fe1rx Does a Deep Dive Into the N54 LPF System

[fe1rx]

This is part 1 of a deep dive into the N54/N55 low pressure fuel (LPF) system. It is motivated by the challenges that have led me to a Stage 2 LPFP module (Precision Raceworks) and a Stage 2 EKPM3 (ET3 Design) to address problems with my low pressure fueling. My car is a 2008 N54 running a Stage 1+ 93 Octane MHD tune, and until recently still had its original LPFP and EKPM2. Declining LPF pressure on track led me first to the Stage 2 pump and improved but erratic and declining pressure led me to the Stage 2 EKPM3. That experience prompted me to gain a better understanding of exactly how the LPF system operates. What follows will be based on internet research, reverse engineering of CAN data, OBDII live data, direct voltage and current measurements and flow testing both in-car and on the bench. The topics consist of the following:

Part 2) OE LPFP Module Construction and Performance.
Part 3) OE LPFP Jet Pump Construction and Performance. The function of the suction jet pumps (1 in the fuel tank module, and one at the fuel pressure regulator) – how much fuel they consume, and how much fuel they pump.
Part 4) OE LPF Pressure Regulator Function and Performance.
Part 5) Used LPFP Performance. Actual flow performance of two 2008 build OE LPF pumps after 17 years in service.
Part 6) Stage 2 LPFP Construction and Performance. How the TI/Walbro F90000274 pump differs in construction from the OE pump, and what its flow potential is when jet pump flows are accounted for.
Part 7) DME and EKPM Communication on PT-CAN. The relevant LPF system data available on PT-CAN and how it is encoded.
Part 8) Electrical Considerations for Upgraded LPF Pumps. The effect of an upgraded fuel pump on current flow, voltage drops within the system and the benefits of upgraded wiring on system function, and of an upgraded EKPM.
Part 9) LPF System Control Laws. How the system is able to adapt to pump degradation over time or to accommodate a significantly larger fuel pump installation.

If I had known I was heading in this direction earlier, I would have logged more data with the OE pump and EKPM2 installed and would have started years ago! Since I didn’t do that, my OE pump measurements were done on the bench. I have the benefit of having 2 OE LPFP modules on hand, both from 2008 N54 135i vehicles. I tested both for flow characteristics, and then disassembled one. Having 2 gave me access to a spare suction jet pump, which was also useful.

As far as units of measure go, I will use a mixed bag. Most fuel flows I will report in liters per hour (lph) because fuel pumps are rated in those units, and they are an easy unit to use for actual flow testing. They are also the native units used in PT-CAN communication. Other sources refer to flow in lbs/hour (lb/h) and US gallons per hour (USgph) so in some cases I will convert to or from those units.

Fuel pressures will generally be in psi because they are familiar and correspond to my pressure gauges and most OBDII live data and pump datasheets, with occasional reference bar because they are the native units used by the DME. The conversion here is simple – the LPF system strives to operate at 5 bar (72 psi) most of the time.

A – PHYSICAL CONFIGURATION

A schematic of the fuel system from the Bentley Service Manual illustrates the physical configuration of the LPF system. It is well known that the low-pressure fuel system strives to operate at a constant 5.0 bar (72 psi), except during engine warmup, where it operates at a higher pressure. The fuel tank is saddle shaped due to drive shaft tunnel. The fuel system incorporates two (venturi) suction jet pumps that transfer fuel within the tank. Item 11 moves fuel from the RH half of the tank into the fuel pump bucket. Item 10 moves fuel from the LH half of the fuel tank to the fuel pump bucket.

One of the jet pumps is integral to the OE fuel pump module. The other is integral to the fuel pressure regulator assembly. We will discover that a single jet pump consumes approximately 55 lph.



Importantly, the jet pumps are both supplied with high pressure fuel at all times (not fuel that has been relieved through the pressure regulator as sometimes reported). All jet pump flow, and any bypass flow from the pressure regulator is delivered to the fuel pump module (bucket) to keep it full, regardless of fuel tank level.

B – ELECTRICAL CONFIGURATION



A fuel pump wiring diagram reveals that the EKPM is not connected to any sensors directly, so any data it needs to operate must arrive via the PT-CAN bus, from the DME. This is useful to know, because all the communication on the PT-CAN bus is readily accessible using a CAN sniffing tool (aside from the fact that it needs to be decoded). Once decoded it can be routinely logged on my dash logger.

(While it would seem logical that a control system that aims to maintain a constant fuel pressure by modulating fuel pump voltage (via pulse-width modulation) would use closed loop control based on measured fuel pressure, this is not how the BMW system works. The actual control laws are open-loop based on a fuel flow request from the DME to the EKPM, and a voltage output by the EKPM based on a lookup table. Importantly, there is a closed-loop adaptation that occurs that allows the system to (slowly) revise the lookup table to accommodate pump degradation over time, or installation of a more capable pump.)

EDIT: with the benefit of hindsight I would rewrite the paragraph above, but I will leave it there for the sake of continuity. While many of the parts of this post were written in parallel, an understanding of the control laws has been the most ellusive. I will attempt to achieve more accuracy in Part 9.

[fe1rx]

PART 2 - OE LPFP Module Construction and Performance

A – LPFP MODULE



The LPFP is housed within a module (bucket). The fuel system ensures that the bucket always remains full, even when the tank level is low. This provides several useful benefits:

- The fuel supply to the pump is assured until all fuel is exhausted.
- A positive head of fuel is maintained at the fuel pump inlet. This is important because the pump has very poor self-priming capability. As the pump is internally lubricated by fuel, it will be damaged if run dry.
- Fuel surrounding (and within) the pump is used for pump cooling.

The 10-part Aeromotive Fuel System Guide provides a useful primer on modern EFI fuel system design practices.

https://aeromotiveinc.com/pages/fuel-system-guide

The section dealing with fuel modules is particularly relevant, and it provides a strong argument against “upgrading” with a bucketless low pressure fuel pump.

https://aeromotiveinc.com/pages/fuel...iable-delivery

 
B – OE LPFP CONSTRUCTION



Unlike aftermarket in-tank electric fuel pumps, the OE pump has an integral jet pump outlet port, which makes for tidy packaging.



The main pump flow and jet pump flows are completely independent, with separate inlets and outlets.



The pump assembly can be thought of in three segments: 1) the housing, containing the brushes and main fuel outlet port, 2) the permanent magnet motor, and 3) the pump. The motor is surrounded by fuel for cooling, and it utilizes an 8-pole carbon commutator. This pump showed no significant wear on the brushes or commutator after approximately 190,000 km in service.



The pump is a regenerative turbine side channel design. The turbine is a graphite composite, as is the lower motor bearing. The flow paths for the jet pump flow and main flow are completely independent. The upper and lower housings are aluminum. This pump has approximately 190,000 km in service and no significant wear is apparent on either the housings or the turbine.



The upper housing is plastic, and it contains the carbon brushes and the upper bearing, which is plastic. This construction requires constant immersion in fuel for lubrication and cooling. The outlet contains a flow straightener. Unlike most aftermarket in-tank fuel pumps the OE pump does not have a check valve at the outlet, or an overpressure relief valve, as both these functions are handled by the separate in-tank pressure regulator module.



The carbon composite turbine has two independent sets of vanes. The vane pitch is variable on both for noise reduction. The turbine requires no thrust bearing as it floats on a film of fuel, with pressures on each side of the turbine balanced by means of bleed holes.



The turbine “blades” are not foil-shaped blades in the sense of an axial flow turbine but are V-shaped buckets that rely on vortex flow to impart momentum on the fuel in the side channels. When the side channels end, the only path the moving fuel can take is out the exit ports.

C – OE BUCKET CONSTRUCTION



The pump is contained in the module bucket, which incorporates a foot valve through which fuel can enter but not leave the bucket. Short standoffs hold the foot valve slightly above the bottom of the tank, allowing any fuel on the bottom of the tank to access the foot valve.

The jet pump is located directly above the foot valve. A small high-pressure jet of fuel into the jet pump creates a suction which entrains the fuel above the foot valve, ejecting a large volume of low-pressure fuel into the bucket. A later post will look at the jet pump flow characteristics in detail.

The net flow into the bucket is easily able to keep the bucket overflowing even at maximum engine fuel consumption.

D – OE LPFP FLOW

BMW does not publish the fuel flow rating of the OE pump and little has been published on the web on the subject. During the early days of the N54 aftermarket development, Shiv Pathak made a couple of particularly informative posts documenting how the fuel system works and providing flow measurements for both a new and a 1-year-old LPF pump. His measurements were made in lbs per hour and were made at non-standard pump voltages. EFI Fuel pumps are traditionally characterized at 12.0 Volts and 13.5 Volts and at a range of operating pressures. As our system is intended to operate at a constant 72 psi and can supply the pump with 13.5 Volts, only that operating point is required to determine the maximum useful flow capacity of the pump.

