DSPs versus FPGA

[Aviador]

Hey. I found this, I think, interesting article. Any day we could see a Scope with FPGA instead of DSP.
http://www.eetimes.com/design/military- ... geNumber=0
Regards

[siriusbliss]

FPGAs are great, but still require a lot of power compared to the more dedicated DSP chips.
Still very pricey too.

But yes, as an alternative to DSPs, there is some great technology there.

G

[jhulk]

fpga audio chips made by fairlight

they used them to create the new fairlight

this audio chip they used to create audio mixers first

then they set about designing filters and osc of the fairlight

what fpga are good for is emulating custon asic chips or cpu or logic chips for memory

pals or any chip that has inputs and outputs as you can use a logic anyliser to record there functions and can program a fpga todo the same functions

on the korg dss1 16mb upgrade we used a fpga as a switching memory page function which enabled us to have the use of bigger samples

a multisample now can be as big as 16 x256k which is about 4mb

so for circuit emulation its better than dsp as with an fpga you can also copy the flaws of circuits which is harder to do in dsp

[Bud Weiser]

FPGAs are great, but still require a lot of power compared to the more dedicated DSP chips.
Still very pricey too.

But yes, as an alternative to DSPs, there is some great technology there.

G

Hammond simulation on FFGA,- latency free and very authentic ...
http://www.keyboardpartner.de/hammond/hoax_en.htm
HOAX 3 is ideal if you own a outdated or trashed Hammond clone offering the manuals/key contacts and sets of drawbars.
A friend of mine just uses the shell of his Hammond/Suzuki XB-2 to run HOAX 3.
Comes with or without Leslie sim.
The Leslie sim seems to be a derivate of Neo Ventilator because Guido from NEO is involved.

RME cards also work w/ a own developement of FPGA and "secure flash update technology".

Some creations work very well using FPGAs, but SHARCs are ideal for synth engines and other tasks in audio world.

Bud

[garyb]

this is stupid. why can't a FPGA chip do a dsp's job?

the reasons for using either depend on the application and the type of knowledge of the guy who has to do the programming. both have advantages. the reason that Fairlight went to the FPGA package was that the DSPs the used at first were purchased from Creamware and they wanted to do the job in house. they didn't have coders for dsps and had to start from scratch, so they used what they knew. after, the marketing department made out like it was a God inspired move.

DSPs are best for certain types of number crunching, but that doesn't mean that a FPGA chip is bad and vice versa.

why the hell would S|C abandon DSPs now and have to start over from scratch, if the final product won't work better, sound better, or be lots cheaper?

[HUROLURA]

You know, there are probably already FPGA on SC hardware. I mean, what would be the big SC marked chip next to the old 21065L DSP on the Xite-1 board if not FPGA ?

Remember the Quicklogic chip on the old Scope PCI: one more FPGA !

They are probably used for host interface but they belong to the system life ...

[garyb]

yeah, this is the problem with technology based products.

everyone spends too much time getting hypnotized by the technology itself. what updates can be made? what specs does it have? how many voices can x-synth do? how about the newest os? get with the times!!!

folks, the point is AUDIO and MUSIC. what does it matter if you have 500GB of ram, if the sound and music that you make isn't any better? the BEST audio gear is often OLD and OBSOLETE, made by a company that went out of business 30 years ago, and that gear is often still used in the same PROFESSIONAL environment that it ever was because it WORKS.

computers suck. they're just a tool, like a hammer. if the new hammer works better, consider buying it. if it's just another hammer and the one you have is still in one piece, you don't need a new one. i'm very happy with my computer and DSPs, but it's because it actually replaces a whole bunch of expensive gear, inexpensively and does that convieniently and with NEARLY the same quality. i have tools because of this technology, that are affordable and super high quality. after that, the work that i do is MY fault. what more is there?

[jhulk]

i agree with gary

im still using p4 winxp comp thats 10 years old now becuase it works and does the same as it ever did and sounds the same as it ever did

and it seems that the new tech at 64bit actually makes the scope dsp systems be it pci or xite perform less

than the p4 that i use

if oit works dont fix it

like gary said if the hammer does the job then keep the hammer

whats the point on upgrading if its still only going to do what the hammer does

the things that fpga do best is make circuits of chips for old gear that the costom chips are no longer made

the biggest use so far has been cpu keyboard scanner chips

as when these go then the synths are dead

but my 2 scope systems they are just that they are in 2 p4 comps and are used like they are sound modules and mixers

my daw comp is an i7

instead of trying to get it all in one comp i just use the best of both

you can do a lot on the old pci cards and winxp32

my latest modular devs experiments have proved this

as i have been testing modular to its limits

22 sample osc max at 4 note poly so thats 88 sample osc if you play 4 notes thats impressive

it runs out of audio connections at 60% on my 27 dsp system at 4 note poly

so you cant even use all dsp power

you can create any synth you want with modular and scope

and only needs a few updates to make it better

but the sound quality from this is ten times better than anything

[Aviador]

