• Protip: Profile posts are public! Use Conversations to message other members privately. Everyone can see the content of a profile post.

Quaife quick ratio manual steering rack

23 May 2020
SF Bay Area
Quaife is making a quick rack kit for the NSX.


[x] Source a stock manual rack
[x] Send rack to Quaife
[x] Get stock rack analysis results & quote from Quaife
[x] Decide on a ratio / rack gain
[x] Minimum 30 orders ready to go
[x] Put 50% of total quote down
[x] Quaife builds the rack kits
[x] Other 50% of total quote paid
[ ] Everyone gets their new rack kit!


The finished rack kits will have a ratio of 15.1:1. See the analysis below.

Ratio: center 15.1:1 - end 15.1:1 (note: the same center & end)
LTL: 2.51 (current 3.25)
15 -> 12.5 20.2% increase in effort
30 -> 24.9 20.2%
45 -> 37.4 20.2%
60 -> 49.9 20.2%
90 -> 74.8 20.2%
120 -> 99.8 20.2%
180 -> 149.7 20.2%
225 -> 183.4 23.3%
270 -> 217 26.6%
360 -> 284.4 33.6%
540 -> 417.6 36.6%

Original post

TL;DR: I have been emailing back and forth with Quaife about making a quicker ratio steering manual rack & pinion replacement. They are interested, but need a stock rack to develop against, plus an order of 30+ racks to make it worth the effort. Read the rest and indicate interest below!

As many of you know, the manual steering rack in the NA1/2 has a pretty high ratio compared to most modern sports cars. The manual says 18.2:1 to 20.8:1 which is something like 3 turns lock to lock.

Quaife makes quick ratio steering racks for a handful of cars. I had one on an autox EF years ago and it went from IIRC 3.6 turns -> 2.8 turns ltl which totally transformed the car.

Quaife does not currently make an NSX rack, but they could if there is enough interest. Apparently their toyota racks were the result of a community-driven effort.

What needs to happen?

1. They need a stock rack + housing with which to develop the rack & pinion
2. There needs to be an initial order for 30+ rack & pinion sets. Are 30 of you interested? The more, the better, obvs

Procuring a stock rack to send to Quaife

Do you have a stock manual rack laying around? Would you be willing to be without it for 1 - 6 months while Quaife tinkers with it? If not, would you sell it to me?

Q&A directly with the Quaife rep

Q: What would ballpark cost for each rack be? (FWIW, EF racks are $230)
A: The price could range between £130-£190+ ($175 - $250) but it depends on how the parts need to be manufactured and the quantities.

Q: What is ballpark development time on something like this? Before 30 racks might be ready?
A: The leadtime would be 4 weeks for the design then 12-16 weeks for manufacture. It may be quicker however.

Q: Could it be a variable ratio rack? Or fixed ratio only?
A: The ratio would be fixed

Q: How would the pre-order process work?
A: We would need to receive the sample rack for us to quote. There is no obligation for you to place an order once we have quoted though. If you was happy then we would request a 50% deposit to proceed with the order.

Q: Will this reuse the factory housing?
A: Yes, this kit would include a new rack shaft and pinion.


* I assume this would be LHD only. I'm not sure how much effort it would be to translate their eventual design to RHD.
* Installation is likely doable by you (I installed my own EF rack / pinion and it was straightforward), or maybe send it to @d1guy when you get your rack rebuilt :)
* We can decide on the ratio once they know the boundaries. I realize there's a tradeoff with steering effort, and some folks feel it's already pretty heavy. The steering doesn't feel heavy to me, but I have never had R compounds on my NSX...

Prior Art

I know there have been some threads on this in the past, including an effort to make a variable ratio rack, but they never went anywhere.

Have there been other efforts I am unaware of?

Hopefully this thread can get it done with your help!
Last edited:
Definitely a good question. I can ask him for sure. Also happy to ask them anything else’s folks are curious about.

Some quick, maybe wrong math on ratio force: if we’re at 18:1 and 3 turns lock to lock, 2.5 turns would be 15:1 or so. Say you put in 100 force units: 100*18=1800, with a 15:1 ratio, you’d need to put in 20% more force or 120*15=1800.
Interested this would greatly improve on track preformance for weaklings like me who don't go to the gym. Especially on long continuous turns where holding the steering wheel takes a lot of effort.
Not to take away from this great effort. I think it's awesome. Have they commented about what the possible increase in force would be to turn the wheel (ballpark)
That's the major concern and knowing the answer to that is required when determining a new ratio for a manual rack. It's even important when changing the ratio of a power steering rack, to make sure the force isn't greater than what the pump can supply.

bogle - don't get hung up too much on the ratio itself, when comparing to other cars or modern cars. The ratio is only part of the steering equation -WHERE the steering rack's tie rod mounts to the KNUCKLE is another significant part.