A full view of Shiv’s investigation can be found in these two posts:

https://www.e90post.com/forums/showthread.php?t=512759

https://www.e90post.com/forums/showthread.php?t=515398

I summarize his findings below adding a conversion to USgph (noting gasoline has a density of 6.0 lbs per US gallon) and lph (noting that 1 US gallon = 3.785 liters):

New Pump Fuel Flow (11.0 V @ 72 psi): 228 lb/h (38.0 USgph, 144 lph)
New Pump Fuel Flow (14.0 V @ 72 psi): 356 lb/h (59.3 USgph, 225 lph)

Used Pump Total Fuel Flow (11.0 V @ 72 psi): 204 lb/h (34.0 USgph, 129 lph)
Used Pump Total Fuel Flow (14.0 V @ 72 psi): 300 lb/n (50.0 USgph, 189 lph)

Using a straight-line interpolation, we can estimate the fuel flow at 13.5 Volts:

New Pump Total Fuel Flow (13.5 V @ 72 psi): 335 lb/h (55.8 USgph, 211 lph)

Used Pump Total Fuel Flow (13.5 V @ 72 psi): 284 lb/h (47.3 USgph, 179 lph)

A point missed in Shiv’s posts is that all this fuel flow is not available to support the engine. Some of it is consumed by the jet pump in the fuel regulator module. For reasons that will become clear in a later post we will assume that the jet pump flow is:

Jet Pump Flow @72 psi: 87 lb/h (14.5 USgph, 55 lph)

The usable fuel flow is the total flow, less this value.

New Pump Usable Fuel Flow (13.5 V @ 72 psi): 248 lb/h (41.3 USgph, 156 lph)

Used Pump Usable Fuel Flow (13.5 V @ 72 psi): 197 lb/h (32.8 USgph, 124 lph)

Following Shiv’s lead with an assumed BSFC of 0.55 lbs per hp hr, these fuel flows will support the following hp:

HP (New Pump @13.5 V and 72 psi) = 248 / 0.55 = 450 hp
HP (Used Pump @13.5 V and 72 psi) = 197 / 0.55 = 358 hp

While these numbers are comparable to those calculated by Shiv, he neglected the jet pump flow which inflated his estimate, while also assuming an unreasonably low fuel pump voltage (11 V) which deflated it. The net effect is that our estimates are similar.

Shiv’s used pump was only one year old, so the degradation in performance in just that time is quite dramatic. Part 5 will document fuel flow measurements from two 2008 build OE fuel pumps, both removed from service in 2025 after 140,000 and 190,000 km, and these pumps exhibited significantly more flow degradation.

Based on these observations, we need to treat the LPFP as a life-limited item, and to consider the OE LPFP as marginal in capacity for any significant power increases.

[fe1rx]

PART 3 - OE LPFP Jet Pump Construction and Performance

A – PHYSICAL CONFIGURATION



Two of the apparent ports on the jet pump at the base of the LPFP are in fact blocked by a steel ball pressed into the plastic casting. Thus, the fuel makes a sharp turn before exiting the nozzle. The inlet flow is entrained by the jet flow.



The nozzle is very small.



The jet pump at the Low-Pressure Fuel Regulator appears to be similar construction, and we assume the orifice size is the same.

Both jet pumps see the full outlet pressure from the LPFP, which under most circumstances is 72 psi. Thus, we expect a fairly constant fuel flow rate through each of the jet pump orifices.

B – FLOW TESTING

For safety, flow testing of a pump base jet pump was accomplished using low-VOC low-odour paint thinner (Stoddard Solvent/Varsol) in lieu of gasoline. The flow characteristics of the two fluids are assumed to be similar.

Fluid was supplied from a LPFP module immersed in a bucket of the solvent and powered by a regulated constant voltage DC power supply. The voltage was adjusted until the desired fuel pressure was applied as measured by a pressure gauge teed into the output line.

Two different tests were performed. The first measured the fuel flow rate through the orifice, which is to say the fuel consumption of the device. The second measured the fuel output of the device, which is a combination of the orifice flow and the entrained flow.

 
C – ORIFICE FLOW TESTING



To accomplish the orifice flow test, the jet pump was placed at the bottom of a large, graduated container containing 0.5 liters of solvent, sufficient to cover the jet pump.



The electric pump module was in a separate bucket with sufficient fluid to complete the test.



The pump was powered by a regulated constant-voltage power supply set to achieve the target fuel pressure. To discover how the fuel consumption varied with applied pressure, the test was performed at 40 psi, 60 psi, 72 psi and 79 psi.

The pump was operated until an additional 4 liters of fluid was pumped into the container, with the time taken recorded.

D – TOTAL FLOW TESTING

To measure the total output of the jet pump (orifice flow, plus entrained flow), the jet pump was placed in the same container the electric pump, but the output was plumbed to a separate empty graduated container.



A large hose provided minimal restriction to the output flow and was directed to the graduated container.



The electric pump was then operated until 4 liters of fluid was pumped into the container, with the time taken recorded. To discover how the total output flow varied with applied pressure, the test was performed at 40 psi, 60 psi, 72 psi and 79 psi.

E – TEST RESULTS

The flow through the orifice as a function of pressure is well approximated by a second order polynomial. Flow at the normal 72 psi operating pressure is approximately 55 lph for a single jet pump. This is about 0.24 US gallons per minute, which is close to the 0.2 US gallon per minute estimate provided by Precision Raceworks – the only credible estimate I have found on line for jet pump fuel consumption.

https://precisionraceworks.com/en-ca...39390226284737



The total output of a single jet pump at 72 psi is impressive at approximately 410 lph. Another way to look at the total output flow is to note that it is approximately 7.5 times the orifice flow. The jet pump turns a low-volume high-pressure jet flow into a high-volume low-pressure bulk flow that is well suited to moving fuel within the tank to the advantage of the electric fuel pump.



F – SIGNIFICANCE

The jet flow represents an overhead on the LPF system, in that main fuel that flows through the jet pump(s) is not available to fuel the engine. As the OE LPFP Module has a separate flow for the module jet pump, only the regulator jet pump is an overhead on that module, which is to say 55 lph. Aftermarket LPFP Modules must use main flow to power the module jet and the regulator jet, so in that case the total overhead is 110 lph at 72 psi.

To provide an example, assuming an aftermarket LPFP Module incorporates an electric pump that can supply 350 lph at 72 psi, the total fuel system will only be able to supply 240 lph to the engine. When considering an upgraded LPFP Module to support increased power output, this overhead needs to be considered.

Of course, this analysis assumes that all jet pumps have similar orifice sizes and flow rates, and to the extent this is not true, the conclusions will need to be adjusted accordingly.

[fe1rx]

Part 4 – OE LPF Pressure Regulator Function and Performance

The question arises, if the EKPM regulates the LP fuel pressure by pulse-width modulating the pump voltage, why does the system require a mechanical fuel pressure regulator?

The answer revealed in this part is that the “pressure regulator” would more accurately be characterized as an “overpressure relief valve”. This is because the regulator set point is significantly higher than the normal 72 psi (5.0 bar) operating pressure.

This type of over-pressure relief function is well described by Aeromotive in Part 4 of their training series:

https://aeromotiveinc.com/pages/fuel...ypes-functions

“On Direct Injected (DI) engines with mechanical pumps this (over pressure due to thermal expansion on shutdown hindering injector function, ed) is typically not a problem, but having excessive fuel pressure on lines and fittings is not a safe condition. Hence, a safety overpressure regulator is built into the modern OEM fuel modules. In the case of most OE fuel modules, this pressure is in the 75-90psi range”

My testing was accomplished on a fully installed fuel system using a Precision Raceworks Stage 2 non-modular bucketed fuel pump, which incorporates a TI/Walbro F90000274 (450 lph) fuel pump, and an OE fuel pressure regulator module. The pump was powered by a regulated dc power supply operating in programmable constant voltage mode instead of the EKPM. This power supply was wired directly to the pump with 10 AWG cables. Fuel pressure was monitored using the on-board fuel pressure sensor, accessed through OBDII live data using MHD. After initially bleeding off the fuel pressure, the pump was run in constant voltage mode at increasing voltages, with the associated voltage and fuel pressure being noted. The conclusions are not affected by the pump being used.

During this testing the engine was not running, so the fuel being pumped was being consumed by the venturi jet pumps and remaining in the tank, until the mechanical pressure regulator reached its set point, at which time it relieved sufficient flow (again back to the tank) to hold the system pressure close to the mechanical set point.

The pressure characteristics reveal a very distinct knee, showing that the set point is approximately 5.45 bar (79 psi). Under normal operating conditions the observed fuel pressure is less than 5.5 bar, meaning that under normal operating conditions no fuel flows through the mechanical pressure regulator and pressure regulation is strictly under PWM control.



As the fuel flow, as represented by fuel pump voltage, initially increases, all the flow is consumed by the jet pumps (blue line). When the fuel pressure reaches and exceeds the fuel pressure regulator set point, an increasing amount of fuel flows through the regulator bypass (red line), clamping the fuel pressure close to the set point. Assuming the EKPM can regulate the fuel pressure within 72 ±5 psi (within green dashed lines), the fuel pressure regulator will never flow any fuel.

To test this hypothesis, we went for a drive and logged the required data (the upcoming Part on PT-CAN communications will outline how that was done). In this case the fuel pump was powered normally by a Stage 2 EKPM3 and the system was well-adapted. The data includes idling, cruise and several full-power 3rd gear pulls. The data represents over 700 data points logged at 1 Hz, so more than 10 minutes of operation. Pressures are logged in increments of 1 psi and voltages in increments of 0.1 Volts, so many of the data points consist of multiple hits.



