22mm Manhattan

Hardware inside and outside of the ECU
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

22mm Manhattan

Post by AndreyB »

Here's a problem:
  • Early 90s cars are getting sparse in junk yards and Frankenso's harness pigtail is getting more expensive - as any other component, pigtail availability is a big deal (technically there are options, but these increase costs or drama)
  • Frankenso PCB is packed with stuff which makes changes like moving to a larger microcontroller harder
  • Frankenso using more generic components instead of integrated chips we use more space comparing with OEM ECUs which means we need a relatively large case
  • there is a desire to make plug and play boards for different connectors
  • I will probably add more arguments later
I would like to suggest modular approach again. More specifically, I would suggest a collection of 22mm tall boards (0,866 inches) all mounted on base board with 90 degree soldered pins. Base board would only have connector and main microcontroller.

Attached is a picture of Honda doing something like this back in the 90s.

22mm vertical boards would fit inside our current case and inside some bosch cases.

Update: this is now https://github.com/rusefi/hw_modular
TLE8888 board: https://rusefi.com/forum/viewtopic.php?f=4&t=1496
EBT chip: https://rusefi.com/forum/viewtopic.php?f=4&t=1521
Attachments
honda_manhattan.jpg
honda_manhattan.jpg (344.65 KiB) Viewed 24141 times
2.54mmPitch SingleRowMale90 DegreeRightAnglePinHeader.jpg
2.54mmPitch SingleRowMale90 DegreeRightAnglePinHeader.jpg (16.77 KiB) Viewed 24141 times
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

I think the question here is, should we consider a modular design? If so than what general typologies should we consider for a modular design.

I'm a fan of an optical back haul. There are fairly low cost fiber optics. No polishing or expensive tools to install just cut them with scissors and insert the plastic fiber optic in the connector. I like how this would make the modules galvanic isolated and could shorten the current loops, also commonly called antenna's. You could also address things via digital nodes, AKA run the fiber how ever you want, and the ECU figured out what is what based on node numbers. I like this kind of back-haul as it allows you to change, reuse or add individual modules as you change your engine setup.
https://rusefi.com/forum/viewtopic.php?f=4&t=412&p=4562

A concern with going Manhattan style, is that just like in the big city, you have thermal issues to deal with. As well there would be concern about vibrations shearing the connectors along the row of pins. Having a pin shear and make an intermittent connection can be a real bugger to debug. Much of this vibration concern can go away if you use those spring style connectors like the header found on the Discovery board. However that's not a good connector for higher amps. I would not use that kind of connection for injectors or similar inductive loads.

I think a modular design could be done with an optical back-haul. However that's perhaps a bit more experimental. I could also see a back-haul done with CANbus which has a proven history. I'm a bit concerned OEM's have engineers who can control the design in a very specific way. This control makes CANbus lower cost in production than an optical bus. However we have random people doing some really odd things, so a CANbus back-haul might not work so well. I tend to think we should consider making modules such that you put a module on your injector, which accepts power and GND, then has some kind of a connector that goes to a brain. This connector could either be a CANbus or it could be a optical.

Hmmm, I know EtherCAT is a robust low latency back-haul for industrial applications. As well Ethernet cables are fairly low cost and easy to install.

So I guess a question to consider is should the modular design be inside an ECU relying on an OEM harness? Or should it connect directly to the injectors and similar devices.

I'm tempted to suggest the OEM harness should be removed from the plan. There are always going to be dramas using the OEM connector. However using more standard connectors like an RJ45, you'll have a long life with out the need for redesign. The problem is that RJ45's can't run injectors, so using such a setup would require the module design to move out of the ECU and become more of a system level modular design.
Welcome to the friendlier side of internet crazy :)
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

kb1gtt wrote:
Sun Dec 23, 2018 11:40 am
there would be concern about vibrations shearing the connectors along the row of pins. Having a pin shear and make an intermittent connection can be a real bugger to debug. Much of this vibration concern can go away if you use those spring style connectors like the header found on the Discovery board. However that's not a good connector for higher amps. I would not use that kind of connection for injectors or similar inductive loads.
My ides is to solder using these angle pins. Actually, I've just realized that double row are also available. I know that our current power supply is pretty heavy so I wonder how many soldered angled pins would be needed to reliably mount our power supply? Unless we go to much smaller integrated chips.

