Next-gen and mini Axoloti hardware discussion

[urklang]

Forming a new thread for discussion of next-gen hardware that started here http://community.axoloti.com/t/where-to-buy-a-axoloti-core-out-of-stock/648/155

Updates on my progress will go here.

[urklang]

Concept sketch I drew for a discussion with @weasel79

[urklang]

My original layout concept from a year or so ago, very alpha, heavy revisions in the works:

Notice that back then I was preserving the separate headphone out. I'm thinking that this will be simplified and merged into a single out that can either be line or headphone per discussion in the previous thread.

BTW, the particular STM32F427 package I used can be easily swapped out for an STM32H7. I'm exploring just immediately jumping to the STM32H7 because it's such a bump up in performance without much price difference. See https://www.st.com/en/microcontrollers-microprocessors/stm32h7-series.html

[lokki]

STM32H7 because it's such a bump up in performance without much price difference.

now that i like!!! 20 times

[Gavin]

Axoloti on steroids, I like that..

[Sonoptik]

STM32H7

Good call & totally makes sense.

+1 for more performance

[urklang]

I really appreciate the support everyone. I'm going to allocate some time in the next week or so to finish the revised design and get a prototype run submitted (probably 5 or so devices). From there I'll have to put some time in to make sure that everything works with the existing firmware/patcher stack. My plan is to sell them as a side project over at my main project's site: http://zrna.org. I'll open it up for pre-orders once I'm convinced that everything is working, which will be after I get and debug the initial prototype run. The approach for version 1 will be to just distribute a separate firmware and patcher release that has any necessary modifications we need for the new hardware while still being compatible with the classic core hardware.

Once we've got that separate stack working I'll be able to start looking at enhancements and some of the next-gen patcher ideas I've been thinking about for a while. That's part of the reason I think jumping to the H7 right away makes sense so we have room to grow and don't feel like we need to revise the hardware right away.

[weasel79]

thanks @urklang you killing it. i agree with whats been said in the other thread, most importantly we need to maintain software compatibility for now, and the same level of plug'n'play-ness. i think we can expect users to plug in a pin header-expansion board for eg. the midi outs, other than that USB and line/hp out needs to be immediately available and supported. also the GPIO pinout should remain compatible for now. love the idea of a split hp/line out. regarding 3.5 vs 6.3 mm, i think we should go with 3.5, but as @tele_player suggested, have a same-layout version of the board without the jacks soldered on, so more experienced people can hack their own wired jack format on it if needed.

one thing that just came to mind, would be awesome to have a stronger output amp to get to studio line levels (+4 dBu). possibly balanced output for a later revision, would need more connectors obviously.

i also like the idea of immediately jumping to H7 but i am worried that this would add a lot of extra developement time, and i think we should most urgently (or also...) try to just make something compatible available ASAP. so if the H7 update would add even 1-2 weeks to it, i would do that as a seperate project. to put it in the thread title words, mini first, next-gen second.

[urklang]

What I might do is run two prototypes simultaneously for the first debug run. They'll be exactly the same design except one will be with the STM32F427 and one with the pin compatible H7. That way I can evaluate things and make the call. If it looks like getting the H7 in immediately is too hairy, we can delay. I don't expect it to be that big of a deal, but there might be some gotchas. In the absolutely worst case we could basically underclock the H7 and make it look as much like the 427 as possible if there are timing related bugs, etc. Then I can cut a firmware release that unlocks it fully later once we're stable.

The main issue I have right now is slow turnaround for PCB and assembly. Just to get the initial prototype from my current manufacturer is about 3-4 weeks. I'm going to start looking into some more rapid turnaround solutions. If anyone has experience in that area, let me know. That delay is what is pushing me to just go for the H7 right away. I can verify that everything works with an H7 development board on the side.

EDIT: we can get the assembly turnaround time way down. To less than a week. It's just going to be more expensive. Looking at a couple of alternative fabrication houses in the US (where I am). I'm going to get the prototype ready for quotation and then I might call on you guys to help with funding so we can do fast turnaround.

[pedroflute]

Please do let us know when ready for pre-orders... Thank you!!