This data confirms our hypothesis that under normal operation the mechanical fuel pressure regulator bypass flows no fuel and all pressure regulation is accomplished by the EKPM and DME. The data points outside our assumed normal operating range represent single data points under transient conditions (e.g., abrupt throttle closure) and are not significant.

As we will confirm in later Parts, the DME requests a fuel flow from the EKPM, not a fuel pressure. The DME knows at every moment how much fuel the engine is consuming (it controls the fuel injectors) so when the operating conditions change it can request an appropriate incremental change in fuel flow. The EKPM controls the fuel pump by providing a specific voltage to the pump. To do this, it monitors the voltage supplied to itself and then calculates the appropriate duty cycle for its pulse-width-modulation (PWM) to produce the required voltage at the fuel pump. The specific voltage needed is a function of the voltage vs. fuel flow characteristics for the system. It learns this through the adaptation process.

That our system strives to operate at a single fuel pressure is very convenient. It allows us to derive the voltage vs. fuel flow characteristic curve for our system. Because the curve is non-linear, we need 3 data points to define it. Two of them can be obtained from the pump datasheet, and the third, we learned from our jet pump orifice flow testing. The datasheet data points are the fuel flows at 72 psi at 12.0 Volts and at 13.5 Volts.



Some interpolation and conversion of units is required, but from the datasheet we get these two points:

305 lph at 12.0 Volts and 72 psi
360 lph at 13.5 Volts and 72 psi

The third point is our measured jet pump fuel consumption for two jet pumps. The voltage at this condition we get from our fuel pressure regulator measurements:

110 lph at 8.4 Volts and 72 psi

Here we confront a fact of EFI fuel pumps – while this pump is nominally in the 450 lph class, it can only achieve that flow rate below about 40 psi. The flow capability of all electric fuel pumps drops with increasing pressure.

Plotting our three data points we get the following:



While the blue line represents the total output of the pump, for the usable flow we must subtract the 110 lph overhead flow required by the two jet pumps. This characteristic curve is what the EKPM needs to “know” to calculate the required fuel pump voltage for a given fuel flow. The characteristic curve for the OE pump is vastly different. The fact that the system can accommodate a wide range of pump characteristics through adaptation is a tribute to its robust design.

[fe1rx]

PART 5 – OE LPFP Testing

This part documents the flow testing of two used OE LPF Pumps. Both were the original pumps installed in 2008 135i cars, and both were removed in 2025 as they demonstrated inadequate performance in service. This was evidenced by an inability to maintain the required 72 psi fuel pressure at the high-pressure fuel pump inlet under all operating conditions.

Flow testing was done on the bench utilizing low-VOC low-odour solvent for safety purposes. This practice is consistent with the Walbro datasheets, which indicates that their test fluid is “142-66”, which is a medium-flash point low-aromatic mineral solvent.



The older of the two pumps “A” had accumulated about 190,000 km and it had a generally dirty appearance, which we attribute to the assumed use of non-top-tier fuel. The “dirt” consisted of a thin layer of soft sediment that could be mechanically removed quite easily. This pump was subsequently disassembled as was shown in Part 2, and despite its age and assumed fuel use showed no sign of internal damage or significant internal wear.

The younger of the two pumps “B” had accumulated about 140,000 km and was pristine, having always been run on top-tier fuel (by the author). This pump was not disassembled.

Part 2 referred to significant performance degradation in LPF Pumps after only 1 year in service (per Shiv Pathak’s testing). Our testing of “A” and “B” reveal significant further degradation with age and use, despite no obvious internal defects or wear being detected when “A” was disassembled. Clearly these pumps are very susceptible to the slightest wear or contamination.

Testing was accomplished with the original filter sock and jet pump in place. The pump output was restricted by a fitting with a drilled orifice and the line was teed to a pressure gauge to measure pump discharge pressure and another tee to a second jet pump (representing the pressure regulator jet pump). The flow from the jet pump nozzle was returned to the source container.



The flow through the drilled orifice was directed to a separate container. This flow represents the usable output of the pump.



The pump was operated until the collection container had collected 3.9 liters and the time taken to achieve this was recorded so that the usable flow rate could be calculated.


The pumps were powered by a constant voltage DC power supply and the voltage was adjusted to either 12.0 Volts or 13.5 Volts as these are the standard voltages used in pump performance charts.



Several different orifices were utilized so that the flow data could be collected over a range of pressures, to ensure that either by interpolation or extrapolation the flow rate at 72 psi could be determined. The orifices were made by drilling and tapping a plastic tee fitting ¼-20, inserting a plastic screw of the same thread, parting off the protruding part and then drilling an orifice down the middle of the screw.



The flow rate at 72 psi and 13.5 Volts is the primary point of interest, as it represents the highest flow rate achievable in service. Because the flow from both jet pumps was excluded from the collected fuel, the rate measured in this manner represents the usable flow rate for the pump.

FILTER SOCK

While the filter socks start out life white, the socks on both A and B were discoloured. Opening the A filter sock revealed no sign of blockage, and the filter material was still freely porous.



PUMP INTERNAL WEAR

I disassembled pump A, and any previous internal images I have posted are of that pump. Inspection of the pump housings and turbine show little perceptible wear. Measuring them shows that the turbine has a thickness of 4.10 mm and the upper pump housing has a depth of 4.18 mm. The lower housing is flat. Wear on the lower housing is less than 0.005 mm and on the upper housing appears to be less than that (both are barely perceptible to a fingernail test for step wear). Thus, the side clearance in the pump is at most 0.09 mm (0.0035”) on a well-used pump (half that on each side of the turbine). This is likely only double the original side clearance, and still a very close fit, so clearly any wear at all significantly affects the pump’s capabilities.



ELECTRIC MOTOR

The motors on both pumps ran smoothly and quietly with no significant vibration. Both the brushes and commutator on pump A appear to be in very good condition. The voltage required to start the pump A turning under no load is only about 1.0 Volts. So electrically and mechanically both pumps appear to be in good condition. The only possible significant degradation appears to be hydraulic (i.e., pump internal wear).



SUMMARY OF RESULTS (USABLE FUEL FLOW RATE)

Shiv New Pump at 13.5 V and 72 psi = 156 lph (100%)
Shiv Used Pump at 13.5 V and 72 psi = 124 lph (79% of new)
Pump A at 13.5 V and 72 psi = 87 lph (56% of new)
Pump B at 13.5 V and 72 psi = 98 lph (63% of new)

[fe1rx]

PART 6 – THE STAGE 2 LPFP MODULE UPGRADE

Vendors (Fuel-It, Precision Raceworks (PR)) advertise the Stage 2 LPFP upgrades as plug and play – a simple matter of replacing the original LPFP Module with an upgraded module containing either a TI/Walbro F90000274 pump (Stage 2) or F90000295 (Stage 2.5). The implication is that the fit is identical to original, and the wiring, fuse and EKPM are all compatible.

The claim is that Stage 2 pumps will support up to 575 hp and Stage 2.5 pumps 650 hp on pump gas. This assumes that our engines require 0.55 lbs of fuel to provide 1 hp for 1 hour. This is a commonly assumed Brake Specific Horsepower (BSFC) for advanced turbocharged engines, with units in lbs/(hp.hr). Accordingly, we can convert these power values into associated effective fuel flow values, by noting that 1 US gallon of gasoline weighs 6 lbs:

Stage 2: 575 x 0.55 x 6 = 52.6 US gph (= 0.9 US gpm = 200 lph = 3.3 lpm)

Stage 2.5: 650 x 0.55 x 6 = 59.6 US gph (= 1.0 US gpm = 226 lph = 3.8 lpm)

The PR website confirms this logic, except their flow numbers for Stage 2 are clearly wrong (let’s call that a typo):



For the claimed supported power levels to be credible, the fuel system must be able to deliver these effective fuel flow rates to the engine. This is not simply a function of the pump, but also of the entire system (jet pump flow and voltage supplied to the pump). We will examine in this Part if those flow rates are credible for a plug and play installation but will leave the EKPM and voltage questions to Part 8.

It should be obvious that adequate fuel flow potential is a foundational requirement, and does not, it itself, generate power. Other modifications will be needed to make full use of the available fuel.

NEW MODULE VS. DIY MODULE

A bare TI/Walbro F90000274/295 pump is much cheaper than a complete ready to go Stage 2/2.5 Module, so you might be tempted to DIY the upgrade. If your modification retains the foot valve jet pump, it should work. This would require teeing the output from the pump and running one leg to the foot valve jet pump and securing the jet pump to the bottom of the bucket. PR does this with their DIY kit:

https://cdn.shopify.com/s/files/1/05...f?v=1624806609

The 274/295 pump fuel outlets utilize an O-ring fitting, while the vehicle plumbing expects a QDM (quick disconnect male, aka Bundy) fitting, so an adapter is required. This needs to be a tee adapter to provide a source for the jet pump. None of the YouTube DIYs I have seen get this right. Something like this might be a good start:

https://performancefuelingsolutions....41508494311489

Then this:

https://performancefuelingsolutions....ducts/small-90

With some suitable hose, to get you to something like PR’s solution:



So, it could be done as a true DIY, but the PR kit will save you a lot of bother with respect to working out the details.

