DIY Miata Fastback

When I built my first fastback, the design was pretty organic, mostly a game of seeing what would fit with the Treasure Coast chop top. At the time it seemed a simpler way to define the roof shape, but it’s probably easier and cheaper to start from scratch.

Treasure Coast chop top.

I started by laying a piece of wood between the wing uprights and playing with the angle. In the pic below you can see I’m using a pair of vice grips clamped to the wing uprights. I moved them up and down until it looked right. I settled on around 15 degrees. The ideal angle is closer to 12 degrees, but that would create problems seeing out the rear window.

I cut silts in the wood to to fit the shape of the chop top. This was too large of a gap to fill a the edges, so I used thin skateboard laminates because I had them on hand and they bend easily.

Mocking up the side shape with skateboard laminates.

I also added a strip of angle aluminum along the body, which acts as the base of the fastback. These rails meant I could no longer use the stock trunk, which became a slight problem when I did the testing at Watkins Glen. Not a functional problem, but I wanted to test OEM bodywork, and that was no longer possible with this modification.

Then I built a wooden “transom” to hold the back end up. I also wanted to taper the sides in at the same 15-degree angle.

Image may contain: car and outdoor
Playing with angles and shapes.

The tapered sides may be a large source of the drag reduction, because the sides of the OEM hard top scoop air into the cabin, while the fastback does not. You can see how much narrower the fastback is by the large gaps below.

Once I was satisfied with the basic shape, I glassed the outside seam to hold it together.

No photo description available.
Fiberglass on the outside seam to hold it all together.

Then I pulled it off and glassed the inside as well. You can see the graphics on the skateboard laminates, which are printed already.

Fiberglassing the seams inside.

I put in three Lexan windows. I filled the bodywork gaps with red race plastic, which was a mistake, as they expand and contract with heat. I’d eventually replace the red parts with sheet aluminum.

Image may contain: car

To help deflect air getting into the cabin, I added small Lexan air dams where a B pillar would be, to help direct air going past the windows along the bodywork, and added vents at the roofline and at the trailing edges of the windows (these are not shown here).

Image may contain: 1 person, car and outdoor
Finished, more or less.

How did it work? Better. In the race, the engine was running badly, to the tune of at least 5 seconds off pace, and yet I repeatedly passed a BMW e30 on the back straight at WGI. He would pull car lengths on me accelerating out of every corner, and I’d reel him in and pass him on the back straight. A year later I’d do aero testing at Watkins Glen and find out that the fastback is even better than I had imagined.

I recently saw a picture of the VW EV and from the back, the top looks somewhat similar to my fastback. I like the box cavity, I’ll do that on the next one.

DIY Cold-side Cowl Ram Intake

Backstory: I added a Megasquirt PNP2 to my 1993 street car in 2017. I replaced the airbox with a cone filter, and ran the IAT sensor through a tube just after the air filter. This allowed me to get rid of the restrictive AFM flapper valve, but I kept the stock crossover tube with the resonator, which is supposed to fill in a midrange flat spot.

From left to right: stock crossover tube, IAT, cone filter.

I was pretty happy with that setup because I picked up 4 mph on the uphill front straight at New York Safety Track. Overdrive Automotive in Johnson City dynoed the car, and it made 106 HP. I don’t know what the car was making before, but this is pretty good considering it’s only a Cobalt cat-back, a cone filter, and a Megasquirt PNP2 running on the base map.

106.5 HP on the base map. Messing around without a wideband O2 made less power.

This winter I added a Raceland header and a Magnaflow direct fit cat. I felt like the intake needed more air than just the turn signal intakes, so I put a dorky hood scoop on the headlight. And because all those intakes on the front were probably pressurizing the engine bay and keeping the radiator from working efficiently, I added a pair of $8 hood vents.

Intakes and vents abound. Probably didn’t work effectively.

The Raceland header doesn’t have a heat shield and I could feel a difference in the engine compartment. I was curious, so I bought a dual element thermometer and put it in different places in the engine compartment. I drove the car around like that, with one sensor at the air filter and the other at ambient temperature. I then put the outside temp sensor in the cowl behind the firewall and the temperature was very close to ambient.

$15 temp gauge was pretty useful.

