The Archives

Browse the content below to find what you're looking for.

First fan blade tests

Wednesday, October 5th, 2016

Initial tests had promise, but still work to do.  Read More

Fan Blade Progress

Thursday, February 18th, 2016

Honestly stuff has been happening… Read More

More winter fiddling

Sunday, November 8th, 2015

I normally learn best by doing… Read More

Fan moulds

Saturday, March 28th, 2015

Once I got into my new workshop situated near the nuclear fusion plant in Oxfordshire, I spent a few months sanding and swearing. Read More

Banana blade

Thursday, December 19th, 2013

I was given a tip by a learned fan design guru, who might have designed fans for the secret military drones keeping you in check. The tip was to add dihedral to the blade, this is when the wing of an aircraft is canted up at an angle, which affects the stability of an aircraft.

Although its got the same name, the effect it has on fan blades is nothing to do with stability. Its to with reducing the stress on the root of the blade, and reducing the amount that the blade will deflect under load. And possibly more important than the latter if i ever planned to sell my blades (which i don’t), it gives the blades a cool 3D curved shape to make it look different to the competitors (doesnt show very well in the CAD screenshots).

This applies only to fast spinning fans and propellers that have a dominating centrifugal load- slow turning wind turbine blades are coned forward to prevent the blades hitting the tower and self destructing.

So how does twisting the blade do this? I will try to explain with a few pictures of the structural model of the blade, when applying centrifugal and aerodynamic loads on their own, and together as the blade would see in real life.

The model is of the whole blade including the skin and foam cores, but they have been greyed out to show only the spar laminate material. Its important that everything is included and the densities and thicknesses are accurate, because for centrifugal loading the weight distribution is key.

Parts sent off for manufacture!

Wednesday, December 11th, 2013

After I showed a few fan gurus my previous post on iteration pU (and after i was nearly done nesting the section for lasercutting), I was given a few suggestions, which really have the potential to make the design a lot better.

-Banana blade: one of my designs had the centre of gravity behind the pitch axis of the blade and i had an almost 30% higher root bending moment which is not good. This is because of the centrifugal forces- imagine spinning around holding onto the handle of something long like a gold club- if you try bend it out of position it will resist that. For iteration pU I moved the spar of the blade back as far as practical (because the blade gets thinner further back, and you need thickness to resist bending), and this reduced the root bending moment a lot. I forgot you can also use this force to your advantage- for instance if you put dihedral on the blade (bend it forwards) then the aerodynamic and centrifugal forces start to cancel each other out! Its not quite something for nothing though, because it means you have to make the blade a more complicated shape. Its not mega obvious in the side by side CAD comparison, but this small change reduced the root stresses significantly ~20%. Which is definitely worth a little wait! The lower the stresses, the less likely any catastrophic failures will be, and the longer the fan will last.

-Blockage effects: the aerodynamic code i am using to design the ducted fan system was never fully finished by the authors, i think because some US government program was cancelled or something (not that im complaining because 10 years later anyone can download it for free). But anyway it doesnt take one effect into account which can result in torque over estimates around 3-5% depending on how many fan blades you have and how thick they are (worse error for more thicker blades- ie high solidity). The basic point is that my blades may be too wide and with not enough angle of attack by a small margin especially near the root. The reason is that the code assumes the blades are flat sheets with no volume, but in reality they have a finite thickness, and as a result they constrict the airflow as it passes through the fan. Imagine a venturi, where the air flows quicker through the tightest part. This venturi effect is very small at the tips, where the blades are very thin compared with the area they travel through, but close to the root where the blades are thicker, and the diameter that they travel through is smaller, the effect may be significant.

Since the pitch is adjustable on my blades, this would be easy to correct, however it means that the blade wont be running quite as efficiently as designed.

Structural design and how to build it

Sunday, December 8th, 2013

For some reason people in engineering call this ‘design for manufacture’ which apparently is another skill you have to learn. Why would you be designing things that are un-manufacturable in the first place? Mongloid business twats dont understand that the whole design and engineering process is fully interlinked from concept to production!