UP THE CREEK WITHOUT A BUCKET

Of course, bucketless “upgrades” are also an option, and are much cheaper. If Parts 2 and 3 did not adequately explain why they are a bad idea, I have failed.

STAGE 2 TI/WALBRO F90000274 “450 LPH” PUMP

The F90000274 is nominally a 450 lph class pump (based on Walbro’s vastly optimistic numbering scheme) but it only approaches this flow rate with unrestricted flow (0 psi). At 72 psi the pump will output about 80 US gph (303 lph) at 12 Volts and 95 US gph (360 lph) at 13.5 Volts. Maximum current draw should be below 19 Amps under all expected operating conditions. While the datasheet contains a lot of data points, only these ones are really of interest in our application.



Of course, we must account for the jet pump flow, which will reduce these values by about 29 US gph (110 lph). This would result in a usable flow of 51 US gph (0.85 gpm, 193 lph, 3.2 lpm, 306 lbs/hr) at 12.0 Volts. This is just shy of the 200 lph we established as the requirement for 575 hp, but achievable if our system can provide the pump with a bit more than 12.0 Volts.

F90000295 “535 LPH” PUMP

The F90000295 is nominally a 535 lph class pump with unrestricted flow. At 72 psi the pump will output about 93 US gph (350 lph) at 12 Volts and 108 US gph (409 lph) at 13.5 Volts. Maximum current draw should be below 22 Amps under all expected operating conditions.



Of course, we must account for the jet pump flow, which will reduce these values by about 29 US gph (110 lph). This would result in a usable flow of 64 US gph (1.1 gpm, 242 lph, 4.0 lpm, 384 lbs/hr) at 12.0 Volts. This meets the 226 lph we established as the requirement for 650 hp, provided our system can provide the pump with 12.0 Volts.

PLUG AND PLAY?

We have established that the supported power level claims for the Stage 2 and 2.5 pumps are reasonable, provided out existing EKPM and wiring can provide these pumps with at least 12.0 Volts. The answer to that is NOT an unequivocal “yes”, but that issue will be looked at in detail in Part 8.

WALBRO 39/50 PUMPS

Both the 274 and 295 pumps are part of the 39/50 fuel pump family, which designates a 39 mm body diameter (same as OE) and a larger 50 mm pump diameter. While the internal details of this pump are not important, they are interesting. While the OE pump utilizes two sets of turbine blades and two independent side channels – one supplying the engine, the other the module jet pump, the Walbro pumps also have two sets of turbine blades, but these feed two parallel side channels that combine and exit at a common port. So, like all aftermarket fuel pumps, the module foot valve jet pump must be supplied from the main fuel flow.

While the OE pump does not incorporate either a check valve or an overpressure relief valve, the Walbro pump does. All systems need these features, so it is prudent for an aftermarket fuel pump to include them, but they provide no particular benefit in our installation because the pressure regulator assembly in the LH side of the tank provides those features. The pump’s pressure relief valves are set well above the in-tank mechanical pressure regulator’s set point of 79 psi, so should never be called to action. A check valve is needed for the LPF system to maintain prime at the HPFP. This function is provided by a check valve in the in-tank mechanical pressure regulator module, and the addition of a second check valve in the LPFP provides no additional benefit.

An interesting look at the internal construction of these pumps can be found here:

[fe1rx]

PART 7 – DME and EKPM COMMUNICATION on PT-CAN

This part identifies the LPF system data transmitted on the PT-CAN bus, the parameter that each CAN ID represents and its encoding. The point is to enable this data to be logged on my AiM dash logger as a means of monitoring the health and wellness of the system as part of a post-track review.

Potentially useful data includes LPF system voltages, currents, duty cycles, flow rates and fuel pressure. All these parameters are correlated – higher pump voltage results in higher pump current, requires higher duty cycles, results in higher flow rates and/or higher fuel pressure. Because of this, careful corroboration of any observed parameter’s function is needed before decoding can be considered final.

Duty cycles need some elaboration. The duty cycle of a pulse-width modulated input voltage is the ratio of on time to cycle time, expressed as a percent. The modulated output voltage has an average value equal to the input voltage times the duty cycle.


Fig. 1: LPFP PWM Output Voltage

The image above is of the LFPP output voltage at idle. The EKPM3 operates at a constant 20 kHz, and at this moment is operating at 62.7% duty cycle. The peak-to-peak voltage far exceeds the input voltage because the on/off transitions result in oscillatory transients that increase the on peak and drive the off peak below 0. The average voltage of 8.23 Volts and 62.7% duty cycle imply that the input voltage is (8.23 / 0.627) = 13.3 Volts (the peak value, relative to ground, without the transients). The EKPM does not see the battery voltage (typically 14.2 Volts) at its input terminals, because of voltage drops in the wiring (a subject we will look at in Part 8).

We have seen in our examination of the LPF pump performance that the flow rate is proportional to the pump input voltage. The job of the EKPM is to provide the required voltage to the pump, and it does this by PWM of the EKPM input voltage. As that voltage fluctuates with electrical load and alternator output, the EKPM must adjust its duty cycle to account for those fluctuations.

After learning what the LPF pump’s flow potential is through the adaptation process, the DME requests some % of that capability by means of what I will call the Normalized Duty Cycle, which is a request to the EKPM. I call it normalized because it assumes that the EKPM input voltage is 14.2 Volts. The EKPM adjusts this request in real time for fluctuations in the EKPM input voltage relative to 14.2 Volts by multiplying the Normalized Duty Cycle by (14.2 Volts / Actual EKPM Input Volts), resulting in the Adjusted Duty Cycle. The EKPM applies some adjustments to this value and outputs the Actual Duty Cycle (we will see if we can make sense of these final adjustments in Part 9).

The AiM logger reads some engine data on the PT-CAN bus via AiM’s decoding of the proprietary BMW PT6 protocol. This protocol applies to various N52, N54 and N55 models so what follows likely applies to all of those also, but my testing has been exclusively with the N54. A couple of useful parameters are available via that protocol, but I have also identified their source and encoding on PT-CAN in this analysis. Useful parameters affecting LPF system function and provided by AiM PT6 include:

“Battery Voltage” (measured across the battery terminals, BMW also refers to this as terminal 30 voltage or Vt30).

The parameter called “Fuel Consumption” (the raw values purportedly in flow units of liters/hour) is actually the Normalized Duty Cycle request made by the DME to the EKPM when expressed as a percentage. The purported flow units imply nonsensical flow rates.

The mischaracterization of parameters will become a theme in this post, as much of the LPF system live data accessible through OBDII is also mischaracterized. While not relevant to the current subject, PT6 has one other mischaracterization in common with OBDII live data - the parameter called “MAP” is not “manifold absolute pressure” from a sensor in the intake manifold but ambient absolute atmospheric pressure from an absolute pressure sensor in the DME.

OBDII live data gives access to various LPF related parameters, most of which can also be found on PT-CAN. ProTool provides a convenient way to access these parameters. An OBDII code reader does not have direct access to PT-CAN. The reader can request specific data on the D-CAN (Diagnostic) bus, that are passed to it from PT-CAN through the Junction Box Electronics gateway in multi-part CAN messages. This bit of hair splitting is necessary to explain why some data available as OBDII live data is not accessible to a data logger on PT-CAN (the message structure is incompatible). It also explains why data accessible and viewable on both live data and PT-CAN may have a different refresh rate on the two streams (they are separate messages containing the same data in different data structures with their own refresh rate constraints). For rapidly changing data, this timing mismatch can result in slight differences in the reported data values from each source at any instant.


Fig 2: OBDII Receives PT-CAN data via JBE and D-CAN

The accessible live data parameters that originate from the EKPM are shown in the next series of images.


Fig. 3: ProTool EKPM Live Data – Pump Control Values

The mischaracterized parameter “Voltage, terminal 30” (aka battery voltage) is actually the EKPM input voltage as measured between the input and output pins on the EKPM, by the EKPM. It is less than the battery voltage due to voltage drops in the wiring on both the power and ground side. Part 8 discusses this in detail.

The parameter called “Voltage, fuel pump” is an accurate representation of the average pulse-width modulated voltage between the EKPM output and ground terminals as measured by the EKPM. The pump operates at a slightly lower voltage than this due to voltage drops in the pump supply and return wires.

The parameter called “Current, fuel pump” is a representation of the fuel pump operating current as measured by the EKPM. It is likely accurate for a standard EKPM however it will be inaccurate with any EKPM that intentionally under-reports fuel pump current (like my ET3 Design Stage 2 EKPM3).


Fig. 4: ProTool EKPM Live Data – Pump Speed Control

Depending on application the EKPM may operate in “pump speed (rpm) control” mode or “pressure regulation” mode. My N54 (and I suspect all applications that incorporate low fuel pressure sensor) operates in pressure regulation mode, so the speed control parameters are identified as N/A by ProTool.


