TLE8888 Ground path.

[LucasFlaneto]

Hi! My name is Lucas and i am currently designing a simple ecu system as my college project (Electronic Engineering),Searching for designs already made to have some kind of start i found that the microrusefi uses an all in one IC for power connections, the decision of going with this route was made.

After reading the TLE8888 datasheet(specifically page 6 and 7) i found that the AGND and PGND pins are internally connected, how did you guys solved the problem of ground loop?
can i simply ignore it and attach both to the ground plane? did you use the ic as the single point connection of the signal ground(battery ground) and the power ground(chassi)?

Sorry for any typing or writings errors, english is not even my second language
If anything was not made clear in this post i am eager to help clarify my doubt

[AndreyB]

I am not the person to respond on hardware but just curios what is your software plan for TLE8888 - you need some code to get outputs to work right? Also you need watchdog code if you go with part number with watchdog.

TLE8888 itself is not a trivial chip so any ECU based on it would not be exactly simple. But you have our code to look at

[kb1gtt]

See Maxim article and other resources found here https://rusefi.com/wiki/index.php?title=Manual:Hardware:PCB_design_rules

[LucasFlaneto]

Thank you so much for this help, there it is something i did not learnt at college, concerning our project we are an 5 person team currently designing the system and surely we will take our peek at your code(and surely you guys will be in a huge chunk of our references page at the article), the current idea is to directly drive the output pins trough the input ones and use the SPI bus for diagnostics only.

[AndreyB]

the current idea is to directly drive the output pins trough the input ones and use the SPI bus for diagnostics only.

direct drive outputs are disabled until you enabled them via SPI
my advice to get part without watchdog, there are three different part numbers available you want -2QK

[AndreyB]

Is there a way to twist your work from "ECU from scratch" into "microRusEFI improvement by adding features XXX and YYY"?

[LucasFlaneto]

the current idea is to directly drive the output pins trough the input ones and use the SPI bus for diagnostics only.

direct drive outputs are disabled until you enabled them via SPI
my advice to get part without watchdog, there are three different part numbers available you want -2QK

got it.

Is there a way to twist your work from "ECU from scratch" into "microRusEFI improvement by adding features XXX and YYY"?

it could be done but our project already has more than 1 year of development and it is a full package of pcbs to handle from engine needs to traction control and ABS systems. This issue was perceived during our final checks in the schematics and pcb to be more safe that we wont order more than 500$ (4 months of minimun wage in our contry) and it not work.

As far as myself, i would be interested in contributing with this and i am sure that some colleagues would too, which ECAD software are you guys using for pcb design? we have experience with kicad and eagle.

Changing the subject, after we get everything to work and get confident in our work, we will be releasing everything in the open source community and everyone will be able to have access to the article and pcb files as long as there is no commercial exploitation of the systems.

[AndreyB]

We use KiCad. There is also an opinion that writing the code is at least same level of effort as producing hardware or maybe even a bit bigger effort.

Where are you with the firmware?

Pictures of your test bench? Pictures of your current hardware? Any schematics you would like to share?

[AndreyB]

https://github.com/rusefi/rusefi/wiki/Hardware

[LucasFlaneto]

We use KiCad. There is also an opinion that writing the code is at least same level of effort as producing hardware or maybe even a bit bigger effort.

Where are you with the firmware?

Pictures of your test bench? Pictures of your current hardware? Any schematics you would like to share?

I only did small programs (<2000ish) lines of code, so i not sure if i could help with firmware (colleagues in software engineering could be interested)on the other hand i have 5 years of pcb design experience and a handful of projects tested and working, regarding this one i never did anything that would cost more than 50$ or take more than a few months to design so in the last 40 days we are in the process of generating 3d models of boards and revising everything as many times as is humanely possible to avoid errors before our first batch is produced.