Since you're only looking to change the rack ratio, focus on the steering reduction and effort increase. A good starting point would be to tell quaife that when the wheel is 180* you want that reduced to 120*; and at 90* steering input, you want it at 75* (for example).
Interested this would greatly improve on track preformance for weaklings like me who don't go to the gym. Especially on long continuous turns where holding the steering wheel takes a lot of effort.

FWIW, the quicker ratio would increase effort. Maybe no bueno if it’s already hard to hold on in those long corners?

Thanks [MENTION=16531]stuntman[/MENTION], the change in angle for a given steering input is a good way to think about it. I’ll do some math with a handful of options, post it here, and ask the quaife rep about them.
did'nt the guys at flag to pole racing want to do a group buy on something similar but the interest level did not meet financial obligation:confused:
FWIW, the quicker ratio would increase effort. Maybe no bueno if it’s already hard to hold on in those long corners?

Thanks [MENTION=16531]stuntman[/MENTION], the change in angle for a given steering input is a good way to think about it. I’ll do some math with a handful of options, post it here, and ask the quaife rep about them.
I look forward to your updates.

did'nt the guys at flag to pole racing want to do a group buy on something similar but the interest level did not meet financial obligation:confused:
I don't know what happened to his variable ratio rack. He was taking money to fund different steps of the development process and the last I posted, his proposed ratio wouldn't make much difference below 90* of steering angle:

FROM THE POLE TO FLAG THREAD: http://www.nsxprime.com/forum/showt...nual-Rack-Feel-a-little-F1-tech-in-your-hands

I think I should clarify and say that I do not recommend drilling holes in the cast knuckle. It was a bit of a brash statement to drive the point of the concept of steering response not solely being attributed to the rack ratio since you need to consider the geometry of the knuckle.

A far better and safer thing to do would be to make a billet knuckle with a shorter steering arm to quicken the steering response without touching the rack ratio. But if you do that, you might as well fix some of the NSX's roll center geometry problems when lowering the car and improve it to get some camber-gain out of it.

I had P2F's proposed VR rack curve analyzed to correspond with the below steering angles witnessed at Buttonwillow Race way, to see what the realistic reduction of steering angle would be on track.

OEM Rack steering input -> P2F Rack steering input

45* -> 44*
90* -> 88* -virtually no steering reduction or weight increase
135* -> 127*
180* -> 163* -17* less steering angle needed but ~33% more steering effort required.
225* -> 197*

The most you are going to turn the wheel on a typical racetrack is 180*. With the P2F VR rack, that is reduced by 17* and will require 33% more steering effort. As you turn the wheel further, the steering reduction is increased at the expense of increasing steering effort. On right switch-backs or autocross, the steering reduction will be a nice improvement, but keep in mind it will be at the expense of more resistance to turn the wheel, which may not be a good thing for some people.

MY STEERING ANALYSIS OF THE NSX vs S2000 vs Elise vs Mustang:

I must admit that I too felt like the NSX's steering was subjectively slower than many other cars. I decided to look at the in-car videos of various cars around Buttonwillow raceway and compare their steering inputs to a few different NSXs of varying degrees of build. I took a screenshot of each car in two different corners at the point of their maximum steering angle.

The cars I looked at were:

-NSX - Mild build: My personal car: 235/40-17, 275/35-18 NT05 street tires, KW V3 coilovers, and NSX-R rear wing
-NSX - Extreme build: FXMD's record holding Time Attack car: 305/35-18 345/35-18 racing slicks, KW 3-way motorsport dampers, MASSIVE aero
-NSX - Moderate build: BellWilliam's car: 235/40-17, 275/35-18 NT01 R-comps, coilovers, big splitter and wing
-2015 Mustang GT - Mild Build

-S2000 - Mild Build: Stock Aero: 255/40-17 Dunlop ZII StarSpec, HKS Hipermax IV SP coilovers
-S2000 - Extreme Build: Evasive's record-setting Time Attack car: 275 street tires? and massive aero
-Lotus Elise - Moderate Build: Manly's record-setting Time Attack car with massive aero
-Lotus Exige - Mild Build: Relatively stock with Hankook Ventus Z214 tires

Turn 2 "Buttonhook" is one of the slowest (<50mph) corners that you'll find on any racetrack. It requires a lot of steering input at low speeds and while trailbraking can reduce understeer, the corner as an up-hill right at the apex that causes almost any car to understeer at the apex. This is a great corner to determine how much steering lock is needed in a technical and quirky haripin.


