PI4, CanBus, LCD displays - automotive design - need a programmer/partner

[ame]

nope. There is positively, absolutely, 100-and-5% (I know per-cent limits it to 100 lol) NO WAY I can use anything else, other than 3x 4" screens.

I am open to using 3x raspberries, one for each screen, just how can I get the canbus to all 3 of them tho? And wouldnt I need 3x software?
Like I said ... whoever feels comfortable programming this should tell me what HE/SHE prefers.

Well, canbus is a multidrop bus, so you could have a can transceiver on each Pi. Or you could have one transceiver on the first Pi and have it communicate to the others. Lots of ways to do it.

Yes, you would need 3x software, but you could have one software image, identical on each Pi, with unique configuration files on each one.

There are many ways to solve your problem, which is why constraints are good. The obvious constraint is 3x 4" diameter 720x720 pixel displays (probably with daylight visibility). I'm glad you won't compromise. Constraints are good. So good I said it twice.

Honestly, I think you're onto something there.
I believe, for simplicities sake the way to go would be with 3x PI Zero 2 (those seem to be readily available) and 3x CAN to PI HAT (also available). Get those displays via HDMI, so that should reduce headache a lot and just put the software on each device. I also like the redundancy - say one display WOULD fail, you can just switch over another one so you have crucial speed/rpm (if speed OR rpm fails).

Question is... would YOU (or anybody) else be down to program this? Seems with 3x PI the software side is a lot easier now... we "just" (ignore my arrogance here) need to read in a few pins and canbus data and make some needles move lol

Right! How hard could it be?



No, I'm not interested. I'm interested in kicking around valid ideas. And I'm interested in where the discussion goes, but you probably need someone local to get the thing going. It could be a side gig for a professional who likes tinkering in their spare time (which covers a lot of people on this site). It's also worth discussing what other people have told you outside of the forum in case they are blowing smoke.

There is also a big gap between hacking something together that seems to work, and building a product, although that doesn't seem to put some people off.

[ame]

I also found this, which might not be available in the configuration you want, but the basic idea as I see it is there is a single HDMI port driving a big virtual display. Half of the display appears on one connector, the other half on another, therefore one screen is split into two physical halves:
https://m.aliexpress.com/item/1005003091353975.html

If there was such a board compatible with the displays you have in mind then one Pi 4 HDMI output could drive two displays, and the other can drive the third display by itself.

Just speculating here. I'm not at all invested in your project, but it does sound interesting.

[Matce]

Well, canbus is a multidrop bus, so you could have a can transceiver on each Pi. Or you could have one transceiver on the first Pi and have it communicate to the others. Lots of ways to do it.

Yes, you would need 3x software, but you could have one software image, identical on each Pi, with unique configuration files on each one.

There are many ways to solve your problem, which is why constraints are good. The obvious constraint is 3x 4" diameter 720x720 pixel displays (probably with daylight visibility). I'm glad you won't compromise. Constraints are good. So good I said it twice.

Honestly, I think you're onto something there.
I believe, for simplicities sake the way to go would be with 3x PI Zero 2 (those seem to be readily available) and 3x CAN to PI HAT (also available). Get those displays via HDMI, so that should reduce headache a lot and just put the software on each device. I also like the redundancy - say one display WOULD fail, you can just switch over another one so you have crucial speed/rpm (if speed OR rpm fails).

Question is... would YOU (or anybody) else be down to program this? Seems with 3x PI the software side is a lot easier now... we "just" (ignore my arrogance here) need to read in a few pins and canbus data and make some needles move lol

Right! How hard could it be?



No, I'm not interested. I'm interested in kicking around valid ideas. And I'm interested in where the discussion goes, but you probably need someone local to get the thing going. It could be a side gig for a professional who likes tinkering in their spare time (which covers a lot of people on this site). It's also worth discussing what other people have told you outside of the forum in case they are blowing smoke.

There is also a big gap between hacking something together that seems to work, and building a product, although that doesn't seem to put some people off.

lol, well not 100% the answer I wanted, but I get you.

The hardware-side is fully covered. I got all the pinouts, the proper connectors (not to damage the OEM harness), etc... the housing is 3D scanned and will be printed (OEM is discon't). Fitment for the displays is already confirmed so the product ultimately will be quite nice.
I totally understand that we could do a full custom PCB with everything we need in one shot, but for a prototype/proof of concept RPI is the way to go, I believe. I also think that RPI should hold up fine over long term - the cars this is designed for dont see more than 2000-3000miles a year anyways.

