid flags, or fewer/more temperature inputs. Swapping without re-test can make the pack-level SOC jump or derating fail.

8.2 Path 1 — Interface / range equivalent

Use this when you want almost the same interface and cell range:

• Planned: TI BQ40Z80 (2–6S)
• Alts: ST L9961 (3–5S, host SOC), NXP MC33772C (6-cell AFE), Renesas ISL94212 (3–8 cells)
• Action: after swapping, re-align register map and verify you still get 1 complete, time-aligned frame.

8.3 Path 2 — Higher-cell AFE used in partial-channel mode

Use this when only higher-channel parts are available:

• Use only the first 4–6 channels of NXP MC33771C (14-cell)
• Use only the first 4–8 channels of Renesas ISL94216
• Use Microchip ATA6870N (6-cell) but populate 4–5 cells only
• BOM: “Device is used in partial-channel mode; incoming test must confirm per-channel sync on used cells.”

8.4 Path 3 — MCU + multi-channel ADC fallback

When no integrated stack gauge is available, build your own measurement front end:

• MCU + TI ADS1015/ADS1115 (for per-cell voltage & NTC)
• Microchip MCU + MCP39F511A (for power/energy channel)
• Melexis MLX90632 to normalize temperature across different cells/modules
• Note: emergency/prototype path only — go back to dedicated stack gauge IC for mass build.

8.5 Validation after any swap

• Run a stack load-step and check all cells move on the same time slot.
• Run a slow thermal ramp (chamber) and check no cell lags behind.
• Read a full frame including unused channels (they must be stable/flagged).
• Compare SOC/power with the original brand’s golden unit.

8.6 Example PNs by path (7 brands)

Charging Coordination Touchpoint

In multi-cell packs, the charger typically looks at pack-level voltage and current, while the stack fuel gauge looks at per-cell, time-aligned data. If the charger thinks it can continue but the stack gauge reports a cell-level problem, the BMS host must give priority to the cell-level fault.

9.1 Why this touchpoint is needed

A multi-cell charger is blind to cell-to-cell mismatch. A stack fuel gauge is not. Therefore, the BMS host must reconcile the two views and enforce “cell-level fault wins.”

9.2 Typical conflict scenarios

• Charger OK, cell over/under temperature: charger sees pack OK, but stack gauge reports Cell3 overtemp → must limit or pause.
• Charger in CC, one cell lags in voltage: charger continues, but stack gauge shows Cell2 not catching up → must slow down.
• Charger wants to terminate: dI/dt or ΔV looks good to the charger, but stack gauge still sees an outlier cell → must delay termination.

9.3 Coordination strategy (cell-level fault wins)

The BMS host polls the stack fuel gauge each frame. If any cell-level violation is present, the BMS host issues a “limit / pause / reduce power” command to the charger. If no violation is present, the charger may continue its CC/CV logic.

• Step 1: Host reads stack gauge frame (cells 1…N + temp + frame-valid).
• Step 2: If any per-cell fault → override charger with limit/pause.
• Step 3: Log the event to the BMS history/log page (link in your site).
• Step 4: If frame is clean → allow charger to proceed.

9.4 Communication requirements

Simple I²C / SMBus is enough to pass “fault / frame-valid / derated-power” to the host. For USB-C / PD sinks, apply the same rule at the policy layer — we do not describe PDO/SRC details here.

Do not let the charger continue when the stack gauge says a cell is out of range.

Validation & BOM Remark

This chapter turns the stack fuel gauge requirements into repeatable lab checks and copy-and-paste BOM text. Run these tests whenever the brand or P/N is substituted, and your buyers will not accidentally choose a pack-voltage-only monitor.

10.1 Validation goals

• Confirm the device is the expected stack fuel gauge family.
• Confirm per-cell voltage and temperature show up in one frame.
• Confirm synchronous change under dynamic conditions.
• Confirm pack-level prediction is present or that the host can compose it.

10.2 Step-by-step validation flow

1. Step 1 — Static frame check: power up, read device ID/vendor ID, read a complete 2–6 cell frame, confirm “frame-valid = 1”.
2. Step 2 — Dynamic load step across cells: apply a stack load / release; all used cells must update in the same time slot.
3. Step 3 — Temperature skew test: heat/cool one cell/NTC and confirm the frame reports it without delay.
4. Step 4 — Pack prediction sanity check: compare IC’s pack SOC / available power with your golden unit or with MCU-composed value.
5. Step 5 — BOM update: if an alternative P/N is used, log it with “retest sync required”.

10.3 BOM remark examples (copy & paste)

10.4 Lead time / alternative phrasing (for buyers)

Use wording that gives purchasing room to act but keeps them inside the technical boundary:

• “If preferred P/N is not available within 2 weeks, use alternative from TI / ST / NXP / Renesas / onsemi / Microchip / Melexis with the same cell count and synchronous frame.”
• “After substitution, re-run validation flow (Steps 1–4).”
• “Alternative must preserve per-cell voltage and temperature capture.”

FAQ (Stack Fuel Gauge, 2S+)

This FAQ is limited to this page and focuses on four topics only: multi-cell sensing, cross-cell sync, pack-level prediction, and small-batch / cross-brand substitution. It does not cover charging algorithms, eFuse topics, or single-cell wearables.