Testing, data collection, but poor data

Discussion in 'Tests and Improvements' started by Dana, Oct 27, 2019.

Help Support HomeBuiltAirplanes Forum by donating:

  1. Oct 27, 2019 #1

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    Yesterday I tested a new prop on my Hatz. The test is inconclusive, with some oddities (more on that later). The test was a series of climbs with first the old, and then the new prop. The flights were recorded with an old Garmin 60CSx which has a pressure sensor, so it's a lot more accurate than GPS altitude. That data was analyzed with "Trackreader", a windows program I wrote myself:

    upload_2019-10-27_15-41-25.png

    At the same time, a GoPro was looking at the instrument panel to record time, speed, altitude, rpm (note the tach display on the phone on the right):

    upload_2019-10-27_15-46-5.png

    Finally, I put the climb date into a spreadsheet. It was a bumpy day, so I really don't trust the numbers:

    upload_2019-10-27_15-48-9.png

    This is the weird part. Climb vs. airspeed should be a smooth curve with a single high point. The dip at 65 knots shouldn't be there, the strange part is that it's there in both tests. The data at 55kts is suspect, too. If I delete the 55 and 65 knot values, I get curves that, although probably not correct (Vy shold be closer to 55kts), look like what you would expect:

    upload_2019-10-27_15-53-43.png

    Clearly I need to do more testing under better conditions, though I probably won't put the original prop back on. Once I believe I have accurate data I expect to send the prop back for rework to have some pitch taken out (I was looking for more climb, hopefully without too much loss of top speed).
     
  2. Oct 27, 2019 #2

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    Update, added some data from earlier flights with the original prop. The green curve is what I'd expect:

    upload_2019-10-27_16-7-7.png
     
  3. Oct 27, 2019 #3

    Himat

    Himat

    Himat

    Well-Known Member

    Joined:
    May 5, 2011
    Messages:
    2,811
    Likes Received:
    653
    Location:
    Norway
    You have got a dip at 65mph, and a low reading at 50mph. To few measurement points to really tell, but to me it looks like there is a higher frequency curve imposed on the climb curve. If the climbs where performed at the same geographical position with the two props, but each of the six climbs I different positions I would say you did encounter a wave pattern.
     
  4. Oct 27, 2019 #4

    bifft

    bifft

    bifft

    Well-Known Member

    Joined:
    Apr 17, 2011
    Messages:
    243
    Likes Received:
    89
    Location:
    Utah
    Doing my climb data also ended up with a lot of noise, probably due to poor pilot technique in addition being on windy days in the mountains. I just took more flights and then did a least squares match to a poly curve.
     
  5. Oct 27, 2019 #5

    TFF

    TFF

    TFF

    Well-Known Member

    Joined:
    Apr 28, 2010
    Messages:
    11,727
    Likes Received:
    3,321
    Location:
    Memphis, TN
    Since you are trying to compare, I would want to eliminate as much surroundings as I could. The day of I would pick a landmark to be the start point, altitude, and use the same direction in the climb. Of course time of day along with temperatures will probably make a difference, but that’s something you have to extrapolate the details.

    I wonder if 65 is the crossover point where drag becomes notable. Below, not much penalty. Above becomes more work.
     
  6. Oct 28, 2019 #6

    Aerowerx

    Aerowerx

    Aerowerx

    Well-Known Member

    Joined:
    Dec 1, 2011
    Messages:
    5,084
    Likes Received:
    1,391
    Location:
    Marion, Ohio
    Are you using a spline fit in the spreadsheet? Not a good idea. You can get some wild excursions like that because it forces the curve to pass directly through your data points. And if the data is noisy...well you can see what happens.

    As bifft suggested, use a least squares fit.
     
  7. Oct 28, 2019 #7

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    I think the main problem was that it was a breezy day, even though I was out over the water it was still bumpy. As you can see, the green curve is much better, that was day when the wind was off the ocean and much smoother. I think it's just coincidence that they both have the dip at the same speed, I was flying in the same general area but not exactly, and several hours apart. I'm hoping to try again this evening, maybe it will be smoother.
     
