DIY DIY EKPM3 Active Cooling for LPFP

The Banshee

Corporal
Nov 18, 2017
175
104
0
Lehigh Valley PA
Ride
2008 335xi coupe 6466 MT
What fault? Do you mean DTC or do you mean in general as in the problem goes away when you shut the car off and turn it back on? Can someone get a log of this with AFR and low and high fuel pressure? Does this happen every single time you go WOT?

If anyone has this happen, check the car for a DTC before you turn the car off. Some DTCs will not illuminate the MIL.

INPA will also report the current EKP utilization from 1 to 100%. It is in "Actuator Controls 1" I think. You can command 0% or 100% fro INPA, or give control back to the DME where it should just report the current utilization. It would be interesting to monitor this when the problem occurs.

I get a fault on MHD that is quickly proceeded by an idrive fault in most instances. It is what Doublespaces has indicated. Mine also runs the higher pressure post code with stock regulators. I find it strange that my car throws this code in the exact same manner with a brand new reprogrammed EKP3. Replacing the module with a new one did nothing.
 

doublespaces

Administrator
Oct 18, 2016
9,303
4,331
0
AZ
Ride
2009 E93 335i
I get a fault on MHD that is quickly proceeded by an idrive fault in most instances. It is what Doublespaces has indicated. Mine also runs the higher pressure post code with stock regulators. I find it strange that my car throws this code in the exact same manner with a brand new reprogrammed EKP3. Replacing the module with a new one did nothing.

In my experience, I've gotten some codes when the tuning is wrong. I think requesting too much fuel, more than can be provided can create issues but I don't have any proof.
 
  • Like
Reactions: The Banshee

KClemente

Sergeant
Nov 26, 2019
284
147
0
Ride
E90 335i
EKP2 active cooling! I bought a set of 8 aluminum heat sinks with adhesive already attached and so far they haven’t fallen off yet
 

Attachments

  • 66E95A05-2F91-43FD-8CE4-280B9582AF8A.jpeg
    66E95A05-2F91-43FD-8CE4-280B9582AF8A.jpeg
    266.9 KB · Views: 343

ajm8127

Specialist
Jul 16, 2020
69
135
0
PA - US
Ride
2007 E90 335i
my car throws this code

I would be interested in the specific code to learn more about what the car thinks is happening.

From what I gather the EKP does voltage and current monitoring of the pump. It also can enter a failsafe state where it turns the pump on 100%.

Normally, the DME requests a fuel flow from the pump and the EKP decides how fast to spin the pump based on the request and a characteristic curves programmed into it to deliver the flow requested.

I've read the DME requests a flow up to 254 liter/hour. If you double the output capability of the pump it will meet that flow at a lower speed so the characteristic curve will be way off.

There are some threads where a user (torr) outlines how to reprogram the EKP with different error limits and characteristic curves.

I am on my phone and I have all of that information booked marked on my computer. I'll try to add links to this post later.

[EDIT]

Discussion on adjusting the coding to change behavior of the EKP module

Discussion on adjusting the EKP coding to raise the current limit to avoid code 2AAE

TIS entry for EKP functional description
 
Last edited:

ajm8127

Specialist
Jul 16, 2020
69
135
0
PA - US
Ride
2007 E90 335i
Data for the EKPM3 HW04 has been collected and is below. Methodology was the same as post 47. In this case the switch IC (BTN7960) has the driver and MOSFET integrated so the gate to source voltage (Vgs) was unable to be directly measured.

EKPM3 HW04 Test Data (BTN7960)Stock pumpStock pump plus additional load
Inductor Resistance7.5 mOhms7.5 mOhms
FET Resistance (Rds_on)6.7 mOhms6.9 mOhms
Current10.85 A17.4 A
Vin13.1 V13.0 V
Vout12.9 V12.65 V
Rise/Fall Time (6.84K Ohms Pin 5 to GND)1.91 us/1.62 us1.92 us/1.69 us
Rise/Fall Time (0 Ohms Pin 5 to GND)1.07 us/0.95 us1.07 us/0.95 us

Three things jump out at me when comparing the results.

