gemellocattivo.com

Which means "Evil Twin". Lets see your projects where you change boring into fun or create the fun from scratch.
It is currently Thu Mar 28, 2024 4:48 pm

All times are UTC - 5 hours [ DST ]




Post new topic Reply to topic  [ 42 posts ]  Go to page Previous  1, 2, 3, 4, 5  Next
Author Message
 Post subject: Re: TMS570 based ecu?
PostPosted: Thu May 18, 2017 9:02 am 
Offline

Joined: Thu Apr 13, 2017 2:10 pm
Posts: 27
We could get good width (to within ~13ns and max of 447ms) - the start would still be +/- a couple of degrees of course.

I'll do some thinking .. I like this platform too much to give it up yet.


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Thu May 18, 2017 9:19 am 
Offline

Joined: Thu Jan 01, 2015 6:47 pm
Posts: 4251
Ok wait.....if you can end a pulse accurately can you use 2 pins? One to start and one to end and do it time based? Don't bother with the angle function at all?


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Thu May 18, 2017 11:37 am 
Offline

Joined: Thu Apr 13, 2017 2:10 pm
Posts: 27
Making precision widths is cake --- we're also not limited to the resolution of the loop period. Special high-resolution counters are used to get sub LRP precision.

For fuel, the start angle is controlled by the angle counter --- once kicked off then width from that start edge is determined using a time counter to achieve the precision pulsewidth scheduled. This can be handled fine as far as I can tell.

Spark is the opposite: start/dwell can certainly be in the +/-2ish deg range .. but it's the hard fire edge that is tied to the angle counter and +/- degree slop isn't going to cut it.

I've got this right, right ?

So how about this; we schedule the spark fire edge slightly early and then we 'ignore the world' for that last couple of degrees and use the core to manually interpolate and fire through brute force. This has the danger of firing early (worst case) where we can't make it in time to do calcs and holdoff the fire event. With some external circuitry we could probably build in safety.

Another option could be to use HET1 for coarse crank decode w/fuel and to schedule spark fire reminders (on the angle clk) back to the core (but NOT do spark) ... on those reminder callbacks (within a 1 or 2 degree window as before), the final spark edge is scheduled on HET2 in the time domain. This would avoid the 'ignore the world' scenario above. Worst case we know that we're within 1 or 2 degrees with the higher chance that we can get better then that ------ HET2 would end up being a simple series of one-shot's that are manually set through the core based upon angle reminders from HET1. Interrupt load would be cake, but not 0. I've always been looking for 0 interrupt solutions.

Maybe to make this even easier, HET1 is dedicated to crank/cam and angle reminder callbacks ... that way we can swap out crank decoder code easily. Dedicate HET2 to one-shot's only (time domain) ... Fuel gets scheduled on a reminder callback and forgotten about. Spark gets scheduled a couple of times: on a start and on a firing window where we do final tweaks on the originally scheduled one-shot. I think channel swaps and configuring get easier in this configuration too.

I wanted to do all of this 'hands off' in hardware, but (worst case) 32 channels scheduled at a 300hz rate with 16 of those generating a second schedule = a 14.4 Khz interrupt rate at 18krpm which is totally do-able. We'd probably have to add on another 1KHz for gap managment on those patterns with multiple gaps.. .. all do-able.

If HWAG ever gets documented, we can re-eval at that point.

I'll sit on this for a little bit and make sure I thought this through correctly.


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Thu May 18, 2017 2:42 pm 
Offline

Joined: Thu Jan 01, 2015 6:47 pm
Posts: 4251
b - crank/cam alone sounds best. Late spark error risk could be tolerated but early spark can not.

I knew you could fix it :)


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Sun May 21, 2017 9:14 pm 
Offline

Joined: Thu Apr 13, 2017 2:10 pm
Posts: 27
Still thinking ..
If I can do this in 64 instruction slots, we'd get ~850ns time resolution which is equivelent to right under .1 deg at 18k (300Hz rev rate).
..
What else is on my mind ?
    The proposed solution exists entirely in the time domain when an angle domain is THE solution.
    Settling again.
    Learning another one-off instruction set (and tool set). This is getting really old, even if it is enjoyable.

FPGA's + verilog/vhdl fixes the above:
    For life.
    Independent of manufacturer and their future plans (no more tpu or atu -> everyone went to Bosch's GTM except TI).
    Skills learned are leverageable in all kinds of areas.
    HARD.
    Surgical solution.
..

The automotive controller world is unfriendly to the diy'er.
Unable to compromise on something so critical, you can already tell which way I lean.


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Mon May 22, 2017 8:14 am 
Offline

Joined: Thu Jan 01, 2015 6:47 pm
Posts: 4251
The pectel ECUs have been FPGA for all time critical stuff for many years. They were still using a 555 I think it was as the main processor last I looked.

The M1 motecs have a FPGA but he assumed it was beeign used for logging and the TPU was doing the engine control...a guy I know sadi his company was contracted to write the TPU code for them so I'm pretty sure they use the TPU.

