Affordable Freescale DSP Board

hmm.. the 56300 series can do 180 mips. this is not more than the emu10k2 isn´t it?
for external controls over serial bus or gpio a microcontroller is needed. this could be a piece of hard work for making a standalone effect-unit or an instrument. for learning purposes on "real" dsp- boards it´s quite good. if anyone plans a commercial release with this little thing it must thought about the consumers "mips- hunger".

greets stylus

 

>it must thought about the consumers "mips- hunger".

no problem here, "mips-hunger" consumer is ready to pay thousands bucks for thousands MIPS.

>for external controls over serial bus or gpio a microcontroller is needed.

aren't most posts in the thread exactly about this?

edit:
and btw. Emu10k2's DSP is only about 50 MIPS (obviously 1024 * 48000, not counting a few additional MIPS on "automatic" circular buffer addressing (e.g. delay lines))

 

of course you were talking about mpu´s.

i´ve found this in the fx8010 datasheet:

"By coupling a highly regular four-operand, 32-bit
integer architecture with independent delay memory
and I/O engines, FX8010 delivers 100MIPS at
50KHZ sample rate..."

fx8010= emu10k1, emu10k2 can perform the double instructions

 

>fx8010= emu10k1

Nope, "FX8010" is a DSP architecture. The DPSs inside Emu10k1, RChip and Emu10k2 are all of "FX8010". (The Emu10k* are "Audio Controllers" and the DSP is only a part of them)

>"FX8010 delivers 100MIPS at"

He might be talking about RChip's DSP (which was 4 times more powerful than Emu10k1's one). (Or he could be just overstating a bit ).

Finally, we don't need to guess or to believe any papers. This can be simply calculated :
Emu10k2: 1024 instructions per sample * 48000 samples per second = 49152000 instructions per second
E.g. 49.152 MIPS
Emu01k1: twice lower (as it has 512 instructions per sample)

 

Err, its more than 3x faster then the emu10k2!
Besides that, it's a more versatile DSP then emu10k.

Duh. Except for a few exotic and expensive DSP solutions practiclly *every* DSP chip needs a host MCU!

As opposed to what? To putting a CL card in a windoze box using a driver and sofware you don't have to write yourself?
If so then Yes, it's *obviously* much harder then that.
In any other case, No, this is considered normal.

And for anything else,....???
What a nonsense!

There are tons of applications and consumers this chip can easily satisfy.
Not every application needs 1000 mips to be satisfactory.

Sorry Stylus but your arguments are pretty lame and frankly I fail to see the purpose of them.

 

don´t panic..

well, let me explain from which point of view i regard it. digital synthesizers have grown up to voice- monsters with 80 voices in a access virus ti for instance. the clavia synths use 4-8 of the 653xx dsp´s. the korg radias is based on a texas instruments TMS320C55x with ~600 mips, i remember. all that stuff has to placed on a hard contesting market. the consumer prefers the machine with the largest power (or a dumping price)- that's a fact. of course other devices like effect- units, mixers don't need the power which modern synthesizers eat. if this is your intention your´re right. most the time i think about synthesizers when i see a dsp..

i really don´t want diminish our work lex. this was only refer up to the (synth)market. (if a commercially release is planed.)

 

Well, i have to agree with Lex. I can't find any purpose of "MIPS penisometria" too
- especially considering the thread title.

 

@lex

I did't know that my point of view would cause so much annoyance. perhaps my conception, regarding marketing a freescale 653xx powered device at present, was somewhat rash. it should be a small remark at the edge only. however the board is nice and i wish much fun with experimenting.

btw. for my own i prefer a sharc.

@max

i agree with you the rchip in emu's mantis system was indicated with 100 mips. and a "mips- penisometria" does exist anywhere out there.

http://freenet-homepage.de/kxdev/docs/original/fx8010-arch.pdf

 

>and a "mips- penisometria" does exist anywhere out there.

doh! as i already mentioned above - it's not a problem at all to buy bigger peni.. erhh - sorry.. more MIPS - but so what?

 

Stylus, are you using babelfish translation to communicate here in english?
I hope so because you don't make much sense in the context of this thread other then
sounding snobbish and annoying.

 

you have a delightful sense of humor. i really hope it's understood now.

 

perfectly

 

Max, in regard to ARM fp support and performance, it seems a few of the faster ARM
chips have build-in hw fpu. (well I'm sure you already knew that)
For example the LPC3180, very expensive and Cirrus EP93x, reasonable priced but there seems to be GCC problems with the Maverick Crunch fp hardware.
You were right, this can all easily become another major PITA
(as opposed to our beloved x86 fp support)
But then again, perhaps GCC softfloat code running ARM @200Mhz covers practicly all host application needs?

 

But then again, perhaps GCC softfloat code running ARM @200Mhz covers practicly all host application needs?

I can only hope so. Eventually, though i'm programming with ARMs for a while now - i've never used FP there - so i can only guess.
By the way, did not you bump any info on how exactly long (in tacts or so) it takes for the softfloat to perform common operations (sin/log)? It would be quite handy to just calc a clue of how many of them one can have per a second on a particular MCU.

 

I've found performance specs (in clock cycles and usec@xxxMhz) for some commercial fp library, but so far not for any free or GCC ARM fp lib.
I agree that would give some form off indication, but I think the best way to find out
is writing some actual code and run it on actual hardware.
Well, that should not be a real problem since many of those ARM boards are indeed relative affordable as well as available.

 

Îòâåò: Affordable DSP Board

btw. - quite amusing news :evil:
Creative started to license EMU8030 chip to third parties.
EMU8030 is basically a emu10k2 with build-in "MCU" (80186 - :duh
OEM - Chips - EMU8030
Curiously they position it as "MIDI Synthesizer".

 

Yep, looks like it's geared towards stand-alone soundfont player applications.
I wunder if there is any demand for such 'ancient' technology these days.
Thanks for this amusing info.

 

Just got my hands on a Tonecore Dev kit. I wish I knew about these 4 years ago
How hard do you think it would be to port of the the kx DSP plugs to this baby? I always loved the sounds of TigerM's verb and ampsim. I'm sure the $20 line6 modules are using better algos but I found a couple of these to have a certain sound that worked really well for certain things in a near-0-latency live environment. Would love to have it available as a pedal!

 

>>]Just got my hands on a Tonecore Dev kit.
New? Afaik these kits can not be ordered outside the US.

>>How hard do you think it would be to port of the the kx DSP plugs to this >>baby?
Well, that depends on your programming skills on both platforms.
Basicly, you will need to port the plugin algoritme to 56K assembly code.

 

Are you talking about the tubedrive? Or what other ampsim type plugin is there?