# Analysis of Wing Tip Design

### Help Support Homebuilt Aircraft & Kit Plane Forum:

#### stanislavz

##### Well-Known Member
Supporting Member
Some graphs. Угол атаки - aoa. К качество - glide ratio. Wing have 1m chord, 6.6 span.

#### Attachments

219.5 KB · Views: 109
• t1.png
76.5 KB · Views: 97
• t2.png
81.9 KB · Views: 77
• t3.png
82.4 KB · Views: 84
• t4.png
78.4 KB · Views: 75
• t5.png
81.4 KB · Views: 75
• t6.png
61.6 KB · Views: 81
• t7.png
88.4 KB · Views: 85
• t8.png
84.6 KB · Views: 80
• t9.png
70.3 KB · Views: 70

#### wsimpso1

##### Super Moderator
Staff member
Some graphs. Угол атаки - aoa. К качество - glide ratio. Wing have 1m chord, 6.6 span.

Upper left appears to be whole wing lift coefficient vs angle of attack - Yes? If no please explain;
Upper right and lower lift are something vs angle of attack, but I can not tell what. Labeling of the abcissa makes no sense on either one, I suspect a column number got in as the Y axis labels;
The lower right appears to be Coefficient of Lift vs Coefficient of Drag - Yes? If no please explain.

#### stanislavz

##### Well-Known Member
Supporting Member

Upper left appears to be whole wing lift coefficient vs angle of attack - Yes? If no please explain;
Upper right and lower lift are something vs angle of attack, but I can not tell what. Labeling of the abcissa makes no sense on either one, I suspect a column number got in as the Y axis labels;
The lower right appears to be Coefficient of Lift vs Coefficient of Drag - Yes? If no please explain.
Hi. It was Cyrylica, Russian

Cx = Cd Cy =Cl
K - glide ratio.

#### wsimpso1

##### Super Moderator
Staff member
Hi. It was Cyrylica, Russian

Cx = Cd Cy =Cl
K - glide ratio.

There is no title for the y axis on any of the plots, which makes reading and interpreting them reliant on assumptions that may or may not be correct. The slide titles are quite obscure to most of us... Please answer the following questions:
• I think the upper left slide is whole wing Cl vs angle of attack. Yes? If no, please explain what it is:
• I think that the upper right slide is L/D (glide ratio) vs angle of attack. Yes? If no please explain it;
• I think that the lower left slide is L/D (glide ratio) vs whole wing Cl. Yes? If no please explain it;
• I think that the lower right slide is whole wing Cl vs Cd. Yes? If no please explain.
Now onto my look at your design study... The tips analyzed appear to come in four span-wise lengths, small (1-3, 6, 8), a little bigger (4-5), much larger (7), and tough to define (9). That brings up some immediate questions:
• With the tips varying in spanwise dimension, did you adjust the span of the straight section?
• Did you keep the straight section the same and allow the wing tips to add or subrtract from total span?
In experimental design, we want to avoid confounded variables. According to wing theory, span loading has a strong effect upon induced drag. Wing taper as we go out the span also is attributed to having effects upon induced drag. The effects of tip span and total span are thus mingled with the effect of tip design, and your experimental design did nothing to allow you to distinguish or separate the effects the variables. That fits the very definition of confounded variables.

Confounded variable is common and has lead to a many design choices in directions away from the desired optimal choice. Let's detail and learn from what we do have. The existing designs can be grouped nicely:
• Narrow tips (they all appear to be the same addition to span) in runs 1-3, 6, & 8 and they seem to plot together. The L/D plots (if that is what they are) show #3 to be the best of this subset, and that is what I would expect from this subset;
• Wider tips 4 and 5. These two plot together with higher L/D, which we might expect with increased length and/or span;
• 7 has a much larger spanwise dimension, is the best performer of the bunch. Again, we might expect it to be so with its increased length and/or span;
• 9 is the oddball. How do we characterize it? Is the tip length the spanwise dimension from the leading edge to the trailing edge?
It may be possible that section 3 (the best of the narrow ones) might be every bit as good as 7 if 3 and 7 had the same spanwise dimension and total span. Likewise, 4&5 might be substantially better if they also matched spanwise dimension and total span of 7. But we can not tell from current runs. And 9 might go from its current "bad" to really horrible by increasing its width...

The designed experiment approach is to identify all of the variables, figure out which variables matter and then experiment in those variables in a controlled manner while fixing everything else as much as possible. Normally to explore a design space you run a factorial, which is all permutations of the available variables. 2 variables and 2 levels each is four runs. 3 variables at two levels each is 8 runs. If you have a bunch of variables, there are balanced fractions that can be run - you accept some confounding in exchange for brevity. I do not believe we need to do that here. I believe that we do need to control for spanwise length and total span.