The temp sensor only reads Celsius, converted to Fahrenheit, this is about a 48 degree difference. And this is running the car, not sitting at idle. A rule of thumb is that every 10 degrees F is equal to about 1% in power, so if I could move my intake from the engine compartment to the cowl, I should be able to get almost 5% more power.

Cowl intakes are nothing new, they take advantage of the high pressure zone where the windshield meets the hood. In the image below you can see the arrows mostly point upward indicating lift or negative pressure, but the cowl has significant positive pressure, which brings in cold air. This is a good place for an intake.

Image result for miata hood pressure map

I didn’t want to re-use the existing intake crossover. While this might be good for midrange, once the engine compartment heats up , the plastic crossover tube is like a sponge and retains heat, which ruins the whole idea of reducing intake temperature. So I needed to replace the plastic with metal. At the same time, I decided I should move everything over to the cold side of the engine bay, away from the headers.

The first thing I did was see if it would actually fit. I would have to remove the charcoal canister and water bottle for sure, and probably relocate the main fuse box. But it looked like it would fit, so I ordered the parts:

I then started making room for the parts, and as I did, I noticed that the alternator wheel had cut a hole in the back of the crossover tube’s resonator. So much for filling in the midrange flat spot! Note that since the MS PNP2 is a closed loop system, this didn’t make the engine run excessively lean, as this isn’t an introduction of unmetered air, as when running a AFM or MAF.

Well that explains the hole in the midrange.

Next I test fitted the parts and found I could snake the aluminum intake tubing underneath the strut tower bar. But only just barely. In the end, I removed the STB, because I’m skeptical of how much it helps, and I wanted a bit more room.

The intake tubing was a bit long, but after trimming the 180-degree elbow, I was able to fit the entire 18″ long tube and filter right up against the cowl. 949 Racing tested different length intake tracts and IIRC a 1.6 made the most power with a 19″ long intake (1.8s should be 21″). Close enough.

Removed the strut tower brace, fits better.

Then it was time to finish up some final details.

  • Cut a hole in the cowl to feed the air filter. Since the hole is slightly above the filter, I made a small shield to deflect air down towards the filter.
  • Put the fuse box bracket and fuse box back in. Can’t believe if fit.
  • Clamp all the hoses. Also strap down the intake from moving, using safety wire. It’s a tight fit next to the throttle wheel, and I don’t want the two to make contact.
  • Tape up all the holes in the cowl area and screen it off from debris.
  • Get rid of the headlight intake scoop and turn signal intakes.

There’s only one final detail, which is to put a new washer bottle on the hot side of the engine. The old airbox location is perfect for a universal washer bottle, which will complete the project. Oh, and I have to dyno it again and report back.

Part 2: Ram Air

That intake lasted about a week and then I decided to change it into a ram air intake. I’d already done one version of this on a 1.8 using the stock airbox and the turn signal. I posted the results some years ago. I measured the manifold pressure using a DIY water manometer (aquarium tubing and a measuring stick), and watched the intake pressure move 4″ of water at WFO. This worked out to about 1% more power at 100 mph.

This time I decided to use the cowl rather than the turn signal for the intake, but once again I butchered a stock airbox to make it work.

Stock airbox inverted, cut to fit and sealed with aluminum tape.

I started by cutting the bottom off of a stock airbox and turning it upside down. I then cut matching (ahem) holes in the airbox and cowl. Effectively, I made the entire cowl into an airbox and I’m using the stock airbox mostly to hold the air filter.

Stock airbox on the cold side using ram air. Washer bottle relocated.

I fastened a clear cover to the airbox, so I can see how dirty the air filter is without opening it. Inside the airbox, everything is sealed up and pressure should build. From experience, I know this will be almost negligible, but it will be fun to measure this and find out.

Third Gear No Brakes

I don’t know a lot of car drills, and in fact I only do one: “Third gear no brakes.” Leave it in third gear (or fourth on a higher speed track) and don’t touch the brakes, that’s all there is to it. I learned this exercise from Keith Code back when I was a motorcycle journalist for Moto Euro. We did an article on the California Superbike School, and Keith made us do this drill for two sessions at Sears Point (Sonoma), in the rain.

Third-gear-no-brakes is a great way to focus on entry speed, and you absolutely have to use reference points. You will eventually scare the shit out of yourself, but after that, you’ll be surprised how fast you can go.

