Home

DO-178C Software Certification: An Engineer’s Guide

DO-178C Software Certification: An Engineer’s Guide

// DCM Aeronautique journal

Whenever a new client asks us whether their airborne code has to meet DO-178C, our answer is almost always yes. Regulators such as the FAA and EASA use this document as the reference for judging airborne software. Strictly speaking, it is not a certificate in itself. It is a structured body of evidence that supports the approval of the equipment or aircraft the software runs inside. If you are approaching DO-178C software certification for the first time, this guide walks through the areas that tend to catch aerospace engineers off guard, drawing on the projects our team has supported over the years.

Why Avionics Software Falls Under DO-178C from the Start

Many small teams begin by asking whether the standard applies at all. A more useful question is what role the software plays inside the flight system. Code that only displays information to the crew, with no bearing on control, carries a much lighter level of rigour. Code that takes part in flight control changes everything, and the documentation load grows with it.

The standard reaches equipment manufacturers, system integrators and airborne software suppliers alike. If you deliver a hardware unit with embedded code, you are inside the scope. Several of our clients were surprised by this. They saw themselves simply as the people who wrote the program, not as the organisation answerable for the whole airborne element. Once the boundary is clear, the rest of the planning becomes much easier to reason about.

What the Standard Asks You to Prove

What the Standard Asks You to Prove
What the Standard Asks You to Prove

DO-178C is a guidance document. It tells you which activities must be carried out and which records must exist, but it does not dictate a single design method. Model-based development, hand-written code or a blend of both are all acceptable. What matters is that you can demonstrate the software does what it must, and nothing it must not.

Three companion documents are worth knowing about early. DO-330 covers tool qualification. DO-331, DO-332 and DO-333 are supplements for specific techniques: model-based development, object-oriented programming and formal methods. You do not need to master them on day one. Knowing they exist is enough to avoid unpleasant surprises later in the project.

Design Assurance Levels from A to E

Design Assurance Levels from A to E
Design Assurance Levels from A to E

This is the concept that confuses newcomers the most. Every software function receives a Design Assurance Level, or DAL, ranging from A to E. Level A is reserved for failure conditions classed as catastrophic, meaning a failure could prevent a safe landing. Level E applies where a failure has no safety effect whatsoever.

Assignment begins with a failure condition assessment at system level. The question is simple: if this function fails or returns a wrong value, what happens to the aircraft? The answer determines the level. A higher DAL brings more objectives to satisfy and stricter independence requirements during verification. Cost and schedule grow along with it, which is why the assignment deserves careful thought rather than a quick guess.

How the Objectives Change Between Levels

How the Objectives Change Between Levels
How the Objectives Change Between Levels

Each DAL carries its own set of objectives, and the gap between levels is considerable. The table below gives a rough picture of how the reference objective count and independence demands shift.

DAL Failure effect Reference objective count Independence requirement
A Catastrophic 66 Most objectives must be independent
B Hazardous 65 Some objectives must be independent
C Major 65 A few objectives are independent
D Minor 28 Almost no independence required
E No effect 0 None required

These figures come from the objective tables in the standard. When you plan a real project, verify them against the full text, because the way objectives are grouped can change how you count them. Our planning sheets are always cross-checked against the source before any estimate reaches a client.

At DAL A, structural coverage is the most demanding obligation. You must show that each statement, decision and condition has been exercised, and for most DAL A code that means MC/DC coverage. It reads like an academic exercise, yet this is where a large share of the effort lands. On one project, our engineers spent close to two months rewriting the test cases for a single small module because MC/DC coverage failed to close on the first pass. The lesson we took from that was to design for testability before the code was written.

The Artefacts an Auditor Will Ask For

The Artefacts an Auditor Will Ask For
The Artefacts an Auditor Will Ask For

Artefacts are the documents and records the certification authority will review. The list looks intimidating, but it follows a logical chain: a plan, then requirements, design, code and tests, and finally a summary showing that everything agrees with everything else.

The first key document is the PSAC, the Plan for Software Aspects of Certification. In practice it is an agreement between you and the authority on how the standard will be met. Teams that draft it carelessly often spend the following year revising it, so invest real time at the outset. The Software Development Plan follows, describing the process you actually run rather than the one that looks good on paper.

Requirements, design and code must support bidirectional traceability. Every requirement should link forward to its design, code and tests, and backward again. This is the gap auditors find most often, and we return to how to close it below.

The configuration index and the Software Accomplishment Summary record what the software achieved against the plan. For a small team taking on its first project, a practical starting checklist looks like this:

  • Has the PSAC been agreed with the authority?
  • Does every requirement carry a unique identifier?
  • Does every test trace back to at least one requirement?
  • Are baselines locked at each milestone?

Tool Qualification Under DO-330

Tool Qualification Under DO-330
Tool Qualification Under DO-330

Many teams leave this until someone raises the question of the compiler. The logic is direct. If a tool could insert an error into the final product, you must show that the tool can be trusted. If it only detects errors and cannot create them, the obligations are considerably lighter.

The standard defines Tool Qualification Levels from TQL-1 to TQL-5. TQL-1 is the strictest and applies to code-generating tools used at DAL A. Full qualification is not always necessary. Checking tool output manually or through independent testing can sometimes be enough, and for a small team that is often the more economical route.

