[sdiy] Electrosmith Eurorack DSP platform
ebrombaugh at gmail.com
Thu Jan 27 01:29:52 CET 2022
I've done one port to RP2040 so far and migrating is definitely a shock
to the system. A few things I've noticed:
* The SDK has a lot of slick features based on cmake that tend to
obscure what's going on under the hood unless you're pretty advanced in
cmake. Digging down into the scripting takes time to absorb all the details.
* The SDK libraries cover a lot of basic stuff you'd want to do, but not
everything that STM32 HAL does. Be prepared to "roll your own".
* Dual core sounds like a very cool thing, and can be useful. Be
prepared to learn a lot of new things about resource sharing and coherency.
* The RP2040 chip has some nice hardware things going on, but they're
rather orthogonal to a lot of the stuff you may be used to on other
architectures. The PIO is fun to mess with, but has some limitations
that can sneak up on you. Don't assume you can route any signal to any
pin. Side pin signals need to be contiguous!
* The RP2040 ADC is a lot noisier than anything on STM32. They claim
12-bits resolution and 10-bit accuracy. Expect less unless you're doing
a lot of integration.
* The XIP flash interface has cache, but you'll find it's pretty
limited. Be prepared to move a lot of critical code into on-chip SRAM.
Dual core code will eat that cache up a lot faster too.
* The 2nd-stage bootloader that RPi provides for the PICO devboard
assumes some unique things about the flash chip it's talking to. If you
don't have a genuine Winbond flash chip you'll need to revert to a
generic 2nd-stage that's got somewhat lower performance, or write your own.
I'm sure there are other issues, but those are the big ones I've hit so far.
On 1/26/22 16:58, Mike Bryant wrote:
> That's exactly what I'm doing at this very moment. Replacing 1 H750 and 4 F730s with 21 RP2040s. Absolute nightmare of a chip - very poor support from RPT compared with STM - but slowly getting there more thanks to people on their forum than RPT staff. And have bought a half reel in advance to make sure I don't get shafted again.
> For anybody else thinking of going this route, be warned that the SDK and the support programs to generate code for the PIOs have been written around gcc features and so it's currently impossible to use a modern compiler like Clang.
> -----Original Message-----
> From: Synth-diy [mailto:synth-diy-bounces at synth-diy.org] On Behalf Of Eric Brombaugh via Synth-diy
> Sent: 26 January 2022 23:18
> To: synth-diy at synth-diy.org
> Subject: Re: [sdiy] Electrosmith Eurorack DSP platform
> Yep. Any STM32 parts more advanced than the low end of the G0/L0 families are basically unobtainium for the foreseeable future. We've had
> F7 parts on back order for the last 9mo with no firm notice of when they'll be available. Unless you're a high volume customer with firm POs at this point you're sucking vacuum or subsisting on inventory you had in Q1/2021.
> Any new MCU-based designs these days are migrating to alternatives.
> RP2040 parts are now available for < $1 in quantity direct from RPi Foundation, and ESP32 variants are cheap and in-stock at all the major distys. Definitely worth considering.
> On 1/26/22 15:53, Mike Bryant wrote:
>> After the AKM fab fire the guy there told me they had several reels of
>> CODECs in stock, so presumably they did the same with STM32s. Wish
>> I’d bought a reel or two now.
>> *From:*Synth-diy [mailto:synth-diy-bounces at synth-diy.org] *On Behalf
>> Of *Peter Pearson via Synth-diy
>> *Sent:* 26 January 2022 22:50
>> *To:* Jason Proctor
>> *Cc:* synth-diy mailing list
>> *Subject:* Re: [sdiy] Electrosmith Eurorack DSP platform
>> I'd like to know where they bought STM32s with the global chip
>> shortage and all that...
>> On Wed, Jan 26, 2022 at 4:40 PM Jason Proctor <jason at redfish.net
>> <mailto:jason at redfish.net>> wrote:
>> Interesting --
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> Selling or trading? Use marketplace at synth-diy.org
More information about the Synth-diy