For example, here are two laps, the red is me, the black is my friend Jim. We are in the same car on the same tires, and he is .2 seconds faster than me. But if you look at the traces, he’s shifting and braking, while I’m staying in 3rd gear the whole time, never touching the brakes.

If you look at the time graph along the bottom, you can see I make up most of the time in the middle of the graph. This is the “knuckle”, a triple-apex corner. I have to shut off the throttle right at the end of the corner, and that long sloping line is me coasting downhill, waiting for the blind hairpin. At the same point, Jim’s trace looks like a mountain, with strong acceleration upwards and hard braking coming back down.

The other place I make up time is in the S-trap, a low-speed switchback just before the front straight. This is at about the 4000 foot mark on the graph, and you can see we’re just crawling here, but I maintain my speed and this makes a big difference.

Jim isn’t a bad driver, he’s taken racing classes, and has raced wheel to wheel. He has strong inputs behind the wheel, and an aggressive driving style. But he slows down too much and my Miata doesn’t have a lot of power to overcome that.

This is what third-gear-no-brakes looks like from the cockpit. It’s not very exciting. I edited out the part where I went three wheels in the dirt!

Pineview Run, 1:28.9 in 3rd gear, no brakes. ’93 Miata, Yok S.Drives

Time Trials on 300 Treadwear Tires? Yes!

Pineview Run holds a time trial series on Wednesday nights, called the Challenge Cup. This is a great chance for non-members to run the track, and for everyone to engage in friendly competition. The series has a unique classing system which uses the UTQG (Uniform Tire Quality Grading) treadwear value as the sole determining factor. The three classes are split like this:

  • Street: 300+ UTQG
  • Track: 200+ UTQG
  • Race: under 200 UTQG

The UTQG rating is supplied by the manufacturers, and is thus total bullshit, especially in the 200 treadwear (TW) category. But everyone knows this, and so it’s still a level playing field. If you care about winning, just make sure you’re on the best tires in the category.

I’m interested in the Street category, mostly for the convenience of it. “Run what ya brung,” is how the saying goes, I’ll call it lazy and be fine with it. Also, my 1993 street Miata has only 110 hp, and this class is about the only place where high-horsepower cars won’t stomp on me.

My daily tire is the Yokohama S.Drive, but I recently purchased some Continental ExtremeContact Sport for racing in the rain. And I also have my RS4 race tires on hand. So I figured I’d take them all to the track in advance of the Pineview Challenge Cup, and see how the tires measured up.

Yokohama S.Drive

My 195/50-15 Yokohama S.Drives have a bit of use on them. The tread depth measures about 6/32″, from the 10/32″ they started with.

Yokohama S.Drive 195/50-15, TR C1M 15×7 +30, 31.4 pounds.

S.Drives are a popular tire on Miatas, possibly because they’ve been around a long time. There’s nothing exceptional about them, except the ridiculous sale price I got. I ordered them online at Walmart, and shipping, mounting, and balancing was free. Out the door they were $50 each, which is insane. I keep looking back for another sale like that, but haven’t seen one. Maybe this was a closeout. I don’t see this tire size on Tire Rack, there’s a 195/55-15 or 205/50-15 instead.

I’ve tracked the S.Drives a few times at NYST and Pineview, and while they are on the slow side, I like the way they communicate. You can hear the howl reverberating around the facility. The S.Drives have a 300 treadwear rating, and these were what I planned to use these at the Pineview Challenge Cup races for the Street class.

I set the S.Drives to 29 psi cold, and on track they come up to about 35 psi. My laps are typically in the low 1:24s, but the track can be a lot faster if there’s some rubber down from other cars. We get a lot of rain in Central NY, so the track gets washed clean quite frequently. On this day my best lap was a 1:24.1, and so right in the expected range for what I consider my control tire on a clean track.

Continental ExtremeContact Sport

The Continental ExtremeContact Sport (ECS) are a newer tire that’s supposed to be a great rain tire, and also good in the dry. I had set them aside as my racing rain tires, not really intending to daily these. They have a higher treadwear rating of 340, but as you probably know, that’s not always a meaningful number.

The ECS also start at 10/32″, and the tires were brand new at the test. The other tires have had some use, so factor that into your armchair calculator.

Continental ExtremeContact Sport 205/50-015, Konig Dekagram 15×7.5 +35, 31.2 lbs.