I am talking (pretty extensively) with a programmer and I already sent him an email regarding 3x PI ZERO 2 + 3x CANBUS ... If he gives the green light, I'm going for it. Ultimately the only way to see if it works, is to give it a shot! lol
I like the "simplicity" of just basically programming 3 systems and installing them "parallel" ... also means every display has full bandwidth, each CPU controls 1 "dataset" ... overall seems to be optimal in terms of performance, rather than stuffing it all in 1x RPI4

[Gavinmc42]

That's Richard's YT ch, he is a real car guy.
And it has taken him some time to do it, which put hims way ahead of me
Mine is just hobby, spare time coding, mostly learning to code on the hardware.
https://github.com/Gavinmc42/GPS-cluster

So far the RN2040 CANbed is proving to have more code space and grunt than the Arduino version.
Probably does have enough cpu grunt to take that CAN data and mux uart it to the display cpus, Pi or Pico.
Might even be able to do one small display with it.

Pico's can be over clocked to do HDMI but max 640x480?
Zero would be better for 720x720.

For instant on, microcontrollers beat the Pi's, for complex displays Pi's are better.

Richard's TVR cluster I think is Speeduino based with two older Pi's.
When doing baremetal(Ultibo) even old Pi's are fast enough, Linux just slows things done and is more complex.
The Arduino/Ultibo IDEs run on Windows/Linux/Pi's.
These are not that hard to learn, compared to Linux

If your gauges are already designed then that's a big chunk done.
I use Lazpaint for my graphics tweaking.
https://lazpaint.github.io/
Handy to do all the layers, multiple backgrounds, text etc.

Do a search on the Pi forums for "Instrument cluster" it will find others doing similar stuff.
You will find me "moving" needles back in 2017 with JavaFX.
Took me an afternoon to learn using Netbeans based on the Stopwatch example.
Old tech now and running on Linux so slower to boot.

The "Pro" way of doing it will be something like Tizen(Linux), CV, AI, ML, NN, C++ and huge software teams
All done with expensive or complex software?
https://github.com/teslamotors/linux

Pretty sure just one CANbed would get the data and be able to split it out to each display.
Uarts should be fast enough.

Wonder is there is any open source CAN displays?

[Matce]

That's Richard's YT ch, he is a real car guy.
And it has taken him some time to do it, which put hims way ahead of me
Mine is just hobby, spare time coding, mostly learning to code on the hardware.
https://github.com/Gavinmc42/GPS-cluster

So far the RN2040 CANbed is proving to have more code space and grunt than the Arduino version.
Probably does have enough cpu grunt to take that CAN data and mux uart it to the display cpus, Pi or Pico.
Might even be able to do one small display with it.

Pico's can be over clocked to do HDMI but max 640x480?
Zero would be better for 720x720.

For instant on, microcontrollers beat the Pi's, for complex displays Pi's are better.

Richard's TVR cluster I think is Speeduino based with two older Pi's.
When doing baremetal(Ultibo) even old Pi's are fast enough, Linux just slows things done and is more complex.
The Arduino/Ultibo IDEs run on Windows/Linux/Pi's.
These are not that hard to learn, compared to Linux

If your gauges are already designed then that's a big chunk done.
I use Lazpaint for my graphics tweaking.
https://lazpaint.github.io/
Handy to do all the layers, multiple backgrounds, text etc.

Do a search on the Pi forums for "Instrument cluster" it will find others doing similar stuff.
You will find me "moving" needles back in 2017 with JavaFX.
Took me an afternoon to learn using Netbeans based on the Stopwatch example.
Old tech now and running on Linux so slower to boot.

The "Pro" way of doing it will be something like Tizen(Linux), CV, AI, ML, NN, C++ and huge software teams
All done with expensive or complex software?
https://github.com/teslamotors/linux

Pretty sure just one CANbed would get the data and be able to split it out to each display.
Uarts should be fast enough.

Wonder is there is any open source CAN displays?

I understood about 30-35% of what you said LOL but if you have done similar work and you have some time, please let me know.

After reading more into it, I think 3x PI Zero 2 + 3x CANBUS HAT is the way to go... cost is neglectable and I can fit it in the housing, no problem.

[Gavinmc42]

3x PI ZERO 2 + 3x CANBUS

More than enough grunt to do it, I like the redundancy option.
A bit of Python code could prototype that software quickly.

Turns out there are CAN displays around, Veethree etc
You could turn this into a product that is useful for more than just one brand/model of car.

Graphic themes are fun to do, did a LCARS, StarTrek display all in OpenVG, Pi Zero did not even get warm.
Plugged into the car media player, powered from the USB port and fed back into the HDMI input
Pi Zero2 not really needed unless you want Linux running on them.

For OpenVG graphics code development I use USBboot mode Zero's, saves uSD card wear and tear.
You could use a Pi with CAN hat USB booting three Zero's and/or use Notro's GUD software.
Might even be able to test it with a PC and USB connected Zero's.

