[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Date Index][Thread Index][Author Index]

Re: DSP ... DJRND2 and TC Dual Delay...



Kims point is great to explain why the cheap multieffect things have short
loop times.

But the new t.c. two seams to be just a delay, so I'm not sure there is a
DSP in it. To comunicate with a cheap CODEC, its almost necessary to use
one nowadays, but t.c. maybe does not use a cheap CODEC :-)

In the t.c. 2290 and PCM42 neither sound nor sound memory adresses are
runing through the processor. It only sets the adresses acording to delay
time and then a counter runs through the whole range and the sound just
goes in and out all the time. Thats why its possible to expand the memory
in the PCM42 by adding  counter chips.

When we started with the LOOP delay, whe thought a lot about using a
ADSP2105, which even in '92 was not very expensive.
Taking chances to speak out Dr. Perilles secret ;-) :
Since the DRAM has column and row adressing, we can send out each part
separately and even save adress register chips.
In the future, it will take memory mapping to organize various loops in the
big memory. Maybe he did that?

Did someone buy the DJRND2? Since at the time, there is no other dedicated
looper product available...
Its made for DJ and maybe less handy for a instrumentalist (you have to tap
tempo before recording, right?), but certainly new ideas would come up...

As I understand it has no MIDI, but it syncronizes to the music by
analizing it, is that correct? Pretty interesting...

How do you say: DJRND2? It looks so twisted...
Has it been discussed on the list before?

>> the reason for this is not the price of memory, but the price of
>> processors. DSP processors usually have small address spaces, so they 
>can
>> only access small amounts of memory. The DSP procs that have large 
>address
>> spaces are very expensive, and not likely to be used in low/mid range 
>audio
>> products. The cheap DSP procs have much smaller memory area (usually
>> requiring expensinve sram memory chips), which is why you see them with
>> small loop times. This is why signal processing boxes are usually not 
>well
>> suited for looping. Looping isn't a DSP function, it needs big address
>> area, good address calculation, and good real time operation. It doesn't
>> need dsp.
>
>For instance, DJRND2 is totally based upon one single ADSP2105 directly
>addressing 14 stereo loops simultaneously from one EDO/FPM 16Mbyte RAM
>module. How is it possible ? => Claimed in my PCT

oh, its not a secret? :-) What is PCT?

>> When a looper function is put into a dsp box, it usually has a
>> small loop time and practically no user interactivity, since the
>> architecture of these boxes is not designed for much user interaction.
>
>Sure ?
>
>> The box is just supposed to sit there running its dsp algorithm on an 
>audio
>> stream, not bounce all around it's memory responding to user inputs. So
>> when you see dsp device touting looping ability, don't get your hopes up
>> too high because it probably won't be that great.
>
>Sure ?
>
>>
>> Loopers usually are based on low-cost microprocessors, which typically 
>have
>> large address spaces even on the cheap processors. The latest low-end 
>procs
>> have built in SDRAM memory controllers and can access 512MB or more with
>> no additional parts. But these procs are not terribly well suited for 
>DSP,
>> which is why most loopers don't have fancy signal processing along with 
>it.
>> These devices are great for having large memory space and being able to
>> bounce all around the address area at the whim of the user. They can 
>have
>> very responsive types of interfaces, where the user can execute all 
>sorts
>> of commands and functions and the looper responds immediately. They are
>> also built for having lots of control input/output, for buttons, knobs,
>> displays, etc. As the cheap procs continue to get faster and more 
>powerful,
>> you'll probably start seeing some more interesting dsp functions in 
>them,
>> but not on a par with something based on a powerful dsp chip. The best
>> approach is to marry a dsp and a microprocessor together, but that 
>drives
>> the cost up.....
>
>Sure ?
>Emmanuel

Maybe you did not find the right tone yet to advert your invention, brother
Emmanuel. I never did either, it takes at least 3 persons for a good
product:
-  the inventor  (sees what shall be working)
-  the tester  (sees what does not work)
-  the salesman  (sees what the public wants)

good luck
Matthias



         ---> http://Matthias.Grob.org