work in progress Common Rail MC33816/PT2001 dev board - 4 injectors GDI

[puff]

lucky you are that it was not rigol
RIP, yet another MRE
though not sure that your reasonoing is valid...

[AndreyB]

With new microRusEFI I can click injectors again! No damage on the PT2001 board. With oscilloscope connected differently here is how it looks like

[AndreyB]

@mck1117 has this current sensing on bench with 4 cell lipo battery and 10A fuse with 1ft of wire. 2A per line.

[AndreyB]

2021 happened and PT2001 availability is now pretty horrible

image.png (53.84 KiB) Viewed 50932 times

[AndreyB]

MC33816 is way more available. Does anyone remember the difference?

image.png (68.44 KiB) Viewed 50932 times

[kb1gtt]

Did we have some small change to the internal firmware? From memory I think there was a small change. Also from memory the current firmware would work on both. Something like if.serieal number blah then X instead of Y.

I think we expect them to be interchangeable.

[Noxz]

There are two different encryption keys (and no way to turn encryption off). iirc, the compiled microcode in git is for the pt2001, but if the history/commits is looked at the mc33816 is there - I don't recall any actual code changes. We may need to make the switch back to the mc33816 code if that is what is currently available.

The/our microcode can be compiled using the development studio for the appropriate chip, I recall replacing the encryption key in a folder on the mc33816 dev studio to generate the pt2001 compiled code because of runtime libraries.

I think we added more sanity checks in the microcode SPI upload during the switch, but overall it was simply just the encryption key change.

[kb1gtt]

Ah there we go just the encryption key. I remembered it was a small thing like that.

[ssmith]

I've looked over the GDI circuit in rusefi/hardware/Common_Rail_MC33816 and compared it to NXP's AN4849 for MC33816, and have some questions/comments. Obviously the one in rusefi/ is meant to be a near copy of the NXP Application Note, so from here on I'll use the rusefi notations for components:

1. Why bother using a high side driver for the fuel pump solenoid? The Link G4+ GDI ecu doesn't have a separate signal for HPFP+, so at least on my car I have it connected to switched +12V (same +12V that the ECU gets). It'd save some board real estate by removing Q11, D16, and four small passives.

2. It seems the large capacitors aren't for boost voltage, but for smoothing. The boost capacitors are C21 and C22, for a total of 780uF (reasonably similar to what I remember seeing in a picture of the VW ECU, but now I can't find it. There's always this pic: https://rusefi.com/forum/viewtopic.php?p=38195#p38195 )
However I think the two 1000uF/80V capacitors C14 and C15 are overkill, though I may be misunderstanding the point of L1/C15. So my real question is what's the point of C14/L1/C15? Is it simply to isolate the boost circuit noise from the rest of the 12V line? Or does the L1/C15 provide some voltage boost over 12V due to ringing with the boost line toggling? If it doesn't provide a voltage boost then maybe the voltage limit of those two capacitors can be reduced from 80V to 40-50V, reducing the size. Maybe they're also assuming bench power supplies that aren't as flexible as a battery/alternator system, so maybe those caps are larger than they need to be.

I guess the VW ECU from the link above does show two inductors and three capacitors, which could map the same setup as the MC33816 circuit (12V cap, 12V to isolated inductor, isolated capacitor, boost inductor, and boost capacitor).

Again for comparison to Link G4+ GDI - they have a separate 12V input pin just for boost voltage, and they are very adamant this be routed directly back to the battery, not through any switches/etc. I think this implies they don't have good isolation / capacitance internally to the board and they're relying on having a separate power line to isolate the noise.

3. How many ounces of copper are you using on the layers? I guess the traces don't drive 10A very long, but they could live at 5A for quite a long time (injector bank duty cycle of 60+% is not uncommon, at least on my car). 5A, 30 degree rise, 1oz gives 1.4mm trace which I guess isn't too bad.

4. I saw comments in this thread about "needing" 80V for some injectors, but IMO the need for boost voltage is just a performance enhancement. Yes you'll reach peak current and thus open the solenoid sooner with higher voltage, but I'm guessing that 50-60V is just fine for most applications. If you want finer control over the opening duration at idle/low load, then drop the rail pressure; let closed loop lambda cleanup the rest.