Anyway… I’ll just put a few pictures up until I have a few spare minutes to add proper words.

There are pages and pages of scribbles and doodles and concepts dating back years from when I started thinking about making my own fan. I didn’t finish the aerodynamics one night and start the structural design the next, there are many iterations between all stages. (the design shown here is iteration U, and i only started using letters a few months ago).

After you get to a decent aerodynamic design, get the geometry into your generic CAD package and get the loads into excel.

That bending moment breaks one of the rules I made when starting this project: the new blade loads cannot be more than the original blades. The Hascon root bending moment was in the region of 100Nm, and in terms of loads acting on the hub, is not as bad as it looks I hope. In hindsight this wasnt a great rule, its pretty obvious that to create more thrust you need more force going through the blade, D’oh!

Anyway next you need to estimate the mass from surface areas, estimated thicknesses, materials. From which you can do basic calcs, and start creating inputs for a slightly better model. (Know that these screenshots are the tip of the excel iceberg).

By the way excel, paper, calculators, books and daydreams are probably where most of the work is done, but they dont really make for interesting pictures. (Also if i could recover from my deep hole of a youtube addiction I could have finished this years ago)

But anyway FE does give you lots of pretty pictures, and they are only that if you don’t know how to use it properly, just like CFD. Thanks to work for letting me use it in my spare time.

The general aim of all models and calcs is to be conservative everywhere, so you have as much safety factors built in when you realise you havn’t taken something into account for example:

-Worst case aero loads are used to get highest root bending moment (1200mm diameter fan assumed)

-Heaviest blade that is likely is modelled at max tipspeed to get highest centrifugal loads

-For the hub model there is a loaded blade in each of the 12 hub sockets which is double what would be seen in reality: 6 blades can absorb 200hp

Above you can see how its necessary to align elements with the direction of the fibres, and to try and make them fairly square and uniform if you are modelling carbon fibre. All of the carbon fibres point along the length of the blade to resist the bending from aerodynamic forces, and the the tension from centrifugal loading (21000xgravity). There are glass fibre braids holding it all together and resisting twisting. The skin is laid on top of all that- the aim is to assemble and infuse the whole thing in a single shot- no tricky trimming and gluing bits together, hopefully it will even be painted in the mould.

Above is how the blade deflects when you assume that the hub is 100% rigid- 9mm seems pretty good. In reality its die cast aluminium, and as i said above my root bending moment is much higher than the original blades so its a good idea to do some checks to see if it will be ok. (Obviously the real test is the first time you blip the throttle, but these tools allow you to at least quantify the problem).

So when a model is built including the hub, the deflections are getting towards double compared with assuming a rigid hub. The model above shows 12 blades, but actually for the real fan i plan to have only 2-6 blades. I am taking the worst case, and also this was the easiest to model (the actual model is only one blade and 1/12th section of the hub with cyclic symmetry. I have no measurements of hascon/ multiwing deflections for high power levels in a 1200mm duct, but i would be interested to hear if you have.

Bear in mind these numbers are all for the worst case- in reality the fan I plan to make and use will be 1100mm diameter which has a 26% lower root bending moment. I have optimised the design for the 1100mm duct, but I thought it makes sense to extend the mould slightly so its big enough for 1200mm blades for the future.

Stress in the hub could be a problem, and is definitely an area I will carry on looking into- the yield strength of most grades of die-cast-aluminium is 320MPa and as you can see there are higher stress regions here. However this doesnt mean it will fail in real life, because this model has simplifications, like not including the little stiffening ribs and fillets that exist on the real part. I can do that later when I am sure this is the final design and i have time to run slower, more complex models. I have also put imaginary bolts on the holes and a tolerance on the blade root so the hub halves squeeze onto the root, both of which may not be quite right.

(By the way if anyone knows what grade the hub material is i would be interested, but without knowing I design for the lowest grade its likely to be).

