DIY DIY EKPM3 Active Cooling for LPFP

[NoGuru]

I've been lurking on this forum for a while, but have been in the community for many years. I was on a different forum a few years back but it devolved into a mess and most people came here. I just took a break from forums for a while. I am also an electrical design engineer so I have some insight here.

From the back of the board you can clearly see where the heat is being generated. For it to get discolored like that it is getting pretty warm. The part in question is a motor driver and that is what is going to dissipate all of the power. Pretty obvious.

Your approach to slap a heatsink on the top of the package is logical, but the thermal resistance from the package die (the silicon die inside the part where the power is being dissipated) is much lower to the board than it is to the top of the package. The part has an exposed pad under it ("pin" 8 in the datasheet) specifically to help get the heat out. You can clearly see a lot of heat is being transferred to the board. That's one of the reasons there are so many vias in the board under the driver. You can see that copper pad on the underside has no connection electrically. It is there to help transfer heat.

I would focus on transferring the heat from the board to the aluminum plate. You can see the part is already designed to do that. Heat gets transferred to the board and then through the gap pad and into the plate where there is a raised area to ensure good contact and a recessed area to prevent interference with parts on the bottom of the board. I would change out that pad with the same thickness pad with a higher thermal conductivity (W/m*k) rating.

Check digikey.com, but watch the "Usage" column. Some of the newer graphite pads are designed to conduct heat in mainly one direction (through plane vs. in plane). Also, graphite is conductive so it would be a poor choice for a PCB. I would just exclude it from the "Type" column and then sort descending by "Thermal Conductivity". Pick one of the right size and thickness.

I would also attach a large heatsink to the aluminum plate and mount the EKP upside down. Orient the fins vertically so convection currents can form. You may or may not need a fan this this solution. Use some non threaded spacers to avoid crushing the plastic case. A new bottom section made from a piece of Al with heatsink fins milled in would be best, but putting a larger heatsink on the aluminum base is second best. Use some CPU thermal compound between the plate and the heatsink along with the new thermal pad between the board the the plate as mentioned above.

Old CPU coolers look like they would be just about the right size to mount on the plate, and they already come with 12V fans. You could attach one with screws to the EKP plate. You should be able to pick up 12V through terminal 30G, the power supply to the module. I think that is switched with the ignition.

I notice the BTN7960 is discontinued. I wonder if pin-for-pin compatible parts exist that have less loss and higher continuous current specs. Those big parts are easy to replace.

[EDIT] BTN8982 replaces BTN7960, is pin for pin compatible, has lower on state resistance and has higher current carrying capabilities. https://www.infineon.com/cms/en/search.html#!view=discontinued&term=BTN7960

WOW, this is such good information, thank you.
Regarding replacing the BTN8982 with BTN7960, how much of a difference would this make?
I mean is this a marginal change or would it make a big difference?

Also this looks like a good replacement pad. https://www.newegg.com/arctic-cooli...mal_pad-_-35-186-163-_-Product&quicklink=true

I have a spare EKPM3 I am willing to test this with.

 

[ajm8127]

Section 5.4.3 explains the resistor that is next to the chip on the EKP. Seems that resistor can be changed to allow more current at the expense of more emission interference.

Changing that resistor should allow more load but may have to shield it better.

The increased emissions could be a pretty big problem considering the high currents needed by the pump and the wires that go between the pump and the module. The wires are where the emissions will emanate from. They will act as an antenna. If the wires were twisted it would minimize the "loop area" and reduce the effectiveness of the wires as an antenna. I am sure there are products like wire loom that are metalized to help shield whatever is inside to further reduce emissions.

I agree increasing the slew rate will cause the MOSFET to switch faster which reduces switching loss in the FETs to help keep the module cooler and help achieve a higher pump current if the temperature of the module is the limiting factor, which I think it is. It's definitely something lo look into.

I see the resistor on the board Filippo posted appear to have "682" written on it, but depending on the brand this may or may not be the value. I wouldn't want to say definitively it is a 6.8K, but that is a standard value. Looks like you reduce the value to increase the slew rate (lower values of rise or fall time, high values of slew rate - the datasheet lists both which was confusing me at first). You could try sticking a DMM across the resistor, but measuring components in circuit can introduce errors in the measurement because you are measuring the resistor and whatever it is connected to.