Cost deserves a candid word. For a small team in Vietnam or the wider region, requalifying a complete tool chain can take months and eat into the budget. Choose tools with qualification in mind from day one, rather than discovering the problem late and being forced to switch midway through a project. A tool that looks cheap at the start can become the most expensive decision in the whole programme.

Recurring Audit Findings and How to Prevent Them

Audit findings repeat in a handful of familiar patterns. That is encouraging, because patterns can be prevented. The most frequent is a broken chain of traceability between requirements, code and tests. Small teams often build the trace matrix at the very end, when details have faded from memory and links go missing. Updating traces as you work is far more reliable, even if it feels slow at the time.

A second problem is weak review evidence for derived requirements. These are requirements that do not come straight from the high-level system but emerge from the design itself. They need a safety assessment and full verification. Without both, auditors will ask pointed questions, and the answers will not always satisfy them. We have seen projects where a single unexamined derived requirement led to weeks of rework near the end of the schedule.

A third is poor configuration management. Without clearly locked baselines, you cannot state exactly which version was tested. We have watched a team rerun an entire test campaign because nobody could confirm which build was which.

Before a certification authority or a DER reviews your work, run an internal audit yourself. Ask someone who did not write the code to act as the auditor, pick a few requirements at random and trace each one from start to finish. If a requirement cannot be traced within ten minutes, something in the process needs attention. That single exercise has caught more problems for our teams than any other practice.

Planning a DO-178C Project in Practice

Successful projects start with correct DAL assignment and a clearly bounded plan. Resist the urge to begin coding straight away. Time spent establishing the DAL for each function pays back many times over, because a wrong assignment can force you to redo large sections of work. A function that looks harmless in a diagram may turn out to share memory or a processing thread with something critical, and that discovery is far cheaper on paper than in the lab.

Next, select a development process that suits the assurance level. A high DAL calls for a tighter process and more evidence, while a lower DAL leaves more room to move. There is little sense in applying a DAL A process to a simple display function, and equally little sense in running a loose process on flight control logic.

Finally, the budget must cover review cycles and internal audits. Many teams estimate only coding and testing time, then get a shock when document reviews are added to the schedule. A realistic plan sets aside substantial time for preparing evidence alone, and it is wise to treat that time as part of the engineering work rather than an overhead to be squeezed.

Where DO-178C Meets Fly-by-Wire in the Real World

Where DO-178C Meets Fly-by-Wire in the Real World
Where DO-178C Meets Fly-by-Wire in the Real World

Fly-by-wire shows most clearly how software classification shapes design. When electrical signals replace mechanical linkages in flight control, the software becomes part of the control path itself. A fault there can affect safety directly, so a high DAL is unavoidable. For a deeper look at how these architectures are assembled, see our piece on Fly-by-Wire Systems in Commercial Aircraft.

One lesson from our projects stands out above the rest: classify functions early. If a low-criticality function shares memory or a processing thread with a safety-critical one, the entire software element may be pulled up to a higher DAL. Separating the architecture from the outset can save a great deal of money. We gathered more practical notes on this in Top 7 Things to Know About Avionics for Engineers, which covers many of the integration questions that arise alongside certification work.

Teams working with precision hardware should also keep the software and mechanical sides in step. Interface control documents, timing budgets and sensor tolerances all feed back into software requirements, and a mismatch found late is among the costliest problems a programme can face.

Questions Engineers Ask Us Most

Questions Engineers Ask Us Most
Questions Engineers Ask Us Most

Does DO-178C apply to every piece of aircraft software?

Not every program on board is treated identically. Software that forms part of approved airborne equipment, however, is usually expected to follow this standard or an equivalent the authority accepts. Confirm your particular case with the authority or your customer before you commit resources, because assumptions made at the start are hard to unwind later.

Can a small team reach Level A without a large budget?

It is possible, though demanding. Small teams tend to succeed when they limit scope, keep the architecture simple and reuse tools that already carry qualification evidence. DAL A costs remain high, so question whether every function genuinely needs that level. Often a smarter partitioning of the architecture reduces the number of functions that must be treated at the highest level.

Do I need DO-330 if I use an open-source compiler?

The answer depends on the compiler’s role in your process. If it could introduce an error into the final product, you will generally need to demonstrate its reliability under DO-330 or verify its output independently. Open-source tools do not arrive pre-qualified, so do not assume they do.

How long does certification usually take?

It varies widely with project size, DAL and the maturity of your existing processes. A small, low-DAL project may take a few months to a year, while a large DAL A program can stretch over several years. Starting planning early keeps you from waiting on evidence you have not yet produced.

How does DO-178C differ from ISO 26262?

DO-178C addresses airborne software, whereas ISO 26262 addresses functional safety in road vehicles. Both rely on criticality levels and matching evidence, but their regulatory frameworks, terminology and certifying bodies differ. Treating one as a copy of the other leads to mistakes, so it is worth learning the specific expectations of each domain before borrowing practices across them.

If you take away only one thing from this article, make it the DAL assignment for each software function. That decision shapes almost everything downstream, from cost and schedule to how testing is organised. Getting it wrong at this stage is usually far more expensive than any single technical error, and it is the step where experienced teams spend the most care.

Got a technical question?

Corrections, topic suggestions and questions about an article all reach the same editorial inbox, and every message is read.