Sizing an airborne acquisition system: a worked example

Four budgets, and they are coupled

Sizing an acquisition system means closing four budgets simultaneously: the bit rate the equipment must sustain, the storage the sortie will consume, the fraction that fits down the telemetry link, and the mass and power the aircraft will accept. None of them is independent, which is why sizing is a loop rather than a calculation.

The single most useful habit is to write the numbers down in that order, in a spreadsheet nobody is ashamed of, and to keep it as a configuration item for the whole campaign. Most sizing failures are not arithmetic errors. They are a number that changed in week three and was never propagated.

Start from the parameter list

The input to sizing is a list of what you intend to measure, with a rate and a reason for each line. Not a channel count, not a product, not a comparison of datasheets: a list. It is tedious to produce and it is the only document that makes the rest of the exercise honest.

Two disciplines pay off immediately. Give each parameter its own rate rather than a group rate — sixty-four temperatures do not need the rate of your fastest accelerometer, and giving it to them can double a bit rate for nothing. And write the reason: "structural fatigue correlation" or "engine surge onset" tells you what rate the parameter actually needs, whereas a bare name invites the safest guess, which is always the fastest.

From channels to a bit rate

The arithmetic per group is a multiplication: channels × rate × bits per sample. Take a mid-size structural and handling campaign as the worked example.

Two things jump out of that table, and both are typical. The analog and discrete channels — the ones the instrumentation engineer spent three weeks defining — account for less than five percent of the bit rate. And the single event camera in intra-only mode is two thirds of the total on its own.

Storage: duration times rate, times honesty

A hundred and twenty-five megabits per second is about 15.6 megabytes per second. A three-hour sortie is therefore some 170 GB. That is the easy part; the margin is where judgement enters.

• The sortie that runs long. Ferry legs, a hold, a second attempt at the point — thirty percent is not generous.
• The sortie you have to repeat, recorded on the same medium because nobody downloaded between them.
• Room to keep the medium comfortably below full, because a nearly full flash device behaves worse than an empty one.

Multiply through and 170 GB becomes a 512 GB cartridge, or a terabyte if the operating concept involves two sorties between downloads. Storage is the cheapest of the four budgets, and the one where buying margin costs least — which makes it the wrong place to be clever.

The telemetry link: designing what does not go down

A range allocation might give you five megabits per second. You have a hundred and twenty-five. So the telemetered set is not a filtered version of the recording, it is a second format, designed on purpose, to answer one question: what does the flight-test engineer need to see in order to judge this point and call the next one?

One point eight megabits per second inside a five megabit allocation leaves real margin, which is the point: the link budget wants headroom, not a full pipe. And notice what the exercise forced — a decision, parameter by parameter, about what is worth watching live. That decision is worth making deliberately rather than by whatever the default configuration happened to contain.

Modules, and the moment the answer changes

Now count hardware. Two hundred and forty-four analog and discrete channels, four 1553 buses, sixteen ARINC 429 buses, four video channels and two audio channels. With analog modules of sixteen channels, discrete modules of sixty-four, bus modules of four to eight channels and video modules of two, the list comes out somewhere around eighteen modules.

A stackable platform typically accepts up to sixteen modules in one unit. So the honest output of the sizing exercise is not "one recorder" — it is two units, and therefore an architecture question that nobody asked at the start: where do they go, and how do they share a time reference?

Mass, cable and power: the budgets that surprise

Here is the calculation that changes designs. Two hundred and forty-four channels, most of them needing a shielded twisted pair and some of them four wires, averaging eight metres of run: call it fourteen hundred metres of cable. At roughly ten grams per metre, that is fourteen kilograms of wiring — several times the mass of the acquisition units it connects to.

Nobody signs off an instrumentation installation on the mass of the boxes. They sign it off on the mass of the installation, and the harness is the installation. Which is why the sizing loop and the architecture question are the same conversation, and why shortening analog runs is worth real money.

Power is more forgiving but not free: a stack of modules and a recorder lands in the tens of watts, which is a supply, a circuit breaker and a heat path. And heat is the term that gets forgotten — an enclosure that dissipates forty watts in an unventilated bay at fifty degrees ambient is a thermal calculation, not an afterthought.

The sizing mistakes that cost the most

• Sizing from the catalogue. It gives you a configuration you can buy and no idea whether it answers the question.
• One rate for a whole group. It is the fastest way to double a bit rate without measuring anything new.
• Forgetting the overheads, then discovering the frame does not close at the rate you promised.
• Sizing storage for the nominal sortie. The one that matters is the repeated one, on the same cartridge.
• Treating the telemetered set as a checkbox rather than a designed format — and then watching the wrong parameters live.
• Leaving cable mass out of the budget until the installation review.
• And no reserve. Parameters are added in week three of every campaign that has ever flown. Size for it.

The output of a good sizing exercise is not a product reference. It is a table of four closed budgets, each with its margin written next to it, and a list of the decisions that were made to close them. That document survives the campaign and answers the question everybody asks six months later: why is it configured like this?