The next thing is to see how the stress in the blade itself is. My current design has a tubular spar running all the way along the blade, which has a 1.5mm thick skin of UD carbon innit. This is the main structural member which runs through the centre of pressure and centre of gravity of the blade as closely as possible, while also running through the thickest part of the blade for max bending resistance. By my reckoning the stresses are manageable even for a lot of fatigue cycles (or hours of blipping the throttle on the grid).

In the pretty picture below you can see the stress along the fibres, positive numbers and bright colours are tension (pulling), negative numbers and cold colours are compression (squeezing). The tensile stress is highest because the blade is spinning around at 2500RPM which makes the weight of the blade 21000 times as much as if it was sitting stationary in your hand.

In another plot below the same result is shown for the model which includes the hub, you can get a good idea why compressive stresses exist: the blade (wing) is trying to bend up due to thrust  (lift). The action of bending means the top of the spar must get shorter (compression) and the bottom must get longer (tension).

Just to highlight why this is a good place to use long fibre composites, which have a lot of strength only in the direction of the fibres, the stress at right angles to the fibres is only about 1/10th of the stresses above.

After you’ve spent a while farting around making changes so many times you cant remember which model was the best, you get organised and write down stresses, and check how they stack up against the strength of the materials:

And we are done (with this iteration)… There are still things that need looking into further (red fail boxes above) and possible tweaks that need to be done (yes I know it needs some sort of blend at the root).

The thing is that i’ll never win any races if i continue iterating fan design and trying to live around my youtube addiction. So at some point you have to call it and get your amazing girlfriend to lasercut sections for the plug, courtesy of UoN Architecture dept.

Prototype fan design

Sunday, September 15th, 2013

It became apparent that the best fan design would be a very tricky shape to realistically make, even if you do have a fancy CNC machine to make the moulds for you. The balance i am trying to find is:

New design must be at least X% better than the baseline = Cool looking sexy shape

> What I want  is somewhere in between>

New design must be realistic to manufacture = Saw it from a 2×4

To get the best theoretical performance you need to design the blade to work under ‘free-vortex flow’. This essentially means that the air travelling through the fan is all travelling at the same speed from hub to tip, and that you have a constant pressure rise at all points on the swept area. If you follow this by the book, you end up with a fan looking like a human powered aircraft propellor with very wide chord at the root and not much at the tip. In a previous post the twisted optimum fan shown is basically this.

The problem with this is actually laminating this structure in as simple a way as possible. With the elegantly sculpted optimum design, I couldn’t think of a way to reliably mold a blade without making it out of at least 2 mouldings and bonding them together. Gluing mouldings together can be done but it is an extra process and one that needs careful control. It is also better practice to have fibres running around the trailing edges of the blades anyway, as is standard for most commercial propellors/ fans out there.

My aim is to infuse the blade in a single piece, which requires the use of a core rigid enough to wrap the glass and carbon around as you put it into the mould. Each blade must be the same weight therefore each core must be identical- they cant just be hand sanded out of a block of foam like ive done in the past. Therefore I plan to use CNC feather-cut blue polystyrene, as model aircraft builders typically use. For this technique the core can only be lofted from a section at either end- the complex optimum fan is out of the question for this then!

So back to the aerodynamics to make the best fan you can that is tapered from two sections. While this sounds pretty crap, this is what happens in real life on aircraft wings. Most airliners have straight tapered wings even though you may be able to get 0.5% less induced drag using an elegant Spitfire elliptical plan form (which is actually not the reason the spitfire wing is shaped like that…). With fans you lose a bit more performance by making this decision, i think its good for 1-2% efficiency loss. But if i can find a straight tapered design that is still at least 5% better than the baseline, im prepared to go for it.

Which brings me to the frozen prototype fan design! Which is the result of many iterations and much frustration and swearing. Is it really worth it for 5% more thrust? Well the more Hp you have the more it is worth i suppose. (I do this because in a twisted way i do enjoy the learning process but also because i want to beat the Tony Gs and Dan Ts of the hovercraft racing world.)