FWIW I gave up on ditching the MC33816 and doing it all in the primary MCU. I would have had to add a quad op amp, added 10 gate drivers, and really would have needed 5 comparators (NXP MKV5* only has 4). But worse I can't fit everything I need in a 100-pin MCU package (with 30 analog inputs, 16 digital inputs, 25 outputs, 8 comm - can/uart, etc), and the other MCU I'm looking at have 2 or fewer comparators. Plus I think if the boost circuit goes wonky you can get blown capacitors/inductors pretty easily (hold down the inductor too long and it'll pass too much current; keep boosting too much and you'll exceed the limits of the capacitor voltage). I guess that's why you fuse the ECU....

[ssmith]

5. Along the lines of removing unneeded components - why not remove Q3 and Q5 too? I.e. the positive rail for the injectors is either 12V - diode, or boost, but never floating? Less heat by running current through fewer components.

EDIT: this (along with the other component removal) would require code changes, as it looks like the sample in AN4849 toggles the high side driver instead of the low side driver.

EDIT: to reduce heat, replace half of D10 with P-ch Mosfet, as illustrated here: https://www.powerelectronictips.com/inexpensive-ideal-diode-mosfet-circuit/

[Noxz]

You are correct that it is a near copy of the evaluation board - we haven't optimized much at this stage..

1)
In [my] diesel setups there are two solenoids to control, one is a high pressure valve on the rail that can simply be held at VBat (and released when pressure builds too much). The other is a linear solenoid valve on the input of the high pressure fuel pump (which is driven off the timing chain), and this is where specifying a certain percentage, easily using the controller by updating a register value through SPI, is advantageous.. I have some info on how I tested the Amps->Flow: https://rusefi.com/forum/viewtopic.php?p=32280#p32280

But we still do not have any sort of PID control for regulating the desired pressure in the rail. This appears to be much more of an issue with diesel where pressures are upwards of 2000 bar, and is much more for an efficiency standpoint.

There are other setups that use a non-linear valve for the fuel restriction, some are timed to the cam position and would require a bit to get interfaced than the simple linear setup.


2)
iirc L1/C14/15 is a pi-network for smoothing, it shouldn't provide any additional voltage. I don't recall any voltage spikes observed on the line. Sometimes the parts are chosen due to availability and price

The voltage rating on the battery and boost caps on the eval board as well as our "0.3" (which is what is in front of me, I forget what is on the BOM) are rated at 100v.

3)
I didn't have the boards made, I forget the copper thickness specified - someone else may need to comment on that. Again, hardly anything has been optimized as we are still in the Proof of Concept stage.

4)
You are correct that the purpose of Peak-Hold is to quickly open the injector solenoid and then keep it open.. The higher voltage/power pushed through the faster it will open. In the case of diesel, recently there is Multi-Injection, so being fast here is very advantageous.

This high voltage is completely configurable through the microcontroller.. I believe I tested it from it's limit at 72v down to 45? It's a single word that you can update in SPI (you then need to cycle the BOOST pin to have the asic reload this value)


This little chip does quite a bit - sure it could be implemented elsewhere but it is a ready to go solution.

Sure, there are places to optimize, but I don't think we are there yet. Ideally a single board with the STM32 would be desired once we have a settled on a feature set and layout.

I personally do not like to look at what other (ECU) are doing if a goal is to opensource, it's better to stay in the "clean room".
I hope to contribute more to the fuel pressure regulation this winter.

All good notes and quite the deep dive..

[ssmith]

I'm on the cusp of deciding whether to build my own board or just add an ion sensing pre-board that sends timing information back to my current ECU, so that's why I'm pushing forward with this now. I'm willing to tailor my board to my own needs (and I thought that was the point of Hellen, too), so if a component is only needed for diesel and not GDI then I can drop it (maybe Hellen will end up with two layouts, one for diesel and one for gas).

1. The variable % voltage stuff sounds like piezo injectors, or continuously variable fuel pump (I forget the term). Both of those were used on BMW N54 engines (gas). Piezo injectors are used on the Mercedes M139 engine (2.0T GDI making 400+hp), otherwise it seems that solenoid style is more common nowadays. Motec actually makes separate ECU for piezo vs solenoid injectors, but can control both hpfp from either ECU, so maybe I'm misunderstanding what is needed from a circuit perspective.

2. Makes sense that the boost caps need 100V rating, it just seems that the battery caps can be lower voltage (and maybe lower capacitance too). I'm trying to figure out how to reduce board space so I can fit all the other junk I need.

4. Interesting on multi-injection. I've heard of multiple spark for GDI but obviously that's not an option for diesel
I was only making the point that using lower voltage is probably adequate given the MC33816 can't handle more than 72V. Maybe the opening time is 10% longer at 72V instead of 80V, in which case just add 40uS to deadtime and move on....