If this were my experiment and knowing what I know about foils, I would make sure total span is controlled and then run narrow and wide tips of each type. That would add eleven tips to be modelled and run.

Is there anything we can do to narrow the search? Well, maybe you make sure that you run narrow and wide tips from the best of each group. This does have the risk of missing something significant in the ones skipped. You would need to model and run wide tips on 3 & 5 and narrow tips on 5 & 7, for only four more. Your knowledge base will still grow strongly. If the picture remains muddy when those are run, maybe you run the rest.

Other options are to only make narrow versions the best of each group. That means only modeling and running a narrow version of 5 and 7, but means skipping the effect of length on everything but 7.

Billski

Last edited:

#### stanislavz

##### Well-Known Member
Supporting Member
There is no title for the y axis on any of the plots, which makes reading and interpreting them reliant on assumptions that may or may not be correct. The slide titles are quite obscure to most of us... Please answer the following questions:
• I think the upper left slide is whole wing Cl vs angle of attack. Yes? If no, please explain what it is:
• I think that the upper right slide is L/D (glide ratio) vs angle of attack. Yes? If no please explain it;
• I think that the lower left slide is L/D (glide ratio) vs whole wing Cl. Yes? If no please explain it;
• I think that the lower right slide is whole wing Cl vs Cd. Yes? If no please explain.
Now onto my look at your design study... The tips analyzed appear to come in four span-wise lengths, small (1-3, 6, 8), a little bigger (4-5), much larger (7), and tough to define (9). That brings up some immediate questions:
• With the tips varying in spanwise dimension, did you adjust the span of the straight section?
• Did you keep the straight section the same and allow the wing tips to add or subrtract from total span?
In experimental design, we want to avoid confounded variables. According to wing theory, span loading has a strong effect upon induced drag. Wing taper as we go out the span also is attributed to having effects upon induced drag. The effects of tip span and total span are thus mingled with the effect of tip design, and your experimental design did nothing to allow you to distinguish or separate the effects the variables. That fits the very definition of confounded variables.

Confounded variable is common and has lead to a many design choices in directions away from the desired optimal choice. Let's detail and learn from what we do have. The existing designs can be grouped nicely:
• Narrow tips (they all appear to be the same addition to span) in runs 1-3, 6, & 8 and they seem to plot together. The L/D plots (if that is what they are) show #3 to be the best of this subset, and that is what I would expect from this subset;
• Wider tips 4 and 5. These two plot together with higher L/D, which we might expect with increased length and/or span;
• 7 has a much larger spanwise dimension, is the best performer of the bunch. Again, we might expect it to be so with its increased length and/or span;
• 9 is the oddball. How do we characterize it? Is the tip length the spanwise dimension from the leading edge to the trailing edge?
It may be possible that section 3 (the best of the narrow ones) might be every bit as good as 7 if 3 and 7 had the same spanwise dimension and total span. Likewise, 4&5 might be substantially better if they also matched spanwise dimension and total span of 7. But we can not tell from current runs. And 9 might go from its current "bad" to really horrible by increasing its width...

The designed experiment approach is to identify all of the variables, figure out which variables matter and then experiment in those variables in a controlled manner while fixing everything else as much as possible. Normally to explore a design space you run a factorial, which is all permutations of the available variables. If you have a bunch of variables, there are balanced fractions that can be run - you accept some confounding in exchange for brevity. I do not believe we need to do that here. I believe that we do need to control for spanwise length and total span.

If this were my experiment and knowing what I know about foils, I would make sure total span is controlled and then run narrow and wide tips of each type. That would add eleven tips to be modelled and run.

Is there anything we can do to narrow the search? Well, maybe you make sure that you run narrow and wide tips from the best of each group. This does have the risk of missing something significant in the ones skipped. You would need to model and run wide tips on 3 & 5 and narrow tips on 5 & 7, for only four more. Your knowledge base will still grow strongly. If the picture remains muddy when those are run, maybe you run the rest.

Other options are to only make narrow versions the best of each group. That means only modeling and running a narrow version of 5 and 7, but means skipping the effect of length on everything but 7.

Billski

Thank you moving this all to new topic.

Yes, it is 4 yes in graphs meaning. On span - it is with wing tips.

On aech drawing is a virtaul ruler with 1000mm scale. Two white strips.

Ok will move all of them to same drawing.

It is not my tests or my design Just a work of ony highly qualified aeri specialist.

