This article was first published in 2004.
The availability – albeit in small numbers – of cars using hybrid
petrol/electric drivetrains creates the possibility of doing unique
modifications – ones that in the history of cars, have literally never before
been done. In this world-first story we modify the regenerative braking
capability on a Toyota Prius to give greater braking capability in light braking
applications. It’s a mod that needed no brake fluid, no rags, created no brake
dust and yet makes a major difference to how the car brakes!
So what the hell is ‘regenerative braking’?
Regenerative braking ("regen") occurs only in vehicles that use electric
power. During braking, the vehicle’s electric motor is operated as a generator,
pushing power back into the battery. (Regen braking is used on electric trains,
where it feeds power back into the overhead wires. And the electric aircraft
pusher shown at left also uses regen braking!)
By using regen rather than friction brakes, the power that’s usually wasted
in heating the brakes gets turned into usable energy that is stored for later
In a pure electric car, the use of regen braking extends the range, while in
hybrid petrol/electric cars, it improves fuel consumption and emissions because
the petrol engine doesn’t need to be used as much.
However, regen braking isn’t strong enough to do all the braking, so cars
with regen braking use conventional hydraulic friction brakes as well. And this
is where it starts getting tricky. If there are two completely separate braking
systems at work, how are they combined in their outcome so the driver has to
press only the one brake pedal?
In the pure electric Solectria that we drove back in 2001 (The Solectria DW Electric Car), the regen braking was operated by the throttle. Push the
throttle past half-way and the car accelerated; lift it above that halfway point
and the car decelerated with regenerative braking. The hydraulic brakes were
operated with a conventional second pedal, being needed in most cases only for
the last few metres before the car came to a complete stop. We thought the
Solectria’s combined regen braking/accelerator pedal worked very well, but it
was certainly a different control approach to conventional cars.
To keep the cars feeling as much like other cars as possible, the hybrids
from Toyota and Honda have both the regen and conventional brakes controlled by
the one brake pedal. In the first part of its travel, the brake pedal operates
the regen brakes alone, then as further pressure is placed on the pedal, the
friction brakes come into play as well. The current Civic Hybrid mixes the two
brake modes together imperceptibly, whereas the first model Prius, for example,
has more of a two-stage pedal.
Incidentally, it’s not only when the brake pedal is being pushed that the hybrids
regen. When the throttle has been fully lifted, a gentle regen automatically
occurs. In the Hybrid Civic it’s like you’re decelerating in (say) third-fourth
gear in a manual gearbox car, while in the Prius models it’s less strong.
The Point of the Modification
The regen in my ’99 Prius has always seemed to me to be less than fantastic.
Firstly – and perhaps because I am running Kevlar pads and slotted discs – the
friction brakes seem to do nearly all the work. When braking at slow
friction brakes can be heard to be working (the hi-po pads have a slightly
graunchy sound), and at higher speeds, the pedal pressures are very much of the
sort that occur with standard hydraulic-only braking systems.
In short, it has never felt much like the regen is doing a lot.
The Prius runs a centre dash LCD that shows how many watt-hours of regen have
occurred each 5 minutes. These are indicated by small ‘suns’ that appear on the
display – each sun being indicative of 50 watt-hours. (So when a sun appears,
enough power has been put back into the battery to run a 50-watt light bub for
an hour – it’s not a trivial amount.) But in my driving, seeing a lot of suns is
rare - indicative that the regen isn’t contributing much. (The battery level
also gives an indication of how much regen is occurring - the battery level can
be seen on the right in this display.)
So it seemed to me that if the regen could be tweaked to do a greater
proportion of the braking work (especially in light braking), fuel economy would
benefit, the braking would be smoother, and it would take lighter pedal
The Prius is a car with stunning engineering.
So it comes as no surprise to find that the system that integrates regen and
hydraulic braking is complex. (Click on the diagram to enlarge it.) In fact, it’s the ABS ECU that handles regen
braking as well as ABS functions, sending a signal to the hybrid ECU to tell the
hybrid ECU how much regen to impose. But how does the ABS ECU know what to do?
Rather than measuring brake pedal travel (which could vary with pad wear,
etc), the system uses pressure measuring sensors to detect master cylinder
pressure. The higher the master cylinder pressure, the harder the driver is
pushing on the brake pedal.
