It is currently Mon Aug 20, 2018 6:58 pm



Forum locked This topic is locked, you cannot edit posts or make further replies.  [ 15 posts ] 
Author Message
 Post subject: Speeduino...$120 ECU
PostPosted: Mon Oct 19, 2015 9:03 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
Anyone seen this yet?

http://speeduino.com/forum/index.php
https://speeduino.com/shop/index.php


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Mon Oct 19, 2015 10:38 am 
Offline
User avatar

Joined: Sun Aug 09, 2015 12:09 pm
Posts: 9
This is..... Interesting.

I'm honestly surprised that poor 8 bit processor can handle an engine up to 9k rpm. It also looks like they use DigitalWrite in interrupts to turn the injectors/ignition on and off which means the scheduling accuracy will be pretty bad (although at least all of the interrupts are the same so maybe the latency will be constant).

IMHO it is scary to think people would run an engine with it. There's a reason silicon companies spend so much money developing complex hardware timers.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Tue Oct 20, 2015 9:49 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
Yes it is scary.

The rusEFI project is using the OS to handle spark/injector timing too I'm pretty sure...but they have a much faster processor.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Tue Oct 20, 2015 11:05 am 
Offline
User avatar

Joined: Thu Mar 26, 2015 11:55 am
Posts: 112
Location: Massachusetts, USA
The current generation of our series of chips from Freescale, 57xx, have some multiprocessor versions. My plan for that is to skip the eTPU and write code to replace the eTPU which runs on a dedicated processor. Basically one processor will do housekeeping and assign the time spacing between events and the other will actually watch the time and then activate the events.

That might be overkill though. If you have a 100 MHz processor and you want timing accurate to 1/10 millisecond, that's a 10,000 instruction window you have to meet. THat seems pretty doable almost any way you go...


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Tue Oct 20, 2015 11:35 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
Horizenjob wrote:
The current generation of our series of chips from Freescale, 57xx, have some multiprocessor versions. .


Just keep in mind that 57xx is NOT the current gen for ECU use. For ECUs they recommend 56xx because these have an eTPU or 2. Not that there aren't other ways to do things, but dedicated time processing HW is a very good way to do it and most chip makers have chips specifically designed and intended for ECU use for that reason.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Wed Oct 21, 2015 1:11 am 
Offline
User avatar

Joined: Thu Mar 26, 2015 11:55 am
Posts: 112
Location: Massachusetts, USA
I understand, but I wonder about that. What I like when I look at those chips is the "end-to-end ECC". If something goes wrong somewhere between the input pins and the output pins on those processors, you know it. With Toyota paying well over a billion dollars in their lawsuit, I would imagine other people are looking at that too.

How could one have trouble trying to decide when to turn on a transistor with these processors? The eTPU is specialized, but it is also slow and simple. The history of computing for many decades is that the general purpose machines replace the special purpose ones.

I'm happy to work with the 5634 now, but I think the real fun will be with the new multiprocessors. It's much more interesting stuff to think about, it gets us native USB and Ethernet. I also think this work would be what can really distinguish o5e.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Wed Oct 21, 2015 9:46 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
Horizenjob wrote:

I'm happy to work with the 5634 now, but I think the real fun will be with the new multiprocessors. It's much more interesting stuff to think about, it gets us native USB and Ethernet. I also think this work would be what can really distinguish o5e.



I have no objection provided we've had a look at how latency issues are controlled without the dedicated timers the eTPU provides. It could very well be that stuff is so fast now usec accuracy is a given...the enginelab unit I'm playing with programs in 0.1 usec steps.....at 10x better than anyone probably actually needs.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Wed Oct 21, 2015 2:09 pm 
Offline
User avatar

Joined: Sun Aug 09, 2015 12:09 pm
Posts: 9
The 57xx processors have a GTM instead of eTPU. It's provides almost identical functionality, but it's designed by Bosch instead of Freescale.

