Turnkey Custom Battery Pack Development: Design to Volume

A custom battery pack programme rarely fails on chemistry. It fails at the seams. The cell supplier ships what was ordered, the BMS house delivers a board that meets its own datasheet, the enclosure shop cuts to the drawing it was given, and the assembler joins three things that were never designed against each other. Each party can point to a document showing they did their job. The pack still does not work, and nobody owns the gap.

That is the case for buying the job as one scope rather than four. This page sets out how a turnkey custom battery pack project runs at DIDIDEDE — what we take responsibility for, where the engineering decisions actually get made, and what a buyer should demand at each stage before releasing a purchase order. If you want the commercial mechanics instead — what to send, MOQ, sample cost, lead time — the custom battery pack OEM process page covers that in detail.

Key takeaways

A turnkey custom battery pack programme means one contract, one technical interface and one party accountable for the finished pack — not that a single company manufactures every component in it.
Most of a pack’s cost is committed the moment the drawing is released — cell grade, busbar material, enclosure process, board specification, labour content. Negotiating unit price afterwards moves only the remainder.
Nine inputs define a project. With fewer than six of them, any quotation is a placeholder that will be re-priced.
A prototype proves assembly feasibility, electrical function and communication handshake. It does not prove batch consistency, vibration life or production yield.
UN 38.3 has to be resolved before volume production, not before launch — it is a transport precondition, and re-testing follows any cell, BMS or structural change.
Five specific points separate “the sample was good” from “the shipments are good”, and all five are auditable.

What “turnkey” means here, and what it does not

The word gets used loosely, so it is worth being precise. In this context it means three things, and each one is testable before you sign:

One drawing authority. A single engineering team signs off the pack drawing, the electrical schematic and the BMS configuration together. If a dimension changes, one party decides whether it is feasible and who absorbs the consequence.
One quality record per batch. Cell grading data, weld parameters, BMS test results and aging data live in one file keyed to one serial number. When a field failure appears eighteen months later, that file either exists or it does not.
One point of responsibility at the volume switch. Moving from 3 sample units to 300 is where projects break. Someone has to own the difference between a hand-built sample and a line-built pack.

What it does not mean: that one company performs every step inside its own walls — mining, cell winding, PCB etching, enclosure forming. Vertical ownership of every step is not where the value sits. The value is that one party holds the interfaces, integrates the cell, the protection electronics, the mechanical structure, the compliance file and the logistics, and answers for the assembled pack as a single system. That is what removes the handoff risk described at the top of this page.

For a buyer, this reframes the due-diligence question. Instead of “do you own a factory”, ask: when the pack fails at the customer’s site, who receives the claim, who performs the root-cause analysis, and who pays for the replacement batch? If the answer is different for the cells, the BMS and the assembly, the scope is not turnkey regardless of what the brochure says.

The real cost of splitting the job: four handoffs

Sourcing the same pack from four specialists is not irrational. Each specialist may be better at its own step than a generalist integrator. The cost is not in the parts; it is in the boundaries between them. These are the four places it shows up.

Handoff	Typical symptom	Who usually absorbs the cost
Cell selection to BMS design	Protection thresholds suit the board’s default profile, not the cell’s actual discharge curve. Pack derates early or trips under peak load.	Buyer, as a redesign after the sample
BMS to mechanical layout	Board footprint and connector height were never checked against the enclosure’s internal clearance. The lid does not close, or the harness chafes.	Buyer, as a second tooling round
Mechanical to assembly	Busbar routing assumes a weld sequence the assembler cannot execute at line speed. Yield drops only at volume.	Buyer, as scrapped units and delayed shipments
Assembly to certification	The transport test specimen differs from the production build in cell batch or weld method, so the report does not cover what ships.	Buyer, as re-testing and held inventory

None of these are exotic failures. They are the ordinary consequence of four parties each optimising its own deliverable. The integrator’s job is not to be smarter than the specialists; it is to hold the interfaces so that no interface is unowned.

Stage 1 — The nine inputs that decide the project

Every custom battery pack project reduces to nine numbers and constraints. They are not a wish list; each one removes a degree of freedom that would otherwise be guessed. Getting them onto one page before engineering starts is the single highest-leverage thing a buyer can do.