Fig. 5: Foxwell EKPM Live Data – Pump Speed Control

Other code readers, like the Foxwell NT510, may display the speed control parameters, even though they are N/A. The “Specified delivery rate, fuel pump” parameter, shown with units of l/h is actually the raw version of the Normalized Duty Cycle Request value (%) mentioned later in this post. In the above instance, 167 represents (167 / 255) = 65.5%, not 167 l/h. This measurement was made at idle, where our fuel and jet pump testing tells us that the fuel flow rate will be approximately 115 l/h to support the two jet pumps and the fuel required to idle. The specified and actual speed parameters have no analog on PT-CAN.


Fig. 6: ProTool DME Live Data

ProTool can access a parameter it calls “Fuel Pump Pwm” (duty cycle). It is actually the Normalized Duty Cycle and does not correspond to the actual fuel pump PWM duty cycle as would be measured with an oscilloscope.

ProTool can access the Fuel Low Pressure Sensor reading, which it reports in units of bar, while calling them psi.


Fig. 7: MDH LPFP Pressure Live Data

MHD can also access this parameter, correctly displaying the information in psi. The Foxwell NT510 reports this parameter in units of hPa (5000 hPa = 5 bar = 72.5 psi).


Fig. 8: Foxwell LPFP Pressure Live Data

I can definitively state that the LPF sensor pressure is not normally present on PT-CAN, and that when it is present, it is not in a form that the AiM dash logger can recognize. That said, it is possible to log LFP sensor pressure on OBDII live data using MHD, and those logs can be exported as .csv and incorporated into PT-CAN logs made with a CAN data recorder for post processing in Excel. Synchronizing the two logs can be done using the rpm trace in both logs.

I have developed a better approach that eliminates the need for separate logs, by decoding the LPF sensor data that an OBDII live data request adds to PT-CAN. This requires initiating a LPF sensor pressure request on OBDII live data prior to initiating the PT-CAN log, but it eliminates the need for a separate live data log, or the need to synchronize separate logs. The method is a bit niche so I will save a description to the end of this post.


DECODING PT-CAN

It should be clear from the above that decoding the available LPF system parameters on PT-CAN is complicated by mischaracterization of what we expect to be authoritative sources of live data. Perhaps we can charitably call it simplification, but to understand the data on PT-CAN we need to find other ways to independently corroborate the parameter function and scaling.

I will start by presenting the end product – location and decoding of the LPF system related PT-CAN parameters I have found, and I will follow that with a description of how I have validated the parameters’ actual function and scaling.

While the PT-CAN protocol is proprietary, some of the encoding is in the public domain thanks to third party reverse engineering. The most complete resource I have found is the .dbc file found here:

https://github.com/BMW-E8x-E9x/opend...mw_e9x_e8x.dbc

That said, this resource does not include decoding of 0x0AA Byte 7, or any of 0x335, which are crucial to the LPF system. I hope that this report will expand the public understanding of those CAN IDs.

0x0AA (170) – “Accelerator Pedal”

0x0AA is a PT-CAN ID that originates from the DME. It reports accelerator pedal position and engine rpm with 16-bit precision, and, important to us, it transmits the Normalized Duty Cycle Request to the EKPM with 8-bit precision. Appropriate to the importance of these parameters, they are transmitted at 100 Hz. Some other status information is transmitted on individual bits within Byte 6, but these are not relevant to us.


Fig 9: Decoding 0x0AA

0x3B4 (948) – “Power Battery Voltage”

0x3B4 originates from the DME. While most of the data we have located is contained in a single 8-bit Byte, which can have hex values from 0 to FF or decimal values of 0 to 255, the battery voltage is transmitted on 0x3B4 at 12-bit precision. The data is stored little endian (least significant hex digits first) so when concatenating the most significant digit needs to be moved to the left.

The scaling factor allows the voltage to be represented in increments of 0.015 Volts. The other voltage parameters related to the EKPM are represented by a single Byte with a scaling factor of 0.1, allowing those voltages to be represented in increments of 0.1 Volts. While 0x3B4 reports with enhanced accuracy, it reports at a leisurely 0.2 Hz (once every 5 seconds).


Fig. 10: Decoding 0x3B4

0x335 (821) – “Electric Fuel Pump Status”

0x335 originates from the EKPM. Whereas 0x0AA represents critical data transmitted at high frequency, 0x335 simply reports the EKPM operating status at a less urgent 1 Hz. The following data is reported, with each parameter occupying a single Byte.


Fig 11: Decoding 0x335

Byte 2: EKPM Output Voltage is scaled 0.1 with 0 offset, providing for potential values from 0.0 to 25.4 Volts in increments of 0.1. This parameter is identical to the “Voltage, fuel pump” parameter available on OBDII live data. I have independently measured the parameter to confirm the function and accuracy as described.

Byte 3: EKPM Amps is scaled 0.1 with 0 offset, providing for potential values from 0.0 to 25.4 Amps in increments of 0.1. This parameter is identical to the “Current, fuel pump” parameter available on OBDII live data. When scaled 0.1 with zero offset it exactly matches the “Current, fuel pump” parameter reported on OBDII live data. As noted previously, to accurately represent the true current with an ET3 Design Stage 2 EKPM3 requires that the scaling and offset be adjusted. When appropriately calibrated, this parameter reads accurately over its normal working range. Specific calibration is likely not necessary for a standard EKPM as the ET3 Design unit intentionally under-reports the actual current to defeat DME overcurrent fault warnings by the DME.

Byte 4: One way to identify a parameter’s function is to log it over a range of operating conditions, then plot it against time, then look for parameters that vary with time in a similar way, based on the operating conditions. From a 22-minute drive, which included a couple of engine off/restarts, I have plotted Byte 4 below:


Fig. 12: 0x335 Byte 4 as %

The shape of this data is very distinctive and if we can find any known data with this same shape from this same drive, we may have a match. As a comparison, below, I have plotted the synchronized MHD low pressure sensor fuel data in psi.


Fig. 13: MDH Fuel Pressure Sensor Log

There is a compelling similarity to these two logs. We might speculate that they represent the same parameter, but with Byte 4 being a smoother more filtered version. We can try various scaling and offset factors applied to Byte 4 to see if we can get the two signals to agree closely enough that we might conclude that Byte 4 does in fact represent the low fuel pressure sensor data. Based on trial and error, the next image shows how the rescaling can be done.


Fig. 14: Does 0x335B4 Represent LPF System Pressure?

The trial rescaling does provide a reasonably close approximation to the LPF system pressure most of the time. Equally clear though, there are occasional large discrepancies and sometimes the two parameters trend in opposite directions. So, the answer to the question posed in figure 14 is “no”. At this point, the true meaning of Byte 4 is unresolved. I will attempt to resolve it in Part 9.

Byte 5: EKPM Input Voltage is scaled 0.1 with 0 offset, providing for potential values from 0.0 to 25.4 Volts in increments of 0.1. This parameter is identical to the incorrectly named “Voltage, terminal 30” parameter available on OBDII live data. I have independently measured the parameter to confirm the function and accuracy as described.

Byte 6: Normalized Duty Cycle represents a confirmation of the Normalized Duty Cycle Request made by the DME on 0x0AA. The EKPM repeats the requested value back as a confirmation of receipt. This parameter exactly matches the request made by the DME on 0x0AA Byte 7, but at 1 Hz instead of 100 Hz. Due to the lower data rate, some nuance is lost, as shown in the data log below, but it is sufficient to let the DME know that the flow requests are being received by the EKPM.


Fig 15: Normalized Duty Cycle Request/Confirmation

Byte 7: Adjusted Duty Cycle is the Normalized Duty Cycle, adjusted for the EKPM Input Voltage. The normalized value assumes an EKPM Input Voltage of 14.2 Volts, and the adjusted value is adjusted up or down in proportion to the ratio of (14.2 / Actual Input Voltage), up to a value of 100%, where the parameter is capped.

By plotting the Adjusted against the Normalized Duty Cycle we can see how they differ. This plot represents approximately 1300 data points collected over 22 minutes of operation. Clearly there is correlation, but some adjustment is needed at times to bring them into strong correlation.


Fig 16: Correlation between 0x335 Byte 6 and Byte 7

Acting on the same data set and by adjusting the Byte 6 normalized parameter by a factor of (14.2V / EKPM V In) at every point, all the data is now very strongly correlated with what we have called the Byte 7 Adjusted Duty Cycle. The fliers in the normalized data are perfectly resolved. This confirms our adjustment hypothesis and confirms the function of Byte 7.


Fig 17: Confirmation of Adjustment Hypothesis and Byte 7 Function

In physical terms the adjustment represents the adjustment to a duty cycle needed to correct the PWM average voltage output for variations in the voltage input relative to the assumed standard 14.2 Volts. Variations occur due to variations in the battery voltage due to loading and alternator output, and due to voltage drops at various places in the electrical system, as a function of current. Large corrections are needed during engine start because the alternator is offline and the starter pulls the battery voltage low and there are a couple of engine starts in this data set.