Maybe I should just buy one from Ebay. I have been thinking of replacing mine anyhow. Are all the EKPM3 modules the same except for the coding?

 

[NoGuru]

The increased emissions could be a pretty big problem considering the high currents needed by the pump and the wires that go between the pump and the module. The wires are where the emissions will emanate from. They will act as an antenna. If the wires were twisted it would minimize the "loop area" and reduce the effectiveness of the wires as an antenna. I am sure there are products like wire loom that are metalized to help shield whatever is inside to further reduce emissions.

I agree increasing the slew rate will cause the MOSFET to switch faster which reduces switching loss in the FETs to help keep the module cooler and help achieve a higher pump current if the temperature of the module is the limiting factor, which I think it is. It's definitely something lo look into.

I see the resistor on the board Filippo posted appear to have "682" written on it, but depending on the brand this may or may not be the value. I wouldn't want to say definitively it is a 6.8K, but that is a standard value. Looks like you reduce the value to increase the slew rate (lower values of rise or fall time, high values of slew rate - the datasheet lists both which was confusing me at first). You could try sticking a DMM across the resistor, but measuring components in circuit can introduce errors in the measurement because you are measuring the resistor and whatever it is connected to.

Maybe I should just buy one from Ebay. I have been thinking of replacing mine anyhow. Are all the EKPM3 modules the same except for the coding?

So far what I have found is they do not need to be coded. EKPM3 just plugs in an works. Now maybe if one was from a model way different it could need to be codded but I don't think it would be too difficult to do.

 

[ajm8127]

The difference between the high side on state resistance is about 20-25% between the parts depending what values you look at (typical vs. max, 25 C vs. 150 C). So that would the same difference in power dissipated in the part: 20-25% less with BTN8982. But this only applies if the pump is being driven at 100%, i.e. no PWM.

Once the pump is driven at less than full capacity power lost in the FET due to the slew rate limiting comes into play and that is harder to quantify. At lower PWM duty cycles, the slew rate limiting function dominates the power dissipated. At higher PWM duty cycles, the on state resistance dominates.

However, the current is highest at 100% duty cycle (pump getting full power) so that's when the most heat will be dissipated. The only way to know would be to quantify the power dissipated by the high side MOSFET when it is switching. It would be easiest to do this empirically because knowing exactly how much power is lost in the FET with slew rate limiting active is not easy. This is something you'd probably need to get directly from an Infineon Applications Engineer assuming we could find one that would tell us.

So I suspect it would make a noticeable difference when driving the pump at 100%, and less of a difference at lower duty cycles. Replacing the part with the newer part and increasing the slew rate work together to make the module more efficient.

One other thing to be aware of if replacing the part is both parts have current sensing. A current flows out of pin 6 proportional to the current in the high side MOSFET. This current flows through a resistor to produce a voltage that is read by the microcontroller on the EKP (probably). If the part is changed and that allows a higher power pump to be used, the current will increase. Eventually, the microcontroller on the EKP might read this current and put the module into a fault condition if the current is too high. Without knowing the specifics of the coding, it's tough to say where this threshold is at. Theoretically, the resistor value could be adjusted to reduce the amount of voltage fed to the micro relative to the current in the MOSFET.

Concerning the pad, maybe Filippo will measure the thickness of the existing pad and that will guide selection of a replacement pad.

 

[S54ca]

I've been lurking on this forum for a while, but have been in the community for many years. I was on a different forum a few years back but it devolved into a mess and most people came here. I just took a break from forums for a while. I am also an electrical design engineer so I have some insight here.

From the back of the board you can clearly see where the heat is being generated. For it to get discolored like that it is getting pretty warm. The part in question is a motor driver and that is what is going to dissipate all of the power. Pretty obvious.

Your approach to slap a heatsink on the top of the package is logical, but the thermal resistance from the package die (the silicon die inside the part where the power is being dissipated) is much lower to the board than it is to the top of the package. The part has an exposed pad under it ("pin" 8 in the datasheet) specifically to help get the heat out. You can clearly see a lot of heat is being transferred to the board. That's one of the reasons there are so many vias in the board under the driver. You can see that copper pad on the underside has no connection electrically. It is there to help transfer heat.