Nominal voltage, and the acceptable operating window around it. Nominal is a label; the window is what the load actually sees.
Capacity or target runtime, in Ah or kWh, with the duty cycle that produced the number.
Continuous discharge current, and the temperature at which it must be sustained.
Peak discharge current, with duration and repetition rate. “300 A peak” means nothing without a time base.
Charge voltage and maximum charge current, plus whether the charger is yours, ours or a third party’s.
Maximum envelope, as a drawing or a measured cavity. Sketches with arrows are acceptable; “about the size of a shoebox” is not.
Connector type and pin-out, or a sample of the pack being replaced. This is where most compatibility surprises originate.
Communication protocol if the host device talks to the BMS — CAN, UART, RS485, or open-loop with a dry contact.
Environment and annual volume: IP rating, vibration class, operating temperature range, cycle-life target, and units per year by year.

With six of the nine, a quotation is meaningful. With fewer, it is a number that will move — and moving a quoted price after the buyer has built a business case around it damages both sides. If you only know part of the specification, the practical starting point is a photograph of the existing pack plus the rating on the motor controller or inverter; that pair usually reveals voltage, connector family and an order of magnitude for current.

Stage 2 — Cell and architecture: the decision you cannot undo

Chemistry is usually framed as a performance choice. In practice it is a constraints choice, and it is the least reversible one in the project because it propagates into the BMS, the thermal design, the certification file and the shipping classification.

LiFePO4 buys cycle life and thermal stability, at a weight and volume penalty. NMC buys energy density and cold-weather behaviour, at the cost of tighter thermal management and more conservative protection settings. For a vehicle where the pack is a large fraction of the bill of materials, energy density often wins. For a stationary or industrial unit where the pack is charged daily for years, cycle life usually wins. Neither preference is universal, which is why a supplier who recommends one chemistry before asking about duty cycle is guessing.

The second decision is string configuration, and it is where two suppliers can quote “the same” pack and ship non-interchangeable products. A 48 V LiFePO4 pack can be built as 15 series cells at 48.0 V nominal or 16 series at 51.2 V. Both are legitimately described as 48 V. One may sit outside a charger’s acceptable window; the other may exceed a controller’s input rating. This is the same trap described in how to read a battery pack specification sheet, and it is worth settling before the drawing exists rather than after.

Parallel count follows from capacity and current, but it also sets the failure mode. More parallel strings means more tolerance for a single weak cell and more work in balancing. Fewer strings means a cleaner BMS and less assembly variation. The right answer depends on whether the pack is expected to fail gracefully or to be replaced outright — a distinction buyers rarely state and engineers always need. The sizing arithmetic itself is covered in how to size a battery pack; the cell options we build around are listed under battery cells.

Stage 3 — Mechanical and DFM: where most of the cost is already fixed

This is the stage buyers most often treat as a formality, and it is the stage that determines the price. By the time a pack drawing is released for tooling, the cost is largely settled: cell grade and format, busbar material, enclosure process, BMS board specification and the labour content implied by the assembly sequence are all consequences of decisions already made. What remains negotiable afterwards is a smaller share than buyers usually expect.

Three decisions carry most of that weight.

Standard housing or bespoke enclosure. Off-the-shelf shells need no tooling and shorten the path to a sample. A bespoke enclosure adds tooling cost and lead time, and it is only worth it when the available cavity genuinely cannot be served by a standard form — or when the enclosure is doing structural work that mounting brackets cannot. Welded stainless construction, as used in our 60 V 40 Ah and 72 V 30 Ah packs, buys ingress protection and mechanical stiffness but is a heavier answer than a moulded shell where weight matters.

Cell layout determines the thermal path. Cells packed tightly save volume and make heat removal harder, which forces the BMS to derate earlier. A layout that saves volume on the drawing can end up with a materially lower continuous rating in practice. The honest trade is usually a slightly larger envelope in exchange for sustained current — and it has to be made on the drawing, not discovered in testing.

Assembly sequence determines yield. Busbar routing, weld access and harness length all assume a build order. If that order is impossible at line speed, the design is not wrong electrically, only unmanufacturable — and it will pass a hand-built sample while failing in volume. This is why a design-for-manufacture review belongs before tooling, not after the first production run.

Two further points are cheap to specify and expensive to retrofit: connector orientation relative to the mounting plane, and service access to the BMS for a firmware update or a fuse replacement. Both are decided in minutes at this stage and cost a redesign later. Packs with high current and long service life, such as the 74.6 V 148.5 Ah NMC build, make the second point concrete: a sealed unit with no service route turns a fuse replacement into a pack replacement.

Stage 4 — BMS and firmware: the integration nobody owns

The battery management system is the component most likely to be bought as a catalogue part and treated as a solved problem. It is not, because the BMS sits between the cells and the host device, and both sides have opinions.

