Rx7 Mods and Tech Info

First the usual discalimer - What I am writing here is based purely on my own speculation and experience. Anything you do to your car is at your own risk. I have already discovered some previously unknown 'features' of the fairly common FCD mod. If I have got anything wrong, please let me know and I'll correct it.

Writen 18/4/99. And yes, it did rain as soon as I'd taken to roof off.
Updated 8/5/99

Cabriolet Roof

Cabriolet Roof


Firstly, don't expect the lid to be 100% waterproof, or you'll be disapointed. Leaks fall into two categories. The annoying dribble you get when braking after leaving the car out in the rain and the 'footwell full of water' syndrome.

The general principle in the design of the roof is not to design for a perfect seal, but rather to ensure that that water which does get in is channeled back out again. Remember this if you start trying to re-model any of the foam rubber parts!

At the back of the doors, where the roof hinges, there are two 'pockets' formed in the fabric to allow the fabric to relax when the roof is folded down. At the bottom, these are drained by plastic tubes. If the tubes are blocked or twisted, there will be leakage. Close inspection when the roof is down will show that this area is prone to fatigue damage. Another area which does not last well is the top corner which is supported by the 'roll bar'. The seams here will eventually split.

Repair of the roof is best done from inside, and easiest if the roof is removed.

First, remove the bolts around the rear edge. Don't worry about the two at the front, they fit into slots rather than holes and can be pulled out.

Remove the trim which covers the motors for the roof, and detatch the webbing below the window. This is fixed with two bolts on each side. From outside, detatch the springs that go to the webbing that rupports the edge of the roof.

Now detatch the arms connecting the motors. Also remove the 3 bolts on each side that fix the hinges to the body.

The whole can now be lifted off (by two people!). Take care not to break the metal strip running around the rear edge.

My fix for the torn pockets was to glue two pieces of heavy duty waterproof material on to the two sides of the pocket. Ideally, this could then be stitched to follow the line of the original stitching but I am planning to glue a piece of vinyl from the outside to form a channel. The glue I used was 'Alphabond AF 178' supplied by Woolies. This is like evostick, but it's very gloopy with excelent wetting of fabric. I haven't had a chance to see how well it works.

Power opener

When my roof decided to go half-way up and stop, I thought that the switch on the dash was probably faulty. The dash is fairly easy to disasemble (without taking the steering wheel off) but the switches are more complicated. The switch for the roof seemed OK (although I didn't take it apart). I did dis-assemble to cruise control switch which had been intermittent. Stretching the springs slightly should fix the problem.

I had expected the switches to be of dubious build quality, but everything seemed fine.

There are two motors, each one has a microswitch for limit sensing. I think one side senses up and the other side senses down. Judging from the number of leads, these are stepper motors.

Each motor has a screw accessible when the spare wheel cover is removed. Turning this clockwise will disengage the motor. To re-connect, turn anti-clockwise till it stops. The motors will snap back as they are turned.

The micro-switches on the motors seem to be the cause of my failure. The left hand one controls closing, and activating the switch brings the motors back to life. I haven't replaced them yet.

Engine and Fuel


The boost sensor is located on a bracket on the suspention turret just behind the air box. It's output is an analogue voltage (approx. 2.3V for 0 PSI and 3.7V for 8psi). An FCD fools the ECU by limiting the output slightly before fuel cut occurs. Fooling the engine management is basically a bad idea. As the boost increases, so more fuel is required to prevent the mixture becomming too lean and causing detonation or pinking. This rapidly overheats (melts) the engine.

A bomb-proof FCD design is presented on the FC3S web site.

It appears that the turbo pressure is controled closed-loop by the ECU.

After installing my FCD and later an industrial grade pressure gauge, I noticed that I was reaching 14 psi at 3-4,000 rpm in higher gears. I'm glad I bought that gauge!

Removing the FCD brought this back to maxing-out at 7.5 psi, although it was possible to achieve momentary pressures of ~10 psi at a gear-change.

This makes the requirements on the FCD more complex. Assume you want to allow Pmax for your nominal max boost, and retain fuel cut at Pcut. For an output of Pmax, the FCD output sould correspond to an un-modified 7.5psi output. Ideally, you would then arrange for an output corresponding to 8.5psi at Pcut. If you clamp the boost sensor output for a Pmax of less than 7.5psi, you may end up with the situation where boost increases to the point of equilibrium. Due to increasing exhaust back pressure, this isn't necessarily the best set-up for power.