    Pops and Basil like this.
  8. Oct 29, 2019 #8

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    Now I'm more confused. Today was a perfect day, nice and smooth, I expected more textbook results. The dark purple line is the new data (new prop only):
    upload_2019-10-28_20-55-7.png

    I added a climb at 45kts this time, on the edge of the stall buffet all the way. It looks a little better, but still not the nice curve I'd expect.

    About all I can conclude is that I've lost about 100-150fpm climb with the new prop, with a 5kt gain in cruise and top speed. Vy is somewhere between 50-55, and Vx is somewhere between 45-50.

    I also looked at glide with the engine idling, Here the data is better (again, blue is the first test, red is today):

    upload_2019-10-28_21-3-6.png

    Min sink is right at stall, which I expected, and best glide angle is
    around 52. This makes sense, as best glide angle should be equal or close to Vy, and min sink should be close to Vx.

    I know the spline fit isn't the best, but as far as I've figured out OpenOffice doesn't have a built in least squares trendline.

    The numbers more or less make sense, agreeing with my less formal testing, and are good enough for my day to day flying, but I was really hoping to get prettier curves for a planned magazine article. I suppose if I repeated each test several times and averaged them I'd get better numbers, but I don't know that I have the patience to do it
     
  9. Oct 29, 2019 #9

    Pops

    Pops

    Pops

    Well-Known Member

    Joined:
    Jan 1, 2013
    Messages:
    7,230
    Likes Received:
    6,117
    Location:
    USA.

    Also confusing to me. I did the same on the SSSC and got a nice line like one would expect with no dips. I did have perfect smooth air in the early morning for the test and like you gave myself a 500' of time to get the airspeed locked on for the 1000' that I was timing with a stop watch. Best ROC was at 42 knots @ 1250 fpm with the oil temps up to the normal 180-190 degs. With the VW engine you start loosing HP when the oil temp gets below 170 degs with the straight weight W30 oil that I use in the VW engine. The WOT rpm's will be about 50-75 rpm less with the colder oil. On a OAT of below 80 degs I always take a long warmup time to get the oil temps up before takeoff.

    This is interesting, since I will be doing the same test on the JMR before long.
     
  10. Oct 29, 2019 #10

    TFF

    TFF

    TFF

    Well-Known Member

    Joined:
    Apr 28, 2010
    Messages:
    11,727
    Likes Received:
    3,321
    Location:
    Memphis, TN
    The question is which prop makes the plane more happy? Speed or climb? Everyone wants everything in a prop. Is the plane more solid with the gain in airspeed, or is the climb really the most important part of what it needs?
     
  11. Oct 29, 2019 #11

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    What makes the plane happy or what makes the owner happy... Of course I want it all, doesn't everybody?:) But it's a biplane, speed is not its reason for being, I'd rather have the climb. In this case I'd say it's the engine that's not happy, never getting to its rated rpm even at WOT in level flight.

    But this thread is more about the data collection itself. I'm going to go back and look at the analysis code today, make sure nothing's wrong there, and also see what the spreadsheet looks like in Excel, which has better trendline tools than OpenOffice... though no amount of massaging can make up for bad data, just look at the climate change debate. :eek:

    I also played with the level flight acceleration technique to predict climb. But having only the camera times and the picture of the ASI to get the speeds from, there the data is even worse.

    I think a big part of the problem is that the speed range of a biplane is so small, the data points are too close together and there aren't enough of them. In a faster airplane like an RV where the curves can be plotted with more points spaced out over a wider range, I think it would work much better.
     
  12. Oct 29, 2019 #12

    Himat

    Himat

    Himat

    Well-Known Member

    Joined:
    May 5, 2011
    Messages:
    2,811
    Likes Received:
    653
    Location:
    Norway
    A short note, those graphs in post#8 look a bit what I work with, coupled systems. I usually have plot of conductance or displacement versus frequency. With two coupled systems I get two peaks. Now, how that relates to climb performance I have to think about. It could be some biplane thing.
     
  13. Oct 29, 2019 #13

    Pops

    Pops

    Pops

    Well-Known Member

    Joined:
    Jan 1, 2013
    Messages:
    7,230
    Likes Received:
    6,117
    Location:
    USA.
    Why not try climbing at 70,65,60 mph with just using a stop watch to see if you still get the notch. I know its hard to get the right air for the test.


    Maybe an area of sinking air in your test ? Test in two different days so not likely.
     