First, as expected the BTN7960 on resistance is higher than the BUK9107-40ATC. The datasheet for the BTN7960 doesn't have a graph comparing the on state resistance to the current through the MOSFETs, but below is a graph comparing on state resistance to source voltage at a range of temperatures. You can see at about 10 volts the graphs flatten out. This is similar to the Vgs characteristics of the BUK9107 in the EKPM2.

1596555899379.png


Second, there was an unexpected increase in the inductor resistance on the EKPM3 versus the EKPM2. When looking at photo of the boards, the inductor is completely different and is much larger on the M2 module. So the fact that the resistance is higher with the smaller inductor is not that surprising. The larger one uses wire with a greater cross section. What this means is the inductor itself will dissipate more power which means it will get hotter. All of this heat adds up. If the board is hotter because of the inductor, then the switch will run hotter as well. The fact that the switch has a higher resistance only compounds the heat issue.

Third, the rise and fall times are ten times as long on the EKPM3 than the EKPM2. The MOSFET is said to be operating in the "linear" region when it is neither fully on or fully off. When it is fully on, it is said to be "saturated". In the linear region, the MOSFET is dissipating more power than it is when it is fully on. This is bad for heat. This rise and fall time should be adjustable by changing the resistor attached between pin 5 and ground of the BTN7960. I measured the value of the resistor and it appears to be 6.84 kOhms. The datasheet says to expect a rise and fall time of 2 us when the resistance is 5.1 kOhms, so that's in the ballpark. At 0 ohms (shorted to ground) the rise/fall time should be 1 us, which is still 5 times slower than the EKPM2. I will retest the module at 0 ohms (pin 5 to GND) to confirm. I'll edit the table above with the results.

One final thing that I noticed when I took apart both modules was the difference in the size of the gap pad. The pad in the EKPM3 is 50% larger than the pad in the EKPM2. This tells me Helbako (the module OE) knew the EKPM3 dissipated more power and increased the surface area of the thermal interface between the PCB and the aluminum base. The comparison of the pads is below. They are approximately the same width, but the EKPM2 pad is 20 mm tall while the EKPM3 is 30 mm tall. The raised portion of the aluminum base is larger to as well by the same amount.

20200804_115108_HDR.jpg


Next up I have the BTN8982 driver to test. This is the pin for pin replacement for the BTN7960. At room temperature it has about 5.5 mOhms typical on resistance. Also it switches much faster. The rise/fall times are 380 ns at 5.1 kOhms and 250 ns at 0 ohms. This is much closer to the EKPM2 MOSFET. The power dissipation with this part should be reduced leading to lower temperatures of the module. The comparison between the EKPM3 with the BTN8982 and the EMKP2 should be interesting.

[EDIT] I updated this post with the rise/fall times with pin 5 of the BTN7960 shorted to ground through a 0 ohm resistor. The rise and fall times were both 1 microsecond, more or less. This is right in line with the datasheet.
 
Last edited:

fmorelli

Lieutenant General
Staff member
Aug 11, 2017
3,748
3,592
0
57
Virginia
Ride
E89 Z4 35i, F10 535d
The comparison between the EKPM3 with the BTN8982 and the EMKP2 should be interesting.
Awesome info! I have a number of BTN8982's ... at this point I'm going to wait out your pushing through the rest of this, and follow your lead on the end result :-D

Filippo
 
  • Like
Reactions: doublespaces

doublespaces

Administrator
Oct 18, 2016
9,303
4,331
0
AZ
Ride
2009 E93 335i
It's party time. I updated post 65 with the rise/fall times with pin 5 of the BTN7960 switch shorted to ground. This is the fastest this part is capable of switching at.

So, basically you've confirmed what I jokingly stated above, the 7960 runs hotter and sucks. As Filippo said, lets see the 8982. Excellent work, this is very informative, thank you.
 

ajm8127

Specialist
Jul 16, 2020
69
135
0
PA - US
Ride
2007 E90 335i
So, basically you've confirmed what I jokingly stated above, the 7960 runs hotter and sucks. As Filippo said, lets see the 8982. Excellent work, this is very informative, thank you.