If the driver is pushing only gently, the piston displacement will be small
and so the hydraulic brakes will be only gently applied. In this situation, the
ECU knows that the driver wants only gentle deceleration and so instructs the
hybrid ECU to apply only a small amount of regen. However, as master cylinder
pressure increases, so does the amount of regen that can automatically be
(In fact, there are four pressure sensors in the braking system and two
pressure switches – but it’s the master cylinder pressure sensor that is most
Modifying the System
So if the amount of regen that occurs is largely dictated by the output of
the master cylinder pressure sensor, what about intercepting and altering this
signal? That way, the ABS ECU would think that there was more master cylinder
pressure than was actually occurring, so resulting in more regen being applied.
Since the actual hydraulic pressure going to the brakes would be unchanged,
there’d be a greater proportion of regen braking in the mix.
The voltage output of the pressure sensor ranges from about 0.4 – 3 volts,
rising with increasing pressure. So if a small voltage could be added to this
signal, the ECU should respond with more regen braking. But if this was done,
would it detect a fault condition? The workshop manual states that a fault will
be detected if the voltage from the sensor is outside of the range of 0.14 –
4.4V, or if the voltage output of the sensor is outside a certain ratio to its
nominally 5V supply voltage. Further, the latter is checked when the brake
switch is off (ie brake pedal is lifted).
In other words, the voltage needs to be within a certain range and in some
cases this is checked with the brake pedal not being used.
Leaving out a lot of the ECU connections, here’s what the master cylinder
pressure measuring system looks like. From top, there’s the voltage signal from
the sensor, the regulated 5V supply to it, the input from the brake light switch
(12V when the brakes are on), and the earth connection. (Note that for the
following circuit, it doesn’t matter which regulated 5V supply and earth
connections on the ECU are used – there are several.)
As indicated above, what we want to do is to lift the voltage output of the
master cylinder pressure sensor, especially at low sensor output levels. This is
easily achieved with a single 100 kilo-ohm pot. If the pot is wired between the
output of the sensor and a regulated 5V, and the wiper of the pot is then
connected to the ECU, the voltage that the ECU sees can be varied from a
constant 5V (not wanted!) right through to the standard signal. If the wiper is
adjusted so that it’s just a little way towards the 5V end, a small voltage will
be added to the signal. Note that to allow for the required fine adjustments, a
multi-turn pot should be used.
But what about the way the ECU checks the sensor output voltage when the
brakes are off? In that case, it might spot that the output voltage of the
sensor always appears to be a bit high. The easy way around this is to add a
relay that bypasses the pot whenever the brake pedal is released. This can be
achieved with a low current SPDT 12V relay. As shown here, whenever 12V is
available on the brake light circuit, the relay opens, sending the signal
through the pot. But when the brake lights are off, the relay closes and so the
sensor input voltage to the ECU is effectively standard. As a result, the system
works as standard until the brake light switch comes on, whereupon whatever
adjustment has been set on the pot immediately comes into
This graph of the input and output voltages at different braking efforts
shows what happens. When the brake pedal has not been pushed, the input and
output voltages are the same. When the brake pedal has just been tripped, the
output voltage rises. The amount that the output is greater than the input
progressively reduces as braking effort increases, until at heavy braking load
the signal is back to standard.
Wiring it Up
With a workshop manual it’s straightforward to find the right wires – but
always double-check with a multimeter and make sure the colour-codes of the
wires match those shown in the manual. The pot and relay (the latter salvaged
from an old ABS controller!) were mounted on pre-punched board.
The first step was to make sure that the relay clicked in and out with
movement of the brake pedal. Next, the pot was set so that its wiper was right
at the ‘sensor’ end and a multimeter was used to check that the signal going
into the ECU was the same whether the relay was tripped or not. The car was
test-driven and the braking system worked exactly as standard... at this stage, just what
Next, the pot was adjusted a little so that the output signal of the sensor
was dragged upwards. Driving of the car then showed a fascinating outcome. When
the brake pedal was pressed just far enough that the brake light switch tripped,
the car would immediately decelerate with regen braking alone. It was quite a
clear and distinct feeling – no friction brakes operating, but the car rapidly
slowing. In fact, at this level of pot adjustment, too rapidly slowing...