I would focus on transferring the heat from the board to the aluminum plate. You can see the part is already designed to do that. Heat gets transferred to the board and then through the gap pad and into the plate where there is a raised area to ensure good contact and a recessed area to prevent interference with parts on the bottom of the board. I would change out that pad with the same thickness pad with a higher thermal conductivity (W/m*k) rating.

Check digikey.com, but watch the "Usage" column. Some of the newer graphite pads are designed to conduct heat in mainly one direction (through plane vs. in plane). Also, graphite is conductive so it would be a poor choice for a PCB. I would just exclude it from the "Type" column and then sort descending by "Thermal Conductivity". Pick one of the right size and thickness.

I would also attach a large heatsink to the aluminum plate and mount the EKP upside down. Orient the fins vertically so convection currents can form. You may or may not need a fan this this solution. Use some non threaded spacers to avoid crushing the plastic case. A new bottom section made from a piece of Al with heatsink fins milled in would be best, but putting a larger heatsink on the aluminum base is second best. Use some CPU thermal compound between the plate and the heatsink along with the new thermal pad between the board the the plate as mentioned above.

Old CPU coolers look like they would be just about the right size to mount on the plate, and they already come with 12V fans. You could attach one with screws to the EKP plate. You should be able to pick up 12V through terminal 30G, the power supply to the module. I think that is switched with the ignition.

I notice the BTN7960 is discontinued. I wonder if pin-for-pin compatible parts exist that have less loss and higher continuous current specs. Those big parts are easy to replace.

[EDIT] BTN8982 replaces BTN7960, is pin for pin compatible, has lower on state resistance and has higher current carrying capabilities. https://www.infineon.com/cms/en/search.html#!view=discontinued&term=BTN7960

Ha, exactly - I was just coming here to post that BTN8982 is a drop in replacement with lower Rds on and marginally lower Rth. It's readily available from Mouser for $5.15.

Combine this updated package with a new gap pad and the t-rise should improve by a fair bit.

 

[fmorelli]

All EKPM3s are not the same, hardware-wise. Firmware revisions exist as well, but right now hardware seems long pole in the tent.

Filippo

 

[fmorelli]

Well ... since we may be on a roll (hopefully not to a dead end. Might a worthwhile alternative be - upgrade the IC (probably needs rework station to pull that off), and go with a known high thermal conductive pad? (as a for instance: https://www.aquatuning.us/thermal-p...ermal-pad-11w/mk-120x20x0-5mm-2-stueck?c=6692).

The existing pad is super thin - 0.22mm. I took some other close-up photo since I have these out.

Back to comments about how to mount the EKP - I believe all cars but the E89 Z4 have the EKP sandwiched behind the seat. But if heat transfer is to the aluminum plate, then the question is what is that mounted too? Or just mount a bigger plate and increase cooling area. Z4 it's in the hatch.

Honestly I hate hate - it kills things. But if swapping the IC, changing the pad, and slamming it back in (maybe with a heat sink on the back of the aluminum or whatever), and then run the no check valve Walbro pump ... that could have plenty of headroom with minimal effort. As is common, figuring out the path is the work.

Filippo

 

[fmorelli]

@ajm8127 glad you are no longer lurking! Thank you for your contribs to this thread ... and welcome to Spoolstreet.

Filippo

 

[ajm8127]

Interesting. I clearly see "HW 06" and "HW 04" on the top plastic covers. I assume because the are all "EKPM3" they are functionally equivalent. I would bet the numbers "4510" and "3409" are date codes in WWYY form.

The main difference I see are the capacitors for filtering. There are fewer of them in the HW06 version.

Can you take full resolution pictures of the top and bottoms of each board? Might not want to embed them inline because of the size, but maybe load them as attachments. If the parts numbers were visible that would be most helpful. I am mainly interested in the inductor (square metal thing) and the microcontroller. It looks like the section with the micro is conformally coated, so getting the part number to show up in the picture might be challenging.