-All of the NSX's required ~180-degrees of steering input at the apex of this hairpin.
-The 2015 Mustang GT also required ~180-degrees of steering input.


-Both S2000s required ~160-degrees of steering input, slightly less than the NSX -by 20 degrees
-Manly's yellow Elise required the same ~180-degrees of steering input as the NSXs.
-The Exige required ~150-degrees of steering input, about 30 degrees less than the NSX. Keep in mind he was turning 2:03s, which was much slower than the rest of the cars in this group other than the Mustang doing 2:06s.


Turn 4 "Grapevine" is a pretty standard corner with very little camber/banking. I chose this corner because it does not require much braking or lifting of the throttle, so the balance and steering inputs aren't affected as much by differences in driving styles and trail-braking, and singles out the setup and balance of the car and the steering inputs needed.


-My NSX required ~95-degrees of steering input
-The FXMD NSX required ~100-degrees of steering input
-William's NSX required ~80-degrees of steering input - quite a bit less than the other NSXs.
-The 2015 Mustang GT also required ~80-degrees of steering input.


-The stock S2000 required ~75-degrees of steering input and was quite loose and oversteering through this corner
-The Evasive S2000 required ~80-degrees of steering input
-Manly's yellow Elise required ~85-degrees of steering input
-The Exige required ~55-degrees of steering input, about 40 degrees less than the NSX. Keep in mind he was turning 2:03s, which was much slower than the rest of the cars in this group other than the Mustang doing 2:06s.

I will let everyone draw their own conclusions on the difference of required steering input between the NSX and a cars like the S2000 and Elise/Exige which are considered to have very quick and responsive steering. The Elise/Exige are the only cars without power steering (like the NSX) but they also have much smaller wheels and tires, and the car is about 1,000lbs less than an NSX.

Keep in mind that the driver's inputs and setup/balance of the car does affect how much steering input that is needed. Which is why I took samples of moderate to extreme car builds, as well as varying levels of driver ability. Overall, I think the required inputs between all of these cars are very similar.

Great, glad you guys are interested, I added you to the top of the thread.

I haven't had a chance to sit down and dig through the other thread (there was a fair amount of analysis) + run any numbers. The math here should be simpler than the variable curve in the other thread. But ideally the result would be a handful of tables at given ratios similar to [MENTION=16531]stuntman[/MENTION]'s steering angle now -> new rack above.

I wanted to verify the current ratio then go from there. Maybe it is definitively stated in the variable rack thread. The manual says it's 18.2-20.8:1, which seems variable. But then the curve chart and a couple posts say 20:1. If it is variable, where is it 20:1, where is 18:1? What's the curve? My assumption would be high ratio at the center, but who knows.

> Visually they couldn't tell the variable section of the oem rack, so if it is indeed a 18.2-20.8 rack, its rather minute.


> We are merely reverse engineer this part to yield a quicker VR ratio as the oem is practically fixed @ 20:1.

The video angle analysis in other cars is definitely interesting.

I wish we could make one off racks in a couple ratios and give them to [MENTION=16531]stuntman[/MENTION] + [MENTION=20915]RYU[/MENTION]..... But alas, it will be difficult to really know the change in feel until it is a thing.
Some calcs! This turned out to be super long. Let me know your thoughts!

First off, there was a really nice PDF doc in that variable ratio thread reverse engineering the existing rack. It lists a bunch of details, the most important being the "Rack Gain" and "Half Rack Travel". You spin the steering wheel 360 deg, how much does the rack move linearly & how much total can it move?

Rack gain: 38.5mm / revolution
Half rack travel: 46.7mm (full travel is then 2x, so 93.4mm)
Turns lock to lock: 2.426 (93.4 / 38.5)


A (maybe not so) quick digression to put some things in context:

Up until digging into this, I thought the ratios quoted for various cars (e.g. 20:1 for the NSX; 14.7:1 for an S2000) worked like gear ratios and could be calculated from, say, rack teeth vs pinion teeth. That is, the rack itself had a ratio built in.

Well, ratios quoted are actually the ratio of steering input degrees to angle of resulting turn of the wheels. e.g. Wikipdia:

> Steering ratio refers to the ratio between the turn of the steering wheel (in degrees) or handlebars and the turn of the wheels (in degrees).

