What do people use small FPGA boards for?

[BigEd]

Off the top of my head, I imagine these small boards might be used for education, or hobby, or possibly for engineering or scientific purposes.

What kinds of designs would be seen implemented on them? Audio? Video? Radio? Process control? (Retro computing, of course, and computer architecture exploration.)

Other than for education purposes, I'm interested because a lot of things can now be done with microcontrollers, so the FPGA board has to be offering something special.

[metaprogrammer]

Motor control, for example. I don't like using MCUs for hard real time, it's much easier with FPGAs. Easier to program, easier to get the right timing, easier to implement slightly non-standard SPI protocols (some encoders are weird).

[Folknology]

Excellent question @BigEd

My feedback from users from the myStorm POV, given the caveat that the bulk of this projects boards have mostly gone into educational applications:

• Retro-Gaming
• Retro-Computing
• Demoscene
• Processor design/experiment
• Microcontroller graphics mini GPU
• Motor/motion control
• Low-res Camera interfacing/processing
• Ultrasonic imaging

Not necessarily in that order..

[lawrie]

Others include:

• Audio synthesizers
• Tools such as oscilloscopes and logic analysers
• AI accelerators (e.g, the Marlann project)
• Custom microcontroller (e.g. using Risc-V)

[r.miller]

Another category is prototyping designs which will be eventually implemented on another platform. The rapid-turnaround open source yosys/nextpnr toolchain, and many accessible pins for attaching auxiliary boards and oscilloscope probes, make the myStorm particularly enjoyable for this purpose. For example, here

I've connected a cheap and cheerful ethernet PHY chip (a LAN8720 on a Waveshare breakout board) and a hex display with a few useful buttons, to develop and debug a MAC design which I hope to adapt later for a different FPGA which is embedded on a RISC-V SoC.

[Folknology]

Other than for education purposes, I'm interested because a lot of things can now be done with microcontrollers, so the FPGA board has to be offering something special.

I think this asks a more underlying question that I have heard a lot which is: I can already do a lot with microcontrollers (I am familiar with them) what would an FPGA (something I am not familiar with) enable me to do that I couldn't already?

[Folknology]

Part of this is just resistance to something new, a new language, new hardware, new discipline. The benefit is of course more powerful tools and techniques and different approaches that enable one to tackle the problems from entirely new perspectives. Also unlike learning a new language FPGA skills can be additive, they enable you to add new skills and tools on top of your existing ones, it doesn't mean 'replacing' those valuable earned skill investments, the new ones can be very much complimentary and this benefit is often missed and skipped over IMHO.

I think both @r.miller and @metaprogrammer are benefiting in this way

[BigEd]

Yes indeed - I'm asking with the aim of getting a story, rather than because I don't get it! A story that will help motivate someone to pick up FPGAs as a tool, and perhaps also help motivate some of the choices as to what's on a small FPGA board by way of interfaces and peripherals.

It's clear that you'd pick up an FPGA board if you wanted to learn about FPGAs. So it's the other cases which I think are most interesting: in what ways do microcontrollers underperform, or FPGAs outperform.

As microcontrollers get cheaper and more powerful, FPGAs have, in a sense, more to prove.

[lawrie]

I am constantly surprised at how small the community that is using FPGAs is, particularly those working open source hardware and tools.

The Von Neumann architecture is such a restricted way of computing compared with all the possibilities that digital logic offers. I suspect we have only got away with it for several decades because of Moore's Law, which is now coming to an end.

Microcontrollers are bad at driving hardware, real-time control and parallelism.

It should be easier not harder to produce solutions with digital logic rather than just software, and the reason it is not is that the hardware description languages and tools are so terrible compared with the software equivalents. That is only now beginning to change.

It is also ridiculously and unnecessarily expensive to produce ASICs, which means people are stuck with slow FPGAs.

[Folknology]

Yes indeed - I'm asking with the aim of getting a story, rather than because I don't get it!

Sorry I know you get it Ed I wasn't aiming the comment at you personally, rather illustrating to those reading the thread that are relatively new to FPGAs and perhaps haven't had that eureka moment yet..

[Folknology]

in what ways do microcontrollers underperform, or FPGAs outperform

For me there are some common problem areas where I reach for the HDL rather than 'C', the most common being more than one realtime requirement that needs servicing (often via ISRs). The moment you use multiple ISRs, 'C' becomes much more difficult a solution compared to HDL, you are no longer in Kansas, sequential models fail here badly, complexity of interactions and in-determinism, parallel HDL wins hands down with simpler models.

[BigEd]

One angle which I think doesn't come up much is that you don't need any money down to get started with FPGA: the first step is to learn an HDL, and you can (and should) be exercising your design in simulation. After building and simulating a few digital systems in simulation, there's less of a worry in implementing on an FPGA to run it in real time. All the tools are free of charge (and many are open source.)

Something FPGAs do offer is massive bandwidth to on-chip memory: it is I suppose an instance of parallelism, as already noted. But it's not just parallel computation, it's also storage.