Protection thresholds should be derived from the cell’s own discharge curve and the load profile, not lifted from a board’s default configuration. A board whose over-current trip sits just above the application’s peak will nuisance-trip on cold mornings; one set generously will let the pack run outside its thermal comfort zone before intervening. Neither failure appears in a bench test at room temperature. The parameters that matter, and how to read them off a datasheet, are set out in BMS specifications explained.

Communication is the second half. CAN is the default for vehicle and inverter integration, but the message set has to be agreed explicitly: which frame carries state of charge, what scale factor, which fault codes, and what the host does when a frame stops arriving. RS485 and UART appear where the host is a simpler controller. Open-loop builds with a dry contact are legitimate for applications that only need a low-voltage cut-off — but they should be a stated choice, not a fallback after protocol negotiation fails.

Firmware ownership deserves a line in the contract. Three questions decide whether a pack is a long-term asset or a locked box: can the configuration be read out and documented, can a parameter be changed without returning the pack, and what happens to the source if the relationship ends? A pack whose behaviour cannot be reproduced is a pack you cannot audit.

Stage 5 — Prototyping: what a sample can and cannot prove

Samples are built and bench tested in about two weeks on a typical project, and a prototype order as small as three units is enough to validate a design. Those three units are genuinely useful, but it is worth being clear about what they establish.

A sample proves that the parts fit, that the electrical architecture functions, that the BMS communicates with the host, and that the pack meets its envelope. That is a meaningful list — most interface errors are caught at exactly this point, which is why building before tooling is cheap insurance.

A sample does not prove batch consistency, because three units can be built from three selected cell groups. It does not prove vibration life, because that requires a shaker and a duration. It does not prove cycle life, because that takes months. And it does not prove yield, because a hand-built sample tolerates assembly steps a line cannot. Treating a passing prototype as a production release is the most common way a good project goes wrong.

The productive use of the sample stage is to define what approval means in advance: which measurements are recorded, against which limits, and who signs. Our own view of that discipline is written up in battery sample approval and incoming inspection, and it applies as much to the buyer’s incoming checks as to the supplier’s outgoing ones.

Stage 6 — Validation and certification: sequence matters more than the list

Buyers usually ask which certificates are available. The more useful question is what the certificate covers, and whether the tested specimen matches the production build.

UN 38.3 is the transport precondition, and it must be resolved before volume production rather than before launch. A pack without a valid test summary cannot legally fly, and sea freight of Class 9 goods carries its own documentation and packaging requirements. Because the test is performed on a specimen, any subsequent change to the cell model, the BMS board or the structural design can invalidate the coverage — which means certification is not a one-off gate but a constraint on every engineering change afterwards. The document set and its dependencies are covered in lithium battery export essentials and in shipping and export compliance.

Market access certificates — IEC 62133, CE where applicable, and whatever the destination market requires for the specific application — follow the same logic. They are applied to a configuration, not to a company. A supplier who offers “our existing report” for a different cell or a different board is offering a document that will not survive an audit, and the cost of discovering that at customs is borne by the importer, not the exporter.

Stage 7 — From pilot to volume: five breakpoints

The gap between a good sample and a good shipment is not mysterious. It narrows to five specific points, and each can be checked with a document rather than a promise.

Cell batch consistency. Ask for the grading data — capacity, internal resistance and open-circuit voltage spread — for the batch that built your units, not a generic specification. A pack built from unmatched cells passes incoming inspection and fails in the field.
Weld and joint stability. Spot or laser parameters should be recorded per batch, and pull-test results retained. A joint that is merely adequate shows up as rising internal resistance after a few hundred cycles.
Component substitution without notice. The most common silent change is a BMS component replaced during a shortage. Ask for a written change-notification clause covering the board, the cell and the connector — and confirm it triggers re-validation, not just an email.
Aging and screening actually performed. Full charge-discharge cycling on every unit is different from sampling a few. The two are indistinguishable on a datasheet and easy to distinguish in a factory visit.
Packaging and declaration consistency. The declared pack must match the shipped pack in mass, cell type and configuration, because the transport documents describe the article, not the product family.

Each of the five is a question a buyer can ask in a single email, and each answer is either a document or an evasion. That makes them a more efficient audit than a general capability questionnaire.

What stays yours

A custom pack programme creates assets: drawings, a BOM, tooling, a BMS configuration, and often a firmware build. Ownership of those assets determines your switching cost, and switching cost is the difference between a supplier relationship and a dependency.