It's possible to use eMIOS or other hardware timers to do fuel and spark control by scheduling events on a counter bus, and connecting the counter bus to the angle clock. A lot of micros now (one by Renesas that someone showed me comes to mind, and I *think* this is the one EngineLab uses) have hardware angle clocks and grouped timers similar to eMIOS (but with many more channels) which are used. Then, it's a sort of hybrid of eTPU and totally software, using hardware to do the output events and software on the main micro to reschedule them (instead of the eTPU microcore to reschedule events).

The MPC5634M is the low-end processor in the range, it's not designed for anything above 4 cyilnders. The next steps up are the MPC5644A (which is available in a 176 LQFP with the same pinout as MPC5643M), and MPC5674F/MPC5676R (which also share pinouts in some packages). These have a lot more IO, more eTPU's, and more memory (and are faster - 5674F is 260mhz and 5676R is 180mhz dual core).


End-to-end ECC isn't something that's necessary in hardware. An OEM ETC system I worked on implemented end-to-end ECC in software for their safety-critical ETC fault monitoring systems - They did a complete CRC check of the flash region at least once per second and kept two copies of all RAM data (the data, and the data's compliment). Every load or store would check or generate the compliment data.


I think there will always be a place for the eTPU or GTM or other specialized hardware for this. The eTPU isn't just for engine control (it's widely used for AC and brushless motor control and less common communication buses). The software interrupt latency is just too high. eTPU is actually quite fast (the total thread latency worst-case is usually a few microseconds, but the IO events are done in hardware and do not incur this latency).


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Thu Oct 22, 2015 1:44 pm 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
you know....I looked for the eTPU and didn't notice the GTM...althogh it looks like only the brand new 5777 has that
http://www.freescale.com/products/power ... t:MPC5777M

and it designed as a motorcycle ECU
http://www.freescale.com/applications/a ... LSMAENGCTR

but it looks crazy capable and they show a price or $110 so no idea why they are talking about lowend applications that couldn't possibly afford to use this chip????

and eval boards start at $295 :cry:


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Thu Oct 22, 2015 4:47 pm 
Offline
User avatar

Joined: Thu Mar 26, 2015 11:55 am
Posts: 112
Location: Massachusetts, USA
Hi apaird, you know a lot about these chips.

Quote:
End-to-end ECC isn't something that's necessary in hardware.


Maybe "not necessary" and "highly desirable" can both be true? There are things that are not required but you would not avoid using if available. I think the "end to end ECC" is in that category. If something is not available you can't use and make due with other things, I think the essence here though is that if it is available it makes the operation of the CPU much more deterministic.

In your description of the ETC system you mention many important features or checks. Toyota got in trouble because their system ( not developed by Toyota ) failed to do some of these things like maintain shadow copies of variables. That is actually supported in hardware on their chip and they failed to use it.

ECC in the processor applies a much more rigorous standard to it's operation though. This is more than just ECC in memory and ECC in cache, it is including data paths and possibly at least some of the CPU operation. I should go read more details...

What are your concerns with the 5634 for running something like a high performance pushrod V8? My interest is not really pollution control engines, but various track and antique motors. Sigh, in this case antique would be motors in cars before I was 35 years old...

FOr that matter also high performance 4 cyl. I have several people lined up to be test cases and one of them has a 14,000 RPM BMW powered D Sports Racer...


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Thu Oct 22, 2015 6:52 pm 
Offline
User avatar

Joined: Sun Aug 09, 2015 12:09 pm
Posts: 9
About 8 bit micros: There's nothing wrong with doing the math on an 8 bit micro. There is a lot wrong with doing an Arduino DigitalWrite in an ISR to schedule ignition and fuel events. Using the timer to directly control the output pin removes any ISR and DigitlaWrite latency from the output, or you can write to the dig register directly instead of using DigiralWrite (maybe in an asm isr routine to avoid saving all of the context).

I don't think end-to-end ECC is necessary for anything O5E would do. It's not even used for the aviation applications I've worked on (MPC5554, which has only ECC on the memories), although they are fully redundant. It's not a reason to switch micros for me.