So many ways to do what you want.



.

[Matce]

3x PI ZERO 2 + 3x CANBUS

More than enough grunt to do it, I like the redundancy option.
A bit of Python code could prototype that software quickly.

Turns out there are CAN displays around, Veethree etc
You could turn this into a product that is useful for more than just one brand/model of car.

Graphic themes are fun to do, did a LCARS, StarTrek display all in OpenVG, Pi Zero did not even get warm.
Plugged into the car media player, powered from the USB port and fed back into the HDMI input
Pi Zero2 not really needed unless you want Linux running on them.

For OpenVG graphics code development I use USBboot mode Zero's, saves uSD card wear and tear.
You could use a Pi with CAN hat USB booting three Zero's and/or use Notro's GUD software.
Might even be able to test it with a PC and USB connected Zero's.

So many ways to do what you want.

Well, I am waiting for someone that wants to knock it out... like I said, there's money in it for sure.
I don't have full understanding of how difficult it is, but in theory it should not be super-hard at this point.

I do not want this to be "universal" - it needs to be specific for this one model car - there is a BIG NEED/WANT for this.

I developed a CAN logger once and we had that thing up and running in 2 weeks spitting out csv data for a nice graphic interface.

[Gavinmc42]

I understood about 30-35% of what you said LOL

If you understood 100% then you wouldn't need anyone
If you understand 30% then you are probably capable of doing it yourself too;)

The csv data format is good, I used that for my GPS Logger and GPS Simulator.
A CAN Logger and Simulator makes testing easy.

Still a few years till my retirement, been thinking retrofitting EV's might be a sideline by then
I would be interested in getting a 4" round LCD with HDMI board and swapping some code for it
Enough for you to do your own mods from then on.

If you can make each display modular then that is a product you can sell too.
Keeping it open source means others can run their code on them.
But it sounds like just doing your own stuff is going to keep you busy.

The Hyperpixel display/Pi combo I have only has a spare software i2c port, not good for uart etc on GPIO.
That's where the HDMI interface is better.

You have got me interested enough in porting my old code to see how it runs on the 4" square one.
https://github.com/Gavinmc42/Speedo_kph
If you look at that code you can see it is not much to make three dials on one display, two more than you need

I prototype with Steampunk themes, if you have a particular style in mind let me know.

[Matce]

I understood about 30-35% of what you said LOL

If you understood 100% then you wouldn't need anyone
If you understand 30% then you are probably capable of doing it yourself too;)

The csv data format is good, I used that for my GPS Logger and GPS Simulator.
A CAN Logger and Simulator makes testing easy.

Still a few years till my retirement, been thinking retrofitting EV's might be a sideline by then
I would be interested in getting a 4" round LCD with HDMI board and swapping some code for it
Enough for you to do your own mods from then on.

If you can make each display modular then that is a product you can sell too.
Keeping it open source means others can run their code on them.
But it sounds like just doing your own stuff is going to keep you busy.

The Hyperpixel display/Pi combo I have only has a spare software i2c port, not good for uart etc on GPIO.
That's where the HDMI interface is better.

You have got me interested enough in porting my old code to see how it runs on the 4" square one.
https://github.com/Gavinmc42/Speedo_kph
If you look at that code you can see it is not much to make three dials on one display, two more than you need

I prototype with Steampunk themes, if you have a particular style in mind let me know.

We can certainly work something out for displays etc but like I said it would be ongoing if/when ppl want custom faces... I'm also willing to learn since swapping canbus inputs shouldnt be much work then if you don't want extra money lol
For now tho, I would pretty much need someone that can get this handled.... I don't know how to reach out to you (cant find PMs) so I'll just put my email here and delete it once you contact me (please) -
thanks!

[ame]

PMs are no longer possible on this forum. You basically have to publish your contact details and hope (as you have done).

You might like to add which country you are in, for the purpose of attracting forum members who are nearer to you. Or you might prefer not to.

[Gavinmc42]

Copied your email down.

I just ported the speedo code to the Hyperpixel.
Made a single dial full size centered.

Just realized 4" diagonal is smaller than 4" round.
Even with reading glasses and a few inches away I cannot see pixels or jagged lines.
4" round at dash distance should be fine?
Brightness might be an issue in cars.

One thing with fixed resolution displays is the size is known.
My OpenVG code scales to any size display and dial size and that won't be needed.
But it does make tweaking the look easier.

[Matce]

Copied your email down.

I just ported the speedo code to the Hyperpixel.
Made a single dial full size centered.

Just realized 4" diagonal is smaller than 4" round.
Even with reading glasses and a few inches away I cannot see pixels or jagged lines.
4" round at dash distance should be fine?
Brightness might be an issue in cars.