More or less. Just reading the datasheets made it pretty obvious. Like I said in a previous post, the EKPM3 looks like a "cost optimized" revision. The more modern switch should help.
 

doublespaces

Administrator
Oct 18, 2016
9,303
4,331
0
AZ
Ride
2009 E93 335i
More or less. Just reading the datasheets made it pretty obvious. Like I said in a previous post, the EKPM3 looks like a "cost optimized" revision. The more modern switch should help.

For someone like myself with little to zero experience with electronics, PCB design etc, it is amazing to me you're able to tell not only by the data sheet but also with your tests. I suppose you have a lot of 'practical knowledge' in this area to draw from so it's good to see that put to use here.
 

ajm8127

Specialist
Jul 16, 2020
69
135
0
PA - US
Ride
2007 E90 335i
I swapped the BTN8982 switch onto my EKPM3 to test the performance of the new part. I kept pin 5 shorted to ground to test the fastest switching speed possible.

EKPM3 HW04 Test Data (BTN7960)Stock pumpStock pump plus additional load
Inductor Resistance7.4 mOhms7.6 mOhms
FET Resistance (Rds_on)5.1 mOhms5.2 mOhms
Current10.95 A17.6 A
Vin13.3 V13.1 V
Vout13.14 V12.85 V
Rise/Fall Time (0 Ohms Pin 5 to GND)505 ns/560 ns490 ns/565 ns

As predicted, both the FET Rds_on and rise/fall times were improved over the BTN7960. The BTN8982 was able to achieve the fastest rise/fall times yet. I had to go back and edit post 47 with the EKPM2 data because I found the rise and fall times with INPA actuating the module were faster than the times with the DME actuation the module with the engine running. For reference the EKPM2 switched with 1082 ns/960 ns rise/fall times respectively under DME control which is the case that is important when the car is running.

Even though the BTN8982 has a lower Rds_on than the BTN7960, the EKPM2 module with the BUK9107-10ATC MOSFET has a lower on state resistance (4.6 mOhms) than the BTN8982. Also, the inductor in the EKPM2 dissipated significantly less power than the EKPM3 design. In my testing this resulted in 63% less power dissipation in the EKPM2 inductor over the EKPM3 design with the stock pump, and 81% less power dissipation in the inductor with a 17 amp load (similar to Walbro 450). It is a function of the square of the current that is why the increase is not linear (ohms law: P = I^2 * R).

The switching speed further handicaps the EKPM3. On this module the switch PWM frequency is 20 kHz. This means there are 20,000 on/off events every second, unless the pump is running at 100%. On the EKPM2, the MOSFET switches at 5 kHz. A fourfold reduction. This also has implications for power dissipation. A higher switching speed causes more transitions per second and all else being equal (the EKPM2 vs. the EKPM3 with BTN7960 and 0 ohms pin 5 to GND) the higher switching frequency dissipates more power. See: https://www.onelectrontech.com/power-mosfet-capacitance-coss-and-switching-loss/

1597493296555.png


What you need to take away from the above formula is the power dissipation in the FET due to switching is found by adding the rise time to the fall time and multiplying that by the switching frequency. So a higher rise and fall times or a higher switching frequency will result in higher losses.

Here is a table of the power dissipation for all tests:

EKPM2EKPM2EKPM3 HW 04EKPM3 HW 04EKPM3 BTN7960 0 ohmsEKPM3 BTN7960 0 ohmsEKPM3 BTN8982 0 ohmsEKPM3 BTN8982 0 ohms
Current (A)10.881710.8517.410.8517.410.9517.6
P_inductor (W)0.5451.3010.8832.2710.8832.2710.8872.354
P_Rds_on (W)0.5561.3290.7892.0890.7892.0890.6121.611
P_switching (W)0.0720.1070.4940.7950.2830.4450.1530.239
P_diss_total (W)1.1732.7372.1665.1541.9544.8041.6524.204

I have also attached a better formated version of this chart. At least I think it is because I used OpenOffice to create it, but needed to convert it to Excel format to attach it because the forum will not accept Open Office spreadsheet files (*.ods). I have no idea what it will look like in Excel.