Or this article:

> For example, if a 360-degree turn of the steering wheel causes a car's wheels to turn 20 degrees, then that car's steering ratio is 18:1 (360 divided by 20).

Sounds basic, but important. This implies that the numbers quoted for various cars have the the knuckle geometry baked in. So an S2000 has a quoted ratio of 14.7:1; using that same rack in your NSX probably wont be a 14.7:1 because of knuckle geometry differences. I suppose this is why quaife never lists the ratio on their aftermarket rack/pinion sets, only turns lock to lock....

In our manual it lists the ratio at 18.2:1 - 20.8:1, which makes it seem like we have a variable ratio rack. But the analysis from the other thread and visuals of the rack / pinion indicate that it is not variable. What's going on?

It seems the variable ratio is the result of the tie-rod end pickup point moving through an arc. Depending on where you are in the rack's travel, a 45 degree steering input will move the rack the same linear amount, but rotate the knuckle by different angles. This will result in less angle on center (20.8:1), more at the edges (18.2:1), creating a natural variable ratio (because circle!)

While I'm sure the knuckle has special magic geometry to shape/position the arc, and the real knuckle doesn't rotate a full 90deg from center, for this we can assume it's just basic circle math. An example you can play with: https://www.mathopenref.com/coordbasiccircle.html

Pretend (0, 0) is where the knuckle rotates, (0, -20) is where the tie-rod picks up.

Turn such that the rack moves 10 units left: the knuckle rotates 30 degrees.


Turn such that the rack moves _another_ 10 units left (full lock): the knuckle rotates 60 degrees.


Changes in steering angle

All that is to say: the ratio doesn't really affect the calculations. It will be an output based on how the new steering inputs generate the same old wheel angles. But for the calcs, we don't really need to care how much a full steering rotation actually changes the wheel angle because for a given linear rack movement the wheel angle will be the same. The rack gain is the only bit we need--yay for that original analysis, for real.

I wrote a script to generate info for a ratio change. Feel free to play with it and check my work. It takes a ratio as input--I know I just said ratio doesn't matter--but only because it's easier to think about and compare than a rack gain value. The math is pretty simple, it's really just scaling based on the old rack gain and new rack gain.

Let's go aggressive to conservative.

Here is the ending rack gain value proposed in the pole to flag thread. It's almost certainly way too aggressive at a 38% torque increase:

$ node nsx/steering-ratio/gen-changes.js 15.04
Ratio: center 15:1 - end 13.2:1
Gain: 53.24mm/rev (current 38.5)
Torque increase: 38.3%
LTL: 1.75 (current 2.43)
15 -> 10.8
30 -> 21.7
45 -> 32.5
60 -> 43.4
90 -> 65.1
120 -> 86.8
180 -> 130.2
225 -> 162.7

$ node nsx/steering-ratio/gen-changes.js 16
Ratio: center 16:1 - end 14:1
Gain: 50.05mm/rev (current 38.5)
Torque increase: 30%
LTL: 1.87 (current 2.43)
15 -> 11.5
30 -> 23.1
45 -> 34.6
60 -> 46.2
90 -> 69.2
120 -> 92.3
180 -> 138.5
225 -> 173.1

$ node nsx/steering-ratio/gen-changes.js 17
Ratio: center 17:1 - end 14.9:1
Gain: 47.11mm/rev (current 38.5)
Torque increase: 22.4%
LTL: 1.98 (current 2.43)
15 -> 12.3
30 -> 24.5
45 -> 36.8
60 -> 49
90 -> 73.6
120 -> 98.1
180 -> 147.1
225 -> 183.9

Above 17:1 as an input, it feels like getting into the sweet spot. Here it's 15% increase in effort (torque), but 25deg less angle to get to 180deg equiv:

$ node nsx/steering-ratio/gen-changes.js 18
Ratio: center 18:1 - end 15.7:1
Gain: 44.49mm/rev (current 38.5)
Torque increase: 15.6%
LTL: 2.1 (current 2.43)
15 -> 13
30 -> 26
45 -> 38.9
60 -> 51.9
90 -> 77.9
120 -> 103.8
180 -> 155.8
225 -> 194.7

$ node nsx/steering-ratio/gen-changes.js 19
Ratio: center 19:1 - end 16.6:1
Gain: 42.15mm/rev (current 38.5)
Torque increase: 9.5%
LTL: 2.22 (current 2.43)
15 -> 13.7
30 -> 27.4
45 -> 41.1
60 -> 54.8
90 -> 82.2
120 -> 109.6
180 -> 164.4
225 -> 205.5

And what we have now:

$ node nsx/steering-ratio/gen-changes.js 20.8
Ratio: center 20.8:1 - end 18.2:1
Gain: 38.5mm/rev (current 38.5)
Torque increase: 0%
LTL: 2.43 (current 2.43)
15 -> 15
30 -> 30
45 -> 45
60 -> 60
90 -> 90
120 -> 120
180 -> 180
225 -> 225
Good work!

