I'm now wondering how the old Perf charts were created.
Further to djpil's comments ..
In days of olde .. when we had local rules (ie pre-Yates Report changes), one could either do one's own thing (ie develop equations or use the standard FT equations) or use the DCA equations (presented in an in-house document to which Industry consultants had access). Needless to say, most folks used the DCA document as that avoided much discussion in the approval process.
The DCA document (I have a copy but can't recall the ID without digging in the archives for it ..) used some pretty simple equations but, in practice, they were of sufficient accuracy that the P-charts were a reasonably good mix of a little conservatism and accuracy.
Later on, the document was revised the better to take account of turboprop performance .. and that document was a different beast altogether .. so, we ended up using the old stuff for pistons and the new stuff for turboprops.
The equations allowed for the interpolation (and limited extrapolation) of whatever data set had been captured .. not too much in the way of extrapolation as that involved a leap of faith .. interpolation was much more defensible. So far as test programs were concerned, we normally tried to get low/high and hot/cold within the realms of reasonable possibility and take it from there. Obviously, the more test data, the more accurate the end charts could be. What variable range ended up in the charts depended a lot on what range of test data could be found. This is no different to the problems faced by Mr Boeing and others ..
At the high end of the spectrum, for data capture, one could do cinetheodolite performance takeoffs and landings which provided a movie tracking record of the aircraft (allowing corrections for camera unit mistracking) and azimuth/elevation data for the camera axes (paper trace .. great fun working with that stuff).
Accurate survey data was required for the runway and camera unit position to do the geometry sums.
From these records, one could spend copious numbers of (relatively) productive hours developing a very accurate time history of the takeoff/landing distance and air distance/height. This allowed speed and acceleration data to be determined to a very good accuracy if you had a need for that sort of stuff.
Unfortunately, back in the 60s/70s, this equipment was pretty high tech (and expensive) so only the military and the likes of GAF were able to use it. Djpil and I spent numerous hours in the Fishermens Bend sweatshop doing this stuff after Nomad flight trials. Also, I had fun doing similar stuff at ARDU for a while.
Once we get out into the GA world, though, no-one was likely to have this kit sitting around waiting for some trials so the local industry, with DCA, developed a much simpler exercise using a reasonable quality still camera to take a series of exposures to get the important sections of the takeoff and landing. Not quite as accurate as the CT .. but much quicker and more than fit for purpose.
Some folk, to save on dollars, did the exercise with constant (gross weight) speeds throughout (reduced the number of flight trials required) and that produced some strange results in the P-charts which were the occasional subject of bar room discussions. For example, one saw landing P-charts where the distance increased with reducing weight (due to float at the end of the landing flare)
On occasions, if the OEM data in the POH appeared defensible, one would negotiate with the DCA gurus of the day and, sometimes, take a few points from the OEM POH data and use those to develop the P-charts.
DCA had a nice little HP plotter gadget to run up the charts .. most of us out in the Industry did it by hand .. same result... the typical P-chart presentation was dreamed up by DCA (Ron S, perhaps ?) and was just one of a range of possible presentations. It has its good and bad points, like any presentation. I did my share of GA P-charts and, I guess, so did djpil
Do keep in mind that this all predated the first of the microprocessor computers which came onto the market in the mid-70s - prior to that, if you could afford it, there was the odd TI bit of kit which allowed some number crunching capability. Needless to say, none of us had private access to the few mainframes around the place.
Spare a thought for us back then .. a computer (with a capability inferior to a modern pocket calculator, along with printers, cost me around $20k in the mid-70s but could we do things with that primitive kit compared to an old (still quite expensive) TI programmable calculator or slide rule. Would you believe I managed to upgrade the system several years later when a 5 kb hard drive came onto the market. (Maybe it was 5 mb .. whatever, it was tiny)