LOGGING RAW PT-CAN DATA

I used a Kvaser Leaf Light V2 interface to access the PT-CAN data. My version of the Leaf Light has a 9-pin D-SUB interface, so I have permanently wired a D-SUB access port into my PT-CAN bus, making hookup very simple. I use Kvaser CanKing software on an old PC laptop to view and log the data. CanKing can log the entire data stream for later review or can filter selected CAN IDs. Logged data is saved as .csv so is readily post-processed in Excel.


Fig 18: Kvaser Leaf Light V2


Fig 19: CanKing Screen Shot


LOGGING PT-CAN DATA on AiM DASH LOGGER

Once a parameter of interest has been identified and been provisionally decoded, it can be incorporated into a custom CAN protocol using RaceStudio3’s custom CAN builder. When that is incorporated into the logger protocol, that parameter is accessible for logging, viewing, math operations, etc. just like any other logged parameter. Logs provide a very useful way to compare a provisional decoding with another trusted source for the same data that is also logged. This permits comparing the two values in minute detail over a range of operating conditions. For some parameters, that level of examination is necessary to validate a provisional decoding. Many parameters fluctuate too quickly to compare two data sources just by eye.

I used this approach to confirm that the 0x335 Byte 5 parameter that OBDII live data purports to be Vt30 (battery voltage) was actually EKPM V In. Logic suggested that the EKPM must adjust duty cycle for its input voltage, and that the EKPM could only report voltages that it was able to measure directly. This strongly suggested that the voltage it was reporting was its own input voltage. To prove this hypothesis, I used an analog logger channel as a loggable voltmeter connected directly to the EKPM input terminals. The channel required calibration, and the addition of an RC low pass filter to remove PWM transients. Comparing the directly logged EKPM V In with the scaled 0x335 Byte 5 over a range of operating conditions confirmed that this parameter was indeed EKPM V In.

To provide a range of data, the EKPM was loaded by means of a switchable resistor load bank in conjunction with the fuel pump so that a range of currents would result in a range of voltage drops in the wiring and thus a range of EKPM input voltages. The steps in the data below represent the switching in of successive incremental loads.

The red trace represents EKP current ranging from 7.6 Amps (just the pump) to 22.7 Amps (pump plus all load resistors applied). EKPM V In is depressed at higher EKPM currents because of voltage drop over the EKPM power and ground wires. The “Voltage Probe” channel was directly connected between the EKPM input and ground terminals, and it perfectly tracks the 0x335 Byte 5 “EKPM V In” parameter, confirming it has been correctly decoded.


Fig. 20: Confirming 0x335 Byte 5 function is EKPM V In

Battery Voltage: Battery voltage is relatively stable, so can be measured with a digital multimeter as a truth value. The AiM PT6 protocol also incorporates a battery voltage channel (“Battery”). This permitted easy validation of our decoding of 0x3B4 battery voltage parameter (“Battery1”) and the AiM value against a direct digital multimeter measurement.



Fig. 21: 0x3B4 Battery Voltage Validation


CALIBRATING THE EKPM CURRENT PARAMETER

My initial plan was to validate fuel pump current using a Hantek CC65 AC/DC Current Clamp as the device is convenient and non-invasive. These devices are very economical and while very cheaply made, mostly accurate. Criticisms are that battery life is very short and the device becomes inaccurate well before the low battery warning light comes on. The clamp has no display – it outputs a mV signal proportional to the current for reading on a digital multimeter. I considered this a feature as that signal could also be logged on my dash.

I learned the hard way that Hall effect DC Current Clamps do not accurately measure currents in PWM circuits due to the transients present. Plan B involved measuring the voltage drop across the circuit supply fuse acting as a shunt resistor. The correlation between drop across the fuse and current is published for standard automotive fuses. My testing was done with a 30 Amp ATO fuse.


Fig. 22: Fuse Voltage Drop Chart

In the above chart, for example, a voltage of 10 mV across the 30 Amp fuse would correspond to a current of 5.076 Amps through the fuse. This method proved to be immune to PWM interference and worked well. To check the calibration at a variety of currents, I used a resistor load bank to increase the current supplied by the EKPM with the engine idling. The load bank consisted of eight individually switchable 6Ω resistors in parallel with the LPF pump. All the running values (blue) represent PWM operation. The single red value was at 100% duty cycle (engine off, via the OBDII pump activation function). That this value also falls on the calibration line confirms that these measurements are immune to PWM transients. The equation of the regression line provides the correct scaling (0.209) and offset (-2.7) calibration factors to be applied to 0x335 Byte 3 to have this parameter correctly display the actual fuel pump current. These calibration factors are unique to the ET3 Design Stage 2 EKPM3, as it intentionally under-reports fuel pump current to avoid fault codes. Having an accurate ammeter is essential for Part 8 looking at wiring considerations.


Fig. 23: 0x335 EKPM Current Calibration (ET3 Design Stage 2 EKPM3)

Directly Logging LPF Sensor Pressure on PT-CAN

By initiating a stream of live LPF Sensor Pressure data on OBDII live data, the DME starts transmitting the pressure data using CAN ID 0x612 on PT-CAN. While regular CAN IDs behave completely predictably and consistently, the 0x6 series OBDII CAN IDs can have a varying number of Bytes and the transmitted data is multiplexed, allowing a single CAN ID to represent multiple Parameter IDs in successive records. In this way, all OBDII live data can be transmitted simultaneously with only a few CAN IDs being needed. 0x612 is used specifically for data originating from the DME. Live data requested from the EKPM is transmitted on 0x617.

The complexity of the 0x612 transmissions become overwhelming if many live parameters are selected, but if only one or two are selected there is a reasonable chance they can be decoded. In the example below, I selected “rpm” and “Fuel low pressure sensor” and used MDH to initiate the data stream and to log it. Conveniently, both parameters are transmitted in a single record.


Fig. 24: Decoding OBDII Fuel Pressure and RPM Live Data on 0x612
(available only if initiated by an OBDII code reader)

Ox612 changes dramatically depending on the live request made, so the above encoding is specific to the request for just fuel pressure and rpm, but it illustrates how live data can be used to insert truth data or otherwise inaccessible data into the PT-CAN data stream

Validating the fuel pressure decoding can be shown by plotting both the MDH live data log and PT-CAN decoded data from the same logged interval on a single graph.


Fig. 25: Validation of 0x612 Decoding of LPF Pressure on PT-CAN

The data matches point for point (approximately 56,000 points), except for 5 fliers that exceed 85 psi in the 0x612 record but do not appear in the MHD live data record. MHD appears to filter out any values that exceed 85 psi.


Fig. 26: OBDII LPF Pressure “Flier Low Pass Filter” Excludes > 85 psi

This data set also allows us to test our Part 4 hypothesis that the mechanical fuel pressure regulator essentially never flows any fuel. By examining the complete data set for instances of the fuel pressure exceeding 80 psi we find that these occurrences account for about 0.07% of the total logged time, so only occur during transient events like throttle closures where LP fuel flow suddenly exceeds demand.

Unfortunately, neither 0x612 nor any of the other 0x6 OBDII CAN IDs can be decoded by the AiM data logger, so LPF pressure sensor data cannot be directly logged on the AiM dash logger.

[fe1rx]

PART 8 – ELECTRICAL CONSIDERATIONS FOR UPGRADED LPF PUMPS

The key questions I want to answer in this post are:

1) Is the original wiring adequate with respect to voltage drop when a Stage 2 (19 Amps) or Stage 2.5 (22 Amps) pump is installed? (No, so I upgraded to 10 AWG.)
2) Is the original wiring adequate with respect to resistive heating when a Stage 2 (19 Amps) or Stage 2.5 (22 Amps) pump is installed? (Yes.)
3) Is the original 20 Amp fuse adequate? (Yes, but a larger fuse is appropriate for upgraded wiring.)
4) Can an original EKPM2 adequately control a Stage 2 pump? (Not in my experience.)

I will show that the OE wiring is inadequate to properly support a Stage 2 or Stage 2.5 LPFP. BMW did a decent job of optimizing the wiring for the original pump that draws 11 Amps intermittently but upgraded pumps that draw significantly more than that will not reach their full flow potential due to excessive voltage drops in the wiring. I will show that upgrading to 10 AWG wiring will support approximately 20% more flow from a Stage 2 or Stage 2.5 LPFP vs. the original wiring, and that this corresponds to support for more than 100 additional hp in both cases. To put it another way, the original wire derates the potential performance of Stage2 or Stage 2.5 LPF pumps by about 20%, which is equivalent to more than a 100 hp penalty.

Excessive voltage drops in the system also force the EKPM to operate at a higher duty cycle to output the voltage needed to overcome that voltage drop. In the case of an EKPM2, it will overtax and already marginal component. My initial testing of a Stage 2 pump with my original EKPM2 immediately revealed this weakness and prompted an upgrade to a Stage 2 EKPM3.

SIMPLIFIED CIRCUIT

The following simplified representation of the LPFP electrical circuit is useful to discuss the various voltages, voltage drops and currents. Voltage drops occur along each leg of the circuit and they reduce the voltage available to operate the LPF pump. I measured the voltage drops using a voltmeter at a known current to calculate the resistance of each leg of the circuit.


Fig. 1: Voltage Drops in the LPFP System

The battery voltage (AKA Vt30) measured between the battery + and – terminals is the highest voltage in the system. The current consumed by the EKPM for its internal function (Io) and bypassing the fuel pump is small enough relative to the pump current as to be negligible, so we can assume that all the current flowing in the EKPM flows through the pump.