I started the Conti ECS at 30 psi, which was definitely too high, as the center of the tire got significantly hotter than the sides. Nevertheless, they were faster than the Yoks, by a full second. I then dropped the pressure 4 psi and gained a half second, putting down a 1:22.6. That’s 1.5 seconds between the Yoks and the Contis. Wow.

The Contis feel a bit vague on turn in, but that could be me just not being used to them. They are also loud, but not quite S.Drive loud. They are on 7.5” wheels because I keep thinking I’ll autocross in the STS class one day. But I’ve said that before, and I just never do it. Anyway, I wonder if an 8″ wide wheel would stiffen up the sidewall some more.

Hankook R-S4

I just finished a two-day aero test at Watkins Glen with my race car and was curious how a 200TW would stack up against the 300s. The 225/45-15 Hankook RS4s are pretty well stretched on a 9″ wheel, but this is what a lot of Miatas use.

Hankook R-S4, 225/45-15, Konig Helix 15×9 +35, 33.4 lbs.

My best time on the RS4 was a 1:21 flat, 3.1 seconds faster than the S.Drive, and 1.6 seconds faster than the ECS. I expected the RS4 to be a good deal faster than S.Drives, what I didn’t expect was that the Conti ECS would split the two almost in the middle.

I’m probably not driving the RS4s to the limit yet, as they are only about half as loud as the other tires, and I expected more talking back. Maybe there’s another second in these, but that’s not enough for me to switch over into the 200-TW category against RE71Rs and Rival 1.5s.

Here’s a video of my first session on the RS4s. The 11″ steering wheel and manual rack make this tight course a bit of an upper-body workout.

Simulating G-forces and lap time

Just for kicks, I want to see the approximate lap time all three tires would do on different tracks, and to get that, I need the lateral cornering Gs so I can plug it into Optimum Lap.

I use an AIM Solo for data, which shows me how much grip the car has in every part of the track. But I don’t drive every lap or corner exactly the same, and the values spike here and there, so it’s not easy to get a steady-state value.

So I plugged my lap times into Optimum Lap and started adjusting the grip values until I got the lap times I got in real life. I started with the S.Drives and called that 1.0g, which is about how much Race Studio shows they grip, and also because it’s easier to view other tires as percentages when you start with 100. This makes the ECS 1.04g and the R-S4 1.09g. Another way of saying that is the ECS had 4% more grip than the S.Drive, and the R-S4 had 9% more.

I wondered what the lap times would be for larger tracks, so I used Optimum Lap to simulate two local tracks, NYST and Watkins Glen. I’ve included my simulated Pineview Run laps, which will let me play with other things like drag, lift, power and weight, at a later date, and find out the differences those changes make (a subject of a future post).

S.Drive 1g1:24.11:23.951:50.702:33.06
ECS 1.04g1:22.611:22.541:49.272:31.41
RS4 1.09g1:21.061:20.941:47.152:29.15

The 3-second delta between S.Drives and RS4s at Pineview Run becomes 3.5 seconds at NYST and 4 seconds at WGI. That’s somewhat surprising to me, I expected a larger gap because the tracks are roughly two- and three-times longer, respectively. But at Pineview Run, you’re on the sides of your tires all the time, so I guess it makes sense.

Conclusions and post-test notes

  • Pineview Run is a really good place to test tires! You can get in a lot of laps to normalize the data, and it’s cheaper than most tracks. I’ll do more tire reviews in the future.
  • This was only my 3rd full day at Pineview Run, and I’m still leaving time on the track. To keep the simulations accurate, I’ll have to fudge the values to reduce track grip in Optimum Lap.
  • Even though the Conti ECS are 1.5 seconds faster than the Yok S.Drives, I probably won’t do the Challenge Cup on the Contis until I wear out the Yoks. Partly because I’m cheap, but I also want to keep the Contis at full tread in case I’m racing in the rain.
  • When the Yoks are all used up, I’ll buy a new set of Conti ECS and use them for everything.

9 Lives Racing Gurney Flap and Dual Wing

In my previous post, NACA Wing Shapes and Airfoil Tools, I compared the 9 Lives Racing “Big Wang” to a NACA 6412 wing. The NACA wing was most efficient in the 5-7 degree range, and would begin to stall above 10 degrees. Based on fooling around with camber and thickness in the NACA tools, I’d guess the 9LR Wang will operate best in a slightly lower range. To make more downforce, you can add a Gurney flap or a second element.

