Next-gen and mini Axoloti hardware discussion

#451

Great questions. I don't know if there can be a definitive answer. These are all just tools that different artists will respond to differently. Few thoughts:

I think in general analog sounds interesting because of imperfection, distortion, non-linearity, saturation, temperature dependency (being slightly off, slight out of tune etc.), also being able to avoid aliasing. Another aspect of it is being able to have continuous voltage control of parameters rather than mapping something to 0-127, etc. I personally feel like I can hear an obvious difference, but it's hard to say if that holds in general. I'm also the kind of person who hears tube guitar amps as obviously better than solid state or modeling. It really depends on what you're doing. It would be interesting to try to A-B test similar filter implementations in both domains, like say a Moog Ladder. To my ears, the difference is obvious.

Some operations just have fundamentally different behavior/meaning in the analog domain: I'm thinking of gain and summing in particular. You're able to avoid losing dynamic range. Also clipping vs saturation: for example, you might intentionally saturate in analog, but you're probably going to try to avoid clipping in digital at all costs.

My sense is that the fundamental differences between analog and digital dominate over component variation effects when it comes to sound. An FPAA implementation is essentially an opamp implementation. An opamp implementation of a filter design is closer to an implementation made out of discrete transistors than it is to a purely digital implementation. At this point we start getting into the weird territory of worrying about "the sound" of an integrated circuit vs a discrete transistor or exactly what transistor material is being used. I think it's clear that analog integrated circuits can sound great; there are a bunch of classic ones. They're very similar in terms of material but different architecturally. I don't think someone would confuse one of these ICs with similar digital algorithms just because they're built out of integrated silicon rather than discrete transistors.

#452

Well I guess I'll just have to try an FPAA out. If they are really more resilient to overdriving, etc, this would be a huge advantage I think. I've been interested in FPGAs for a while and this has really piqued my curiosity in their analog analog.

#453

So what is the ball park ETA on your suped up akso and their eventual FPAA expansion boards (so I can manage my excitement)?

#454

A question that I'm obsessed with:
Would it be posible to program thoses fpaa with the axo interface , using objects specialy dedicated (like LPF, vca, osc...)? That would be the Graal to me...

#455

Software-controlled true bypass or dry/wet signal path would be very cool!

#456

The ability to run at 96kHz instead of 48kHz would allow crisper and brighter sounds without aliasing.

#457

dream on.
high definition audio seems to be a heresy

#458

I already have coded oversampled and downsamplig objects but they are high on CPU for the basic Axoloti. This variant will allow to use them in polyphonic patches.
The possibility to switch to 96kHz would be even better as there will be no need for the downsampling object.

#459

Very interested in this.

Any idea when a possible release would be?

what would cost be?

#460

Any chance you could add an ICE40 FPGA to the new board? Or some other FPGA, but the ICE40 has been reverse engineered and has an open source toolchain.

#461

Currently, I'm playing with a Xilinx Artix7 based board (the Digilent Basys3) connected to my classic Axoloti through SPI and to an I2S adafruit audio breakout.

Xilinx provides a free version of their Vivado dev environment (Vivado HL WebPACK device limited) capable of VHDL and systemVerilog simulation and circuit synthesis.

Even if FPGA are really great, they are quite difficult to handle if you want to do things efficiently as you have to design almost everything...
There is a lot of things to master : tight pipelines, clock domains, synchronization, finite-state machines, multiport RAMs, DSP blocks...
Assembly code is a high level language once you've ventured in FPGA territory.

P1030933.JPG2160x1216 559 KB

#462

Cool. I'm also getting my feet wet. For the ice40, since it is all reverse engineered, they have ice storm, a totally opensource tool chain, and then even icestudio, a graphical patching environment with subpatches, etc. It is a lot like axoloti, except the code blocks are filled with verilog rather than c++. If it were brought into the axoloti environment more synthesis related functions could be encapsulated as higher level objects that interact with the same memory as the axoloti arm. This could be really powerful. though of course the ice40 is not the most powerful fpga... It might be a good direction nonetheless.

Edit: also since it is opensource, you can reprogram the fpga on the fly, right from the microcontroller, from what I gather.

#463

Question: could you not also just instantiate a Delta sigma each on the fpga and tap the signal directly? And at really high bitrates?

#464

I plan to use the Basys 3 and its Artix7 as a very specific and optimized sound engine and i really like it having 90 DSP slices ( 18bits x 25 bits multipliers compared to 16bits x 16 bits multipliers of the ICE).

Ideally, it would be great to have a high level modular system that would deal with all the detail stuff (pipeline sequencing/scheduling...). And sure, an open platform would be great for that purpose !

Sadly, the Artix7 is not (yet) fully inverse engineered (project X-Ray)
https://symbiflow.github.io/

Sure, it is possible to do some PWM or sigma delta modulation.

But, really nice DACs exist and I2S is quite easy to handle.

Most converters can handle 96kHz which is enough for the algorithm i want to implement.
Some I2S converters are capable of 192kHz sample rate.

#465

This is great. In your current application how are you using the axoloti and fpga?

I'd live to add the fpga to my axoloti but id like to transfer ideally axoloti tables to the fpga. I guess you would have to sample the axoloti pins and load them into the fpga. Have you tried this? Transferring memory or structured information to the fpga rather than control signals and switches?

#466

As you can see on the picture at the moment I experiment very basic stuff. I try to use the SPI to communicate between the axoloti and the Fpga.

#467

There's been a bit of banter about unrelated FPGA stuff but not much about the new design.

I'd love to hear about any progress or eta on the new core.

@urklang?

#468

Hi there
The idea of the Axoloti Mini Is very nice.
Could you please provide updates of the project in the first message?
(Yeah i got lazy about reading the whole thread)

#469

My understanding from Urklang's Discord channel is that the prototype boards are back from the manufacturer and in the hands of testers. I think they mentioned a month of testing before taking preorders assuming no further revisions? Something like that. Seems like its not too far off.

#470

Just found this thread as I've been away. The Axoloti mini is a very exciting prospect and I look forward to ordering a couple of them.

Because of all the chatter about VAT and FGPA's its hard to locate information about the actual board on this thread. So forgive me for asking, will this board work with the same Axoloti editor and be able to use the older firmware? I assume it will be compatible with the Axoloti Core, but I have to ask as there are some posts on this thread suggesting a different processor and so on.