Has anyone had any success with non-RPi boards like the Radxa Zero 3W or Orange Pi Zero 3?

Hi,

My project is growing arms and legs and am at the stage of looking to build this at scale. It’s working well based upon a Rasberry Pi Zero 2 but as most people may know I’ve hit the wall of being able to source these in bulk (say a run of 100x). While I have tried, I thought I’d found a supplier (rapidonline) that had 600 in stock, after I bought three I checked a couple of weeks later and they had gone.

Anyway I was thinking I might have more luck with alternative boards such as the two listed in the subject. I’m happy doing an amount of work to the library to add support for these if push comes to shove, but wanted to know if anyone had any experience running the hzeller library successfully (in terms of performance) on non-RPi hardware (which understandably it is not designed to do).

Thanks all, apologies for the vague question, and if anyone has an interest in this it’s possible I’ll roll my sleeves up and get a couple of these boards and try to make it work.

Ross

(apologies this is purely for academic\discussion puposes!) :slight_smile:

My project has taken some turns and while it’s fun I had hoped to settle on a single SBC solution rather than supporting everything. I’ll list the options below but trying to create a saleable product based upon the Pi is a really interesting paradigm given the availability! Although there’s a degree of technical reworking given I’ve initially based things on the hzeller library it’s totally doable, but if it’s of interest here were the potential solutions driven annoyingly by availability alone!

  • Support for Pi 3/4/Z2W as standard (no change).
  • Support for Pi 5 using led-rp1-rio option (requires testing).
  • Support for CM4/CM5 (will need to rework driver board as custom carrier with SD card slot).
  • Create RP2040/2350 based coprocessor driver board (like the Interstate 75W) with custom firmware to control DMA loop, receiving frame data over SPI from any capable SBC.
  • Any other options people have gone with? (performance is critical, and the SBC does some heavy API lifting so needs to be SBC based)

The last two options present the best in terms of availability, the chips for the driver boards are in good supply so fabrication isn’t a problem, the former (CM4) is sold in blocks of 250 at most places, the latter opens up any capable SBC with SPI support to run the canvas bit.

The Pi Zero 2W was the best option though as they were so cheap (I used the form factor for my custom PCB’s thinking this was the way!), but understandably they disappear whenever stock materialises.

The best approach which I then settled on was the CM4, this is available in abundance, has support for quite some time, and is rock solid, and works without testing with the hzeller library. Compared to squeezing every last cycle out of the PiZ2 it made sense to have some headroom, plus the carrier board design lets me consolidate some other bits I wanted to include (audio in\out, 5w amp, accelerometer, fuel gauge) but you need to include an SD card slot! :slight_smile:

honestly all I can tell you is I don’t think anyone has hit the walls you’ve reported.

You’re going to be pioneer here and you’ll need to find out your best solution.

Thanks Marc, yes I’m nearly done with my project and will start to fab the carrier board today, will keep my fingers crossed!

Hi @cubinarium , I am curious as to what your project is?
Also, I take it this “carrier board” is your own design?
Judging by the 2 IDC connectors, I take it that it has a matrix driver on it?
Personally, I would love to get my hands on something similar to this - my project has a bunch of extra perhiperals that greatly increase the size of my control box (My project is a wearable t-shirt, similar to Marc’s), and it would be cool to consolodate them onto one board.
At the moment, my control box (Which mounts to my belt) contains a voltage regulator (Regulate down from 4S 21700 battery voltage to 5v) and an ATX power switch (Currently using a Petrockblock) as well as an Electrodragon matrix driver board.
I wonder if you are willing to share, at least enough (eg the CM4 connectors, plus the matrix driver) to serve as a basis for other people’s projects?

Hi Clive,

Absolutely, my plan was to open all of this stuff up once I’d gotten over the Kickstarter hurdle. I’m on the final straight to getting the marketing done, last few renderings and then it’s good to go. I’ll post the link once I’m there (a few days at most).

