This article was first published in 2005.
Last week in The Story of Turbo'ing a Hybrid Prius, Part 3 we ended with two major problems affecting the
turbocharged Prius: high load mixtures were completely inconsistent, and at full
power, the electronic control system would momentarily shut the throttle.
Solving both problems took over two weeks of frustrating work.
So what was making the electronic throttle momentarily close itself at full
power? If in fact it was because the hybrid system couldn’t cope with the extra
power, the effect would be most pronounced when power was at its maximum... say,
on a cold night. And that’s just what proved to be the case. When the weather
provided me with an 11-degree C evening, I was able to get the throttle-shutdown
to repeatedly occur. (But what about those inconsistent full-load mixtures?
Well, if the air/fuel ratio meter was watched like a hawk, it was possible to
get in peak power runs with the mixtures satisfactorily rich. You just couldn’t do it
So the obvious answer to the automatically closing throttle was to lower
boost at the very top end, so reducing power. Since the greatest driving
improvement over standard occurred due to the mid-range boost, dropping boost a
bit at peak power would make very little difference. But how to decrease
boost? With the ex-Subaru IHI turbo’s wastegate connected, 7 psi was the minimum
boost available – and that’s what I was running. (The boost level had risen from
6 to 7 psi with the redesigned intake plumbing.) Modifying the turbo wastegate
actuator to achieve a lower boost was possible, but it would mean taking off the
turbo to access the wastegate actuator. Which in turn would mean removing a
front driveshaft, taking off the exhaust, draining the ‘gearbox’ of oil, and so
on. A lot of work.
Instead, was it possible to bleed off some of the boost? Already in the
system was a GFB blow-off valve running a shortened internal spring – a
modification performed on the valve when it was fitted to the supercharged car
as a recirc valve. In this design of valve, boost pushes on the base of the
piston, trying to open it against both the internal spring pressure and the
pressure in the boost-sensing hose. If the boost being fed in via the hose was
reduced, it was likely that the internal piston would be pushed open, so
allowing some boost to escape. By placing a bleed solenoid in the boost pressure
feed hose, the pressure in this hose could be regulated. By using the
Independent Electronic Boost Control (IEBC) kit, the action of this solenoid
valve could be mapped. (See
The Independent Electronic Boost Control, Part 1.)
The system was quickly set up and the solenoid valve opening set to 0 per
cent at lower injector duty cycles. (The IEBC kit sets its output duty cycles on
the basis of input injector duty cycles. That is, any relationship between
injector duty cycle – ie engine load – and output duty cycles can be set via the
hand controller.) At higher engine loads, the solenoid was gradually brought on
line, until at very high engine loads, it was completely open, so acting as a
bleed of the boost pressure in the feed hose to the blow-off valve.
The mapping of the control system was initially done very coarsely but the
system soon showed it was possible to drop boost to 5 psi at the top end of the
engine power band. This stopped the auto throttle shut-down procedure – the
system was sufficiently sensitive that even a 2 psi boost decrease made enough
Hmm, OK then – it was easy enough to stop the auto throttle shut-down with
very little loss in performance. But what about these bloody mixtures?
Not Solving the Air/Fuel Ratio Problem
It’s worth recapitulating what had so far been done to try to sort the high
A Simple Voltage Switch (SVS) kit had been fitted that allowed the two oxygen
sensors to be disconnected on the basis of measured airflow meter output, so
forcing the car into open loop. In the standard car this had resulted in the
mixtures automatically going very rich – too rich in fact. The Digital Fuel
Adjuster (DFA) kit had then been used to intercept the signal coming from the
standard airflow meter, which allowed these ‘oxy sensors disconnected’ mixtures
to be leaned out a little. However, once engine power had been increased by the
use of the supercharger and then – subsequently - a turbo, these ‘oxy sensors
disconnected’ mixtures had become leaner and leaner, and so for forced
aspiration, the DFA had been used to richen the mixtures (ie increase the level
of the airflow meter signal).
However, mixtures at full load were still too lean.
To ensure there was sufficient fuel flow and pressure, a new in-tank pump had
been installed, together with a new adjustable external regulator (a Malpassi
rising rate design) and a new external fuel filter. Installing the external
pressure reg had required fitting a new return line to the tank. This system
delivered plenty of fuel and allowed the adjustment of fuel pressure. The
injectors from a Corolla had also been trial-fitted, but despite coming from a
much more powerful engine, at the smaller duty cycles of the Prius system, had
proved to flow less fuel than the standard injectors.