[lokki]

i am very happy to do beta testing if you need any help with that

[Tim052]

The company i get most of my boards made with are launching a low cost SMT assembly service meant for qualitys netween 1-30 units with a 24 hour turn around and setup costs starting at $7 you have to use their set part libary but they seem to have most stuff including the STM32 lines. Based on past experience i've found them to be crazy quick at producing black less than 3 days from ordering to my door in Australia and really nice boards. Defs worth investigating once they launch.

https://jlcpcb.com/smt-assembly

[jaffasplaffa]

Sounds like a very nice idea, to make an updated version. Kudos to you guys if you make it happen:

A suggestion:

It would be very freaking nice if both ram and sram is increased significantly. Like a lot. Or even make it possible to change it, with a socket for them, so people can choose the size they want?

Anyway, I just think a big update on the ram and sram is a must, cause they are pretty much maxed out here in every single patch.

So if everything else is increased maybe 2 times, the ram and sram probably has to be quadrupled, at least. Or maybe even 8 times more Some "ram headroom" is a must, hehe

[jaffasplaffa]

one thing that just came to mind, would be awesome to have a stronger output amp to get to studio line levels (+4 dBu). possibly balanced output for a later revision, would need more connectors obviously.

That would be very nice, for both in/out.

i also like the idea of immediately jumping to H7 but i am worried that this would add a lot of extra developement time, and i think we should most urgently (or also...) try to just make something compatible available ASAP. so if the H7 update would add even 1-2 weeks to it, i would do that as a seperate project. to put it in the thread title words, mini first, next-gen second.

The more the merrier. A few more weeks in my book is worth the wait for a more updated device.

[tele_player]

Since the stm32 is a micro controller, not a general purpose CPU, the fastest ram can’t be expanded, and tends to be small.
H7 has more ram than F4, if I’m not mistaken.

[urklang]

This is the 427 package I was using in my original design: https://octopart.com/stm32f427igt6-stmicroelectronics-27553389?r=sp&s=wivp1CD1RfmBPuqo-Qcohg

Here's the H7 in the same package: https://octopart.com/stm32h753iit6-stmicroelectronics-87680940?r=sp&s=m9D4TVreQiO2df1YHqMDHA

Gives you a sense of how close the prices are; look at the distributors table specifically. So with the H7 we'll jump up to 1MB SRAM from 256K.

I'll look at expanding SDRAM. I don't expect that to be a big deal just to get working but it introduces the possibility of subtle compatibility problems if we're trying to keep the entire stack working for new and classic hardware. Again, what we might have to do in the beginning is to sort of artificially limit the expanded hardware so it looks as much as possible like the classic hardware and then unlock it all later when we're convinced that things are stable.

[urklang]

On assembly, I'm looking at working with these guys: https://www.protoexpress.com/quickturn-pcb-assembly/
They offer 5, 7, 10 and 15 day assembly turnaround. I'd like to make the jump to a USA-based fab house if it doesn't end up being super expensive. It can be challenging dealing with the language barrier with some of the Asian fabs even though in my experience they do great work.

[jaffasplaffa]

Gives you a sense of how close the prices are; look at the distributors table specifically. So with the H7 we'll jump up to 1MB SRAM from 256K.

That sounds great!

I'll look at expanding SDRAM.

Cool that is great too. I think it is very important, cause I think the ram is where most patches comes in short, there is simply not ram enough on Axoloti as it is now, to do some really advanced stuff. I have never hit cpu limit but ram I reach limit all the time.

And thanks again for taking this up guys. A more powerful Axoloti has been a dream of mine for a very long time. I think it has been for many. I wish I could help, but my programming skills are too limit to dive into such a project.

[lokki]

Gives you a sense of how close the prices are

actually the H7 is cheaper!

[weasel79]

Again, what we might have to do in the beginning is to sort of artificially limit the expanded hardware so it looks as much as possible like the classic hardware and then unlock it all later when we're convinced that things are stable.

I really like this idea, given it is not too hard to implement. throw in a gig of SDRAM but limit it to 8mb until we have the firmware/compatibility issues worked out. if thats possible with limiting the H7 to F427 standards, awesome!! just so it won't hold back the availability of basic working axoloti hardware.

also yeah if it doesn't mean unrealistically high prices i would also LOVE to have the assembly done in the US or EU, for ecological/moral reasons...