My interest in the FPGA technology applied to the audio has emerged after reading card information from Fairlight CC-1 as replaced and several cards with a total of 64 SHARC DSPs on a single FPGA.
So I took some time researching this technology. All for intellectual curiosity
I am a happy user of three cards (two Pulsar Pulsar II and I) and I think are among the few products that after over 13 years in the market are able to function in modern equipment, in my case a Core Duo2 Ultimate Win 64 bits. I doubt that a Protools overrated the same old can operate a computer and operating system present on both Mac and PC.
I think they call scheduled obsolescence ....

[Bud Weiser]

everyone spends too much time getting hypnotized by the technology itself. what updates can be made? what specs does it have? how many voices can x-synth do? how about the newest os? get with the times!!!

Well, that´s a relative thing.

1.)
Thinking about updates was introduced by the software industry allowing releases of unfinished products early and let the user wait for full functionality.
Not possible in the times we buyed expensive hardware in shops, not online.

2.)
The demand of max polyphony of synths came up when MIDI was introduced and MIDI multi mode entered the MIDI standard and because hardware technology enabled getting several sounds out of a synth simultaneously.
This became a important factor for live performers as well as the "MIDI Studio" enthusiasts starting w/ Commodore computers and later the ATARI.
As a result, computers became the center of music PRODUCTION (not performance) and keyboard arrangements became not only more complex (if done right) but replaced real musicians, 1st the drummers when drum machines came out, then bass players in addition when the hardware sequencers came out and so on.
In the end, the only ones hard to replace were vocalists and guitar players because that was almost impossible to sample w/ great result,- but the main goal was, replacing real players, complete studios, the engineers, assistants and everything generating production costs.

I´ve seen that coming early 80th already, the evolution of this technology being a job killer for all the skilled and being a job generator for those not having skills and learn programs to compensate.
It´s exactly what the industry wanted,- selling gear to the masses, like the consumer electronics market, telling ´em everyone can be a producer, a star or whatever shit.
And it worked !!!

But, when it came to put the computer music productions on a stage for a tour because the artist became famous or sold many records suddenly,- well, all the polyphony was necessary to replicate all these and very often musically sensless tracks done by sequencing, an it required 2 or 3 keyboardplayers too.
Well, we did it without backing tracks and sequencers live and I liked that was possible w/ good players,- but polyphony is necessary to do that unless you go w/ 2 trucks for keyboards alone or become a slave of backing tracks.

folks, the point is AUDIO and MUSIC. what does it matter if you have 500GB of ram, if the sound and music that you make isn't any better?

Right and good question,- but today requires definitions of good sound and good music.
How much good sound and good music do you hear today and is that, what most people consumate as "music",- good ?

RAM doesn´t produce music at all,- so if you don´t need much samples to replace real players/ instruments, you won´t need much RAM.
As a reminder,- Joe Zawinul did excellent string arrangements in musical context playing a Prophet T-8,- not any sample.
The others use tons of samples, sound like shit and simply show lack of knowledge about how to arrange for classical instruments.
So, we´re back to the skills mentioned above.

But times changed,- don´t forget listening experience and habit.
What people often consumate today isn´t music,- it´s noise but defined as some kind of new music,- and it can only be done w/ a computer as the center of the studio as well as all the VST crap out there being satisfying for the crowd buying/ consumating that music.

Produce quick and cheap, you need a computer (and OUR program) only,- that´s the message being sent out from the industry since decades.
And it works !!!

the BEST audio gear is often OLD and OBSOLETE, made by a company that went out of business 30 years ago, and that gear is often still used in the same PROFESSIONAL environment that it ever was because it WORKS.

YES, YES, YES and YES again !

And this gear was made for the skilled, not any masses of pseudo producers, pseudo engineers and pseudo one-finger keyboardists because not anyone of these idiots had a chance to enter a stage or studio for a professional production in the 70th/80th.

computers suck.

YES !!!
I hated my Commodore, my 1st ATARI, all the Macs and my PCs,- they all sucked and suck up today.
I hate, XITE-1 is addicted to such a piece of shit !
I´d wish, XITE-1 had a OS in ROM and some RAM to load the OS from ROM as well as the devices, I switch it ON and work w/ my connected hardware instruments as a performer and record in VDAT if necessary, just w/ a harddrive or any media connected.
XITE-1 is such a good piece of gear, excellent sounding and worth it´s price just only as a summing machine and as a mixer,- the devices come in addition.