As the for your project, that sounds amazing, I’ve seen Marc’s stuff and it’s really cool. If I can help in any way just shout - as I mentioned above being dicked about by Fiverr was the best thing to have happened, as now designing circuit boards is my new favourite thing! That CM4 carrier board is probably the most complicated one I’ve done, and been procrastinating over the fabrication until it was perfect, but as of Wednesday I’ve bitten the bullet and did it, so middle of next week I’ll be able to test this and if it’s good then you’ll have a working basis in this PCB - but if you wanted to tweak it I can help.

It’s great fun, but can be frustrating, like those CM4/5 mating connectors go in and out of stock at JLC, the ones I have at the moment give 4mm clearance which is why I was able to fit the register shifters needed for the active driver hzeller made, and I did enjoy manually routing these.

The cool thing about the CM boards is that you get everything, not just the standard I2C, GPIO etc but also the USB stuff so it ends up being really compact. The CM4 runs a bit cooler when at rest compared to the PiZ2 in my experience. Also, and while I haven’t tested it yet, the board is compatible with the CM5 as there’s not PCIe involvement.

I also designed a power board that slotted into this carrier board which handled the power path (PSU+BAT), boost (as I was using 4.2v lipo’s), charge, and a mosfet for a jumper to shut the battery off. While I still need to fiddle with that (the boost board is tricking the charger to always be chargin’) you could do something similar, any extra wiring, switch, and your buck convertor could all fit on it and plug right in.

If the CM is the route you want to go down I can definitely help (and definitely provide the KiCad files, and TBH the original is GPLv2 so I’m obliged to do this anyway :wink:), even designing a low profile curved belt control box to be 3D printed would top this off!

Anyway sorry for the long response, I’m clearly excited about this stuff, I’ll drop the link in a few days to the project and will provide the files\updates when this thing is delivered (the SD card bit scares me, there’s so much that could go wrong with this board!), which had a couple of alterations (picture below, shows the clearance of the mating connectors also, and threw in the picture of the power board that drops into this too). The only additional bits on this carrier are accelerometer, ADC (using that instead of the more expensive lipo fuel gauage), the audio and amplifier chips. Although the accelerometer chip isn’t being fabricated this time around as it’s leadless (which is interesting in itself) which means it needs x-raying, and is more expensive at JLC.

If you’re interested in getting in KiCad there’s three plugins which absolutely blew me away for importing the symbol and footprints, for mapping to JLC components and the autorouting, I can talk about this stuff until the cows come home!

Ross

Thanks Ross, I look forward to it.
FYI, I have a thread for my project here.

I did design a curved battery box, but TBH I found that it did not really help to keep it low profile - I tend to keep the battery pack at the back, and it’s a lot flatter there.

A curved electronics box for the project in it’s current state would probably be very complicated for a curved case as there are so many components in there:

Pi 4 / Active-3 board stack
Petrockblock power switch board
Regulator
4 input buttons on a veroboard strip
1 power button on a veroboard strip
1 panel mount male XT60 socket (Power in)
1 panel mount female XT60 socket (Power out to panels)

Plus associated wiring to connect everything together - I have plugs pretty much on every connection (And these are inline plugs - ie wire soldered where it needs to be each end, then male / female plugs in the middle), because there are bits everywhere in the case, and otherwise it’s a PITA to get at any component I need to.

Which is why the idea of a custom board appeals so much - in theory I could get it down to one board - maybe keep the separate regulator (As these have a habit of wearing out) and button assemblies, but I think it could be compacted down considerably.

Hi Clive,

If you did go down the custom circuit board route, be it for the CM4/5 (carrier) or Pi4/5 (hat), you can carve a section off for power which might simplify a lot of the wiring. What you could do is have a carrier board with the IDC connectors one side, and the power path on the other. The power area would allow you to dump XT30’s for the battery and for the panels, then it would handle the switching using a mosfet so you can have a simple electronic switch. That way you’d only need to house the batteries, this board\hat\carrier, and the cables going to the panels, makes the belt design easier to manage.