We do not like to rely on magic custom chips on the bigger boards due to lack of availability concurs. One of the reasons I suggest making our PCB from pieces is so that we can replace one custom chip with another custom chip easier.
Attachments
2.54mmDoubleRowMaleRight AngleHeader.jpg
2.54mmDoubleRowMaleRight AngleHeader.jpg (127.69 KiB) Viewed 24127 times
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
ZHoob2004
contributor
contributor
Posts: 153
Joined: Sun Apr 03, 2016 7:11 pm
Location: Tucson, AZ

Re: 22mm Manhattan

Post by ZHoob2004 »

kb1gtt wrote:
Sun Dec 23, 2018 11:40 am
Or should it connect directly to the injectors and similar devices.
Am I wrong to feel that such an approach would be a bit of a step backwards in usability? I've always thought one of the great features of rusefi is the ability/expectation to use the existing harness and sensors, as opposed to megasquirt which often suggests that sensors and valves be replaced with GM components and requires either a new harness or modification of the existing one.

At the same time, what about shifting frankenso towards a more MS-like architecture? Essentially, choose an existing connector with expected long-term availability and design around that for the full-featured ECU. This would be equivalent to an AEM infinity or some other universal performance ECU. Include every feature that anyone can think of and it's up to the user to figure out how to connect it. If someone wants to use their OEM harness, they need to source a jumper harness or cut/splice the connector. Otherwise, generic harnesses could be produced and then cut to length. This would be especially good for efi conversions where no harness exists, as well as adding new technologies to existing engines.

As for those wanting the ECU in an OEM case, perhaps a return to Frankenstein is in order. Frankenstein could be mounted in the case either loose with jumpers or as a sort of "shield" to a custom adapter board. Perhaps certain components like the power supply or canbus, could be moved to the adapter board to help shrink the controller.


TL;DR: I propose a compromise modular design where commonly used components are all on one universal board, which is then connected to a case-specific board with additional application-specific electronics and connecters.
ZHoob2004
contributor
contributor
Posts: 153
Joined: Sun Apr 03, 2016 7:11 pm
Location: Tucson, AZ

Re: 22mm Manhattan

Post by ZHoob2004 »

For the universal connector, may I suggest the Ampseal 776164 series? This is the connector used on the microsquirt, and would allow us to use the existing harnesses available for it. Unfortunately, we would likely need to use 2 connectors to get as many connections as we want, and so we would want to use different keying, meaning the microsquirt harness would only provide a shortcut to half of a harness (unless one wanted to buy 2 and replace one of the shells)
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

ZHoob2004 wrote:
Sun Dec 23, 2018 6:34 pm
TL;DR: I propose a compromise modular design where commonly used components are all on one universal board, which is then connected to a case-specific board with additional application-specific electronics and connecters.
ZHoob2004 wrote:
Sun Dec 23, 2018 6:51 pm
For the universal connector, may I suggest the Ampseal 776164 series?
This gets us to https://rusefi.com/forum/viewtopic.php?f=4&t=464 - maybe better move the discussion of a better universal case and universal connector there.
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

kb1gtt wrote:
Sun Dec 23, 2018 11:40 am
I think a modular design could be done with an optical back-haul. However that's perhaps a bit more experimental. I could also see a back-haul done with CANbus which has a proven history. I'm a bit concerned OEM's have engineers who can control the design in a very specific way. This control makes CANbus lower cost in production than an optical bus.
AndreiKA says you want to look into ATA6616C System-in-Package (SiP) product, which is particularly suited for complete LIN-bus node applications. But it uses QFN packages only so not DIY for you here.
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
stefanst
contributor
contributor
Posts: 703
Joined: Wed Feb 17, 2016 12:24 am
Location: USA 08530