If I think about what a p.ex. NEVE summing box costs, limited to 16 channels, vol / pan / cue and inserts only and 2 or 3 stereo outs, 1 transformer balanced,- the XITE-1 is a steal.

Now, 2 DAW computers failed here again since a week and I´m too busy w/ other things to repair and I´m also too lazy to do it because I hate this f**king computer component crap failing all the time.

I own gear you mentioned above, being 30-40 years old and NOW it´s time to do some service and we do.
We do because this gear will work for another 2 decades again then.

they're just a tool, like a hammer.

I need the hammer for THAT tool !!!

... after that, the work that i do is MY fault. what more is there?

So true !
Do I need a hi-end computer and "featuritis" sequencer ?
No. I can play.
When I play a track and it sucks, I sucked and need to practise, then play again and until it´s correct.
O.k., it won´t work for the drums when I have to do that and don´t have a recording booth and a kit being set up miked.
So,- sometimes I´d need a sequencer, but what kind of ?

Oh well, yeah ... I forgot,- there´s my old and trusty Yammi QX-1 fully working but collecting dust, my ATARIs which never failed within 3 decades and Emagic Notator SL I did orchestral arrangements with already in the past.
There are these samplers in SCOPE,- I hope for optimization and more sample formats,- at least EMU in addition to AKAI.
I´m fine w/ EMU and AKAI for drums, even in mono and I think most of the modern bedroom producers will be fine too with that because they like the crappy analog drum machines like TR808, 909 and such.

Well, I´d wish we had a fully standalone working XITE machine,- like one of these 30 year old pieces of studio gear mentioned above.

Bud

[garyb]

it was the best of times, it was the worst of times....



naturally, i use computers and it's great. i just don't want to be confused about what's important. it's definitely not the technology, even though the technology is super cool!

[Zimbo]

I think there are more reasons than experience with the One or other to choose what to use for wich purpouses. There a Big differentes in how this different Kind of Processors are programed.

Programming an fpga means more or less describing a Hardware circuit- that of cause can be realy efficient on some Tasks while difficult on others. Generally i think fpgas are Good for Emulating Things Existing in analog Domain. One stregh of fpgas is that they can realy to a lot of things in true paralel. also everytime you Startup an fpga it can get a completely different structure (as a funny example it might be possible to Emulate a dsp on an fpga) whats cool on the One Hand but also gives you Limits... Every Peace of Gear i know thats Based on fpga without the combination with dsps comes with a with a more or less fixend functionality once loaded- there are mixingdescs for example with Lots of routing and a Lot of instances of eqs, copms, whatever- but this is ALL THERE ALL THE TIME
So if i stay on mixingdescs as an example- switching of every eQ section on the Board WILL NOT FREE ANY PROCESSING POWER. Undertand what it means? For something like a scope Environment you use Daily all the device you want to use/ connections you want to use have to be there all the Time from the beginning... There are some audiomatrix Things in pa World wich have a Huge amount of flexibility (hundreds of Objekts/devices) and routing (similar to scope Minus Musical Aspekts) a result of using fpgas is that once you change anything ( ad a device, Change the routing, whatever ) you have to load the whole Thing again while nö Processing is possible at all.... So a dsp its much more Dynamic on Sending power for a process when needed...

Thats because of you have to Think about an fpga Programm the Same Way than Hardware- so if there is nö kable betqueen two Objekts you cannot connect them without Changing the "Hardware Description" of cause there can be a routing object where everything Else is connected to, than you can Route arround as much as this object allows... But you have know what you want to be able to do before. Dont missunderstand: its Not a Problem to Change any Parameter in realtime (this boxes often have connections to wire analog Pots, faders and switches to it and specify what they do) also tweeking eveything from the Computer or some remote is Not aproblem- but its only possible whats defined to be possible when Loading the Programm on the fpga- as fare as i know its Not possible to load or Exchange Parts of the Code while Running. That also has streng points- i have Never Seen an fpga chrashing or Produce errors- if it Runs Fine now it will Run Fine with a loaded Programm it will do so as Long as it dies.

Thinking scope again- imagine running on fpga everything you might gonna use within a Session has to be loaded and connected from the beginning- that would Change the whole behavior and possible Workflow.

In the end fpga is a whole different ballgame and there is another Thing that i think an sdk as you know it for scope would Not be possible that easy for a fpga Based system. As metioned before- you can find fpgas on some sc- Boards too, also uad Users fpgas alongside the dsps- so i Thing in Most cases developers do know why they use either this or that- if it can be done Equaly Good on the One or other than it might be decided by what the one who has to to is more experiencent/comfortable with...