regarding our bench we us the college one (we made some connections with the "higher rank" academics and got a small lab four ourselves). there we have 2 oscilloscopes, 2 arbitrary function generators,2 power supplys, one lpkf machine for simpler single sided boards and the huge amount of ~80$/month as "research aid"

regarding hardware there are some pics of the 3d models, the small board is one of the daughter boards that goes in the wheels, it has 2 VR sensor inputs and 2 analog 0-5v that are connected to brake fluid pressure sensor and brake fluid temp sensor, we don have the 3d model yet but there is also an mlx90616 ir sensor for disc temp, all daughter board communicate via CAN bus to the ECU. The bigger its the ECU itself.there are more models of daughter boards(etc and more) but my dual core 1.1ghz notebook is almost killing itself just opening fusion360. (my excuses for the unfinished components models as black or red squares in the pcb)

[kb1gtt]

Does your school have FEA software? It would be cool to run simulations on electrical components, and example circuits. I currently want to know more about how thermal relief on via's work, and when they should be used.

[kb1gtt]

I think that MAXFem could be used to generate the heat generated profile, then perhaps it could also be used to show how the heat moves around on the PCB. Perhaps CAELinux could do it. Both potential solutions are to big for me to take on these days. But perhaps you have some more time than I have. Or perhaps you can work int into your school studies.

[LucasFlaneto]

Does your school have FEA software? It would be cool to run simulations on electrical components, and example circuits. I currently want to know more about how thermal relief on via's work, and when they should be used.

yep, i primarily done the pcb itself but i dont have much FEA experience, everything except the uC, the TLE8888 and the Bluetooth is in the process of eletrical simulation.The priority right now is to see if we can put logic level power mosfets in the coil drivers, FQD12N20LTM was the best candidate so far as the rds at 5V was very low, that would remove the need of heatsinks commonly found in the igbt designs, the Achilles' heel of the FQD12N20LTM is the 200V of breakdown voltage, we don't now if 200V peak at the primary could be sufficient for a good spark, we are using a cb600f engine and original ignition coil.

Because of financial constraints we are following the philosophy of "overthink and simulate everything 3 times"
before the virus outbreak we had the task of maintenance of previous designed systems and creation of new ones, now with the classes and maintenance out of the way we are going full throttle at R&D, we had the schedule of 3 years to accomplish this challenge (most of the team before this initiative knew nothing more than Kirchhoff law and had zero ECAD experience) but now i feel confident that we can do this before 2021, we cant evolve if we keep in our comfort zone.

I think that MAXFem could be used to generate the heat generated profile, then perhaps it could also be used to show how the heat moves around on the PCB. Perhaps CAELinux could do it. Both potential solutions are to big for me to take on these days. But perhaps you have some more time than I have. Or perhaps you can work int into your school studies.

I think i forgot to properly explain, I am the electric and electronic lead of an FSAE team , my division has 5 electronic engineering students, the team has ~30 members, mostly in the mechanical engineering. They are the guys with huge experience in FEA software, right now i am trying to communicate them what problem i need solved (what is cº/W in the datasheets, etc) honestly, i myself would use app notes to eyeball everything heat related but we must justify every aspect of the design to the competition judges and more data = more points.

[AndreyB]

From what I read so far I would worry about software to run this great hardware.

[LucasFlaneto]

I will check CAELinux, as it is free and and seems to be on par with commercial options,i would like to thank you guys for the help, we develop some animosity with the mainstream because everyone we reached in the mainstream communities talked down on us saying that we would never be able to accomplish this feat and only 20+ years experienced engineer should attempt to do those kind of things. Reading this forum i found here to be very welcoming and motivating as self-learning and cooperation seems to be the pillar of this community, i assure that every major update will be shared here with you guys and the idea of forking rusefi software or porting it to our ECU following a public release is generating very healthy conversations among our software engineering partners.

[kb1gtt]

the Achilles' heel of the FQD12N20LTM is the 200V of breakdown voltage, we don't now if 200V peak at the primary could be sufficient for a good spark, we are using a cb600f engine and original ignition coil.

I do a lazy ballpark approximation like this.
0.001 inch gap ~= 1,000V, so 0.035 gap is aprox 35k volts.
Coil is 100:1-ish ratio. so 35kV secondary is aprox 350V primary.