Before custom circuit design my projects used to look like the images below, and while it was so much fun (the collection of little AliExpress boards feature a charge, boost, accelerometer, mosfet for switching, ideal diodes for bat\psu power path, fuel gauage and XT30’s) I couldn’t do that at scale (I’m hoping to get a thing on Kickstarter)!

Buy yeah, suffice is to say custom circuit design changed my life :grinning_face:

Ross

1 Like

I am not really an electronics guy, not 100% sure what the mosfet is for?
I am guessing it’s so that when the pi is off, you cut the 5v to the panels?

That is already handled by my powerblock - When I press the button on the powerblock, the LED starts pulsing, and when it detects that the Pi has finished shutting down, it kills the 5v power out (Which the panels are also connected to)

I can very much recommend one. It’s a very nice feature if you are looking to make things consumer grade.

BTW, I found panel mount XT30s: https://www.aliexpress.com/item/1005004917020972.html

I doubt very much that they are genuine Amass tho

They do male and femal, but TBH I would not really class the males as panel mount, they have no through holes. For my use tho, it’s XT60 into the control box, which houses the regulator, then XT30 out to the panels, so I only really need the female panel mount XT30s

Yeah mosfets are just switches really, they have a gate pin which if it gets a little voltage will either send power (can be up to 6A) out of the drain legs. The consumption of fets are super low which is why phones with no discernable physical switch can last for months without charging while ‘off’. Yeah you can see that square IC in the powerblock is a mosfet, they’re super cheap too.

Those XT30’s are interesting, I tried to make room on the board for a couple, and even in the carrier board wanted to have more so power can flow through it (even though for a carrier board that’s a bad idea) but their footprint is a bit bulky and didn’t want it to be bigger. Well I’m in my overdraft now but JLC have said my carrier boards are on the way so maybe by friday I’ll find out if I’ve wasted $100 on five boards :rofl:

Finally got the pre-launch site for my project done, let me know your thoughts! (sorry for the spamming, have dropped this in as a new topic also, hopefully admins don’t mind, it’s certainly RGB matrix related!!)

https://lucentcube.com

Sorry what is the CM4, could you provide a link?

The CM4 carrier turned up and it does have an issue, I guess I should be thankful as the important stuff works (SD card, ACT light, HUB75 ports) but annoying the CM108B audio chip isn’t being enumerated by USB… I’m working through this now but OTG is grounded and the USB D+/- is correctly wired with the same series resistors I’ve used on the stand-alone audio board. I’ll get this fixed, but ultimately it’s a working CM4 carrier active controller board (can just DNP the audio stuff and it’ll work fine for everything else).

Although if it were for just HUB75 connectors I’d tuck the SD card under the CM4, even though it’s shielded the hirose mating connectors I’m using give a massive amount of room (CM4 is 6mm off the carrier board).

Ross

Such a banana, had the HS hirose schematic upside down, luckily the only bit I tap from that side is USB, so it’s back to refabbing after payday!

I have five of these boards if anyone wants them, two have the IDC connectors on them, the others are clean, for a two-channel CM4 carrier this works totally fine.

Ross

Just a quick update after my testing, the active controller carrier board is solid (USB audio and accelerometer), and works with CM4 and CM5, however I hadn’t realised how HOT the CM5 gets, I presume this is the case for the Pi5 also! It just can’t be run without a heat-sink as it climbs to 80’c really fast and then throttles it (I got about 50% CPU with a small heatsink).

Although the cool thing everyone knew is that the hzeller library works perfectly with the RP1 flag without any other changes for the CM5 (Pi5) - it’s RIO not PIO yet, but that’s fine for me for now. Worked so well I’m now worried CM5 might become my baseline!! TBH the only reason I got a CM5 was to A) test it actually works with the carrier, and B) I needed the performance to emulate Amiga demos (confirmed flawless). :rofl:

Please note that @kingdo9 has a branch with both IO support for rPI5.
And I’m glad that in your case you at least the CPU power in the Pi5 instead of using ti because 5 is bigger than 4;)

1 Like