Fig. 2: PT-CAN Measured Voltage and Current Parameters

Because of the resistance in the wiring, the current flowing in the circuit results in a voltage drop in every wire through which it flows.

The entire LPFP circuit is wired with 2.5 mm2 metric primary wire, which does not correspond exactly with an AWG size. The wiring is often described as being 14 AWG as this is the closest North American equivalent, but the difference is materially significant, as the characterization results in an over-estimate of the benefit of a wire upgrade to 12 AWG.

Potential wire size upgrades include 4.0 mm2, 12 AWG and 10 AWG. The two systems can be compared as follows (typical values):

14 AWG (1.88 mm2) => 3702 CMA => 3.21 Ω / 1000 ft
2.5 mm2 => 4934 CMA => 2.32 Ω / 1000 ft
12 AWG (2.96 mm2) => 5833 CMA => 2.04 Ω / 1000 ft
4.0 mm2 => 7894 CMA => 1.43 Ω / 1000 ft
10 AWG (4.73 mm2) => 9543 CMA => 1.27 Ω / 1000 ft

Measuring wire resistance over installed lengths with a conventional ohmmeter function on a digital multimeter is not accurate because the resistances are so low, but measuring the voltage drop across the wire with a substantial known current flowing through it works well. Connectors, crimps, fuses, etc. contribute to the total circuit resistance. This point is illustrated by the measurement described in Fig. 3. The resistance can be determined by using Ohms Law (R = V / I) in conjunction with measured voltage and known current.


Fig. 3: Measuring Wire Resistance using 10 Amp Current

The measured resistance of the conductor alone corresponds to a resistance per unit length of 2.6 Ω/1000 ft. This is too low for the wire to be 14 AWG (3.2 Ω/1000 ft) and is close to the rating for 2.5 mm2 wire (7.6 Ω/km = 2.3 Ω/1000 ft. The ISTA wiring diagram in Fig. 4 (2008 N54 335i) confirms this wire size. I am flogging this issue to death because the wiring is widely reported to be 14 AWG which makes the readily available 12 AWG upgrades look much more beneficial than they actually are.


Fig. 4: ITSA Wiring Diagram (N54 335i)

ORIGINAL WIRING

I calculated the various circuit resistance values by means in situ measurements with the engine running. Current was measured using the calibrated PT-CAN current data as described in Part 7. Voltage drops were measured using a digital voltmeter from point to point.


Fig. 5: Measured Resistances for Original Wiring

The resistance of the ground wire is surprisingly high, given the proximity of the ground lug to the EKPM. Clearly this wire does not run directly from the EKPM to the ground lug as the measured resistance corresponds to about 4 ft of wire. There is a splicing of ground wires within the harness in the luggage compartment, so I assume the EKPM ground wire takes that convoluted route to the ground lug just above it. Running a direct wire to this lug takes only 18” of conductor and is very worthwhile.

I confirmed these resistance values using the PT-CAN data parameters accessible on my data logger (described in Part 7). I added a resistor load bank (consisting of eight individually switchable 2Ω resistors in parallel) in parallel with the LPFP to allow me to examine circuit currents from about 7.5 Amps to 18 Amps. Fig. 5 shows the load profile applied in Amps, along with the fuel pressure observed during the test. The load was applied in eight 1.3 Amp steps but unloaded in four 2.6 Amp steps. The fuel pressure was stable with only a small transient response to the unloading.


Fig. 6: EKPM Current Loading Profile at Idle

During this testing, the LPFP voltage and current remained constant and sufficient to maintain stable idle at 72 psi LPF pressure. The additional current flowing in the circuit bypasses the LPF pump and flows through the resistor load bank. The effect on the measured parameters is shown in Fig. 7.


Fig. 7: Voltage Response to Current Loading

The red line represents battery voltage, which declines only slightly as the pump load increases, as the alternator is easily able to supply the current required to maintain the battery voltage. The green line represents the voltage at the EKPM input terminals. It declines with increasing current due to the resistance in its power and ground wires. The blue line represents the EKPM output voltage. It increases slightly as the increased current flowing through the LPFP supply and return wires results in an increased voltage drop in those wires. The EKPM adjusts its duty cycle as required to provide the additional output voltage needed to maintain a constant pump voltage.

It is interesting to plot this same data against circuit current, in addition to the duty cycle parameters. Fig. 8 shows this with the original wiring.


Fig. 8: Original Wiring - Voltage and Duty Cycle vs. EKPM Current

Because R = V / I, the slope of the voltage vs. current regression lines represents resistance values. The slope of the green line represents the sum of dVp and dVg from Fig. 1. The calculated value of 76 mΩ closely matches the measured value of (55 + 20) = 75 mΩ from Fig. 5. The slope of the dark blue line represents the sum of dVs and dVr from Fig. 1. The calculated value of 25 mΩ closely matches the measured value of (12 + 12) = 24 mΩ from Fig. 5.

The light blue line represents the normalized duty cycle. The increasing voltage drop in the pump feeders resulting from increasing current requires the DME to request additional duty cycle to maintain a constant voltage at the LPFP terminals. The flow rate and fuel pressure remain constant.

The purple line represents the adjusted duty cycle. As the increased current pulls the EKPM input voltage below 14.2 Volts, the EKPM must adjust the duty cycle upward to achieve the desired output voltage implied by the normalized duty cycle.

The olive line represents the actual duty cycle. It incorporates the learned flow characteristics of the installed pump through the adaptation process. Luckily, our pump is strong and this results in a reduction from the adjusted duty cycle.

The loss of LPFP capability in response to the voltage drops can be framed in a two equivalent ways:

1) The decline in EKPM input voltage with increasing current reduces the maximum voltage that can be supplied to the pump and so reduces its flow potential.
2) The increase in adjusted duty cycle in response to the voltage drops at increased current reduces the available duty cycle to accommodate actual increased flow requirements.

10 AWG WIRING

To reduce the voltage drops in the circuit, and to increase the flow potential of the installation, I have replaced the wiring with 10 AWG wire, with the power connection going direct to the battery. I left the original EKPM ground wire in place but supplemented it with a 10 AWG wire directly to the adjacent ground lug. For comparison purposes I have measured the circuit resistances as previously, with those results shown in Fig. 9.


Fig. 9: Measured Resistances for 10 AWG Wiring

I tested this in the same manner as previously, with the results shown in Fig. 10.


Fig. 10: 10 AWG Wiring - Voltage and Duty Cycle vs. EKPM Current

The slope of the green line representing the sum of dVp and dVg has a calculated value of 23 mΩ, which matches the measured value of (20 + 3) = 23 mΩ from Fig. 9. The slope of the dark blue line representing the sum of dVs and dVr has a calculated value of 13 mΩ closely matches the measured value of (6 + 6) = 12 mΩ from Fig 9.

The light blue line representing the normalized duty cycle rises with increased current as the DME requests additional duty cycle to maintain a constant voltage at the LPFP terminals. The flow rate and fuel pressure remain constant, but the EKPM must work harder to achieve this as we increase the current, but less so than with the original wiring. As with the original wiring the battery voltage is negligibly affected by the modification.

The purple line representing the adjusted duty cycle is only slightly steeper than the normalized duty cycle because the EKPM input voltage (green line) is never far from the ideal 14.2 Volts. Accordingly, the 10 AWG wiring allows the full potential of a Stage 2 or Stage 2.5 LPFP to be realized without running out of input voltage or duty cycle.

VOLTAGE DROP ACROSS THE EKPM / EFFECTIVE EKPM “ON” RESISTANCE

In Figures 5 and 9 I have indicated that the effective resistance between the EKPM V In and V Out terminals as 10 mΩ. This is a consequence of the on resistance of the half-bridge MOSFET switch, and the resistance of internal conductor traces and solder connections within the EKPM. As a result of this resistance, even at 100% duty cycle, there will be some voltage drop across the device. This is designated dVe in Fig. 11.


Fig. 11: Duty Cycle Calculation Accounting for EKPM Internal Voltage Drop

As a first order approximation it is typically stated that duty cycle equals (V Out / V In) in a PWM circuit. A more accurate representation is obtained by accounting for the voltage drop cross the switch as shown in Fig. 11. Where I have plotted true duty cycle in the previous graphs, it has been calculated in this manner.

THEVENIN’S THEOREM

The LPF circuit can be simplified to a voltage source and resistor in series with the LPF pump according to Thevenin’s theorem. It is useful to consider the case of 100% duty cycle. The voltage source simplifies to the battery voltage, and the series resistance is the sum of the resistances shown in Fig. 5 for original wiring (0.055 + 0.010 + 0.012 + 0.012 + 0.020) = 0.109 Ω, and the sum of resistances shown in Fig. 9 for 10 AWG wiring (0.020 + 0.010 + 0.006 + 0.006 + 0.003) = 0.045 Ω. The total voltage drop between battery and LPFP at 100% duty cycle is then this resistance times the fuel pump current. Based on this we can calculate the potential benefit of 10 AWG in terms of maximum LPFP voltage for a variety of scenarios.

