• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Introducing DSPi | A powerful, user friendly and open source DSP for less than a cup of coffee

Attachments

  • Screenshot_20260509_160145_AliExpress.jpg
    Screenshot_20260509_160145_AliExpress.jpg
    136.3 KB · Views: 99
Talking about that, I see Ali is flooded with the 124 pin version (STM32H723ZGT6, Z for 124 pins instead of V for 100 pins that @Weeb Labs linked) that is also cheaper. Any stm32 nerd knows if it can be easily made work with whatever comes out of this project?

Example:
 
Last edited:
An STM32H7 module is around $30. The chip is about the price of ADAU DSP chips.
I think that price is really outlier


Talking about that, I see Ali is flooded with the 124 pin version (STM32H723ZGT6, Z for 124 pins instead of V for 100 pins that @Weeb Labs linked) that is also cheaper. Any stm32 nerd knows if it can be easily made work with whatever comes out of this project?

Example:
They should be compatible
 

Attachments

  • IMG_2202.jpg
    IMG_2202.jpg
    58.9 KB · Views: 61
  • IMG_2203.jpg
    IMG_2203.jpg
    63.5 KB · Views: 64
A quick update on the STM32H723 port. This board offers approximately five times the DSP headroom of the RP2350. With 40 PEQ filters, loudness compensation, crossfeed and volume levelling all active, CPU utilisation is just 10%. The STM is quite a monster of an MCU!
Wow, that sounds really interesting! Could you tell us more about the other capabilities of this chip? What are its additional advantages compared to the RP2350?
 
Talking about that, I see Ali is flooded with the 124 pin version (STM32H723ZGT6, Z for 124 pins instead of V for 100 pins that @Weeb Labs linked) that is also cheaper. Any stm32 nerd knows if it can be easily made work with whatever comes out of this project?

Example:
The Z and V variants are fully code compatible. Most of the others (H725/730/743/753/747/750/7A3 etc) are also compatible with minimal accommodations. About 70% of the codebase is portable by design, so it can support many MCU platforms with relatively little work.

Wow, that sounds really interesting! Could you tell us more about the other capabilities of this chip? What are its additional advantages compared to the RP2350?
Some of the advantages are quite technical in nature. For instance, the STM has hardware peripherals that are specifically intended for use with I2S and SPDIF, which drastically simplifies clocking and master/slave relationships. It also has integrated networking/MAC, hardware PTP, hardware accelerated graphics output and double precision float.

The STM is actually capable of supporting AES67, which the RP2350 simply can not do.

Wow! Eager to try the next beta with LG Sound Sync support!
Fantastic job @Weeb Labs!
If you're interested in testing, I have attached a build with Sound Sync enabled by default. I don't have access to my LG C7, so have no way of verifying functionality at the moment. All that you should need to do is enable it on your TV and switch to SPDIF input. :)
 

Attachments

If you're interested in testing, I have attached a build with Sound Sync enabled by default. I don't have access to my LG C7, so have no way of verifying functionality at the moment. All that you should need to do is enable it on your TV and switch to SPDIF input. :)
Do you have one for the 2040 too? Only have one of those for now.
 
Will it be possible to connect the PCM1808 system to provide an analog input as an option, which communicates with the RPi via I2S?
 
Will it be possible to connect the PCM1808 system to provide an analog input as an option, which communicates with the RPi via I2S?
Yes. I2S input will be available in the near future. I have a branch with a working implementation already!
 
I did an experiment today. I've been building a slightly upscale PCM5122 board to see if I can improve the rather lackluster Adafruit board. I wanted to test my first
spin with a balanced output by having DSPi just invert the right channel to see what synchronization was like and how it came out. Here's a spectrum showing the balanced vs others and then some charts comparing just Adafruit vs MZ (me).

My first spin-> (there's another one coming with labels, without connectors, and some more improvements) and that splash of solder is me trying stuff out : )
1778432425736.png


The balanced is great so I didn't bother with the two sweeps for balanced since it does just as well as on the Spectrum test and this reduces chart complexity. My current spin is a bit high 2nd order harmonic (unbalanced) but noticeably better on the sweeps where the fruit struggles.