Re: 22mm Manhattan

Post by stefanst »

From a VERY general standpoint, the best solution with the highest adoption-rate would be one where we have a custom solution for every car in that car's original ECU case with the stock connector in the stock location. Most rusefi users will want to have an installation in their car as easy as possible. From personal experience I know that I'd rather be sitting at my workbench, routing wires on the jumper board in the rusefi, than trying to route new wires in my car. So, the more we can keep stock inside the car, the happier I would be.

How can we get close to that? Of course, if you have a '89-'00 Miata, you're already 99% there. I guess if you have one of the Toyotas (or Hondas?) that currently donate their ECU cases and pigtails to the rusefi cause, you're also golden. But anything else will require some more adaptation.

If we look at the MS-Pro concept, there is maybe something for us to learn here. It's a board with a full feature set, sufficient to operate 95% or so of cars that are looking to go aftermarket ECU. This module can either be had in a bespoke enclosure with high-quality, even water-proof, connectors for people who don't mind doing some re-wiring on their car. That's usually racers. Or it can be placed on a motherboard that can be custom designed for different cars/ vehicles as a direct-ish OEM replacement. With this design you have the best of all worlds.

I agree with the general idea of a more modular design. It doesn't have to be 100% modular- because, as @ has mentioned, this may create some trouble with cooling and large currents going through connectors. I would also vote for soldering the modules for production, instead of spring-contact connectors.
Maybe we can have a certain number of high-current low-side drivers on the main board for the true high current circuits and all the rest can comfortably fit on modules. All the inputs are low power, high-side drivers are usually low power, etc, do no worries there.

Your standard 0.025" square post header can take about 5A per pin, so we should easily be able to stand up our 5V power supply. Maybe use 2 contacts for each- 12V in, GND and 5V out.

For the I/O module connection to be universal you would need to supply to each module the following (I think):
Common:
- 12V
- 5V
- 3.3V (could be created on the module from 5V when needed, if we want to use less connections)
- Power GND
- Signal GND

Specific
- connection to brain
- connection to jumper rail / main connector

So the power supply module would have to be special, but most I/O module connections could be fairly generic, maybe just offering a different number of I/O. Have modules with 2,4 or 8 I/O. Also have some modules with more power GND connections for injectors / igniters or other high-power low-side shenanigans.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

Hmmm, it seems like there are likely different groups of people. AKA racers probably don't care much about replacing a harness, and going full custom, such that they can simplify things and they know exactly where the wires are and how they are routed. Then there are weekend warriors who want to plug and play. It's fast, simple and who cares if there is an extra weight in the harness.

I'm not sure I see both preferences playing nice. It might mean it would be best to have different hardware platforms. My guess is that we'll have a connector with a board. That board's dimensions will be based on some existing shell. All power devices and all inductive loads will be soldered to this base board. If the lower energy signals do not fit all on that board, then you start adding headers and stacking boards to get additional features, like GPS or accelermeters, or who knows what extras can be added. I think the PNP's are going to be limited to an OEM case, either that or spend some extra money on making a case which fits the existing mounting holes.

As for racers who are concerned with weight, a plastic wire is lighter than a copper wire, and I think optical would prove to be more robust than electrical wires. Hmmm, we are also neglecting motorcycles. They don't have room for large ECU boxes. AKA it's preferable to not have a 6 or 8 cyl ECU in a 2 cyl bike, as the extra space for the extra cyl's is not desirable.

Hmmm, perhaps build the ECU as small as possible, then use an adapter to connect it to a variety of connectors.

I think the big question, if we do a more modular layout, would that effect how the software is developed? If so how would it effect the software?