Of course your kick down ratio in your coil will vary, I think it's really more like 114, instead of 100. As well your plug voltage will vary based on things like compression, and some other factors, but for easy to remember numbers, the above lazy approximation should be safe from a design stand point, but not to excessively safe. Also keep in mind if your 0.035 gap opens to 0.045, your primary will be closer to 450V. Many ignition drivers clamp at some voltage. If your driver clamps at say 400V, your clamp will absorb the spark energy and get hot. Typically you'll notice miss fires when under load. If you ignore this, it's common people overheat the driver and cause the clamp to fail. When the clamp fails your primary voltage will jump back up to 45kV. This often causes your coil to break down internally and destroys your coil. When your coil fails, then you get miss fires that you can not ignore any more.

[LucasFlaneto]

Also keep in mind if your 0.035 gap opens to 0.045, your primary will be closer to 450V.

https://pt.slideshare.net/ThiagoHuari/manual-de-servio-cb600-f-hornet-ignicao(portuguese)

few minutes ago i found the maintenance manual of the cb600f (engine donator) and in the ignition part of it there is a section that says that if the primary voltage peak is above 100V (the table in the second page) the coil is working in good condition, i think that higher volume bike engines uses a higher winding ratio than the car ones (maybe because of the dual spark design?).

Still not convinced though, i need more sources to confirm this, as this sparked another doubt, what is the goddamn ratio of this coil!? 1:700 at least for the two 35kV sparks of the approximation.

[Dron_Gus]

After reading the TLE8888 datasheet(specifically page 6 and 7) i found that the AGND and PGND pins are internally connected, how did you guys solved the problem of ground loop?
can i simply ignore it and attach both to the ground plane? did you use the ic as the single point connection of the signal ground(battery ground) and the power ground(chassi)?

Current version of MicroRusEFI has singe ground connection (but through two connector pins to increase max current). This connection is used for both ECU power and GND return for TLE8888 and other high current low side outputs.
We also have two GND "outputs" for sensor grounding. But now we use single plane for all ground. It is not possible to do advanced GND routing on two layers.
And yes. All TLE8888 grounds connected to same ground.

[kb1gtt]

Please define advanced ground routing. I think you are saying adding barriers in your ground plane to create barriers for analog and digital. Or power and signal, etc.

Per the Maxim article noted below, it is highly recommended to avoid ground plan barriers. It is very rare that you need to split the GND planes. I do not like the separated GND plane under the analog stuff. It's my opinion that you should steer your ground current(s) via placement and that creating a barrier is only making the antenna's larger and more effective.
https://rusefi.com/wiki/index.php?title=Manual:Hardware:PCB_design_rules

Each GND pin is capable of handling the full current. The extra GND pins are just for redundancy. They do not increase the current capability of the PCB.

Some day I'd like to see this put on a bed of nails or a flying probe, or similar and measure the voltage gradient across the PCB. We could potentially make a colored image, similar to a thermal camera.

Are there noise issues that you are concerned with?

[Dron_Gus]

Please define advanced ground routing. I think you are saying adding barriers in your ground plane to create barriers for analog and digital. Or power and signal, etc.

Separate ground for
1. ECU power
2. hi current low side output
3. analog inputs
4. shielding
At least all this types of ground signals present on my OEM subaru ECU. Each of them has its own connection point to chassis/engine.

Per the Maxim article noted below, it is highly recommended to avoid ground plan barriers. It is very rare that you need to split the GND planes. I do not like the separated GND plane under the analog stuff. It's my opinion that you should steer your ground current(s) via placement and that creating a barrier is only making the antenna's larger and more effective.
https://rusefi.com/wiki/index.php?title=Manual:Hardware:PCB_design_rules

With current MRE component placement this is not possible. Currently sensor ground connected to gnd plane in a middle of gnd return path of TLE8888. So hi current from TLE8888 will shift all sensors ground.
Instead of short app notes I prefer books like https://en.wikipedia.org/wiki/The_Art_of_Electronics

Each GND pin is capable of handling the full current. The extra GND pins are just for redundancy. They do not increase the current capability of the PCB.

Some day I'd like to see this put on a bed of nails or a flying probe, or similar and measure the voltage gradient across the PCB. We could potentially make a colored image, similar to a thermal camera.

Are there noise issues that you are concerned with?

I'm not. I have my own PCB
I bet we will see noise on adc inputs related to hi current outputs signals. If we someone will do such measurements on MRE.

[kb1gtt]

I recall the calculated lift of the GND plane for like a 10A current was around 0.001V. That amount of lift of the GND pin is negligible from the ADC stand point. It would be interesting to measure the lift and check it against the real world.

That book seems quite old, it was written long before we had the test equipment and simulations we have now. I'm less interested in the length of the article, I'm more concerned with the content of the article. A couple pictures of a simulation can quickly show me what others could write in an entire book.

[Dron_Gus]

I recall the calculated lift of the GND plane for like a 10A current was around 0.001V. That amount of lift of the GND pin is negligible from the ADC stand point. It would be interesting to measure the lift and check it against the real world.

Does this include voltage drop on ground wire from ECU to chassis/battery negative terminal?

That book seems quite old, it was written long before we had the test equipment and simulations we have now. I'm less interested in the length of the article, I'm more concerned with the content of the article. A couple pictures of a simulation can quickly show me what others could write in an entire book.

Physics did not change a lot since this book was published.
Problem with app notes and guides is that it is set of short rules. That you should follow without understanding idea behind them.
For example https://www.autodesk.com/products/eagle/blog/8-pcb-grounding-rules/ . I think Autodesk knows something about PCB routing.
But rule #3 contradicts rule #2. Same for #2 vs #8.

[kb1gtt]

Why would AutoDesk know anything about electrical? Do they offer any electrical PCB design software? I know about the fairly recent routing options for Inventor. AutoDesk has been primarily Architecture and mechanical.

How do these contradict?
#2 – Never slice up your ground layer
#3 – Always provide a common ground point

If you had multiple ground points, that is the same as a splice, just not as obvious as it's a splice that is on on your PCB. You should have one ground point. In this case that GND point is the ECU's plane. The battery, sensors, everything is all connected to this one GND point.

I don't see anything in #8 that says you should have a barrier. It says to keep analog and digital isolated. This is done by knowing your current loops and placing the chips. The currents will flow through the copper in the analog area instead of the copper in the digital area. They also suggest keeping the one GND point in the middle of the analog and digital, so that the current do not over lap. I think their verbiage and picture are not as clear as they could be.

Altium knows about PCB design, they have some good simulation tools.
https://www.altium.com/pdn-analyzer/features

[Dron_Gus]

Why would AutoDesk know anything about electrical? Do they offer any electrical PCB design software? I know about the fairly recent routing options for Inventor. AutoDesk has been primarily Architecture and mechanical.

Yes. They are selling EAGLE. That includes auto-route by the way.

How do these contradict?
#2 – Never slice up your ground layer
#3 – Always provide a common ground point

Because star grounding != fill everything with one ground plane.
Also star grounding assumes separate ground trace from each block to central ground point. Take a look at picture in this article.

If you had multiple ground points, that is the same as a splice, just not as obvious as it's a splice that is on on your PCB. You should have one ground point. In this case that GND point is the ECU's plane. The battery, sensors, everything is all connected to this one GND point.

No. Everything in car connected to ground at the battery. You should think about star routing within whole car, not ECU only. You do not ground headlights to ECU. So why ECU should be grounding point for high current injectors and small signal sensors?

I don't see anything in #8 that says you should have a barrier. It says to keep analog and digital isolated. This is done by knowing your current loops and placing the chips. The currents will flow through the copper in the analog area instead of the copper in the digital area. They also suggest keeping the one GND point in the middle of the analog and digital, so that the current do not over lap. I think their verbiage and picture are not as clear as they could be.

Yes. Keep analog and power grounds isolated and connect each to battery separately. Sorry, but you'll have to have two grounds inside the ECU.