My FCD consists of a 3.3V zenner diode and two BAS16 diodes connected in series with a 1k resistor. The BAS16 is a surface mount general purpose signal diode. All the semiconductors have 100nF 1206 surface mount capacitors soldered across them - probably OTT but it might help against static and interference pickup. The whole assembly is connected in paralel with the boost sensor output.

Changing the 1k resistor to 3.3k results in the wastegate limiting boost to 9-10psi but I can still occasionally get fuel cut.

I haven't managed to test this setup in a static manner, but the output voltage at 0psi is unaffected, and the dash gauge is almost unafected.

Detonation should be less likey to occur with higher octane fuel (hence some turbo cars require 'super unleaded', but the Rx7 is designed to run on standard american gas, which is 88 octane. In the UK, unleaded is 95 octane so a small reduction in fuel flow may be ok. Next project is an O2 sensor meter. This should provide an indication of how lean the engine is running.

Lambda Sensor

Much has been written about the lambda sensor and it's replacement, and also air/fuel ratio meters.

To allow the fuel system to run closed loop, and effectivly re-tune itself on the fly, the exhaust gas is measured to determine the amount of oxygen in the exhaust gas.

Complete combustion will occur with a 14.7:1 ratio of air to fuel (by mass?). This statement obviously incorrect as it doesn't define the fuel, but the correct ratio gives 100% combustion with no residual oxygen. For best power, a richer mixture is required. A richer mixture is also required for the engine to run when it's cold. If the mixture become too lean, especially if the inducted air is hot, the chances of detonation are increased.

The lambda sensor is an oxygen concentration meter. It doesn't measure fuel at all, but it can tollerate being placed in the exhaust gas stream. Lambda is defined as the excess air factor with lambda=1 being the stociometric ratio (nominally 14.7:1) With lambda less than 1, (i.e. rich mixture) the output is >800mV. At lambda=1, the output changes rapidly. For lambda=1.2 (3.3% O2) the output is approximately 25mV.

These graphs are taken from the Bosch 0 258 104 002 data sheet.

The Rx7 uses it's lambda sensor for idle and light load conditions. At least that's what everyone else says. Under heavy load, the system goes open loop due to the response time for the sensors. Fuel is despensed based on temperature, boost, rpm, throttle opening, etc. This is fine until you decide to go for more boost, install an FCD and end up with the ECU thinking it's got 7psi boost whilst the motor is receiving 12psi boost and almost twice as much air. Yes, we all know the system is conservative and uses more fuel than is neccesary to prevent dettonation with 95 octane, but it's nice to be sure.

There is a digital A/F ratio display design published on the FC3S web page, but I've gone for the analogue version. Take a 5V regulator, a reasonable op-amp and a meter. Three 100k resistors make the opamp into a gain=2 non-inverting amplifier. Add a suitable (32k) resistor and trimmer in series with the output and connect the meter so 1/2 scale corresponds to 800mv input. Find the ECU under the carpet on the passenger side footwell (under a metal plate about 1' square) and connect to the centre connector, 4th from the right on the bottom row. If you've got this far you should be able to confirm that this sits at about 1V diring idle.

The important thing to remember once this gauge is installed is that there is significant variation in lambda sensor output under different conditions. This gauge is intended to show when the motor is running under too-lean conditions under load. Don't expect it to show an idle mixture of lambda=1. You should expect to see lean running when you ease off, and more rich under load. Limited tests on my car show full-scale (very rich) under acceleration when 'reasonably warm'. Minimum reading was about 40% with 50%=800mV.

Boost Gauge

Being an electronics engineer by trade, I don't trust these electrical gauges. Especially after I fitted an FCD, so I've plumbed in a mechanical pressure gauge.

Intercept the small rubber tube which goes to the pressure sensor (small box behind the air box on the driver's (UK) side). This comes from a T connector, and it is easy to fit another T with some windscreen washer pipe. I currently route this in to the interior around the door...

The gauge came from RS and reads -15/0/15 psi (in BAR) I forgot to check the size - 100mm is 4" across!

Main Rx7 page

Site Index

Main Rx7 Page Past Club Events Photos from club events
Le Mans Trip, 1999
Trax '99 at Silverstone
My 1st RX7 (for sale)
My Convertable Rx7 Turbo II
Rx7 Stats
Modifications and Tech Info
My bit on Road Safety
Site Statistics
RX7 Owners Club (UK)
Suppliers in the UK

  • rx7-uk@rx7-uk Mailing List List of members cars Links

    back Sean's home page

    This document is copyright Sean Houlihane,1998-2001, and may only be reproduced in full with acknowlegements, or referenced as a link.