As for Manhattan style, that depends on who ever is designing a board. They would need to figure out how to cram what ever they want to cram in the box.
Welcome to the friendlier side of internet crazy :)
User avatar
mobyfab
Posts: 139
Joined: Tue Oct 29, 2013 10:09 am
Location: Versailles, France

Re: 22mm Manhattan

Post by mobyfab »



Just split or even stack the boards like some commercial PnP ECUs :)
One generic with the mandatory stuff, one per vehicle type with specific stuff.
Done cleanly this would have no effect on software.

Stuff like GPS/Accelerometers/LCDs should have no place inside an ECU.
Just connect external modules via CAN.

Motorcycle ECUs are very tightly packed, 4 or 6 layers pcbs, and are all potted because you can't fit heatsinks anyway.
But some brands (Honda, Yamaha...) use the same ECUs for their complete range of 4cyl bikes, connectors/pinout included.
ZHoob2004
contributor
contributor
Posts: 153
Joined: Sun Apr 03, 2016 7:11 pm
Location: Tucson, AZ

Re: 22mm Manhattan

Post by ZHoob2004 »

mobyfab wrote:
Mon Dec 24, 2018 2:22 am
Just split or even stack the boards like some commercial PnP ECUs :)
One generic with the mandatory stuff, one per vehicle type with specific stuff.
Done cleanly this would have no effect on software.
This is essentially my idea but said a lot more succinctly. And as a way to avoid a significant divide between PNP and racers, the "racer" setup can be the same mainboard but with a more full-featured (or even optical) other half.
mck1117
running engine in first post
running engine in first post
Posts: 1493
Joined: Mon Jan 30, 2017 2:05 am
Location: Seattle-ish

Re: 22mm Manhattan

Post by mck1117 »

What is the ACTUAL use case for having on-board modularity? Once the main board has a standard set of functionality to run a fuel injected, spark ignition engine (not DI, not diesel), I don't see a need for additional hardware outside of us (developers) prototyping new features.
kb1gtt wrote:
Sun Dec 23, 2018 11:40 am
I could also see a back-haul done with CANbus which has a proven history. I'm a bit concerned OEM's have engineers who can control the design in a very specific way. This control makes CANbus lower cost in production than an optical bus. However we have random people doing some really odd things, so a CANbus back-haul might not work so well.
I think a CAN bus would work very well. Every hardware CAN implementation I see us using is a proven hardware implementation, so in reality it's pretty hard to "break" the bus in a way that causes a failure. It's lower cost to us too, because CAN transceivers and stacks are mass produced and common, unlike a custom optical solution.
kb1gtt wrote:
Sun Dec 23, 2018 11:40 am
I'm tempted to suggest the OEM harness should be removed from the plan. There are always going to be dramas using the OEM connector. However using more standard connectors like an RJ45, you'll have a long life with out the need for redesign. The problem is that RJ45's can't run injectors, so using such a setup would require the module design to move out of the ECU and become more of a system level modular design.
Half agree. I do think moving away from OEM connectors is a wise choice. I don't think things like injectors or ignition should ever move off of the main board. They require synchronous operation with the crankshaft, and any inexpensive bussed solution (optical or electronic) will make this infeasible.

Functionality that requires crank-synchrony should be on the main board, but anything else could get moved off. That list is pretty short: I think analog inputs (at least for MAP/MAF) injectors, ignition, knock sense, and CAN is really all the main board needs.

Want an accelerometer/gps/LCD/knob/extra sensor? CAN bus. A bunch of those also don't make sense to have physically near the ECU. GPS goes on the roof, accelerometer bolted to the floor in the middle of the car, and LCD near the driver.
mck1117
running engine in first post
running engine in first post
Posts: 1493
Joined: Mon Jan 30, 2017 2:05 am
Location: Seattle-ish

Re: 22mm Manhattan

Post by mck1117 »

kb1gtt wrote:
Mon Dec 24, 2018 12:25 am
I think the big question, if we do a more modular layout, would that effect how the software is developed? If so how would it effect the software?
I've been doing some prototyping on the side, abstracting out the concept of a sensor to just something that can provide a floating point number that might mean something. Where it came from could be swapped out at runtime, so if you're using an internal analog O2 sensor it could use that, or be a "synthetic sensor" that came over CAN bus or something.
User avatar
kneelo
Posts: 39
Joined: Sun Jun 10, 2018 6:36 am
Location: Western Australia