The practical position is that the design and the tooling belong to the buyer who paid for them, and that this is stated in writing before the first invoice rather than after a disagreement. The commercial detail of how we handle that is on the OEM process page under “your design stays yours”. What is worth adding here is the engineering corollary: a drawing you own is only useful if it is complete enough to be re-quoted elsewhere, which means tolerances, material specifications, weld schedules and the BMS configuration file — not a 3D preview.

Who this fits, and who it does not

A turnkey custom programme tends to suit

Buyers with a defined end product and a volume that can absorb non-recurring engineering cost across the first year or two.
Teams without in-house battery engineering, who need one technical counterpart rather than four.
Applications where a pack failure is a warranty event with a named party, such as mobility, robotics or stationary storage.
Programmes where the pack must evolve — capacity changes, a new connector, a second market — and the change has to be validated rather than improvised.

It tends not to suit

Buyers who need a standard product shipped this week. A catalogue pack, including the standard housings we list, is cheaper and faster, and customising it adds cost without adding capability.
Volumes so low that engineering and tooling cannot be amortised. A 20-unit programme is usually better served by adapting an existing platform than by starting from a drawing.
Buyers whose procurement rule requires the pack to be built on a line they own or directly audit, with no integrator in between. That is a legitimate requirement, and it is incompatible with the single-scope model described here — better to know at the outset.
Programmes where the battery is an afterthought to a design already frozen, and no envelope change is possible. The engineering value of this model comes from influencing the cavity; if the cavity is fixed and too small, no process fixes chemistry.

How a project starts

A one-page specification is enough to begin. In practice the fastest start is a message containing: the application, nominal voltage, target capacity or runtime, maximum envelope with a drawing or photograph, continuous and peak current, connector details or a photo of the existing pack, communication protocol if any, operating environment, and expected annual volume.

From there the sequence is feasibility and cell selection, BMS and communication definition, mechanical design, quotation at your volumes, then a sample built and bench tested — typically around two weeks. Standard housings require no tooling; a bespoke enclosure is quoted separately and only when the application needs one. Prototype builds start at three units.

If half of that list is still undecided, send the half you have. The specification gets sharper through the engineering conversation, and a project that waits for a complete brief usually waits longer than one that starts with a photograph and a motor rating.

Frequently asked questions

Does turnkey mean you manufacture every component yourselves?

Very few suppliers, if any. Cells come from cell manufacturers, BMS boards from electronics suppliers, enclosures from fabricators. What is consolidated is the design authority, the quality record and the warranty responsibility for the assembled pack. The test of the claim is not a factory tour but a single answer to “who handles the claim if this pack fails”.

Do we have to pay for tooling to get a custom pack?

Only if the enclosure is bespoke. Standard housings need no tooling, and many custom projects are a custom cell configuration, BMS and connector inside a standard shell. Tooling becomes unavoidable when the available cavity cannot be served by any standard form, and it is quoted as a separate line so the decision stays visible.

What is the minimum order quantity?

Three units for prototype and sample builds. Volume MOQ depends on the cell, the enclosure and the certification scope, and it is stated with the quotation rather than quoted as a general rule.

Can you build to an existing design we already own?

Yes — this is a build-to-print engagement rather than a development one, and it is usually faster. It requires a complete drawing set: tolerances, materials, cell specification, weld schedule, BMS configuration and connector pin-out. A 3D model alone is not a build-to-print package, because it does not define the things that determine whether two units behave the same.

Can we use your existing UN 38.3 report?

A report covers a specific tested configuration. If your pack uses the same cell model, the same BMS and the same structural design as the tested specimen, the coverage may extend to it. If any of those differ, a new test is required, and a supplier who says otherwise is handing you a document that may not survive scrutiny. This is checked case by case before quotation.

How long does a turnkey project take end to end?

Sample build and bench testing run about two weeks on a typical project. The full path to volume adds design iterations, tooling if the enclosure is bespoke, and certification, all of which vary with the application. Rather than quote a single number, the honest approach is a dated milestone list with the certification step shown explicitly — that is usually where schedules slip, and it slips for reasons that are visible in advance.

Start a custom pack project

Send the application, the envelope and the current requirements — a photograph of the pack you are replacing is often enough to begin. You will get a feasibility view, a cell and architecture recommendation, and a quotation at your volume, with tooling and certification shown as separate lines rather than buried in the unit price.

Contact DIDIDEDE to start, or read the OEM process and MOQ details first if you need the commercial mechanics.

Turnkey Custom Battery Pack Development: From Specification to Volume Production