How a Gurney flap works depends on who you ask. Some sources say that the flap changes the location where the air above and below the wing meet, making that point further away from the leading edge. Effectively, the Gurney flap makes the wing chord (front to back length) longer.

And other sources say that a Gurney flap works by keeping airflow from separating at higher angles of attack.

Image result for gurney flap

Either way, Gurney flaps allow you to make more downforce. Typically Gurney flaps are about 1-5% of the chord length. Given the 9LR chord of about 10″, the 1/2″ flap is 5%, and this is on the large side. 9 Lives makes flaps in 1/4″, 1/2″, and 3/4″. This chart breaks it down nicely for you.


What the chart doesn’t list is the lift-drag ratio, but that’s easy enough to calculate from the 9LR numbers. I’ll use just the 100 mph values and make my own chart (below). When I then sort by L/D ratio, you can see that the single-element wing at zero degrees is the most efficient, with a 14:1 lift/drag ratio.

If you crunch the numbers further, you’ll see that the most efficient way to adjust rear downforce is changing the wing angle in the 0-5 degree range. If you need more downforce, keep the 5-degree angle and add progressively larger Gurney flaps. Finally, if you still need more downforce, increase the wing angle until it stalls.

But aero isn’s just about efficiency, but balance. Let’s say you have an airdam and splitter making 200 lbs of downforce at 100 mph (which is what the Hancha Group CFD showed in Miata Airdam and Splitter). If you want the same downforce at the rear, you can generate that by either running the 9LR wing at 10 degrees AOA, or at 5-degrees AOA with a 1/2″ Gurney flap. The latter does so with 10% less drag.

More wings, more downforce

Time attack Miatas use wide and complex splitters, with multiple dive planes and surfaces to create more front downforce. And they turbo or swap the engine, so they don’t care so much about drag. For this specialized application, they need more rear downforce, and a multi-element wing is one way to achieve that.

Multi-element wings effectively increase the camber of the wing, and can therefore be used at a higher angle of attack. The more wing elements you add, the more downforce you get, and with that comes more drag and reduced efficiency.

But I’ll just comment on the dual-element wing. The placement of the second wing is critical, which should create a convergent slot (larger in the front, tapering to the rear), to accelerate air.

According to Katz, in Race Car Aerodynamics, the main wing should be run close to zero degrees, and the second wing at an angle up to, but not exceeding 40 degrees. 9LR just released a dual element wing that can be added to their standard wing and their CFD also shows that greater than 40 degrees is a mistake.

The CFD L/D ratio of 2.7:1 isn’t very good. In fact you can see that the single wing with large Gurney flap makes the same downforce as the dual wing at 35 degrees. And it does so with 1/3 the drag. Perhaps there’s something wrong with the values in the chart. 9LR claims up to 10:1 L/D ratio, and so something is clearly off. I’m also wondering if better numbers will come when they optimize the wing angles and slot position.

In Competition Car Aerodynamics, McBeath also examines dual-element wings. The dual element wing made of 60% more downforce than the single element wing, which is pretty close to the 9LR values above. So maybe the L/D ratio is correct, and there’s just a lot of drag when you add a second wing?

NACA Airfoil Tools and 9 Lives Wing

Airfoil Tools is a really neat site. It serves as a catalog of existing airfoil shapes, and allows you to compare them, simulate different speeds and angles of attack, and draw and plot your own wing shapes. Naturally most of these wings are designed for airplanes, but some have been used on cars (upside down) to create downforce. Cars don’t go nearly as fast as planes, and so the ideal shape of a car wing is different, and optimized for Reynolds numbers on the lower side of the spectrum.

The Reynolds number isn’t easy to explain, so I won’t. But it’s worth noting that the airfoil tools calculator has values that can be used to simulate car speeds. For example, a wing with a 10″ chord that is traveling through the air at 67 mph, and at 68-degrees F has a Reynolds numbers of around 500,000. Doubling the speed doubles the Reynolds number. You’ll only need this number if you’re using Airfoil Tools, otherwise forget it.

