Getting It Right...

Short Brothers in Belfast had an extraordinary history for such a relatively small aircraft manufacturer. Between 1940 and 1980, they designed and built over 25 different aircraft types, a high proportion of which were research aircraft for government contracts. And there were some really innovative designs, too – like the world’s first fixed-wing flat-riser VTOL jet, for example……

One of the relics of that era was the instrumentation system that we used in my time at Shorts on the older test aircraft. This was a fifties-vintage photographic trace recorder, using a hundred feet of one-foot-wide photographic paper on every run, which was wet-developed after every flight. The processing costs for this kit must have kept Kodak in business for years!

The data was recorded onto the paper by a range of light points, powered by rotating galvanometers, each individually calibrated, all assigned a separate operational range on the available one-foot width of paper. It was a tedious and specialist job to set all this up, and each change of configuration required a full manual re-calibration. The nominal maximum data recording available was twenty-four parameters, but that only worked if all parameters were stable or moved in lock-step.

I initially specialised in aircraft handling qualities, so squiggles were my thing – which meant a lot of lines moving all over the place on the record. In reality, we could only set the system up to use twelve parameters at a time for my flights, since any more than that and the data traces at my test point would converge into a chaotic mess. Specifying which twelve really focused the mind as the effort required to set it up was pretty high. Specify the wrong twelve and you’d earn an earful from the instrumentation guys who lived in “The Huts” for the wasted effort. You quickly learned to Get It Right.

The recorder was started by the guys prior to flight (usually after a bit of fettling), and ran consistently throughout – we were never sure if it would restart after being turned off in flight, so we didn’t. To conserve paper, it was run at slow speed for transit flying, and set to high speed for the duration of the test. If you missed setting “fast” for the test, it usually required a re-flight as the condensed data traces were just too difficult to read. And the test points were identified manually by a personalised morse code blip on a digital event button. It really was voodoo to set it up and make it operational – and the Instrumentation Group was a real bunch of artisans in keeping it running. Failures were surprisingly infrequent, given the age of the system – testimony to the skill, talent and dedication (or desperation, depending on your point of view) used to keep it running.

Once the roll of paper was processed and delivered after the flight, the real fun started. After getting past the fading smell of the photochemical developer, the drum of paper was loaded onto a roller, placed at one end of a board and tacked onto another roller at the other end. We then manually wound our way through up to a hundred feet of squiggly lines on paper, looking for the occasional blip where the test data was recorded on “fast”. Once you had the test point, you then had all twelve parameter time histories from the flight test point laid out in front of you for assessment – the unvarnished truth for you to exploit. You selected your data line of interest, and you measured the height of each of the parameters on the trace from the datum at the bottom of the paper. If the datum wasn’t there – it was a little quirky on one of the recorders – then the flight was worthless. Once you had the measurement, you then had to look up the manually-drawn calibration chart and find out what the value was. For each point. On each parameter. For every test.

It was a fabulous education on the difference between information and data.

After a couple of sessions grinding through a flight, you rapidly developed your own appreciation for the interaction of the parameters during a test. Which, in turn, meant that you could quickly identify - just from the patterns - what was good, bad or indifferent. Which meant you were able to filter out the rubbish test points quickly. Which meant that the measured data became a quantitative justification for the analogue pattern presented.

It was an awful lot of effort for just twelve-odd parameters. But I became surprisingly good at interpreting these squiggles quickly and identifying and extracting the necessary information rapidly. Once you had a technique and recognised the patterns, it wasn’t really as painful as it sounds. Tedious, yes - painful, no.

I had five years and three flight test programmes to perfect this art. We certified aeroplanes very effectively with this system. It was solid, it was repeatable, it was reliable (sort of) and it worked. Which is why we were still using it after all this time. A lot of people trusted it – and the guys in the “The Huts” to keep it working.

Ten years later, it took an excellent and insightful flying instructor in Wichita to educate me how to read aircraft instruments in a cockpit properly, and stop trying to visualise the traces I thought they were generating. I finally put this old habit to bed, and learned not to read the value the instrument needle was pointed at (which I’d been doing for my entire VFR experience) but to start reading what the instrument was really saying. And it worked. Herb, thank you!

To this day, I still view the engineering world in primarily time histories. They speak volumes. Properly correlated, they can transmit huge amounts of information instantly through a small image. Poorly correlated and they can be mischievously misleading. Sifting information from any data set really is an artform, occasionally requiring a high degree of integrity to tell the correct story with the right quality and resist the temptation to fudge. And quality of information beats quantity of data every time.

Digital instrumentation on my subsequent programmes delivered the vastly more data than the squiggles on those vintage traces – to three decimal places, and a lot more efficiently, if not always cost-effectively.

Same information, though. But instantly visible without the noise.

More doesn't necessarily mean better.