As we showed in Part 3 the LPF system must provide approximately 110 lph to support the pump module and regulator module jet pumps. Thus, any flow penalty due to reduced maximum pump voltage needs to be considered against the usable flow. In Part 6 we established the usable flow rates for the Stage 2 and Stage 2.5 LPFPs at 12.0 Volts and 13.5 Volts. We can use this to estimate the flow penalty associated with the calculated voltage drops. We also calculated the relationship between flow rate and hp. All these concepts are combined in Fig. 12 to calculate the potential benefit of a change to 10 AWG wiring with respect to maximum LPFP voltage via a reduction in total voltage drop, and the associated benefits in terms of fuel flow rate and hp potential.


Fig. 12: Potential Benefits of 10 AWG Wiring with Stage 2 and Stage 2.5 LPFPs

We conclude that the change to 10 AWG wiring will provide the potential for about 19% additional usable fuel flow from either a Stage 2 or Stage 2.5 LPFP over what it could provide with original wiring and that this corresponds with the potential to support at least 100 additional hp. Perhaps a better characterization would be that retaining the original wiring cripples the flow potential by over 100 hp.

ACCEPTABLE VOLTAGE DROPS

While there is no universally accepted rule of thumb for acceptable voltage drops, the aviation standard from FAA AC 43.13-1B states that for a 14 Volt circuit, the maximum allowable voltage drop between the bus and the equipment ground is 0.5 Volts for continuous loads and 1.0 Volts for intermittent loads. Fig. 12 shows that the intermittent standard is not met with original wiring but is with 10 AWG, for both Stage 2 and Stage 2.5 pumps. The original wiring woefully fails to meet this standard.

WIRE CURRENT LIMITS (AMPACITY)

All wires have some resistance, and when current flows through them they heat up proportional to the current and their resistance per unit length. The only way to reduce the heating is to provide better cooling, which is generally not practical, or to use heavier gauge wire (numerically smaller AWG or bigger mm2, which has lower resistance). Alternately we can use wire with a higher insulation temperature limit and allow the wire to run hotter, which is generally not desirable. Standard automotive primary wire is typically rated at 105°C. Wiring used for motorsport upgrades is typically either TXL (125°C rated) or M22759/16 (150°C rated). While higher temperature ratings provide an additional margin of safety against wires overheating and insulation failure, running hot risks damage to electrical connectors.

A wire’s resistive heating current limit is referred to as its ampacity, and it is not a fixed value. How hot a wire will get at a given current depends on the ambient temperature and whether the wire is alone in free air or part of a bundle, and how much current the other conductors in the bundle are carrying. There are empirical methods that can establish current limits based on these and other factors. FAA AC 43.13-1B has a method that can be used, but the science is inexact and highly dependent on assumptions, so a healthy margin is prudent. Wire data sheets often provide a current rating, but we need to read the fine print to understand the assumptions involved. The data sheet for 2.5 mm2 metric FLRY B automotive primary wire shows a current limit (ampacity) of 38 Amps, assuming a single wire in free air at an ambient temperature of 40°C and the wire at its temperature limit of 105°C (221°F). While the cooling assumptions are not conservative the Stage 2.5 maximum current of 22 Amps is an intermittent load condition. For these reasons I have concluded that the ampacity of the original wire is sufficient for a Stage 2.5 pump.


Fig. 13: Metric Primary Wire Data Sheet

Fuses don’t overheat/blow instantly, so the 20 Amp fuse protecting the circuit (which is to say the wires) is unlikely to blow at 22 Amps intermittent. Increasing the fuse rating to 25 Amps would be acceptable based on the wire’s ampacity but provides no tangible benefit.

UPGRADED WIRE OPTIONS

In terms of voltage drop, both 12 AWG and 10 AWG are improvements, preferably in TXL or M22759/16 for their higher temperature ratings. Metric primary wire is also an option, but much harder to find. 4.0 mm2 would be a good choice because it is the largest size that the required connector pins will nominally accept. 10 AWG requires that any pins be crimped out of spec, which you may not be comfortable with (of course solder is an option the experts abhor).

Arc Terminator’s aftermarket pump upgrade harness (there are others too) is only 12 AWG because of the crimping issue.

Considering just the wire, and ignoring the benefits of shorter direct wiring to the battery or shorter grounding, the larger wire sizes provide the following voltage drop benefits:

2.5 mm2 = 2.32 Ω / 1000 ft -> status quo
12 AWG = 2.04 Ω / 1000 ft -> about 12% less voltage drop than status quo
4.0 mm2 = 1.43 Ω / 1000 ft -> about 38% less voltage drop than status quo
10 AWG = 1.27 Ω / 1000 ft -> about 45% less voltage drop than status quo

PUMP WIRING OPTIONS

The combined length of the pump supply and return wires from EPKM to pump is 6 ft. A 12 AWG TXL upgrade harness is available from Arc Terminator and others for those disinclined to make their own and would provide some improvement, particularly if paired with an upgraded EKPM ground wire.

The fuel pump connector end of this harness uses 2.5 mm Series crimp sockets. The largest compatible sockets are compatible with 2.5-4 mm2 wire (p/n 61131376206). 10 AWG is strictly slightly larger than 4 mm2, so testing for acceptable crimps is essential. As I don’t know the tensile strength specification for this style of crimp pin, I referred to the specification for AMP Mate-N-Lok 0.140” diameter pins (document 108-1032), which specifies a tensile strength of 65 lbs. I found a crimp tool/cavity combination that achieved a pull-out load of about 60 lbs (with just the conductor crimped, not the insulation), which for 10 AWG wire is a bit less than ideal, but acceptable (in my application, in my engineering judgement) given the benefits. I lack the technology to do an official test at the 1 inch/minute rate but used a tensile load cell and a steady pull by hand. For a commercial solution, using 4.0 mm2 metric primary wire would be a better solution in terms of achieving fully qualified crimps, and it would be objectively (26%) better than 12 AWG in terms of voltage drops.


Fig. 14: Mate-N-Lok Crimp Tensile Strength Requirements


Fig 15: LPFP Pin Test Crimp with 10 AWG M22759/16 Wire

I bought extra pins so that I could do a couple of trials to figure out the best crimp tool for the job. I used 10 AWG M22758/16 wire, which has a CMA of 9354 and a metric area of 4.7 mm2. I crimped the 2.5-4.0 mm2 rated sockets using a Delphi 12085271 ratcheting crimper. 10 AWG TXL wire has a slightly higher CMA of 9704 and a metric area of 4.9 mm2. I doubt that difference will materially affect the ability to crimp these pins on TXL, but I haven’t tried it.

BMW sells a universal connector housing (p/n 61138352390), which can be used for either the pump or gauge connection if your housing is cracked. The housings can be de-pinned relatively easily, with the correct tool. I could not find such a tool in my collection of Amazon cheapies, or my Schwaben kit. A good tool can be made from stainless steel hypodermic tubing (0.180” OD, 0.160” ID, McMaster Carr p/n 8987K9).


Fig. 16: 10 AWG LPFP Harness Installed at Pump End

The EKPM end of the harness uses ELO-Power 2.8x0.63 contacts. These crimp contacts are also available for wires 2.5-4.0 mm2 (6113913612), which would work with either 12 AWG or 4.0 mm2 wire. I suspect 10 AWG would be a problem with these connectors. As my ET Design Stage 2 EKPM3 has screw terminal blocks, I did not need to use this connector, which allowed me to easily go to 10 AWG at the EKPM. Depinning the EKPM connector housing does not take any special tools.


Fig. 17: ET3 Design Stage 2 EKPM3 Installed with 10 AWG

The EKPM draws no standby current unless woken up by the CAS. For that reason, it can be wired direct to the battery without introducing a parasitic load.


Fig. 18: EKPM Direct-to-Battery Wiring Through 30 Amp ATO Fuse

My EKPM main power wire follows the main harness back to the battery compartment and then connecting to battery + through a 30 Amp ATO fuse. Littelfuse makes a nice stackable fuse holder housing (p/n 178.6150.001) that will accept a variety of crimp pins depending on the wire size being used (p/n 178.6116.6001 works for 10 AWG). Again, the Delphi crimper worked well, but I did need to do a bit of fiddling to get decent insulation crimps. Spare pins are your friend when finding a workable crimp setup for a previously untested configuration, as pull testing is always prudent. I took the opportunity to replace a previous pigtail fuse holder used by my trailer module by creating a two-fuse stack. By following the existing wiring routing for protection, rather than taking the most direct route for minimal length, the total conductor length from battery to EKPM is about 7 ft. I installed a 30 Amp fuse as the upgraded wiring will easily support this.


Fig. 19: ATO Fuse Holder

The EKPM connectors are a known weak point, and the ground pin in particular has been known to overheat. For this reason, ET3 Design added a terminal strip to their Stage 2 EKPM3. While it looks a bit clunky, it is very practical, and it supports wires up to 10 AWG. That feature prompted me to adopt the ET3 Design unit when upgrading. As I left my original power wire in place in the main EKPM connector, I removed the glovebox fuse 70 to prevent the new fuse from back feeding the fuse box through that wire.


Fig. 20: Original LPF Fuse 70 Removed