Open source player?

[speachy]

are there any plans for rockbox to support risc-v, if so maybe this would be viable? - https://www.gigadevice.com/product/mcu/risc-v/gd32vw553kmq7

Rockbox has been natively ported to four distinct CPU architectures, and the hosted port can run on anything that is capable of running Linux or SDL.  So there's no inherent reason why a RISC-V port can't be done, other than the willingness of a suitably skilled someone to undertake the effort. 

However, it's not going to be to that particular gigadevice SoC - It only has 320KB of RAM, placing it in the same general class of shovelware SoCs like the RKNano and ATJ212x/215x/216x series, sporting an order of magnitude less RAM than is needed for something that would have the features users expect from something called "rockbox".

While we already support slightly-feature-limited low-res monochrome targets with only 2MB of RAM,  The full feature set (with a color screen) requires at least 4MB of RAM, and ideally over 8MB for the full set of plugins.

[spaceship9876]

I saw in a different post that psram would work with esp32 chips. I have found a couple of risc-v based chips that use psram, would these be suitable?

Bouffalo Lab BL606P: https://en.bouffalolab.com/product/?type=detail&id=16
https://files.sekorm.com/opt/fileStore/cms/nps/editor/20220905/1662390034905031533th.jpg
http://www.semiinsights.com/uploadfile/2021/1217/20211217044459863.png - supports 4-32MB PSRAM

Bouffalo Lab BL702: https://en.bouffalolab.com/product/?type=detail&id=4
BL702 datasheet: https://github.com/bouffalolab/bl_docs/blob/main/BL702_RM/en/BL702_BL704_BL706_RM_en_1.2.pdf

Bouffalo SDK: https://github.com/bouffalolab/bouffalo_sdk

I'm also wondering if rockbox could run on a 16bit microcontroller.

[speachy]

Bouffalo Lab BL606P: https://en.bouffalolab.com/product/?type=detail&id=16

Given the asymmetric nature of the processor cores (one native 64-bit, one native 32-bit) I don't know if we can utilize both simultaneously, but even limited to a single core it should be more than powerful enough for an uncompromised rockbox por.

Bouffalo Lab BL702: https://en.bouffalolab.com/product/?type=detail&id=4

This one is _possible_, but the processor is pretty slow -- we're likely to struggle with many of the codecs.

I'm also wondering if rockbox could run on a 16bit microcontroller.

Every Rockbox port has been to a 32-bit processor, so I'd expect that a port to a 16-bit processor would be quite perilous due to deeply-embedded assumptions like "int==32bits" and the expectation a linear 32-bit address space.  Plus, there would be the inherent performance impact (you'd probably need a dedicated mp3/etc hardware decoder to handle the audio output in realtime).  It's probably _doable_ but... why?  Especially when 32-bit MCUs can be had for pennies.

[saratoga]

Unless your goal is to make your life as difficult as possible, pick a CPU with either a few MB of SRAM or the ability to add an extra dollar or two of DRAM to the board.

[spaceship9876]

it looks like we might have a perfect chip for rockbox if risc-v support is added: https://www.cnx-software.com/2024/01/11/spacetouch-spv60-ai-audio-processor-features-a-400-mhz-andes-d25f-risc-v-core/

[saratoga]

it looks like we might have a perfect chip for rockbox if risc-v support is added: https://www.cnx-software.com/2024/01/11/spacetouch-spv60-ai-audio-processor-features-a-400-mhz-andes-d25f-risc-v-core/

I don't think that SoC makes any sense. It's a way faster CPU then you need, has a bunch of useless DSP/AI hardware and only has a tiny amount of memory without a real memory controller.  You'd be making your life extremely difficult and probably be limited in what you could do with it.

[Trzyzet]

Hi,

I started to think about that kind of project, a Rockbox dedicated, open source player and I found something which could be a very nice start.
https://www.cnx-software.com/2022/12/02/pine64-ox64-sbc-bl808-risc-v-multi-protocol-wisoc-64mb-ram/
Nice CPU, plenty of RAM, I2S support, low power draw and a lot of modern things like bluetooth and wifi.

With some effort it can be a fully featured DAP similar to a Surfans F20 and others with same CPU.
MIPI sounds like a nice, 3.5 inch 640x480 screen, with I2S there can be support for HiRes.

The problem might be a CPU architecture but there should be a linux for this board and for Linux, we can run something like a hosted port.

[speachy]

https://www.cnx-software.com/2022/12/02/pine64-ox64-sbc-bl808-risc-v-multi-protocol-wisoc-64mb-ram/

It's a nice SoC, that's for sure -- especially the embedded DRAM.  Taking the hosted port approach would be the path of least resistance, and there's enough RAM to still have plenty of resources for Rockbox.  But while that Pine64 board might make for the basis of an early prototype on a breadboard, the real hurdle is the physical enclosure. That dictates the design of the PCB and the other major components (screen, battery, etc)

...the tooling to make the case can be pretty expensive (and paid up-front), but you amortize it over your production run, so the more you make, the lower the overall per-unit cost.  But the larger your production run, the more money you need up-front to pay for components.  Economies of scale on component pricing don't really kick in until you're looking at 1000-10000 units.

...If we had some folks that speak one of the common Chinese languages, we might be able to negotiate with an existing OEM to produce us a palette-load of an existing player design, contingent on supplying complete board schematics and all relevant source code.  Or even just purchase a pile of empty cases for us to design our own PCB into. 

I wonder if it's feasible to design a new PCB to go into a classic iPod enclosure.  Actually it would probably be cheaper to buy a pile of those shovelware atj2127-based players, throw the PCBs out, and put a new PCB in there instead...

[Trzyzet]

I've already started some basic PCB design. It will be similar to tangara from this thread - PCB with PMIC, buttons, charge/data USB port, socket for LCD and you just need to solder the ox64. I'm opting for a removable battery like Nokias BL-xx
When I finish the basics I will share kicad project on github. Maybe someone will have some time and will to improve it. Also, there is a few different BL808 based modules, I'm sure there will be something more like ESP32 size.

I was thinking about 3D printed cases. Even PLA will be looking factory-like when you will smooth it with acetone.

The only thing about SoC which hurts me really bad is lack of USB 3.0 support. USB 2.0 is so outdated these days it just hurts.

[Trzyzet]

I will put this as new post, it is quite important.

To make it able to build it by a hobbyist from scratch, as many components as possible must be an off shelf ones.
BL808 used in that risc-v board is a version C which is lacking MIPI support for a display. Only a non-existent D version has them.
Still, no pain because we can use just an SPI display. I have an open question to community, which one we should use?

I'm opting for something ILI9341 based, because they are widely available but they are more expensive than SBC itself, which is kinda weird.
If someone knows some LCD with resolution at least 320x240 and they are cheaper than the ones I mentioned, I will be happy to use it.