The pot was then adjusted back a little and further driving undertaken.
As finally set, the very light pedal brake pressure voltage at the ECU input
is lifted from 1V to about 1.15V – just a 15 per cent increase at this end of
the sensor’s output. This results in a clear deceleration when the pedal is
lightly pressed, and much stronger regen than normal as the pedal is pressed
harder. At high braking efforts, the behaviour of the car is near standard –
it’s in very light braking where there’s a clear difference.
But I’ll be frank – I am not sure that everyone would like the end result.
Why? Well, that takes some explanation.
With a purely hydraulic braking system, if they so desire, the driver can
brake extremely gently. In fact, perhaps only imperceptibly slowing the
car. However, with the modified regen system, the minimum braking decel that can
be achieved is now noticeably more than the minimum with a purely hydraulic
system. That is, if you want to slow down, you put your foot on the brake pedal.
If you don’t want to slow down, don’t brake – cos if you do, for the same light
pedal pressure you’ll slow at a quicker rate than in a conventional car.
So extremely gentle braking, of the sort perhaps where in a conventional
system you’d just barely be touching your foot on the brake pedal, is no longer
possible. Once that switch is tripped, braking is occurring. This is because
when the relay is activated by the action of the brake pedal, the ECU
immediately thinks that the brakes are being applied more strongly than they
really are – and so gives you the extra regen braking.
But it’s important that this minimum braking threshold not be overstated.
Testing with an accelerometer showed that when the throttle is lifted (without
any braking occurring), the deceleration from 100 km/h is about 0.1g. When the
brake pedal is pressed at its modified minimum value (that is, the relay is just
tripped), the deceleration increases to 0.11g. At 60 km/h, a throttle lift
results in 0.08g deceleration, while minimum braking results in 0.12g decel. (As
a guide, about 0.8g indicates very hard braking.)
This graph compares the figures - a throttle lift typically giving about 0.1g
decel, minimum regen braking about 0.11g, moderate regen braking about 0.2g, and
a hard emergency stop resulting in braking of (at least) 0.8g.
Logging the input signal from the sensor and the modified output signal to
the ECU shows that over a 6 minute hilly urban drive, the average value of the
sensor signal was 0.507V and the average value of the modified output was
0.562V. However, as the lower trace shows, the modified output of the sensor has
a lot more ‘area under the curve’, a better indication of the changed feel on
And the benefits?
Firstly, the regen braking is clearly doing much more of the braking work.
Or, to put it another way, much less energy is being wasted in the brakes. This
can be both seen in the display of watt-hours regenerated (there are more suns
than usually achieved on this stretch of road), and also in the feel of the car.
The regen braking is smooth and effortless, slowing the car substantially before
the brake pedal is moved a little further to activate the hydraulic brakes and
bring the car to a halt.
It’s hard to assess overall urban fuel consumption, but in some 5 minute increments it has improved by 30 per cent.
The wear of the hydraulic brakes and pads will also be substantially
lessened. (These brakes still get used in anger down long hills, though – as the
battery fills, regen progressively reduces its braking action and the hydraulic
brakes then do more and more of the work.) The regen also switches off at very
low speeds, so the discs still stay shiny.
Sure, we know that for nearly everyone reading this story, the described
modification will be irrelevant – they won’t be able to do it on their own car.
But firstly, this mod shows that whatever the direction that future cars takes,
tweaks will always be able to be performed.
And secondly, if you’re driving a conventional car, start thinking: why am I
paying more at the bowser because my car wastes all braking power in heat? The
latter is a damned good question!
Delete the Relay?
So since there is a ‘step’ change in the voltage that the ECU sees from the
sensor once the relay is tripped, is it possible to delete the relay and always
pull this voltage upwards? It might set a fault code, but does it give more
The answer is ‘no’. We cannot be sure, but it appears that the ECU looks at
the brake pressure signal from the sensor when the brake pedal switch is not
triggered and uses this as the starting value. When the brake pedal is pushed it
then appears to check the difference between the starting value and the ‘brake
pedal on’ value and sets the regen according to this difference.
What accounts for this theory? Well, with the voltage boosted all the time,
there appears to be no difference at all in the strength of the regen