I was looking at one near me in PA that claims to be out of an F chassis that has a build date of 2712. It has a completely different HW rev number: 000.001.001

BMW F30 F32 F10 F12 F25 F26 3 4 5 7 X3 X4 FUEL PUMP CONTROL UNIT EKPM3 7301554 | eBay

Find many great new & used options and get the best deals for BMW F30 F32 F10 F12 F25 F26 3 4 5 7 X3 X4 FUEL PUMP CONTROL UNIT EKPM3 7301554 at the best online prices at eBay! Free shipping for many products!

www.ebay.com

I see 1614 7229173 is the current one RealOEM lists for my 10-2006 E90 build. It matches the one on the right, but I would be surprised if the one of the left or the one from ebay couldn't be used, though they might need coded.

For the pad, I would assume that is 0.25mm thick nominal. Digikey has a lot with 0.5mm thickness with thermal conductivity around 16 W/m*k. The thinner ones seem to be significantly less, of course they are thinner, but to match (produce same delta T) 16 W/m*k at 0.5 mm, a 0.25 mm pad would need to be 8 W/m*k (I think, I am not a mechanical engineer). These are below 6.

I thin this is a good candidate:

TG-A1660-40-40-0.5 t-Global Technology | Fans, Thermal Management | DigiKey

Order today, ships today. TG-A1660-40-40-0.5 – Thermal Pad Gray 40.00mm x 40.00mm Square from t-Global Technology. Pricing and Availability on millions of electronic components from Digi-Key Electronics.

www.digikey.com

 

[veer90]

very insightful thread exposing all the weak points and inadequacies in the EKP. now we know why some people are getting limp and fuel pump malfunction even stock, cus the board is simply getting too hot and getting cooked.

THIS IS WHAT HAPPENS WHEN YOU LET BEAN COUNTERS AND NOT ENGINEERS DESIGN STUFF ARGH

 

[fmorelli]

Well ... @veer90 ... except none of those things seem to happen when it's pushing the lowly stock lpfp .

@ajm8127 glad to send you full blown pics (just PM me), but I not sure why you'd want to compare the older capacitor-laiden LPFP with the newer one.

Here's a quick chart on the differences between the two IC's. The grey rows are identical.

Filippo

 

[ajm8127]

I am starting to think the only reason EKPM3 was created was for cost reduction. Let me explain.

I pulled the EKPM2 out of my 10/2006 build E90. It is a HW rev 5.0 and has a build date of week 33, 2006 (BMW PN 16147180426). Below is a picture of the top layer of the PCB. Excuse the blur, the near range camera in my phone refuses to focus, and I refuse to replace the whole phone because of that problem.

The green box is the micro controller, NXP MC9S12D64. On the bottom of the board there is a CAN transceiver, NXP A82C250.
The blue box is a DC motor driver from a company called Austria Micro Systems. There are no references to it on their website and I couldn't find an actual datasheet, only a single page features list that is mainly a marketing document. The part number is AS8446, as shown.
The yellow box is a N-channel MOSFET, ST BUK9107-40ATC. This is the power switch.
The red box is a dual Schottky diode, ST STPS3045C. This protects the FET when it switches off and the magnetic field in the pump motor causes a voltage spike that could exceed the FET's maximum Vds.

Here is a diagram from the AS8446 datasheet. The only difference in implementation is I doubt the LIN bus connection exists. All comms happen via CAN through the micro controller most likely. Also the EKPs have a big differential pi filter built in for conducted emissions compliance. That is what the big capacitors and inductor are for. The diagram only shows an input capacitor. In the EKPM2 the inductor is wound on that ferrite rod at the top of my image above. In the EKPM3 it is the square metal box around the middle of the board.

So Tr1 is the BUK9107-40ATC and D1 is STPS3045C.

What I find interesting is this MOSFET is rated for Rds_on of 14 mOhm max at 175 ° Tj and 5 V Vgs, and 5.2 typ 6.2 mOhm max at 25 °C and 10 V Vgs. Both of these ratings are at 50 amps! The part has a maximum constant drain current of 75 A which is limited by the package. I included the typical Rds_on value at 10 Vgs because the FET is driven by a driver(AS8446) with a charge pump so the Vgs is going to be closer to 10 than 5 V in normal operation.

