Hello, sorry for the late show-up today
Let me first answer gimmeapill's first message of the day:
- any word yet on the SOC / chipset and operating system?
Yes, we are using a Cortex M7 at 400 MHz (STM32H7). This enables us to keep power consumption low and is sufficient for most presets, from our tests. In the cases where presets are too demanding (e.g. some guitarix ones), the user can either choose to decimate by M=2 (i.e. an effective 24kHz sampling rate) for this particular preset, or to connect a raspberry pi to the box (there is a dedicated raspi extension connector on the PCB). The CortexM7 recognizes this and offloads processing to the raspi's (single or quad) cores.
We are not using any operating system for latency reasons. We program the CortexM7 (and the raspi) bare-metal, with guaranteed real-time constraints. Which takes me to your next question:
- on the audio side, what are you (ideally) targeting when it comes to sampling frequency and round trip latency?
I saw mention somewhere on the site that there's the potential to go down to 1ms and that indeed picked my curiosity...
The sampling frequency is fixed at 48kHz, but we may make it configurable in the future if needed.
The buffer size is configurable in the code and can be as low as 4 samples (0.1 ms), so we pretty much control the latency. There is little CPU overhead in the audio routing and effects' processing time is generally proportional to the number of samples (bare loop unrolling etc), so that we are rather flexible with latency.
We need to do more testing to find the optimal buffer size vs effects' CPU consumption but 32/64 samples feels like a good value for now.
So latency over cable (jack/xlr combos) should not be a problem, but then Linux-based solutions don't have a problem with this either nowadays. Now, another reason why we went bare-metal is for wireless communications. In this case (only), we need to add 10ms buffering for over-the-air packet loss concealment, so that we get pretty close to the dreaded 15-20ms limit
We have a good trade-off at the moment with unnoticeable latency in our prototype tests, but we need more testing in real time conditions, with the real antenna inside the hardware's enclosure, external WiFi interference, line-of-sight obstacles, etc.
- "3W amplifier" -> do I understand that this could easily be modded into a mini combo?
Yes indeed, the (class D) amplifier is exposed through a screw terminal block to which you may connect a 3W - 4 ohm speaker. One may additionally connect a few physical knobs that the system may exploit. For the DYI-minded
Another use-case we tested for the amplifier is to feed its output to a surface exciter fixed to the back of your (classical) guitar. This turns the box into a sort of ToneWoodAmp for an extra 5~10€.
... if this all makes sense?