End view on blade showing all the sections if the blade as it will be made

What I have come up with is a blade that will look something like the larger wing-fan blades. The two main design points are:

(in an 1100mm duct spinning at 2510RPM (145m/s tipspeed), using a 12 blade hub)

  1. Around 100Hp with 3 blades
  2. True 200Hp with 6 blades                   (with room for more)

Compared to a multiwing 5Z blade, the new blades should produce 5% more thrust for a given Hp. As well as this they are larger, so require fewer blades to get the same power through, saving weight and $$$.

So details about the new blade are shown below:

This plot shows some general performance figures for the fan operating with 3 blades taking around 100Hp

From the plot above you can see that the lift coefficient of the blade is not linear, due to the nature of a straight tapered fan. The Cl at this point is 1.44, which is near the limit of this airfoil’s performance, its may be necessary to go upto 4 blades, which is easy with the 12 blade hub.

To check if this is alright, you need to do a quick analysis of the blended airfoil section to the radial station where Cl=1.44. At this point (about 50% blade span) the section will be a mix between the sections at the root and tip (the parent airfoils).

Plot showing the parent and blended airfoil sections as they lie on the blade

When you create a section by blending two parents, which is what esLOFT does for you, you don’t necessarily get good children (funny joke about your brother etc…) so you need to check using airfoil analysis programs that the section does what you expect.

A visual check to see if the blended section leading edges have come out OK

So anyway, we take station 5 coordinates and input them into JavaFoil, which will give you a polar of the airfoil, as long as the relevant Mach number and Reynolds numbers are input for that section at that operating point.

But will it be ok running at Cl=1.44, which is pretty high? Im not sure…

Below are some pictures of the blade after it has been lofted in CAD from the sections taken from DFDC and esLOFT (and many spreadsheets, 2D airfoil tools hundreds of text files etc.). Ater you’ve done it a few times it only takes a few hours to get a design from a set of requirements to a CAD model which is pretty good. All of these tools are freely available too (apart from the CAD) all you need to do is give a few hundred hours to learn how they work (and how to run linux).

Javafoil polar showing section 5 performance- looks similar to original airfoils which is good.

Now you can have a nice solid model, but the hard bit is still to come… Making it from real materials

In this picture you can see the prototype in green and the 5Z in blue

As for actually making a blade you can bolt into the hasconwing hubs, tune in next time!

More than paper performance

Thursday, July 25th, 2013

Using any design method for complicated problems such as fan design, it is possible to come up with a number of solutions that appear to give good results. The problem is that a lot of these are just ‘paper performance’ and dont necessarily work in real life.

These arise because all design methods have to use simplifying assumptions to make them manageable. Even using the latest technology, Rolls-Royce does not create a model to work out the airflow around an entire jet engine taking everything into account. Instead the problem must be broken down into lots of bitesize pieces and analysed one at a time. Typically you have one model dealing with the aerodynamic side, and another with the structure which you need to go back and forth until you find what you need.

The new fan design (aero)

Monday, April 1st, 2013

(For structural design click here)

The following is an article i wrote for the famous hovercraft club mag, ‘Light Hovercraft’:

Fans, what’s wrong with what we have and what performance gains could you get with the hoverclub’s own blade design?

The short answer is not much. The best improvement in thrust for a given power I’ve found so far is around 7%. These £15 lumps of plastic are actually pretty perfect for our needs!

I set myself a threshold of 10% improvement to get me in the workshop making the moulds, but since that’s not looking likely I will share my design with anyone interested in making them. This is still a work in progress and there is (or will be) much more information on my website: (…one day i will post a walkthough of my whole process from excel spreadsheet to DFDC to CAD to handcalcs and FEA)

How did I come up with this?

To arrive at this number I have chopped up multi-wing blades, made computer models of them, and run them through the fan design program which I talked about in an earlier article (Ducted Fan Design Code by Marc Drela, free online). This is the baseline fan, which produces X thrust for a given duct diameter, power and fan rpm.

I then design a fan specific to a hovercraft’s needs, which produces Y thrust at the same duct diameter, power and fan rpm. Then I compare its results to the baseline in the same program, to find out the relative improvement (Y/X-1)%. I’m not writing any thrust values here because the model I am using does not take into account duct blockages from engines/ guards/the hull. It’s a waste of time comparing values from a real thrust test with numbers from a computer program which does not take important things like this into account. The important thing is the difference between two fan designs when compared on a level playing field.