Way back in the early days I bought a phytec 5554 SOM....its got a FPGA on it.

No OEM used a FPGA that I know of......there must be a reason, but that reason could simple be cost. Or the reason could be something else......I friend told me about a plant control system that was FPGA based that went down for no appearent reason and the company canned it and band FPGA based systems for all their plants world wide. But that is probably much more complex than we'd be talking about so the risk probably doesn't apply.

I see freescale is now NPX. What looks to be the latest MPC57xx is still eTPU based.....have you read when the GTM change is coming?

I don't know that there is any magic about beign angle based vs time based? There is so convience to thinking in angles for some things but in the end the chip is always actually working in time...so really the only question is where is the angle to time conversion done right? A chip with an "angle clock" has a converted hardwired into it vs coding in the conversion.....hardwired is faster I'm sure but the conversion only needs to be done when updates are being sent so I don't know that it really matters does it?


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Mon May 22, 2017 1:26 pm 
Offline

Joined: Thu Apr 13, 2017 2:10 pm
Posts: 27
I still have my phytec som. The fpga was hooked into the mem bus and they were using it for i2c operations (as a simple demo), but you can use it for anything really. Bosch uses an FPGA in their MS6 - they even give you the entire state diagram in their model documentation. But as an OEM, I don't know --- they certainly do a lot of volume --- I'd guess that fpga mfgrs can't meet the volume and/or single sourcing an fpga makes the oem shaky about putting their eggs in one basket. The fpga mfgrs are probably unwilling to sign a volume contract w/an oem who will beat them over the head with it if they don't meet production/longevity ... they get plenty of easy money from open/loosy-goosy military/aerospace contracts.

Actually nxp was bought by qualcomm. This is kinda my point - tech rolls forward so fast that it's impossible for 'one-man bands' like myself to keep up without having the carpet ripped out or being forced to adopt their tools/methods to play their game. Yeah, nothing has happened yet, except for everyone transitioning to GTM. I think qualcomm has a pretty good history of poor documentation and NDA's needed to get anything useful too.

GTM is here .. now ... Renesas(RH850), Infineon(Aurix), Freescale(5777M). TI doesn't technically make a powertrain controller like the other three which may be why they're not bothering to pay Bosch for the IP and just continuing with the HET. It took a while before they even jumped on the ARM bandwagon by buying Luminary. I'm predicting that if they want to compete in powertrain with an ARM solution (surprisingly there are none), then they'll break down and put the GTM on their roadmap.

Automotive powertrain workflows have already converged on simulink/matlab for the firmware side(even Bosch has now switched from ETAS ASCET to simulink) and it now looks like we're approaching the age of commodity IP (GTM) for the hardware side. This workflow is too heavy and inappropriate for a diy'er.

The only commodity thing that I like is ARM's since you can get them everywhere and anywhere.

I just want to take my ball and go home .... go buy an EL ecu and play with models.


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Mon May 22, 2017 2:26 pm 
Offline

Joined: Thu Jan 01, 2015 6:47 pm
Posts: 4251
essess wrote:
I just want to take my ball and go home .... go buy an EL ecu and play with models.


That is not a bad answer. It is a bit pricey but the tools are pretty nice really....you don't have the control you get when you have access to the low level stuff but you do get to just say..Jim, this isn't working right when you find you need access to low level stuff.

You want to call it off then and not the EL world?


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Mon May 22, 2017 7:40 pm 
Offline

Joined: Thu Apr 13, 2017 2:10 pm
Posts: 27
I'd like to see if anyone else has some ideas/inputs about all of this. I'm not too enthused at the moment. This is also the reason why I keep coming back around to attempt again every few months --- I'm attracted to the really hard stuff.

The TRK board is still on the table .. as before, nothing lost .. it's easy to just go back to that. Then I can get fedup with it and keep ping-ponging between the two boards and in another decade, have something done! :D


Top
 Profile Send private message  
 
 Post subject: Re: TMS570 based ecu?
PostPosted: Tue May 23, 2017 8:09 am 
Offline

Joined: Thu Jan 01, 2015 6:47 pm
Posts: 4251
ping-ponging is a sure path to nothing finished.

The TRK is very functional and probably best from a pure engine control standpoint which is why I was so entrenched...but I think your point about tools and useability is spot on so a less good technical path that makes the project more user freindly and lower cost or entry is probably the best which is what sold me on the 570 path.

The HET code is the key though. I would spend no time or though on anything but that. If that can be solved, the 570 is a clear winner. Fuel and spark need to be updated every 10msec...25msec would do. If you can make the timers start/stop thinks accurately and accept updates atevery 10msec....its solved. Nothing more than that is needed other than maybe PWM on the fuel for lowZ injectors but that's optional.

It sounds like you already have a path to solving the crank decoding in the other HET? Maybe test that be sure its good then try spark which is probably the biggest challenge? If you get that working decent....its a project! If not, well it was a good idea.


Top
 Profile Send private message  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 42 posts ]  Go to page Previous  1, 2, 3, 4, 5  Next

All times are UTC - 5 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 4 guests


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

Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group