[Noxz]

The point of our current board was to give a slightly reduced footprint+parts over the evaluation board (which is tailored for the KL25Z host microcontroller board) while still retaining most functionality, and into the hands of more people.

I personally would want a slightly customized version as well, maybe the 6 cyl variant but only in order to gain a workable channel for the high pressure solenoid, otherwise I'd need a separate driver and just hold it at VBat (which is more or less what my test setup is doing).. We actually dropped the "Fuel Pump 2" from the eval board since we never saw a case where we have to switch between two VBat solenoids - and unfortunately both can not be driven at the same time from the chip as they are on the same channel.

5)
I believe Q3 and Q5 (sorry, I did a reinstall of my computer and I don't have my dev environment set up enough to actually pull up the schematic, so this is by memory and also looking at the boards [0.3] in front of me) are the VBat fets.. Basically the chip is setup into channels, and 2 injectors per channel. There is a fet for the Boost and the Battery voltage for each channel, and then a low-side fet for each injector, so you have 4 total per channel as the two injectors share the same Boost & Battery fets.. and I think the chip does some fault tolerance with this additional control.


I think piezos require a much higher voltage, like 120v+ last time I read? I am unsure if they need a peak-hold topology. I guess we'll venture into that land once someone gets a car that uses them.

And understandable that you want to use the least space as possible.. Funny that the size difference from the commercial eval board was quite impressive to us. I am sure use of both sides could be done better.. having only the two inductors on the bottom side (well, and diodes and other SMDs) is not exactly optimal neither, and if it is on your desk/bench not mounted to anything it wobbles..

[mck1117]

1) Like Noxz said, some diesel fuel pumps need the boost voltage to open fast enough, but most GDI pumps do not, and can be controlled lowside-only.
2) C14/16, L1, C15/17 form an input filter to prevent boost converter switching noise leaking out of the ECU. These are about conducted mode EMI. Even a good separate wire to supply the boost converter does nothing to contain EMI. As for voltage rating, the supply wire for this may be too high of impedance to try and effectively clamp a load dump event, which means the caps have to take the full brunt, with the input filter keeping the input voltage on the chip low enough to not vaporize it. I actually don't think they're very far oversized.
3) it's 1oz copper
4) nobody makes a 60v rated cap, so you use 100v caps. Perhaps more importantly, you WILL get switching transients far above the target voltage, and you really don't want those transients poking holes in your dielectric in the cap. Look inside an OEM ECU some time, you'll see much higher voltage ratings than you think are required.
5) Q3 and Q5 are about improving efficiency. They do mention a design without those transistors in an app note, but it loses you an extra diode drop on the high side, which is a heat vs. cost tradeoff that you may or may not want to make.

Piezo: Piezo injectors are just a stack of piezo elements that gets longer when you apply voltage. It's functionally a giant capacitor. I seem to remember you just drive them with a half bridge: short to the boost rail to turn on, short to ground to turn off. They draw zero current steady state.

[ssmith]

5)
I believe Q3 and Q5 (sorry, I did a reinstall of my computer and I don't have my dev environment set up enough to actually pull up the schematic, so this is by memory and also looking at the boards [0.3] in front of me) are the VBat fets.. Basically the chip is setup into channels, and 2 injectors per channel. There is a fet for the Boost and the Battery voltage for each channel, and then a low-side fet for each injector, so you have 4 total per channel as the two injectors share the same Boost & Battery fets.. and I think the chip does some fault tolerance with this additional control.

I'm not sure how much fault tolerance it really buys you; older port injection engines that I've looked at have had relays to control the +12V to the injectors, but this is just a mosfet. If you don't trust the low side mosfet to turn off, do you really trust the high side mosfet to turn off?

5) Q3 and Q5 are about improving efficiency. They do mention a design without those transistors in an app note, but it loses you an extra diode drop on the high side, which is a heat vs. cost tradeoff that you may or may not want to make.

There's a diode from the +12V to inj+ path anyways, D10 and D11 (well, half of them anyways). That's why I'm not sure what the point of the mosfet is. I wasn't able to find anything else directly referring to those two mosfets in AN4849.

Though I later realized that the code they supply enables the lowside mosfet for a given injector, and then toggles the high side. If you get rid of Q3/Q5, you have to change the code to hold the high side and toggle the lowside mosfet. For all I know that isn't easy/possible, or perhaps it has other ramifications, like an inability to measure current through the sense resistor (though their sample code doesn't do that when the injector is off, instead it waits a certain amount of time).

