From Brian Willoughby Sent Sat, Jun 16th 2018, 03:00
From Brian Willoughby Sent Sat, Jun 16th 2018, 03:00
Yes, as a hardware platform where it would be possible to substantially = change the character of the sound system architecture every few years, = an FPGA system would certainly be capable of that. It wouldn=E2=80=99t = be so much of a plugin system as a total system overhaul. i.e. No mix = and match of 3rd party features unless a lead programmer coordinated all = of the various IP cores to make sure that they worked together. = There=E2=80=99s no API for FPGA logic glue. Just for the sake of being a Devil=E2=80=99s Advocate (I always am, it = seems), it=E2=80=99s fairly common knowledge that hardware design is = cheaper and faster than firmware design (not that it=E2=80=99s = particularly cheap, or fast). This is true across several technology = areas, not the least of which is music and audio. I=E2=80=99m a huge fan = of Dave Smith, but it=E2=80=99s no secret that his company is sometimes = criticized for bugs in the firmware. I don=E2=80=99t hold that against = them too strongly because I know that they=E2=80=99re creating new = hardware at full speed, and there just isn=E2=80=99t enough firmware = talent to keep up. All of that is my lead-in to suggesting that FPGA = design is an order of magnitude more difficult than firmware design. I = don=E2=80=99t see a lot of quality FPGA designs coming out for such a = magic platform unless someone somehow designed the perfect hardware = platform that never changed, so the developers could have enough time to = do a complete job. These days, hardware is still improving at such a fast rate that = programmers can=E2=80=99t quite keep up. At the very least, customers = generally aren=E2=80=99t happy working with last year=E2=80=99s = hardware, much less 20-year old hardware. That said (and ending my DA mode), Metric Halo Labs just announced today = that they=E2=80=99re upgrading the decade old FireWire audio interface = to support USB and Ethernet connectivity, all within the original case = and analog components. The rarity of the situation that someone could = have top-notch digital audio conversion in a platform that lasts for = decades is probably the best illustration of why it=E2=80=99s not a good = bet to place too much hope into =E2=80=9Cupgradable=E2=80=9D hardware, = much less if those upgrades depend on third parties. Brian p.s. Now I=E2=80=99m just hoping that the Kyra comes to a local synth = shop where I can twist the dials and hear it in action. On Jun 15, 2018, at 7:33 PM, annika morgan <xxxxxx.x.xxxxxx@xxxxx.xxx> = wrote: > As far as arista switches go you can=E2=80=99t be just any random = yahoo developer fresh out of a coding boot camp and build say a = multilayer load balancer into the arista FPGA. Either you=E2=80=99ve got = the money to hire people with FPGA expertise (google or Facebook) or you = work with one of arista=E2=80=99s ecosystem partners to help you design = your solution. It=E2=80=99s definitely NOT a general purpose solution = where you write some portable C+ then compile it and expect it to work.=20= >=20 > I was thinking about it from the perspective that you can invest into = a hardware platform, hang onto it for 15-20 years and every 5 years be = able to upgrade/change/swap out the application running on your FPGA. = =46rom an =E2=80=9Cat scale=E2=80=9D perspective it=E2=80=99s pretty = compelling and it would be interesting to see if there=E2=80=99s any way = that model could work in commercial music applications as well, but = thinking about it general purpose computing and simple analog designs = are probably way more sensible well into foreseeable future.=20 >=20 > On Fri, Jun 15, 2018 at 7:14 PM Brian Willoughby = <xxxxxx@xxxxxxxxxxxx.xxx> wrote: >> How, exactly, would an FPGA support plugins? >>=20 >> The only FPGA system I know of that supports 3rd party software of = any kind is the Audinate Dante Brooklyn II platform. But that=E2=80=99s = not really an FPGA plugin system, it=E2=80=99s a FPGA that implements a = virtual processor that runs Linux. The 3rd party software is nothing = more than a Linux program, compiled for the virtual processor, and = stored in an attached Flash file system. The 3rd parties do not have = have access to custom FPGA design. >>=20 >> Usually what happens in 3rd party situations is that the original = developer packs their FPGA design into an IP core and licenses that to a = 3rd party. The 3rd party then creates a new, complete FPGA design around = the IP core. That=E2=80=99s basically the opposite of a plugin. It takes = significant skills to create a complete FPGA design, even with IP cores. >>=20 >> In other words, you can=E2=80=99t load partial programs into an FPGA. = They=E2=80=99re always built from scratch. Note the comments in the = Wired article about Microsoft, and how they had to completely reprogram = the FPGA every time they changed what they wanted to do. >>=20 >> Those Arista switches don=E2=80=99t allow plugins or partial = programming, do they? Then again, there are SmartNICs with MIPS cores = and ARM chips running Linux, but that=E2=80=99s programmable software, = not 3rd party FPGA programming. >>=20 >> Brian >>=20 >> p.s. What is a =E2=80=9Cplug-out=E2=80=9D anyway? I started hearing = that term in audio hardware, and it sounds like marketing nonsense. >>=20 >>=20 >> On Jun 13, 2018, at 6:35 PM, annika morgan = <xxxxxx.x.xxxxxx@xxxxx.xxx> wrote: >> > FPGA seems promising for 3rd party plugin architectures, thinking = similarly to how Arista switches work. (I=E2=80=99m a network = engineer..) >> >=20 >> >>> On Wednesday, June 13, 2018, Niall Munnelly = <xxxxx.xxxxxxxx@xxxxx.xxx> wrote: >> >>>> Announcement after announcement. >> >>>> Novation=E2=80=99s =E2=80=9CPeak=E2=80=9D has an FPGA = architecture, as, I suppose, do Intellijel=E2=80=99s biggest modules. I = know there are more. >> >>>>=20 >> >>>> People have been calling FPGA as the wave of the future for = years; it=E2=80=99s nice to see it in the present. >> >>>>=20 >> >>>> For a little more background about FPGA (non-musical uses, but = still interesting)=E2=80=A6 >> >>>>=20 >> >>>> = https://www.wired.com/2016/09/microsoft-bets-future-chip-reprogram-fly >> >>>>=20 >> >>>> Sent from a mobile device. Typos and probably bad ideas. >> >>>>=20 >> >>>> On Jun 13, 2018, at 2:12 PM, David Messenger = <xxxxxxx@xxxxxxxx.xxx> wrote: >> >>>>> = https://www.keyboardmag.com/gear/waldorf-music-announces-kyra-fpga-powered= -synth >> >>>=20 >=20