I think the 17:1 looks like a good option that will give you room to go up to the 18:1 if the steering effort is too high, or down to the 16:1 if it's not (and if an even quicker ratio is desired).

wow! That's great work and thanks for explaining the true meaning of the ratio. I had assumed the same as you (i.e. the knuckle wasn't a factor).

Again, w/o any measurements taken, simply just seat of the pants observation, this is why I feel like the AP1/CR rack in my 91 NSX seems to feel just right from a steering ratio standpoint. It doesn't feel twitchy like my previous AP1 nor previously owned AP2 as well. I just assumed this was due to the NSX longer wheelbase but perhaps this is more due to the knuckle differences between NA1 and AP1/2/CR. Is a faster EPS assist like the AP1/CR worth it to loose some feel from a manual rack? to me, yes... to others, maybe not.

Well, @stuntman has a well calibrated butt dyno but we've been missing each other lately. One of these days he can drive my NA1 with the AP1/CR rack. Who knows.. he might hate it!

With that said, I was recently on Latigo Canyon road in Malibu. For those not familiar with that stretch of road, it's very tight, but flowly with good grip. Even just driving well w/in my limits I found the car much much easier to drive in than my best recollection of the manual rack.

I'm well aware of confirmation bias.. so will leave it up to Billy to decide one day. It won't be perfect, but it's been a positive improvement for me.

This is great work @bogle - thank you
Last edited:
Thanks guys, I agree the 17:1 could be a good starting point. It's tough to know exactly what a 22% increase feels like.

Next step is to procure a factory rack and send to over to Quaife. If anyone has one they'd be willing to sell me, let me know!

A couple more options:

17.35:1 produces some pretty even numbers at a 20% increase

$ node nsx/steering-ratio/gen-changes.js 17.35
Ratio: center 17.4:1 - end 15.2:1
Gain: 46.16mm/rev (current 38.5)
Torque increase: 19.9%
LTL: 2.02 (current 2.43)
15 -> 12.5
30 -> 25
45 -> 37.5
60 -> 50
90 -> 75.1
120 -> 100.1
180 -> 150.1
225 -> 187.7

And a 25% increase:

$ node nsx/steering-ratio/gen-changes.js 16.65
Ratio: center 16.7:1 - end 14.6:1
Gain: 48.1mm/rev (current 38.5)
Torque increase: 24.9%
LTL: 1.94 (current 2.43)
15 -> 12
30 -> 24
45 -> 36
60 -> 48
90 -> 72
120 -> 96.1
180 -> 144.1
225 -> 180.1
I would be interested in checking one of these out. This will be a complete replacement or drop in the original rack?

If it can come in less expensive then the OEM rack and solve the knock bushing issues that would be a hit as long as force is not crazy and it’s still drivable.
I would be interested in checking one of these out. This will be a complete replacement or drop in the original rack?

If it can come in less expensive then the OEM rack and solve the knock bushing issues that would be a hit as long as force is not crazy and it’s still drivable.

Great, added you to the OP.

This would reuse the factory housing. The kit would probably look like the civic kit:

<img src="https://user-images.githubusercontent.com/69169/101668280-2e3e1480-3a05-11eb-8a9c-398c117480ce.jpg"/>

I had this kit on my 88 EF a long time ago and I remember the install being pretty easy. I assume this housing is not so different. But another option would be to send the Quaife kit to nsxrackrepair when having your rack rebuilt.
https://www.amayama.com/en has new OEM racks in stock. I recently got one from them from bout $300 less then what it cost rebuild one from nsxrackrepair especially when you factor in shipping to and from.

So if you need a new rack to send Quaife the link I posted above has really good prices. Just punch the part number in and they can get them. I have gotten a lot of parts from them they are almost always cheeper. Didn't really take to long to get to.
Sweet yeah, I’ll look into a new rack from amayama. I definitely love their prices.

I’ve also been in contact with nsxrackrepair and they may have a used rack I can buy to send to quaife. If that falls through, a new rack might be the best option...