Adjustment of the fuel pressure was then carried out which resulted in
adequately rich mixtures when the oxy sensors were switched out. In this
approach, the DFA did little – mixtures were adjusted by altering fuel pressure.
The Malpassi reg was plumbed so that it saw boost but not vacuum. In this way
the off-boost fuel pressure remained constant (as it does in the standard
system) but rose when on boost. This resulted in full-boost mixtures which were
satisfactory – at least at first.
At this stage the airflow meter sensing element was installed in a
custom-made body with a 31 per cent larger cross-sectional area. The DFA was
then used to lift the output voltage value of the larger airflow meter to
achieve correct closed-lop mixtures. New free-flow intake and intercooler
plumbing was also installed at this time.
Next, the mixtures became erratic at full-load. At times they were correct
(as set by fuel pressure and the DFA) and at other times, incorrect.
Furthermore, in successive full-throttle events they could be seen to be heading
back to stoichiometric, even with the oxy sensors disconnected.
An attempt was made to intercept the oxygen sensor outputs, using the DFA to
modify their levels at high loads and leaving the system always in closed loop
(ie oxy sensors connected). However, the car ignored this and maintained 14.7:1
The oxygen sensors were then disconnected at all loads, and an attempt made
to tune the mixtures throughout the load range with the DFA. (This was done with
great success on a Maxima V6 turbo, where the air/fuel ratio could be maintained
very accurately at all loads despite the lack of an oxygen sensor feedback
loop.) However, on the Prius it became clear that with the oxygen sensors
disconnected, it isn’t just the airflow meter input that is used to set
mixtures. This could be clearly seen because even with the engine held at one
load site (ie one airflow meter output voltage), the mixtures would gradually
slide back from being rich to 14.7:1. (This is what so strongly suggests the
presence of a look-up table that compares measured engine power with expected
fuel injector pulse width.)
So despite trying each of the following, consistent full-load mixtures could
not be obtained:
the airflow meter signal with the DFA
the oxy sensors in and out on the basis of throttle position
the oxy sensors out as fuel pressure was progressively increased
the oxy sensors out and intercepting the airflow meter signal with the DFA
the oxy sensors out, increasing fuel pressure and intercepting the airflow meter
signal with the DFA
the oxy sensor signals with the DFA
the oxy sensors at all loads and then setting the mixtures via the DFA working
on the airflow meter signal
It’s no understatement to say that by this stage I was pulling my hair out!
I then decided to go back to an approach tried a long time previously – one
Electronic Fuel Pressure Increase. This involved
switching a restriction into the return line from the regulator and so boosting
fuel pressure in one hit. My thoughts were these: if the increase in fuel
pressure occurred only when the oxy sensors had been switched out, the ECU
couldn’t be aware of the change. Therefore, even at the standard look-up table
pulse widths with which it was triggering the injectors after the oxy sensors
were disconnected, the mixtures would have to be richer.
(This approach contrasts with using a rising rate pressure reg that increases
fuel pressure progressively when on boost. Taking the rising rate fuel pressure
route, and switching the oxy sensors out only when boost over say 4 psi, means
that often the system is in closed loop with slightly heightened fuel pressure.
If the ECU can indirectly learn fuel pressure by looking at the injector pulse
widths required to maintain an air/fuel of 14.7:1, then maybe the ECU can pull
back all injector pulse widths to compensate for this increased fuel pressure.
In other words, despite not having a fuel pressure sensor, the ECU can probably
still calculate fuel pressure, and compensate for it. I don’t want it to do
I then set up the system to allow the fuel return line from the pressure reg
to be restricted, increasing fuel pressure. By closing a solenoid, fuel is
forced to flow through a ball-valve, which comprises a restriction that can be
varied in flow. Shut off the ball valve and when the solenoid closes, the fuel
pressure will rise to the maximum the pump can flow (which is normally limited
by an internal pressure relief valve). And in fact testing soon showed that to
get adequate fuel pressure, the ball valve did have to be completely closed –
ie, the pressure reg’s return shut off. This resulted in an immediate increase
in fuel pressure of 20 psi – from 50 to 70.