Re: 22mm Manhattan

Post by kneelo »

mck1117 wrote:
Tue Dec 25, 2018 1:18 am
What is the ACTUAL use case for having on-board modularity? Once the main board has a standard set of functionality to run a fuel injected, spark ignition engine (not DI, not diesel), I don't see a need for additional hardware outside of us (developers) prototyping new features.
Agree with this... need to think about developers independently to consumers.

My assessment is the project is developed enough to start producing a number of different production type hardware configurations with a fixed feature set (Prometheus has been the first of these to appear). I think haltech is a good example to look at in terms of their model lineup ecology if we take that direction.

I have been working on a schematic for a full featured 8 injector / 8 ignition board for a little while and should be ready to post sonething in the next couple of days.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

Cool 8 cyl with included ignition should be very interesting to the V8 crowd. When you do this, could I suggest trying to make it 70's V8 friendly? AKA some kind of screen like the LCD screen, such that the old folks who don't know what computers are, can program it with out a PC. The PC turns off many folks, who prefer the carburetor engines. I suspect that programming with out a PC would be something of interest to that crowd.

Let me know if you would like a review of your schematic.
Welcome to the friendlier side of internet crazy :)
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

kneelo wrote:
Thu Dec 27, 2018 10:57 pm
I have been working on a schematic for a full featured 8 injector / 8 ignition board
Isn't Frankenso a full featured 8 injector / 8 ignition board? :)
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

mck1117 wrote:
Tue Dec 25, 2018 1:18 am
What is the ACTUAL use case for having on-board modularity?
Being able to easily produce PnP boards for original cases.

While one universal mega-board with a bunch of connectors is A solution for easy PnP, just a mega board would be quite big. I am looking for more purpose-builds boards to be smaller to fit inside original ECUs.
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
User avatar
mobyfab
Posts: 139
Joined: Tue Oct 29, 2013 10:09 am
Location: Versailles, France

Re: 22mm Manhattan

Post by mobyfab »

I disagree.

A "mega board" has almost no uses besides development or purely universal, and is certainly not suited for PnP.
What would be the point in having 8 coil and injector drivers with full circuits for a 4 cylinder engine with distributor? Not even talking about the lcd/gadget stuff or cost.
Even if you don't populate everything you waste a lot of space, will leave floating I/Os and complicate board routing for no reason.
You won't be able to use OEM connectors without a ton of wires or even at all.

You need a base board with the *generic* stuff you *absolutely* need in all cases, then a vehicle specific I/O board that contains the OEM connector and required peripherals/circuits.
Basically: power supply, mcu, analog filtering, maybe some buffers and transceivers. You export all the I/O to the edge of the board, even if unused.
Some engines will have direct coils that require IGBTs, some logic coils that require a tiny BJT, etc. You get the idea.

You want a universal ECU? Just make an I/O board that uses most/all the features of the base board.
It's easier to upgrade or to iterate for development since you can make I/O boards for testing.
People can make their own I/O boards based on your specifications, without soldering complex parts. (if they bought a soldered base board)
Want to port to a new MCU? Make a new base board, which could use sockets for nucleo/discovery.
You can even make an I/O board for HW simulation.

See my picture above.
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

mobyfab wrote:
Fri Dec 28, 2018 12:41 am
You need a base board with the *generic* stuff...
Basically: power supply, mcu, analog filtering
which power supply? we have a not cheap and lots of components power supply on the Frankenso, or is it some TI chip, or should it be an Infenion chip? Chips come and go. Suddenly it would be nice for power supply to be a module.