In Competition Car Aerodynamics, , Simon McBeath frequently uses the NACA wings as examples, such as the NACA 6312,. In NACA 4-digit wings, the first digit is the camber as a percentage, the second digit is the location of camber in relation to the chord length, the third and fourth digits are the percent thickness of the chord. So, a NACA 6312 wing has 6% camber, the max camber occurs at 30% of the chord length (from the front), and the thickness of the wing is 12% of the chord length.

McBeath discusses these form factors and how they allow a wing to generate more downforce.

  • More camber creates more downforce, but too much and you get separation underneath at the trailing edge.
  • Moving camber rearward generates more downforce, but further forward is more efficient for low-drag, low-angle applications.
  • A thickness of around 12% is quoted as being best for flow separation, but a thicker wing doesn’t create much more drag at car speeds, and could be better when run at a high angle of attack. The author suggest around 16-18% thickness as being ideal.

First I want to give a shout-out to 9 Lives Racing who makes a very strong, economical, and proven wing. I don’t know if it originated from a NACA profile, and I’ve tried to find a NACA shape that’s the same, but I can’t find an exact match.

For fun, I’ll plot a wing that looks similar. Start at the NACA 4-digit airfoil generator and plug in some numbers. At first I’ll try 8% camber, 50% chord position, and 12% thickness and I get a wing that looks kind of like the 9LR wing (although 9LR has more camber).

Now I want to take McBeath’s advice and increase the thickness to 16% and I’ll add as much camber as the tool allows (9.5%).

To my eye it looks chubby, especially the leading edge. It’s a lot of effort to build a wing, and it would take a lot of trust to build one based on an online calculator and other people’s suggestions for what might work. A couple proven wing shapes I’m interested in building are the Eppler E423 high lift airfoil, and the Chuch Hollinger CH 10-48-13 high lift low Reynolds number airfoil. And then again there’s the smart move, which is to buy the 9 Lives Racing wing and move onto other projects.

Another thing I discovered using the Airfoil Tools page is that mosts wing shapes operate best in a fairly narrow range of angle, and that this angle of attack is dependent on speed. Let’s see what the Airfoil Tools page can show us on how wing angle affects downforce and drag.

First I’ll select a wing, and the NACA 6412 is as good as any for this demonstration. Next I’ll set the Reynolds number from 50,000 to 200,000 and the NCrit range between 5 and 7. Click Update Range and look at the resulting graphs.

First up is Cl vs Alpha. Cl is the coefficient of lift, which is downforce to you and me, and Alpha is the wing’s angle of attack. The more wing angle, the more downforce, up until about 9 degrees. After that flow separations occur and the wing creates less downforce.

Next take a look at Cd vs Alpha, which is drag vs wing angle. This wing has the least drag around zero degrees. There are three lines in these graphs that represent different speeds, but you can see that drag goes up quite a bit after 10 degrees.

The next one I’m interested in is Cl/Cd vs Alpha, which is how efficient the wing is at different angles. The wing is most efficient between 5-7 degrees, depending on speed.

So you might think to set this wing at about 6 degrees, so that the wing performs at its best lift/drag ratio over a range of speeds. However, this isn’t necessarily the best setting for your car.

  • More downforce, at the expense of more drag, is almost always faster. You can read more about that in Downforce vs Drag.
  • Having an over abundance of rear downforce will make the car understeer at high speed (safer), which means you can tune the car for oversteer at low speed (funner). In the end, the best setting is how you want to adjust the balance.

If you look at the open-air CFD data for the 9 Lives Racing wing, it also operates in a similar range for angle of attack, but with more camber, it’ll probably stall earlier than the 6412 wing. As you can see in the chart below, the 9LR wing makes a tiny bit more downforce at 10 degrees than 5 degrees, but with a lot more drag. It may be starting to stall at 10 degrees, and the downward wash of most rooflines makes the 10-degree setting a bad one.


Also included in their chart is data for Gurney flaps and dual-element wings, which is the subject of the next post.

Splitter Length and Side Fences

In this post I’ll examine several aspects of splitter design, starting with length. Some racing organizations regulate the maximum size of a splitter, for example:

  • Supermiata S2, NASA ST6 – airdam, but no splitter
  • Supermiata S1, NASA ST5 – 4″ max
  • Champcar – 12″ max (wow)

In Race Car Aerodynamics, Katz states that the splitter length should be double the chord length, the chord being the distance from the splitter to the ground. So if your car has 4″ of ground clearance, the splitter should be 8″ long. That seems rather long to me, but this rule of thumb may depend on the shape of the nose, and for a vertical airdam, perhaps shorter is ok.

