Fligt Testing: Saw-tooth Climbs and Descents

Today was a maintenance day, so I changed the oil on the plane and generally poked and prodded to make sure everything was ok.  Tomorrow I'll be doing longitudinal, lateral, and spiral stability.

Meanwhile I started crunching the data from the previous two flights, which were saw-tooth climbs and descents.  Again following the protocol the climbs and descents are paired in airspeed so that it can be flown as a climb, then push over the top, pull the power to idle, and descend at the same airspeed.  The test order is randomized to minimize error caused by up or down drafts changing throughout the test, or weight loss or gain from burning fuel or refueling from creating a systematic error in the data.  The flight was broken into two, where I refueled in between.  Total flight time for these tests was something close to 3 hours.  

Again I am somewhere between mediocre and awful at holding my airspeed constant, but the data seems to have saved me, mostly.  I parsed most of it manually, but I think I can build in some automation into my spreadsheets to do it automatically.  Generally I chose 7500 ft DA as the crossing altitude for collecting data.  I picked up points from 7000 to 8000 and averaged the airspeed in that window as well as the vertical speed. I have data for most of them starting much lower and going much higher, so I could generate similar curves for different altitudes.

For the climb data, I got this:

Basically it is a very flat response curve, with some noise in the data.  Generally you couldn't go wrong climbing anywhere from 85 KIAS to something like 115 KIAS.  It looks like I have Vy climb at 90 KIAS, but I should probably repeat the data set between 75 and 95 to make certain that curve is correct. I have not corrected my pitot error, so these numbers are slightly high, but they generally agree with the published climb performance.  

If anything my climb rate is a little low.  I attribute that to the lack of wheel pants as well as the fact that I appear to be over-propped a bit.  WOT and best-power mixture generally were giving me about 2500 RPM at 7500 ft DA.  I'm going to leave the prop for now, and see how it does once the wheel pants are on.  My engine should be putting out a bit  more power than a stock 180HP parallel-valve O-360, but my Lightspeed prop is purported to be a 65.5" x 89".  The published data was done with a 3-blade Performance Prop that was 64"x76".  

The descent data looked very similar, although I think maybe I was better at holding a fixed airspeed on the descents.  Maybe because I wasn't worried about looking at CHT and oil temperatures on the way down.  It looked like this:

Here it is, again, pretty much by the published numbers.  It looks like I need to do some descents all the way down at stall speed to see if I can actually see the inflection point.  I'll be interested to see if these curves move up a bit once I have the wheel pants on it.  I suppose first I'd have to make a set of them...

Pitot-Static Calibration

Flight testing continues, and I was finally able to sit down and do a data reduction of my pitot-static calibration flights.  I am using my EFIS (a Garmin G3X) to record data for me once per second.  I fly roughly cardinal headings in a box at a target airspeed, and the whole flight was done at 3,500 pressure altitude (density altitude was closer to 4500).  I found I was piss poor at holding an altitude, heading, and airspeed all at once.  Actually, I don't think I managed to hold any of them even one at a time. With the magic of the dense data recording from the EFIS and a spreadsheet, it basically didn't matter, as long as I wasn't really bad at it.

After the flight I was able to sort and eliminate data.  If the 3 data points ahead and behind of a point were not in a heading range of less than 3 degrees, an airspeed range of 3 KTAS, or a vertical speed of less than 100 FPM, the data point was thrown out.  So, each point I was sure that I had 7 seconds of relatively steady data.  Then I sorted the data by indicated airspeed, so similar speeds on different headings could be used to do the calibration.  I used a modified version of this spreadsheet to do all of the trigonometry for me.

I was amazed that taking data points that differed sometimes by seconds and sometimes by nearly the duration of the flight the resultant wind vector was very nearly constant.  The range was 237 to 253 degrees from 12.4 to 16.9 knots.  The reported wind for the day was 280/13G17.  The G3X records magnetic heading without correction for magnetic variation (that is a separate data field) which is -14, which would put the reported wind and the calculated wind very close.  During the flight the G3X calculated a wind vector, which it does by doing vector math with the magnetic heading, the ground track, GPS ground speed, and TAS.  The G3X reported wind was 243 degrees at 16.9 knots.

What I was left with is this:

Basically my pitot-static & OAT system are optimistic.  Starting at 85 KTAS it is indicating 2 knots high (meaning the G3X says 85, but really it is flying at 83), and out at 183 KTAS it is 13 knots high (meaning it says 183, and really it is going 170).

My aim being to minimize the error, I think I have a few things I could try.  First, this could be an error of the OAT probe.  If it was off by a few degrees then the TAS calculation would similarly be off.  I have two OAT probes, and could compare the two, or I could get a third temperature probe of known calibration and try that.  I could also re-do the data analysis, calculating TAS from IAS and OAT, instead of taking the TAS number from the G3X, then fudge the OAT to see how much error would be required to create this.  That is a simple exercise and maybe I'll do that.