One thing with fixed resolution displays is the size is known.
My OpenVG code scales to any size display and dial size and that won't be needed.
But it does make tweaking the look easier.

The OEM is 4" round .... if you can shoot me an email, I can send you pics and more details
The display they are making will be OEM specification for brightness, etc ... has all been taken into account!
thanks!

[aBUGSworstnightmare]

What is your overall display active area requirement for 'three Circular displays' in a row?
Why not simpley take a candy bar display which comes closest and then mask it off by the hood (which is for sure over/around the dials)? Then there is only one CM4 to drive that on 4lane DSI (there is no DSI on a zero and only 2lane on a standard Pi (which limits your available resolution, but still good for 1280x800 i. e.).

Talking about Corvette dials? There's been a guy here on the Forum doing exactly that.

[ame]

What is your overall display active area requirement for 'three Circular displays' in a row?
Why not simpley take a candy bar display which comes closest and then mask it off by the hood (which is for sure over/around the dials)? Then there is only one CM4 to drive that on 4lane DSI (there is no DSI on a zero and only 2lane on a standard Pi (which limits your available resolution, but still good for 1280x800 i. e.).

Talking about Corvette dials? There's been a guy here on the Forum doing exactly that.

Tell me you didn't read the thread, without telling me you didn't read the thread.

[Gavinmc42]

Most dial clusters I have pics of are one or two main large dials and a few smaller ones on the sides.
Would the Hyperpixel 2.1" round LCD one be too small for the side dials?

Waveshare also have some small HDMI LCDs, 2.8, 3.2, 3.5" Rect but small.
HDMI is good because just about any Pi software will be able to drive them.
Been thinking they might be handy for some CANbus/Robot battery management systems.
CANbus is popping up in non car apps now .

[thagrol]

With all due respect,

So no respect at all then...

you have not even the slightest idea what you're talking about in the "automotive field" ... I work in that field for 21 years now and built more of those cars than you have seen.

And just how was I suposed to know that based on your posts?

This is for a VERY specific car, the output is on the CAN-BUS (from a standalone ECU), not OBD - there is no compatibility issue with "other" cars or "placements' because its for ONE car only.

Again, how was I supposed to know that. Not something you mentioned before now.

With that said - since you are absolutely clueless in that field

You don't know that though you may think that.

, please don't lecture me or draw idiotic conclusions what type of "I have a fun idea" guy I am.

Given that almost complete lack of information on yourself and your proposed project combined with the number of folks I've seen on here over the last ten years who "fun ideas" but no knowledge of how to impliment them or their consequences, what was I supposed to think?

Just to set you on the right direction here. If you WOULD want me to build your car, you're looking at a 75.000USD bill + parts - and people pay that. That the kinda of "fun idea guy" I am.

I have NO idea about software development, I don't lecture you. You have NO idea about the automotive/tuning industry

Like I said above. You don't know that.

- extend the same courtesy, would ya?

Lastly... there is ONE person - WORLDWIDE - that restores those clusters currently. That's me. And I charge 1100USD. So there ya go. And supplies are running out - hence the idea to go digital. and yes .. that is "GUARANTEED DEMAND" because I have customers lined up around the globe waiting for replacements.

Again, you're confusing the generic with the specifc. And assuming I know more about you than you have posted prior to this point.

So AIUI, you have a business that builds $75,000+ custom cars to a niche market. For one base model you're designing a replacement instrument cluster that will sell at an estimated $2500. For the same model, you can refurbish the old cluster for $1100.

Yet you are on here asking for free tech help and support from unpaid volunteers. Go find a local(ish) consultant and pay for their services.

I'm sorry that you found my previous post insulting. Unlike with your reply that was not my intent.

Oh, and don't worry, I won't be bothering you again.

[aBUGSworstnightmare]

What is your overall display active area requirement for 'three Circular displays' in a row?
Why not simpley take a candy bar display which comes closest and then mask it off by the hood (which is for sure over/around the dials)? Then there is only one CM4 to drive that on 4lane DSI (there is no DSI on a zero and only 2lane on a standard Pi (which limits your available resolution, but still good for 1280x800 i. e.).

Talking about Corvette dials? There's been a guy here on the Forum doing exactly that.

Tell me you didn't read the thread, without telling me you didn't read the thread.

Good luck with your project (as you either need to use a bulky HDMI interface board with each display or you need to find out some more info on your display).

As you were not willing /able to consider a different approach later one seems the best for you.
Add a DPI (or a DSI) to HDMI bridge and you'll have the missing interface for connecting your third display too. Might give you some Challenge too.

[ame]

This appeared on Hackaday over the weekend:

https://hackaday.com/2022/07/09/can-per ... /#comments

Interesting, and could be useful.