I suppose integration too: there will be cases where any other solution would need a handful of chips, where an FPGA can hold the whole design.

[Folknology]

Another low hanging fruit is cases where a microcontroller doesn't have a specific peripheral to interface/handle connectivity with another actuator, sensor or device. Often you will see 'soft' solutions running within the microcontroller steeling major chunks of the available cycles, squeezing the embedded programmer into difficult corners and compromises..

[Folknology]

I am constantly surprised at how small the community that is using FPGAs is, particularly those working open source hardware and tools.

I agree and our new Icicle board is aimed at growing that community by building a better on ramp..

The Von Neumann architecture is such a restricted way of computing compared with all the possibilities that digital logic offers. I suspect we have only got away with it for several decades because of Moore's Law, which is now coming to an end.

This is very true but I have always believed you can have both 'Von Neumann' and HDL working together heterogeneously, that is why we always included a hard core in addition to the FPGA fabric. The problem in the past has always been how to make that accessible. Icicle will be the first step in that direction which will mean some changes in approach, a more focused development setup in order to accomplish that.

[BigEd]

There is the possibility that some people who would benefit from using an FPGA have already discounted the idea, having misunderstood what an FPGA is, or what it can do. The reason I say this is because from time to time people in the retrocomputing or retrogaming communities will describe FPGAs as "emulating" hardware: that because an FPGA is "programmed" it must offer comparable facilities to a microcontroller. I think this is a big mistake, and I mention it just in case someone can think of a way to get over it.

One idea would be to find and use some descriptive phrase, such as "configurable custom logic" which captures the idea that half a second after power up the FPGA actually becomes the machine that you described in HDL.

[Wells]

I’ll chime in, and give a hobbyist perspective
(And I’ll add that this is my first post in a couple years, life intervened but has let up for the moment:))

As a Jack of all trades (and master of none) I got into Arduinos a decade ago without a specific use, but as the saying goes when holding a hammer every problem looks like a nail, I found uses in many of my other hobbies. Plus they’re neato. I used to build solid state Tesla coils, and dabbled in holography for example.

As an exercise I cobbled up a brushless motor and was using the micro to do the commutation. This led me down the path of seeing how fast I could switch, and I learned about bit-banging, as “digitalWrite” was hopelessly slow. I soon reached the limit of what the chip could do, which while reasonably fast, completely consumed the chips capabilities. That’s when I found out about FPGAs, and suddenly many of my nails needed a new hammer

So basically the parallelism and speed is what I’m interested in. I haven’t done much yet, and un-learning the sequential mindset that has defined all of the programming I’ve ever learned has been a challenge, but once I started messing around the possibilities are exciting.

I doubt my uses will become as complex as vintage machine emulation or video processing, but when I was making holograms I built an analog PID controller for a fringe-locking mechanism, and could have def. used an FPGA. Arduino was abandoned early on, not nearly responsive enough.

Hope that adds to the story

Wells

[dsp8bit]

Hi @BigEd,
I'm more related to the design of digital signal processing and telecomm design. So, I can tell you that FPGAs boards are getting more affordable and accessible for students, hobbyists and engineers in general. The question that you shares is a big one. Another application where is necessary to bring in the open source community (specifically for Lattice FPGAs, and then, opensource tool chains) is the Software defined radio based projects.

I'm interested in developing dsp ipcores for telecom, right now i'm focusing in Altera and Xilinx boards (as the Nuand bladerf). However, i would be great to have a SDR platform based on a lattice fpga. I like so much the BlackiceMx board due to the dsp possibilities. It is the ARM-FPGA based board more affordable nowadays.

In summary, it is kind of difficult to promote the use of FPGAs when a FPGA is put to compete with a microcontroller with ONLY IO controller applications and other peripherals. We need to explain more about the difference in those technologies.

On the other hand, @Folknology it would be nice to have another version of the BlackIceMX board with more power with ECP5 and bigger DRAM. (This is a letter for Santa claus jejeje)

Cheers,
dsp8bit

bladeRF 2.0 micro - Nuand

[Folknology]

myStorm ECP BlackEdge board is well underway..

[cdohl]

Hi
I'm new to FPGA programming and thought of a small project to get me started. In physics labs we very commonly use digital delay generators. These generate pulses at arbitrary time and have a specific duration. While some applications need very high time resolution, the 100MHz of the Lattice is already pretty useful for many applications. There are a number of companies selling these for several 1000$ for 4-8 channels (BNC or SI). I've finished a first working version of an open source delay generator and placed it on Github. It has very few options as compared to the commercial ones, but could be configured to have many more channels. The drivers and the electronics for trigger input are still in the make. Yet, maybe some of you find it already interesting and can use it as a start for create something useful out of it. Here is the link to Github. Btw, it is implemented on a Blackice II, but should be simple to port to Blackice MX. The STM32 does the serial to SPI translation and a GUI is available with a Jupyter Notebook.
Cheers, dieter