The EKPM3 with a Walbro 450 will dissipate up to 5.154 watts. That is about double the EKPM2. When the rise/fall times are about the same the EKPM3 with 0 ohms dissipates about 4x more when switching than the EKPM2 because of the four times higher switching frequency. Even the EKPM3 with the BTN8982 and twice as fast rise/fall times dissipated more power then the EKPM2 because of the higher switching frequency.

The total power dissipation figure in the chart above represents a worst case dissipation where the switch is mostly on with just enough time to switch it on and off again. It would be operating at about 98 or 99% effort (duty cycle) in these cases. In most cases dissipation will be lower, but assuming the duty cycle is the same, the table can be used to make direct comparisons. Also these figures only take into account power lost in the inductor and the switch, but that is going to be the vast majority of the power.

Please, if something is not clear, ask a question.

The next step is to see if I can code the module to reduce the rise/fall times on the EKPM2. I think I have found the data values used to program the AS8446 MOSFET driver in the trace file. I should be able to edit the values and code the module with the edited trace file to speed up the rise/fall times (in theory).
 

Attachments

  • EKP Test Data.xls
    11.5 KB · Views: 32
Last edited:

fmorelli

Lieutenant General
Staff member
Aug 11, 2017
3,748
3,592
0
57
Virginia
Ride
E89 Z4 35i, F10 535d
Brilliant, @ajm8127. I need to sit and read this evening to digest, before questions.

In the meantime let me help on the XLS. You can embed Google Docs here. So I uploaded your XLS in Google Drive, converted to native Sheets, and here it is.

Filippo

 
Last edited:

NoQuarter

Major
Nov 24, 2017
1,662
1,066
0
Indiana, USA
Ride
Z4 35is, 535xi, X5 35i
And the switching is to control temperature of the components? While also being a contributing factor to the temperature?
 

fmorelli

Lieutenant General
Staff member
Aug 11, 2017
3,748
3,592
0
57
Virginia
Ride
E89 Z4 35i, F10 535d
So I am still gronking the write-up and the spreadsheet. I made a variety of changes to the spreadsheet (which I hope are correct) and help clarify the compare/contrast on the data.

So here are my questions:
  • Why did BMW move to the EKPM3? Are there benefits to which we are unaware, in the above analysis?
  • Is the EKPM2, actually, best suited to pursue? (you mention decreasing rise/fall times ala programming?)
  • Might there be some dialog on the desirable performance behaviors and plus/minus? (not to be in minutiae ... but enough to understand what/why we care about some things. Heat is obvious, other things less obvious to me).
I'll stop there :). I'll also note that I'm planning to move to the Walbro 535LPH F90000295, which @martymil has been vetting out successfully. If you have not seen this, @ajm8127 , you may find it interesting: https://www.spoolstreet.com/threads/ti-automotive-walbro-274-vs-285-vs-295.5882/post-95536

Filippo
 

ajm8127

Specialist
Jul 16, 2020
69
135
0
PA - US
Ride
2007 E90 335i
And the switching is to control temperature of the components? While also being a contributing factor to the temperature?

In order to control the speed of the motor a common technique in digital electronics is pulse width modulation, or PWM. The micro controller on the EKP reads a request from the DME (over the PT-CAN bus) and decides how much effort to use to control the pump (0-100%) based on the characteristic curve coded to the EKP. This effort is called the "duty cycle". Because the microcontoller can only turn the switch on or off, commanding a variable effort is achieved by cutting up the on/off periods into fractions of a second. This produces the PWM frequency, 5 or 20 kHz in this case.

At 100% duty cycle, the switch is on indefinitely. At 50% duty cycle the switch is on and off for the same amount of time. The PWM period is the amount of time of one on/off cycle and the PWM frequency is the inverse of the period. Higher frequencies can be advantageous because they reduce ripple current and are often above the range of human hearing. Lower frequencies are typically used in higher power situations because there are fewer transitions which reduce switching loss, as seen above.