1778432528984.png
1778432550536.png
1778432572918.png
1778433126543.png


No guarantees that these are accurate. It's tough to measure the DSPi because I do it through the windows sound system which adds some inaccuracy and randomness. I also messed up the intermod data tables - they're both for different runs of the Adafruit PCM5122 board - but the chart is correct.
 
Last edited:
I received an STM32H723 a few days ago (thanks to project supporters) and have pushed an initial port. This board offers many advantages over the RP2350 and is only slightly more expensive on AliExpress. Among those advantages are a faster CPU, much higher resolution feedback, hardware SPDIF RX and a double precision FPU.

To be clear, this port is a side project. While it would be nice to fully support the STM32, the Pico is my priority. :)
Very interesting!
3 month ago (before DSPi had surfaced) a got a similar board from WeAct Studio, with audio in mind:
I picked this one because it is the fastest (600 MHz) and has an extra USB HS port (thinking UAC2). The chip is STM32H7R3Z8J6.

STM has a deal with DSP Concepts, allowing to use their Audio Weaver framework and UI for free. That board seemed to be a cheap experimental platform to bring it up. For mainly lack of time and also STM exposure, I got as far as having PlatformIO as an IDE and blinking a LED, before it went to the drawer.. :facepalm:

Up to now, I am tinkering with Analog Devices Sigma DSP chips and Sigma Studio, but that's barealy maintained any more, even ST switched to Audio Weaver for their later Sharc DSPs.
(Still, I prefer a "real" DSP over an MCU, since it processes sample-by-sample with minimum latency, has specific hardware like audio grade PLL and ASRCs.)

So by now, there may even be 2 options to use the board, great!
 
Last edited:
Just a little update. LG Sound Sync is now confirmed to be working! :D

image.png


Through my testing, I learned that older LG models such as the B7 make use of a slightly different control bit configuration. Support for this has been added via protocol auto-detection.
 
Just a little update. LG Sound Sync is now confirmed to be working! :D

View attachment 531490

Through my testing, I learned that older LG models such as the B7 make use of a slightly different control bit configuration. Support for this has been added via protocol auto-detection.
I'll test it as soon as family life lets me breath :D
Thanks for this feature, really appreciated !
 
I've been thinking about the STM port and my 2c would be...

It's easy when using processors to get into a mode where bigger and faster processors seem clearcut. They code more easily, you don't have to be nearly as careful at optimizations, there's often more cache space, ... but with all that said, one reason this is such a fascinating product is the price of entry.

I look forward to possibly building active speakers that I can give as presents (or maybe sell off and on) knowing they are easy and cheap to build and maintain. Finding a random ali STM32 board is not always the best thing when you need to replace a processor. Raspberry Pi provides support (like accurate doc) and longevity for a seriously crazy budget price.

So, I'm not opposed to the STM32 but not really a fan either.
 
I've been thinking about the STM port and my 2c would be...

It's easy when using processors to get into a mode where bigger and faster processors seem clearcut. They code more easily, you don't have to be nearly as careful at optimizations, there's often more cache space, ... but with all that said, one reason this is such a fascinating product is the price of entry.

I look forward to possibly building active speakers that I can give as presents (or maybe sell off and on) knowing they are easy and cheap to build and maintain. Finding a random ali STM32 board is not always the best thing when you need to replace a processor. Raspberry Pi provides support (like accurate doc) and longevity for a seriously crazy budget price.

So, I'm not opposed to the STM32 but not really a fan either.
I think that good thing about DSPi is that it can be used on both pico and stm32 with a not too drastic modifications. It's a good thing to have a choice, so that users who want it as a dsp crossover just for their active speakers builds can use cheap RP2040 while users who want to use it for more advanced and resource heavy stuff can use more expensive STM32 boards .
 
Back
Top Bottom