Which MCU? stm32f4 is not latest and greatest any more, stm32f7 is the new kind and we should start moving towards stm32f4. This rusEfi project here is now five years old, even MCU got old :(

So far 12 channels of analog inputs with three four-channel op-amps is the only for sure part :) But technically most users can live with 8 analog channels while MCU technically supports 16 channels, so looks like an 8 channel module which you either use once or twice :)

Just to pour gas into this fire: do we consider 0603 DIY grade? :)
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
puff
contributor
contributor
Posts: 2961
Joined: Mon Nov 11, 2013 11:28 am
Location: Moskau

Re: 22mm Manhattan

Post by puff »

let me in with my few 5 paise :D
1. no, we don't consider 0603 upgrade, unless the board is manufactured automatically. BUT you have all your sales stats. if its not a diy kit, but a fully assembled unit, then it could be even smaller than 0603 ;-)
2. it seems, power supply chips don't come and go that fast.
3. they seem to have similar pinout and require similar additional parts (capacitors, inductor). btw, isn't stm32f7 pin-compatible o stm32f4?
4. modular design adds much more possible fault points (I've seen plenty of boards with damaged soldering of through the holes components that are under the strain)
5. all in all this manhattan idea seems to add complexity in terms of development, rather than benefis.
6. speaking of jared's idea about lcd - don't think it makes any sense: we are getting older, and there aren't that much people lef who are afraid of computers. moreover, I don't imagine how to configure rusefi with all its tables using a 4-line lcd and a joystick. I'd say, since we don't need LCD for initial debugging purposes (ability to see error code in case of hard fault), we could save a bunch of pins if we move it (lcd+joystick) to a separate devices operated via can.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

russian wrote:
Thu Dec 27, 2018 11:32 pm
Isn't Frankenso a full featured 8 injector / 8 ignition board? :)
He noted he was including the igniter, where Frankenso includes an external igniter driver. For engines before about the 90's and after the about 70's, an integrated igniter is handy. However most modern car's use pencil coils and do not need an igniter in the ECU.
Welcome to the friendlier side of internet crazy :)
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

puff wrote:
Fri Dec 28, 2018 9:18 am
6. speaking of jared's idea about lcd - don't think it makes any sense:
I once was helping a hill climb event known as "climb to the clouds". I was one of the check points fairly close to the finish line. One of the competitor's cars which passed my check point, had the engine die, and he ended up coasting in reverse and parked in my check point. He was given 2 tries to make this climb, he failed both times. He forgot his back up programming cable, and his primary cable got damaged. His entire problem was that his ECU would protect the engine and shut down if the engine got to something like 230F. He was really close to the finish line and did not care if he sprayed some water as he crossed the line. He didn't care if he destroyed the engine. He wanted to disable that temperature limit. It also didn't help his mental state of this google software developer, when he rolled into my check point, where I noted Frankenso has an on-board LCD screen which can allow track day adjustments without the need for PC. He was already upset being a programmer by trade and being hosed because he couldn't fix the program. I can see people with a need to adjust things like a fuel injector's flow, alarm set points, etc. Programming the fuel tables is not likely, but there are many individual pieces of data which could be handy to adjust on the fly. After all, an LCD screen is common on your dash for normal use. It would be handy if that dash could adjust the ECU's parameters. In this case, I pushed for that LCD to be on the ECU. Some day I envision it being part of the dash display.
Welcome to the friendlier side of internet crazy :)
User avatar
mobyfab
Posts: 139
Joined: Tue Oct 29, 2013 10:09 am
Location: Versailles, France

Re: 22mm Manhattan

Post by mobyfab »

russian wrote:
Fri Dec 28, 2018 3:31 am
mobyfab wrote:
Fri Dec 28, 2018 12:41 am
You need a base board with the *generic* stuff...
Basically: power supply, mcu, analog filtering
which power supply? we have a not cheap and lots of components power supply on the Frankenso, or is it some TI chip, or should it be an Infenion chip? Chips come and go. Suddenly it would be nice for power supply to be a module.