But the fact that the error grows over the scale (3% at the low, 7% at the high) indicates to me that it is an error of the static system.  My static ports are not per plans.  I don't seem to have a good picture of them, this is about the best I have.  Basically they are aluminum, roughly 1" round, and create a step that is about 0.080" from the fuselage, with a chamfered edge.  There are two, one on either side of the fuselage, and they are T'ed together.  Because the airspeed indicator is  indicating too high, I tend to believe the static port is measuring a pressure that is too low.  To correct it I am going to try adding layers of tape behind the actual port, in an effort to raise the measured static pressure and reduce the error.  If that were to work I would remove the static probes and machine a similar sized step into them.  

Stay tuned for the next episode of Knuckleheads Pretending to be Flight Test Pilots, where your intrepid pilot/engineer/dufus seeks to determine if flutter testing should be in KIAS or KTAS, right here on the Cozy Forum...

Modified Hendricks Manufacturing Canopy Latch

I have been messing with our canopy latch for WAY too long, and now that it is installed and functional, I thought I'd share the result.  I bought a Hendricks Manufacturing canopy latch.  It is a right side version, installed upside-down.  

I very much disliked the fact that there is no real hard stop or over-center that keeps the latch closed.  So, I completely re-engineered the internal handle to provide these functions.  I changed the overall range from 90 degrees to 115 degrees to accommodate the geometry of the stock Cozy latch hardware.  You can see photos of the external installation, the internal latch handle with a 3D printed prototype, and the SolidWorks model of it here:

http://tinyurl.com/CozyModifiedCanopyLatch

So, now I have it finally working, after one plastic prototype and three aluminum CNC'ed versions.  I posted a short video of it in action here.  Evidently it was really cold in the hangar, my hands turned blue...

It was a lot of work, but now that it is done, I do like how it works.  It is very clean on the outside, and on the inside it is REALLY obvious that it is open or closed.  I still have to install the key locking plate and the canopy warning switch, but for now it is off to the anodizer.  

Moving from Petaluma to San Carlos

After waiting nearly six years on the hangar waiting list, a spot finally came up at San Carlos airport.  The rules of the list are once you're offered a spot you have 30 days to get a long list of documentation to the airport manager, and have a plane in the hangar.  That meant that despite the fact that the plane wasn't quite ready to fly to the new home, it would have to get there somehow.  

After researching how folks had moved their projects, looking at a few builder websites, and considering my options, I chose to move it "flat" on a truck, rather than a trailer.  The main spar of the cozy is roughly 11' 6" wide with the strakes in place, which is wider than legal pretty much everywhere.  Your choices are to get the wide load permit, rotate the whole thing about 45 degrees to meet the width requirement, or point it vertically with the spar running longitudinally on the axis of the truck/trailer.  I researched each option and found that the easiest way was to get the wide load permit.  In California (and probably most other states) a wide load permit for under 12 feet can be had with little hassle, and no need for a pilot car.  I applied for the permit by fax (what decade is this?!) to CalTrans, and with only a minor amount of back and forth I was granted a permit for the sum total of $16.  

A local truck rental place has a cab-over 16' stake-bed truck with a lift gate for $50/day plus $0.30 per mile and fuel.  The lift gate turned out to make this a very EZ task, and I would highly recommend finding one if you need to move a plane.  The truck vs trailer decision was made because the truck would get the strake way up high, over the top of even the bigger SUVs on the road.  If I had a trailer with the strakes at car level I would have been paranoid of someone not paying attention and hitting it.  As it was I stayed to the right side of the lane, especially when big trucks were passing by, and there were no clearance issues.  

The permit conditions specified which roads I could drive on and when.  This turned what would be 65 mile trip into a 120 mile adventure.  I was required to have wide load signs on the front and rear of the truck.  Those were easily sourced on ebay for $30.  

We started early in the morning, figuring the first we could possibly be on the road was 9 AM per the permit curfew.  We set the lift gate roughly halfway down, and used two motorcycle ramps to roll the plane up to the lift gate.  Once it was on the lift gate, we chocked the tires, lifted the gate the rest of the way, and rolled the plane into place.  Literally 5 minutes after we started this process the plane was on the truck.  I was amazed.  The wheels were chocked by 2x4s screwed to the deck of the truck.  We ratchet strapped the axles down, and put a 3" wide ratchet strap over the top of the canard mount to hold the nose.  All ratchet straps were covered with towels and corners were covered with cardboard to protect the plane.  It ended up taking us almost exactly 3 hours to get everything on the truck and secured to my satisfaction.

The drive to San Carlos was completely uneventful.  I had my friend drive my car behind just to keep an eye out for anything shaking loose, and it took us the better part of 2 1/2 hours to make our way down to San Carlos.  Unloading was the reverse of loading, with the exception of the fact that we figured out that if you were willing to let the plane's nose get pretty high (which meant I was holding a fair amount of weight down to keep it from falling on my brand-new rebuilt engine!) that we could actually lower the lift gate right to the ground and skip the loading ramps altogether.

I am sure there are a million ways to move a plane, but this one turned out to be really EZ.