Still - question here is - which one gives similar result with less area overall.

5 one was tested and it is used in be-200 aircraft.

#### Jay Kempf

##### Curmudgeon in Training (CIT)
Supporting Member
Looks like 7 adds quite a bit of area or is it normalized to the same area as the rest. 7 would win for higher CL before giving up and for lower drag than all the others from the research I have done. These planform changes if the airfoils and twist are optimized and a straight trailing edge or even a bit of sweep in the outer panel shed the most controlled tip vortex. Put a winglet on it and it gets a little better. But tiny tip chords have small Re so it all has to be done right.

#### wsimpso1

##### Super Moderator
Staff member
Thank you moving this all to new topic.

Yes, it is 4 yes in graphs meaning. On span - it is with wing tips.

On aech drawing is a virtaul ruler with 1000mm scale. Two white strips.

Ok will move all of them to same drawing.

It is not my tests or my design Just a work of ony highly qualified aeri specialist.

Still - question here is - which one gives similar result with less area overall.

5 one was tested and it is used in be-200 aircraft.

Stanislavz,

Really tough to measure the white part with a ruler shorter than the white part and offset from it.

Wait a minute, all 9 samples, with wingtips, have the same span? This is a lot of difference for only tip geometry…

Oh, this is somebody else’s work who does not understand confounded variables and experimental design.

If the runs we have are it, the best tip is clearly 7. Who cares if it has more area, it is still the highest L/D over the range of climb and cruise Cl. This tip type or others of similar schemes have been widely used and work well.

I personally would love to see 3 proportioned like 7 modeled and run for comparison.

My airplane? Looks sort of like 3, with a span 0.375 of tip chord and an upswept outer edge. Look in post 39 here:
Do we see why I might be interested in 3 fattened up a bit?

As for the tips with extensions of the trailing edge, they are trying to do the same work on the air with a lot less contact. That’s tough to achieve.

There is a lot of experience indicating that all other things being equal 3 is a pretty good base. Lots of later Mooneys and Lancairs and Glasairs have wing and tail tips treated this way. Lots of canard tips and racers this way too.

Billski

#### stanislavz

##### Well-Known Member
Supporting Member
Stanislavz,

Really tough to measure the white part with a ruler shorter than the white part and offset from it.

Wait a minute, all 9 samples, with wingtips, have the same span? This is a lot of difference for only tip geometry…

Oh, this is somebody else’s work who does not understand confounded variables and experimental design.

If the runs we have are it, the best tip is clearly 7. Who cares if it has more area, it is still the highest L/D over the range of climb and cruise Cl. This tip type or others of similar schemes have been widely used and work well.

I personally would love to see 3 proportioned like 7 modeled and run for comparison.

My airplane? Looks sort of like 3, with a span 0.375 of tip chord and an upswept outer edge. Look in post 39 here:
Do we see why I might be interested in 3 fattened up a bit?

As for the tips with extensions of the trailing edge, they are trying to do the same work on the air with a lot less contact. That’s tough to achieve.

There is a lot of experience indicating that all other things being equal 3 is a pretty good base. Lots of later Mooneys and Lancairs and Glasairs have wing and tail tips treated this way. Lots of canard tips and racers this way too.

Billski
I will put all of them on same image with same scale.

#### Victor Bravo

##### Well-Known Member
Supporting Member
Number 7 will be the overall winner for use on low speed light aircraft.

#### Bill-Higdon

##### Well-Known Member
Number 7 will be the overall winner for use on low speed light aircraft.
Some designs similar to # 7 carry the tip to smaller end like the Dornier 228 & 328 or even a point

Last edited:

#### llemon

##### Well-Known Member
I'd like to know the wetted area of each tip.

#### wsimpso1

##### Super Moderator
Staff member
I'd like to know the wetted area of each tip.
Might be an "interesting to know" point, but wetted area is a step towards estimating drag. We already have whole wing lift and drag...

Billski

#### llemon

##### Well-Known Member
Maybe I'm missing something, but he is changing the tips while keeping the span constrained. This changes the area of the wing, both projected and wetted. But what is he using for calculating CL and CD, chord x span area or projected?

I know this probably seems pedantic but the other day I was reading an old Stan Hall article in SA that was all about wingtips. This is what he wind tunnel tested;

Which has slightly different spans and projected/wetted areas for each one, making direct comparisons hard.

I've tried testing out different wingtips in OpenVSP with the span constrained and keeping the wetted or projected areas constant but I don't trust the results I get from it.