A few fans are shown in a Pintail duct below:

Pros Cons
5Z (baseline design) Available now at good retailers! Designed to cool your office
Prototype (5%) 5% more thrust and could be more resistant to fan stall. Fairly easy to make without CNC moulds Not quite pushing ‘rip your face off’ performance
Optimum (7%) 7% more thrust, looks cool Difficult to make and may not work well off design point

What are the changes from the multi-wing blade?

To get the 7% improvement, there have been two main changes to get from the multi-wing to the new blade:

  1. The airfoil section of the blade is changed to a much less draggy one, giving about 5% improvement. Because the multi-wing is an injection moulded part, it has to have a fairly thick trailing edge which gives it pretty crap performance. Fortunately for multiwing, it turns out that the airfoil you use has surprisingly little effect on overall fan performance (1/2 of the airfoil drag does not equal double fan performance).

2.   The twist and chord of the blade is altered so that every station along the blade is working at its optimum angle of attack. This gives the extra 2%, and you only get the full benefit of this effect at one fan RPM and fan loading. If you start changing the pitch, diameter and number of blades to suit a different engine/ duct, you may lose some of this performance.

Shape distributions of the three fan blades shown above

This is where some sort of practical experience is required; there are many combinations of twist and chord distribution that can give similar performance, but may give a fan with a stupid amount of twist or a foot long chord at the root. I have been reverse engineering as many different fans and propellers as I can lay my hands on, to understand how to make something that won’t just work on paper! For a few of the captivating nuances of the various approaches i have tried see this page. For a shed builder it may be necessary to sacrifice a bit of performance to get to a blade without complex changing shapes like the optimum fan shown here:

F2 thrustfan model

Most of the models I have been running are for the thrust fan on my new F2 GSXR 750 using the following design space: 2400 rpm, 229Nm torque, duct diameter 1100mm (Pintail section), 6 blades. I have also tried fans in the formula 3/50 range and found similar results in terms of thrust improvement (but you only get all the benefits if you design for a specific set-up). Flow straightener angle is tailored for each case, but does not affect performance that much.

What about fancy pictures from CFD?

The methods I have been using so far do not take into account details like sweep, dihedral and wavy surfaces like you may have seen on some ‘revolutionary’ fan designs. For this you need to use CFD (Computational Fluid Dynamics), which apart from giving you pretty pictures, is only really chasing the last 1 or 2 percent of performance. For a fan or propeller, simpler models which only require  excel and patience capture the basic physics very well, CFD is really only the icing on the cake. I have some experience using CFD for a rotating fan and my opinion is that you can only go into this sort of detail if you have a few months spare and some very experienced friends to help you (not to mention a powerful computer with the software installed).

Looks good, but what does it all mean?


-My 7% result may be pessimistic due to the fact that the multi-wing blades deflect slightly in use. The model assumes the blades do not twist or bend like they do on the hovercraft. If the new blades were in a stiffer material and didn’t twist under load, they might perform better. The only way to know if this fo sho, is to make it and test it!

-Increased thrust per horsepower may not be the only benefit to a re-designed blade; it may be possible to delay fan stall by operating at lower lift coefficients. Proof is in the pudding!

-Because the airfoil shape does not have a massive effect on the overall fan performance, it may be possible to get good performance from a flat plate type fan, which would be much easier to make.

-Noise: a more efficient fan may be quieter; again a case of actually measuring it to see if this is true.

-Like any ‘system’ the performance depends on how everything works together. No matter if you have the best fan in the world, you won’t get much thrust if you obscure the inlet with a massive engine (why did you buy a GIXXER then!?!). Everything needs to be right including the duct and cover, hull shape, straighteners, centre cone.

I’ve spend the last 5 years getting to this stage so I probably will have to go ahead and make some prototypes to get some closure; does anyone fancy helping out?

Next time: how to design prototype blades and make them without proper moulds.