Actually I might have found one reason (p6 of AN4849):

During the Bypass phase, all the low-side and high-side switches are turned off. The current decays through the injector, the
diode connected to ground, and the diode connected to V BOOST for a fixed time (t BYPASS ).

If you get rid of Q3, then the inj+ voltage will always be 12-ish volts, which means inj- will elevate higher than if inj+ was at 0-ish volts (due to the other half of D10). I wonder if that could pump too much voltage/current/whatever into Vboost.

Similarly, Vboost could be elevated during hold. By anchoring the low side, you only have to discharge current through the diode into the ground rail, rather than anchoring the high side at 12V and having the low side elevate above VBoost.

[ssmith]

4) nobody makes a 60v rated cap, so you use 100v caps. Perhaps more importantly, you WILL get switching transients far above the target voltage, and you really don't want those transients poking holes in your dielectric in the cap. Look inside an OEM ECU some time, you'll see much higher voltage ratings than you think are required.

No, but 63V is pretty common. I don't have any OEM ECU I'm willing to open, but I have two TPMS boards, one with a 35V cap and an older one with a 50V cap.

[Abricos]

16-17 VOLKSWAGEN JETTA 1.4L CZTA /ENGINE

IMG_20211202_221654.jpg (2.76 MiB) Viewed 50662 times

IMG_20211202_221708.jpg (2.83 MiB) Viewed 50662 times

IMG_20211202_221722.jpg (2.63 MiB) Viewed 50662 times

IMG_20211202_221730.jpg (2.79 MiB) Viewed 50662 times

IMG_20211202_221744.jpg (2.35 MiB) Viewed 50662 times

IMG_20211202_221753.jpg (3.65 MiB) Viewed 50662 times

IMG_20211202_221809.jpg (4.17 MiB) Viewed 50662 times

IMG_20211202_221901.jpg (4.79 MiB) Viewed 50662 times

IMG_20211202_221914.jpg (3.59 MiB) Viewed 50662 times

IMG_20211202_221923.jpg (3.12 MiB) Viewed 50662 times

[Abricos]

IMG_20211202_221935.jpg (3.92 MiB) Viewed 50662 times

IMG_20211202_221950.jpg (3.18 MiB) Viewed 50662 times

[ssmith]

Awesome thanks...

Guess I should take it as a hint that if the OEM needed both sides of the board, then maybe I should look for a bigger case.

Hey look, a BGA ecu! Where's the BGA thread...

[kb1gtt]

Stackup can be found here.
https://jlcpcb.com/quote/pcbOrderFaq/PCB%20Stackup

From memory I think it was 1oz external and 0.5 oz internal.

[AndreyB]

We had MC33PT2001 availability scare a couple of weeks ago but looks like distributors are getting trays of 160 chips! I'll take the risk with "three weeks from now" cheaper mouser option while keep at eye on how those 160 in stock at digikey would go.

https://github.com/rusefi/rusefi-hardware/issues/8

image.png (2.29 KiB) Viewed 50224 times

[AndreyB]

Let's make a GDI injector driver with nice package. More or less same thing as bosch 7506280

https://github.com/rusefi/rusefi-hardware/issues/9

[AndreyB]

aux analog inputs, aux high current low side outputs... https://github.com/rusefi/rusefi-hardware/labels/GDI

at least no LCD screen so far.

Totally

[AndreyB]

Surprise surprise BUK9230-100B is not available and this board used many of those. Was 47A really needed? Would VND14NV04 12A do or is it a wrong kind of part?

[mck1117]

No, you cannot use protected drivers for this board. They must be plain unprotected N-channel mosfets. The protection is internal to the PT2000/1.

[kb1gtt]

I recall some concern about the transition time as well. AKA faster or slower devices may cause issues. I recall this concern was an unknown, so if you make a change I suggest to make low QTY order and test before committing to larger QTY.

[AndreyB]

I will order whatever i am told to order it would be helpful if anyone gives me an option or three

[AndreyB]

qty30 PT2001 have arrived

[AndreyB]

GDI4 which is PT2001+stm32F103 on the same board is stuck more or less the same way as https://rusefi.com/forum/viewtopic.php?p=44162#p44162 just using https://github.com/rusefi/rusefi/tree/master/misc/stm32f1_test_project codebase. Need help or moral support

[AndreyB]

Progress is happening with GDI4 board!

Maybe last open question https://github.com/rusefi/rusefi-hardware/issues/33 need help