#### Jay Kempf

##### Curmudgeon in Training (CIT)
Supporting Member
Before winglets there was Schuemann which is in the direction of #7. You need to understand both to fully optimize lift and drag for a given wing. Schuemann's work analyzed the pressure distribution on the upper and lower skin and combined that knowledge carefully arranged planform to minimize spill of high pressure from under the tip.

3D pressure differentials create flows that make tip vortices and those are closely related to span loading or lift distribution. Winglets manage the stuff that planform can't. Not a patch or a gimmick, just exploiting those pressure differentials all the way to a small chord sharp swept tip that wants to shed those vortices the way you want to shed them instead of leaving it to chance. Effective span is the distance between the tip vortices. The farther outboard you can drive them and the smaller they are (least amount of energy -- vorticity --) the more efficient the span.

#### Sraight'nlevel

##### Well-Known Member
Number 7 will be the overall winner for use on low speed light aircraft.
How about sorta mixing the 3 and 7 ?

#### Jay Kempf

##### Curmudgeon in Training (CIT)
Supporting Member
With the advent of really available CFD, CNC, FEA one can easily build a wing skin with no breaks with constant changing curves from root to tip into a winglet. Before affordable CNC you would have to do templates and carve a plug to shape manually. VERY hard to get left and right mirrors perfect. And then there's twist. Now it is all done in CAD ahead of time and just cut into the mold directly. So yeah, blending a #7 like approach with #3 like tangent blending is very much the right place to go. Most of the wings that I have worked on of late are that, a straight section for a while and then a blended long tip section so that the two panels approximate an elliptical distribution of wing area (which in the end isn't necessary) if you get all the profiles and incidence angles right across the whole span. Not hard to do once you have a math model you can tailor to dial in the wing stations. Taking Re into account adds some complexity. Modifying a base airfoil for local flows, Re and center of lift adds even more complexity. But if you just want to put this in the context of wing tip you can call the outside say 30% of the semi span a wing tip if it helps.

#### wsimpso1

##### Super Moderator
Staff member
Maybe I'm missing something, but he is changing the tips while keeping the span constrained. This changes the area of the wing, both projected and wetted. But what is he using for calculating CL and CD, chord x span area or projected?

Is that really a problem? The total span gives us our distributions of lift, shear, bending and torsion, so same overall span goes equivalent there. Any circular or elliptical tip loses 22% of the raw area, so they are pretty much equivalent on area. As for the other tips, as long as we treat them the same mathematically, our basis for a design decision among the tips is not biased by using one area number for them all.

I know this probably seems pedantic but the other day I was reading an old Stan Hall article in SA that was all about wingtips. This is what he wind tunnel tested;
View attachment 126928
Which has slightly different spans and projected/wetted areas for each one, making direct comparisons hard.

I've tried testing out different wingtips in OpenVSP with the span constrained and keeping the wetted or projected areas constant but I don't trust the results I get from it.

Are you saying that because you got results you can not trust with OpenVSP, we maybe should not trust these results either?

Heck, the several tips tested are not comparable as is - there are three different break points between wing proper and the tip treatments.

#### WonderousMountain

##### Well-Known Member
Is that really a problem? The total span gives us our distributions of lift, shear, bending and torsion, so same overall span goes equivalent there. Any circular or elliptical tip loses 22% of the raw area, so they are pretty much equivalent on area.

It might be a better experiment, to compare based on S area.

Span is a useful metric for storage, but limited as predictor of
surface friction. Skin normally comes in fixed thickness, even if we change medium, Gauge thickness will likely be equivelant for each new wing trial.

If weight is to be constrained, surface area, not Chord, ought be prioritized. My own make is likely to keep 300 gram fabric regardless of any further mods. So it is a fair suggestion.

Now my favorite tip would be a taper to a point, and if we are keeping area constrained to the faulty, but elegant elliptical wing, the straight section would be 4b, while the Wing Taper would be 3b for a sum of (4+4+3)/14 Fun. and nearly Pi/4 .

I suppose this Zero condition WingTip proposal would seem more reasonable on a short wing sport type plane.

#### Sraight'nlevel

##### Well-Known Member
All in all...could one make a straight wing AC better ( better L/D ) than a tapered one....by using just a clever wing tip on a straight wing ?

#### stanislavz

##### Well-Known Member
Supporting Member
All in all...could one make a straight wing AC better ( better L/D ) than a tapered one....by using just a clever wing tip on a straight wing ?
Yes, 20% of span in 1/2 taper as per Strojnik wing. Which i a simplification of Schuemann which is used still today in most of gliders (two joined tapered sections)

Question here is - how worse is 5 variant (stepped wing tip) compared to 7.