In Competition Car Aerodynamics, Simon McBeath cites a CFD study done on a NASCAR model, using splitters of 2″, 4″, and 6″. In this case the ideal splitter length was 4″, producing the most downforce, and best L/D ratio.

  • The yellow line represents the total downforce, and you can see that the 100mm splitter is just slightly higher than 150mm.
  • The blue line shows downforce increasing fairly linearly up to 100mm, and then leveling out at 150mm. So you don’t want to go longer than 6″ (at least not on this stock car).

In the NASCAR CFD study, the splitter added 10% more downforce than using the airdam alone. A Miata isn’t a stock car, and this is all calculations, so YMMV. This is in contrast to the Hancha Group’s CFD work, who’s theoretical airdam produced 34% more downforce with a splitter than with just an airdam. In my real-world testing, I found a 4″ splitter added .38 to the front coefficient of lift (it increased downforce substantially over an airdam alone) and decreased coefficient of drag by .01. More downforce, less drag, so do it.

The same NASCAR CFD study found that extending the splitter rearwards underneath the car had further benefits, and the longer the better. This isn’t surprising, because it’s effectively creating a flat bottom. The interesting part was that lengthening the splitter rearwards made the splitter less effective, because of a build-up of pressure in the engine bay. Overall downforce did increase, but this was because of body interaction. The CFD model was revised to add vents in the hood, and then the splitter and underbody panel both made more downforce. Hood vents are not just for cooling!

Another unexpected result comes from Competition Car Aerodynamics. In Chapter 9 McBeath explains the wind tunnel work done at MIRA using a championship winning Integra Type R (which always makes me want to go all Dor-Dori and shout “Inte-R!” in a Japanese accent). In the wind tunnel, they experimented by putting different ends on the splitter, with ramps of different sizes, and with and without a vertical fence on the end. Each time they measured the result for drag and lift.

The configurations were as follows, and correspond to the image below:

  1. Shallow ramp.
  2. Baseline configuration with no ramp or fence.
  3. Medium ramp.
  4. High ramp.
  5. High ramp with vertical fence.
  6. Vertical fence alone, no ramp.
  7. Vertical fence, shallow ramp.

In the image above, the total amount of front downforce is the yellow line, and configuration 6, a simple vertical fence, is the winner. Adding a small ramp (configuration 7, which is pictured above), or large ramp (configuration 5) in front of the vertical fence actually reduced downforce and increased drag (the black line). Who’d have thunk it?

McBeath doesn’t reveal the actual numbers (it was a private test, they hold the cards close to their chest), but he did say that the best configuration reduced total drag on the car by 4.8%, and more significantly, total downforce increased by 50%!

I’ll conclude this post with some generalizations about splitter design:

  • You can make your the splitter as long as your rules allow, but the longer it is, the more it will affect the front/rear balance. Also note that it may work just as well at a shorter length. If you want to choose a length and not experiment with it, 4″ seems a safe bet.
  • Extend the splitter rearwards as far as the rules allow. But note that this may increase pressure under the hood, and then hood vents may be necessary.
  • Extend the rearward edge of the undertray as close to the leading edge of the front tires as possible. You can generate downforce from wheel wash.
  • If the undertray curves upwards at the rear, it will accelerate the air in front of it, creating more downforce and drag. McBeath quotes a value of about 4% increase in downforce and 1.5% increase in drag. For me, the effort wouldn’t be worth it, and I’d just use a flat undertray and perhaps rake it slightly.
  • Your splitter may create over 200 lbs of downforce, and so if you can’t stand on it, it isn’t strong enough.
  • Many people use birch plywood for the undertray and splitter, but mahogany marine plywood is better. I suggest Okume or Meranti BS-6566, it’s more weather resistant than birch and 33% lighter. I’d go with 9mm-15mm thick depending on if it’s just an undertray or splitter.
  • The front edge of the splitter should be radiused on the underside to avoid separation of flow. Sharp on top, radiused on the bottom.
  • Add vertical fences on the sides of the airdam to shield the tires and increase downforce.