Which MCU? stm32f4 is not latest and greatest any more, stm32f7 is the new kind and we should start moving towards stm32f4. This rusEfi project here is now five years old, even MCU got old :(

So far 12 channels of analog inputs with three four-channel op-amps is the only for sure part :) But technically most users can live with 8 analog channels while MCU technically supports 16 channels, so looks like an 8 channel module which you either use once or twice :)

Just to pour gas into this fire: do we consider 0603 DIY grade? :)
Any generic 5v LDO that works in automotive environment such as an LM2931 which has been there for 18+ years. (on the I/O board)
For the few cars that need 5v for gauges, an SMPS like the one I showed last time (TPS54340)
On the base board itself, separate low noise 5v->3.3v LDOs for digital and analog close to where they are needed.

https://www.st.com/content/st_com/en/support/resources/product-longevity.html
STM32F4 will be there for a long time (2028+), then there is F7 which is pin compatible, then another one will be launched, etc. You'll just need to put a new chip on the base board or make a new one if it's not pin compatible. Plenty of time to plan ahead.
There are many MCUs that can replace STM32s. That's the beauty of choosing something that's made for general purpose.

0603 is fine for hand soldering in my case, it's up to you.

As for the LCD, sure it can be useful for some very specific scenarios like kb1gtt described, but an ECU is not a dashboard, it has absolutely no reason being in there.

Stacked boards on AEM's infinity: (which is actually built by another company)
mck1117
running engine in first post
running engine in first post
Posts: 1493
Joined: Mon Jan 30, 2017 2:05 am
Location: Seattle-ish

Re: 22mm Manhattan

Post by mck1117 »

We're getting ahead of ourselves-let's not conflate the two problems of deciding upon design goals with developing solutions to those goals.

Since the stuff being discussed in this thread has lots of overlap with the "let's choose an enclosure" thread, I've started a new thread for determining what we actually want out of the next hardware iteration.

It's located here:
https://rusefi.com/forum/viewtopic.php?f=4&t=1463
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

mobyfab wrote:
Fri Dec 28, 2018 10:55 am
As for the LCD, sure it can be useful for some very specific scenarios like kb1gtt described, but an ECU is not a dashboard, it has absolutely no reason being in there.
I see a blinking LED for basic diagnostic as really handy. Heart beat is handy, as it lets you know the MCU is pumping bits. Then of course the heart beat with a magic decoder is also really handy as it allows some level of knowledge about what bits are being pumped. However the decoder is a pain, and you are very limited in the data you can get out of a blinking LED. Granted LED's are cheap, so why not toss in one to get this diagnostic feature and when you have no other options, deal with the pain's of the decoder ring. However the LCD's are less than $5, so why not remove the decoder, and get lots of diagnostics data easily and quickly. Well, if you have an LCD, why not allow adjustments to the parameters.

When there is almost no PCB real-estate costs and almost no monetary costs, it's easy to put on an LCD. As well if you're a racer and weight is a concern, you can remove the LCD. You could also remove the weight of the LED. However you're more likely to save more weight by going on a diet, or you are more likely to get a competitive advantage by dimpling your car like a golf ball. I see more reasons why you should have an LCD then to not have it. For those that don't want it, they can remove it. Any how, that's why it's there on Frankenso.

I see the LCD as a handy diagnostic tool, which can also be handy if you're in a pinch. I see it as negligible costs, so I added it to Frankenso. I also added EGT's but those don't seem to be as valuable as I had thought they would be. EGT's are really handy, but they don't seem to get used. If I were to spin a board from scratch again, I might put things like LCD and EGT's on some kind of remote wire. Those didn't get the popularity that I thought they would get. Or who knows, perhaps those are handy features and they simply don't get advertised. People don't talk about spoons very much, but if they were removed from your life, you would likely be un-happy about it. Perhaps we simply take them for granted, perhaps they aren't really that important. As for if they should be on a new spin, that's up to the developer to decide. What ever motivates them is what will put features on that spin. I put them in the design, as I didn't want to pigeon hole the design, and desire features which weren't there just because they weren't being though of in the early stages.
Welcome to the friendlier side of internet crazy :)
User avatar
mobyfab
Posts: 139
Joined: Tue Oct 29, 2013 10:09 am
Location: Versailles, France

Re: 22mm Manhattan

Post by mobyfab »

Errors should be handled in the tuning software on a laptop.
A couple LEDs for a basic functions are fine, but not an LCD.