There is a proportional relationship between Rds_on and drain current. As drain current increases, so to does Rds_on The BTN parts in the EKPM3 have Rds_on rated at only 9 amps. So these aren't apples to apples and the BTN parts will have a higher Rds_on at higher currents. The FET in the EKPM2 is rated at nearly the same Rds_on as the H-bridge in the EKPM3, but the EKPM2 part specifies that rating at more than five times the current.

The reason I think the EKPM3 is a cost reduction revision is because they move the driver, current sensing and power switching into a single part, the BTN series. The newer module uses newer parts (see surface mount caps and inductor), but I think the FET in the old module will outperform the BTN series H-bridge in the new module. Or maybe a part went obsolete, so it was time to redesign the module.

From what I see there doesn't appear to be any reason to believe the EKPM3 will outperform the EKPM2.

I would love to hear your experiences with the EKPM2 and EKPM3. Did you originally have an M2, did you switch to an M3? Why did you switch? Maybe I am overlooking something. All I have done is analyze the parts on the boards. I plan on doing some testing on my EKPM2 this weekend to learn more about the module's efficiency.

 

[NoGuru]

My car (2011 335is) came with an EKPM3. I have since added a Walbro 535, and a Walbro 450 on a 15psi Hobbs.

What I have heard, is if you upgrade the LPFP on a EKPM2 it is prone to fail. I have no proof of this just what I have read.

Since I have this setup with a blower style fan attached I have not had any issues, even after doing about 30 runs at the drag strip last weekend.
Has my cooling helped? Maybe, maybe not but I like doing this for science. So for science I will try and get a Temp probe out and see temps with the fan unplugged vs plugged in if I can find the time.

 

[ajm8127]

I have read reports of the EKPM2 not being able to handle a 450 as well. That seems to be conventional wisdom. But there are some threads that seem to indicate the EKPM3 really can't either.

In this thread @Jeffman was having a problem with low fuel pressure when cruising or idling. He replaced his EKPM3 with a BMP4 which is designed for aftermarket fueling solutions. That seemed to fix his issue.

Should I replace my old EKPM3 module with a new one?

My 2008 335xi E90 came with an EKPM3 module. In the last year or so my LPFP pressures no longer stay at 72psi when cruising or idling. Instead pressures drift down into the 60s, and then into the 50s, and then into the 40s as I drove along and the engine heats up. The engine otherwise feels...

bmw.spoolstreet.com

It just seems like the EKPM3 isn't a reliable solution for a Walbro 450 either. I think that is why people started to get curious about active cooling a EKPM3.

 

[gmagnus7]

At least in the CPU and VRAM world, changing from a lower thermal conductivity paste/pad to a higher one can be the difference between a computer that can overclock, and one that can't.

Wouldnt be surprised if that fixes the problem - assuming that's the actual issue. Look for any 17 W/mK thermal pad, there's a few companies that make them. If youre unsure of the thickness necessary then go for one that can compress and is slightly larger in size (within reason). Not all of them compress well, and some are one time use.

 

[NoGuru]

Would be nice to see the board of the BMP4.

 

[veer90]

Would be nice to see the board of the BMP4.

They don't want you to lol it says opening the case voids your warranty

 

[veer90]

Well ... @veer90 ... except none of those things seem to happen when it's pushing the lowly stock lpfp .

Not true. I've seen reports here and there of fuel pump malfunction popping up on idrive for cars with a stock lpfp and aging EKP. Usually they're very confused because the car only throws limp mode or just dies after 15-30 mins of driving and is fine after shutting it off for 10-15 mins and trying again.

Sounds like overheating EKP to me.

 

[Jeffman]

My lpfp pressures are still under 72 psi, especially when it’s warm outside. I also have the original lpfp sensor installed which may be the source of the problem / sensitive to temps. I have a new one to install to check - will do the next time I’m taking off the intake like when walnut blasting. But overall I think my pressures are slightly better with the BPM.

 

[NoGuru]

They don't want you to lol it says opening the case voids your warranty

How long is the warranty, maybe someone out of warranty can open theirs.