And what were the mixtures like when the oxy sensors were switched out and
simultaneously the solenoid was closed? Finally – finally! – they were
consistently close to what was desired – around 12.5:1 at full load. I rebuilt
the underfloor bracketry to include the solenoid and performed a final on-road
But again the mixtures at full load were
WTF was going on this time? Watching the fuel pressure gauge provided the
clue: even with the return line shut off, the fuel pressure was now rising to
only 60 psi – 10 psi of high load fuel pressure had been lost. Repeated testing
showed that with the fuel pressure reg’s outlet closed off, the fuel pressure
could jump to anything from 60 to 80 psi...
Basically, it appeared that the pump couldn’t cope with the increased fuel
pressure requirements. I checked the voltage at the pump to make sure that it
was getting full battery voltage - it was. This meant another pump was needed –
but the in-tank pump had already been replaced and getting a physically larger
pump in place would be near impossible. The only alternative was to add a second
in-line fuel pump – one that could consistently produce fuel pressure of at
least 70 psi. And to regulate this fuel pressure – which is much safer than
using the pump’s internal pressure relief valve – a second regulator would be
The New Fuel System
Many turbo and supercharger upgrades use a second fuel pump in series with
the first, with the second pump activated only when the extra fuel is needed. In
this approach, the first pump flows fuel through the second pump which is
usually switched off. However, it is better to use a bypass one-way valve around
the second pump so that the fuel can freely flow past it. When the second pump
switches on, the extra pressure it generates closes the one-way valve.
The complete fuel system is shown in this diagram. The in-tank pump feeds the
second pump with the one-way valve providing the bypass. The two pressure regs
are plumbed in parallel, with a solenoid valve positioned in the feed line of
the lower pressure reg. When the fuel pressure needs to be increased, the second
pump is switched on and the solenoid closed. This forces the second reg into
action and the fuel pressure is then regulated at the higher value. It’s a
complex and relatively expensive system but it’s the only approach that gives
two fixed (but adjustable) fuel pressures, with the second pressure much higher
than conventional fuel systems normally use.
The system was installed under the car and testing carried out. Initially I
just used a manual switch to turn on the new fuel pump and turn off the solenoid
(the two steps that result in high fuel pressure). The oxy sensors were being
disconnected by a Simple Voltage Switch working on the airflow meter output
voltage (ie when airflow reached a certain value, the oxy sensors were
disconnected) and when I heard that relay click, I manually toggled the fuel
pressure increase switch. And watched the fuel pressure gauge. And watched the
boost gauge. And watched the MoTeC air/fuel ratio meter....
On-road testing showed that the system worked – consistent rich mixtures on
high loads with no apparent learning around them occurring. I then added two
relays – one the changeover from solenoid-off-to-fuel-pump-on, and the other a
power feed relay for both the solenoid and the pump. A second relay was also
added to the Simple Voltage Switch to trigger the fuel pressure changeover, so
that the oxy sensor switching and fuel pressure switching occurred
I then decided to see what would happen if I swapped the switch-over input
signal from being triggered by the airflow meter to being triggered on the basis
of accelerator pedal position. Initially, this seemed to work even better – at
full throttle, you always got rich mixtures, irrespective of engine airflow –
but then I found if the Prius was booted at full throttle from a standstill, the
mixtures returned to stoichiometric. (How many times has this return-to-stoich
now occurred?!) It appears the car needs to be running in closed loop for at
least a little while before the switch-over occurs.
High, High Fuel Pressure
The fuel pressure required to consistently gain a full-load air/fuel ratio of
around 12:1 – 12.5:1 is significant – about 85 psi. All the fuel hoses at the
rear of the car have been replaced with hose good for 290 psi, and I intend to
also replace the underbonnet injector fuel rail feed hose.
But with the high load air/fuel ratio problem solved, and the throttle
shut-down problem consigned to the history books, I could finally see the light
at the end of the tunnel: touch wood, that was it for the big problems. (Yes, it
takes real skill to get three clichés into one sentence. Was this car really
driving me mad or was I always like this?)
Now we could drive the turbo Prius, confident in its mixtures and
performance. Hmmm, performance... so
what’s it like? And what’s happened to the fuel economy? Next week, in the final
in this series, we’ll find out.
Did you enjoy this article?
Please consider supporting AutoSpeed with a small contribution. More Info...