An heartbeat LED and an error LED would be sufficient.
There's an error? Plug your laptop to troubleshoot it.

There's no PCB real-estate issue for now, but there will be at some point.
There's also the STM32 I/O pins that you use, and most importantly the code to handle the LCD, font etc which has absolutely no place into an ECU.

If you really want an LCD, you could make a separate module that reads data from a serial/can/spi port, the LCD logic being handled by the module itself. You can even add a TFT lcd with a keypad that you would place somewhere on your car or keep in your toolbox.

I think all in all this is part of a bigger problem: we're trying to do too many things instead of doing them well. (kiss: keep it simple stupid)
It's fine to add features, just don't put everything into the ECU.
Last edited by mobyfab on Sat Dec 29, 2018 7:46 pm, edited 1 time in total.
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

I believe that breaking ECU into multiple smaller PCBs which need to be soldered together could help multiple people contribute. At the moment we have half a dozen people developing or trying to develop or claiming to be trying to develop their own complete PnP boards.

I hope that splitting the effort into way smaller units shareable between different people increases the changes to have at least partial usable results.

Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
puff
contributor
contributor
Posts: 2961
Joined: Mon Nov 11, 2013 11:28 am
Location: Moskau

Re: 22mm Manhattan

Post by puff »

why hasn't that guy on climb to the clouds event just disconnected the temperature sensor or even connected a resistor instead? :D
I agree completely with mobyfab in everything.
I like the idea of moving lcd screen (perhaps with a joystick) to can - this would allow to place it almost anywhere.

Noone knows what makes people develop their own boards - some just like learn something new. others haven't found the already existing soluions. someone else just have some peculiar requirements. anyway, it doesn't mean they would start contributing.
User avatar
kb1gtt
contributor
contributor
Posts: 3758
Joined: Tue Sep 10, 2013 1:42 am
Location: ME of USA

Re: 22mm Manhattan

Post by kb1gtt »

He was running hot, when he disconnects his sensor, it believes it's cold, and runs rich. AKA he wouldn't make it to the top as he'll boil off all his water. During the event he didn't have time to add a resistor. It was back to back runs. Also if he wasn't using a re-flashed EVO ECU, it would have been much easier to deal with the program. Hind sight is always 20/20. When I saw him he was also guessing it was a temperature limit. All he knew what the if he sat fore a while, it would start again and go. He was guessing it was cooling down and allowing him to start again. If he had some diagnostics, he could have had a better guess. Of course some kind of small wireless feature to the skynet would have been a handy option. Lots of things could have gone better for that fellow. At the end of the day, it was a low budget effort and his team appeared to just be him. A bigger budget or better team would have likely helped him allot. I think it was just him, and just his wallet.

What about some kind of purchasable industrial stack? Would that perhaps help or is space to much of a concern?
https://en.wikipedia.org/wiki/PC/104

I recall Freescale has some kind of stack, but I think it's more intended for robots.
Welcome to the friendlier side of internet crazy :)
User avatar
AndreyB
Site Admin
Posts: 14292
Joined: Wed Aug 28, 2013 1:28 am
Location: Jersey City
Github Username: rusefillc
Slack: Andrey B

Re: 22mm Manhattan

Post by AndreyB »

Is it time to agree that we probably would not change each other's views on LCD and to stop discussing the LCD? :)
Very limited telepathic abilities - please post logs & tunes where appropriate - http://rusefi.com/s/questions

Always looking for C/C++/Java/PHP developers! Please help us see https://rusefi.com/s/howtocontribute
Post Reply