Why are DPS's used in Modelling or audio interfaces

Stagepig

Stagepig

Member
Just curious after reading that the Axe Fx's DSP Chip seems to be discontinued and the ax8 and fx8 using an cheaper model with lower MHz.
While checking out other modern Modelling products before deciding for fractal I heard the Helix Soundchip is also pretty old and not the cool new stuff. The Helix DSP and the Fx8/AX8 chips seem to be the same ones so that would make them old too.
In addition to that AVID does a lot of marketing about their great actual dsps but an eleven rack is also pretty old and the last AVID Dsp form seemed to last 14 years or so I at least heard.

So what are the advantages of using dsp in the first place and why use old ones then?
I tried reading into this topic but it seems at least as interesting as it is complicated
 
A DSP usually provides enormous compute power for typical signal processing algorithms (filters, FFT, and such) at a very low power budget. They're rather bad at general purpose tasks like running an operating system. You select the DSP that fits your needs, unlike desktop processors that need to run all sorts of user installable BS that you need enough headroom for, DSPs are part of appliances with a clear set of performance requirements that usually doesn't change over time.
 
Last edited:
:tearsofjoy: oh that's what they are doing?:wink:

Ok price point seems to be a convincing point. But with an 3000 bucks + price would the difference between dsp and for example FPGA be that relevant for the end price?
And why are DSPs used in high end products as well where price shouldn't be deciding. There are a lot of audio interfaces and stuff like that too that even state what cool DSPs are used in them (AVID) but then built 14 year old stuff into their units.
I'm not against old tech in general but in Digital Technology 14 years are a really long time
 
Do you know the difference between DSPs and FPGAs ? DSPs run your programs. FPGAs implement functionality in hardware, you design the hardware logic, let it translate into a binary image by the tool chain and stuff it into a little SEEPROM from where the FPGAs loads it to implement that hardware design, it's a way to implement algorithms in hardware. FPGAs are potentially faster than DSPs doing the same thing, but comparatively limited in what they can do.

Changing from one processor architecture to another requires a major investment. All your development tools will have to be replaced, the code library you improved over the years is rendered useless, you will have to educate your developers to use new tools, concepts, languages and optimize code for the new architecture, tons of testing to achieve the same quality you already had with the old processor. You don't do this unless you're forced to because the processor you use is discontinued. I bet this is what FAS is currently doing and why we haven't seen any new modelling products lately.

In hardware design you have a clear spec that you work against. What's the max. power draw, what are the cooling requirements, price, horse power. You never buy a faster processor if you don't need it because you will have to be able to satisfy the higher power demands and cooling requirements which will increase the overall price even further. If I would want to build another MIDI footswitch I wouldn't think about trying to integrate a top-of-the-notch Intel processor because an 8-bit AVR will do just fine and cost only 1% of the Intel based solution which in the end reduces the product price and increases your chances of having a successful and competitive product.
 
Last edited:
That's the kind of info I was looking for thanks for that!

Is the comparison to a MIDI switch fitting? The Midi switch has a limit in it's functionality given by the number of midi channels and messages possible with midi protocol. The Axe fx might have it's limit also but you can't reach it because long before that the chip reaches maximum and has to force disable your shunts

Looks like this works just as everything in the market stick with what you have until it ceases to function. When things reach the point where this happens there will be a lot of bad products with new tech before it gets good. That's a sad thing in my eyes. Look and electric cars same thing.
 
In case of the AX8 the design constraints were having to stick with passive cooling in a closed box. This limits the processor speed and overall power draw. In order to be able to use the same code on AxeFx and AX8 Cliff decided to use the slower processors of the same family accepting the reduced horse power. The AxeFx has been around a lot longer and was based on one of the most powerful DSPs available back then I assume. Back then only few people were willing to spend 3 grand on a modeller so no one can blame Cliff for having chosen dirt cheap hardware.
 
Another reason is that DSP's typically have a market time of 10 years or better. When designing a new product that you plan to develop and support for a few or several years , then factor in the time its takes to go from concept to market which could be a year or two. You need something that's going to be available for at least a decade to be safe. CPU's generally have a market time of 2-3 years at best. then something newer replaces it.and your dealing with new chip sets ect ect.
 
You must log in or register to reply here.
Back
Top Bottom