Freescale is proposing the 5777 as the upgrade for all of the 5600 series powertrain chips, including the 5676R. So it would do low and high end, although low end would probably stick with the 5634M/5644A for a lot longer. GTM is very similar to eTPU in concept but not derived from it at all.

The only concerns with the 5634 are running out of IO pins, especially on the 144 pin package. A V8 uses 2x of almost everything compared to an I4, so you need 8x spark, 8x fuel, double the diagnostics (if you have a diag pin for 2x drivers, now you have 4x more diag pins between fuel and spark), double the O2 inputs, and sometimes more sensors (e.g. bank1 and bank2 MAP). The micro is very capable, but 144 pins isn't many. The 5634 also has a very few eMIOS channels and many of them aren't fully capable so the eTPU might be loaded up doing generic PWM and stepper motors and such.

If you're designing a PCB for the 176 pin package anyway, the 5644A is a drop-in replacement with more complex IO (24ch eMIOS among others), a cache/better MMU, and more memory. If you're okay with BGA, going to the BGA 5644A or 5674F has a LOT of IO.

Most of the ECU's I work with (for Formula SAE among other things) are much more heavily IO limited than computation limited, IO type/quantity and the sampling/pulse scheduling limitations are always a concern of mine more than the processor core itself.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Fri Oct 23, 2015 12:10 am 
Offline
User avatar

Joined: Thu Mar 26, 2015 11:55 am
Posts: 112
Location: Massachusetts, USA
Quote:
It's not even used for the aviation applications I've worked on


Well it's neither here nor there for the current discussion really. It's on the new chips and unless they move the old designs to the new processes, the old stuff becomes more expensive. So you spend more money on parts and you have to put more effort into software and various certifications become more difficult.

Your comment on the 5644A makes sense. I started with a layout for the 176 pin package because I wanted to be able to upgrade to the 57xx series. The 5634 didn't seem reliably available in small quantities though, so I changed over to the 144 pin package.

I'm using the Freescale support chips for the coils and injection ( MC33810 ) so I'm reading most of the diagnostic stuff over SPI. I haven't run out of pins yet. I want to get a handful of boards built so we can start getting practical use of the o5e stuff. I'll put a little block of testpoint vias for extra signals somewhere though.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Sun Oct 25, 2015 2:23 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:52 am
Posts: 304
Location: Over here, doing 'over here' things.
Off topic much, are we?

_________________
/me goes off to the corner feeling like Jerry Springer with a mullet.

My O5E candidate: 1982 Honda CX500TC motorcycle.


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Thu Jun 23, 2016 12:21 am 
Offline
User avatar

Joined: Fri Aug 16, 2013 12:04 pm
Posts: 57
Location: Jersey City, NJ
mk e wrote:
The rusEFI project is using the OS to handle spark/injector timing too I'm pretty sure...but they have a much faster processor.

No: not using the OS to handle spark/injector.
Yes: using one 1MHz timer to handle spark/injector timing.
Yes: 16MHz processor.

_________________
another open ecu


Top
 Profile  
 
 Post subject: Re: Speeduino...$120 ECU
PostPosted: Thu Jun 23, 2016 8:53 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
russian wrote:
mk e wrote:
The rusEFI project is using the OS to handle spark/injector timing too I'm pretty sure...but they have a much faster processor.

No: not using the OS to handle spark/injector.
Yes: using one 1MHz timer to handle spark/injector timing.
Yes: 16MHz processor.


I thought when we chatted you were running everything through the OS...I must have misunderstood...so you're not using the OS to read and react to the timer?

Maybe I misunderstood the speeduino approach too then.


Top
 Profile  
 
Display posts from previous:  Sort by  
Forum locked This topic is locked, you cannot edit posts or make further replies.  [ 15 posts ] 

Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Theme designed by stylerbb.net © 2008
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
All times are UTC - 5 hours [ DST ]