    Last edited: Oct 29, 2019
  14. Oct 29, 2019 #14

    Tiger Tim

    Tiger Tim

    Tiger Tim

    Well-Known Member

    Joined:
    Apr 26, 2013
    Messages:
    2,905
    Likes Received:
    1,630
    Location:
    Thunder Bay
    Could it be that there’s something vibrating at 65 knots that increases drag? Wing wires humming, maybe? Do you have arrows installed where the wires cross?

    Your article could yet happen, it might just be about a different subject.
     
  15. Oct 29, 2019 #15

    Aerowerx

    Aerowerx

    Aerowerx

    Well-Known Member

    Joined:
    Dec 1, 2011
    Messages:
    5,084
    Likes Received:
    1,391
    Location:
    Marion, Ohio
    Its been a while, but let me see if I can remember my measurement sampling theory....

    Assuming the variations are random, you need a minimum of 4 samples to get an idea of the "true" value. That means 4 separate tests at each climb speed.

    But 10 tests at each speed would be better. Or more than that if you can.

    Do them on different days, different winds, temperatures, etc. Eventually the variations will average out and you will see the "true" curve.

    Remember you are dealing with air here. A chaotic constantly moving substance, so you will never get the same answer twice in a row but, as I said, they should average out.
     
    Himat likes this.
  16. Oct 29, 2019 #16

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    That's effectively what I'm doing, when I select a range on the elevation profile it gives me the time and altitude change between the two selected points, and calculates the resultant averaged rate of climb. It highlights the data points in between but doesn't use them in the calculations.

    The strange part is the size of the variation in what should be a smooth curve. If my airspeed is varying by, say, +/- 2kts, the instantaneous rate of climb could vary by as much as 75 fpm depending where on the curve it is, which is the size of the variation in seeing, but I doubt very much that my average airspeed over a minute varies anywhere near that much from the target speed. But maybe it does.
     
  17. Oct 29, 2019 #17

    Himat

    Himat

    Himat

    Well-Known Member

    Joined:
    May 5, 2011
    Messages:
    2,811
    Likes Received:
    653
    Location:
    Norway
    Some more exotic possibilities for the unexpected data sets:
    • A nonlinear pressure altimeter together with angle of attack static pressure variation.
    • Interaction between the low and high wing that make good and bad speed points for climb rate.
     
  18. Oct 29, 2019 #18

    Pops

    Pops

    Pops

    Well-Known Member

    Joined:
    Jan 1, 2013
    Messages:
    7,230
    Likes Received:
    6,117
    Location:
    USA.
    Both good points.
     
  19. Oct 29, 2019 #19

    Dana

    Dana

    Dana

    Super Moderator Staff Member

    Joined:
    Apr 4, 2007
    Messages:
    8,687
    Likes Received:
    3,064
    Location:
    CT, USA
    I considered the first (as the recording device is inside the cockpit measuring the pressure there) but don't think that's it, as I'm only looking at the change in altitude; the actual altitude doesn't matter. Whatever variation or error there is should be relatively constant during each climb segment.

    The second, maybe... some odd thing that causes extra drag around 60-65 knots? It just doesn't seem likely, and doesn't show in the older green curve.

    I was able to put a polynomial trendline in using Excel (turns out open office doesn't have that feature), and the resulting curve looks very much like the shape I would expect, even if the numbers aren't correct (which is where I got the 2kts = 75fpm value).

    I'm thinking about a different approach, using the instantaneous rate of climb between each subsequent data point and plotting it against GPS groundspeed to get a short segment of the curve, then offsetting that to the user entered airspeed. Gotta figure out how to do that in software now.
     
  20. Oct 29, 2019 #20

    Himat

    Himat

    Himat

    Well-Known Member

    Joined:
    May 5, 2011
    Messages:
    2,811
    Likes Received:
    653
    Location:
    Norway
    Your last paragraph then make me question, how good is the airspeed calibration?

    The solution is probably what Areowerx suggest, make enough tests and average. Computing both “instantaneous” and start to stop climb rate may give some information too.
    A few more questions. How much difference is there in pressure and GPS altitude? Is GPS altitude good enough to have a check on pressure altitude?
     

Share This Page



arrow_white