NAM A2 on Daisy Seed

Guitar Pedals
1

A Daisy header for the recent A2 NAM variant (the standard format from now on, I think). tried to optimize speed and balance SRAM/DTCM so it can fit in a multi-effect and not just as a NAM processor. Tested with BOOT_SRAM, BlockSize=48. It loads a built-in JCM model and enabled by defualt. CPU is higher than the nano model (53% vs 40%). The A2 has about doubled the weights compare to the nano model, but uses embedded friendly activation so overhead is manageable and sound quality is far better. Natually this supports only the A2 lite variant.

NamA2Daisy.zip (18.8 KB)

static __attribute__((section(".dtcmram_bss"))) nam_a2_daisy::A2Jcm2000Daisy48 nam;

init:

   // Load and prewarm the model before audio starts.
    nam.load_default_model();

audio callback:

nam.process_block_48(input48, output48);

2

Was your nano test also at 48 samples? I’ve only seen tests on Daisy where it needed like 128.

3

it was 48 samples. I posted it on this forum somewhere. I think it can run with smaller block without major increase in CPU. Anyway Tone3000 are porting the models to A2 so I don’t think nano would be relevant.

4

Thanks! I started a branch/module for the bkshepherd pedal [DRAFT] NAM A2 module by xconverge · Pull Request #94 · bkshepherd/DaisySeedProjects · GitHub

There is no source/attribution in the header, is this file your work? Am I free to use it?

5

The file is original work (AI assisted). Free to use. You can just point this forum thread as reference in the header. I hope to have some time to make a video of this model running on my JamMate. It can read native NAM models directly from the SD card or stream them in from the BLE app using the Tone3000 API. Model audio quality is much better than the nano variant. Just notice where you place the buffers because it can impact the CPU load.

1 Like
6

Thanks! I have it running, sounds good to my not-picky ears :person_shrugging: