a small question about emu10k's delay line~

JoshuaChang · May 7, 2011, 08:54 AM

does the code below means 2 x 1sample delay?

Code:

    idelay write w1 at 0x0;
    idelay read r1 at 0x1;
    idelay write w2 at 0x2;
    idelay read r2 at 0x3;

stylus02 · May 7, 2011, 11:19 AM

Quote:

Originally Posted by JoshuaChang

does the code below means 2 x 1sample delay?

Code:

    idelay write w1 at 0x0;
    idelay read r1 at 0x1;
    idelay write w2 at 0x2;
    idelay read r2 at 0x3;

yes, this could be a stereo delay, because you write 2 audiostreams independently into the delay ram. both are delayed for 1 sample. "w1" means the write- and "r1" the read- adress of audiostream nr. 1. the delay ram circulates through it's predifined range and picks up a sample when it meets the read adress. this process corresponds to the command- page processing speed of the dsp 48000 times each second.

Russ · May 7, 2011, 11:23 AM

Yes (assuming that those addresses are not changed later by DSP instructions used by the plugin).

JoshuaChang · May 7, 2011, 11:49 AM

but it use 4sample tram size?
each delay line will use (actual delay size+1) tram size?

Max M. · May 7, 2011, 12:12 PM

Quote:

but it use 4sample tram size?

Well, yep. Technically it could be:

Code:

idelay write w1 at 0;
idelay read  r1 at 1;
idelay write w2 at 1;
idelay read  r2 at 2;
itramsize          2;

But this would be a harder to understand code (like it would require always reading from r1 *before* writing to w2 - while it's not), so it's just easier/more readable/safer to keep some gaps between lines.

In fact, internally - it's even more complicated due to xtram latency and dependencies on a position of the corresponding code in the code memory. There're some hidden flags and complicated conditions to set them which may to work not always correct due to lack of documentation - so it's just... yes, "safer" to use some gaps just to keep yourself away from unexpected mysterious bugs/issues.

stylus02 · May 7, 2011, 12:35 PM

what max showed is a "trick" to save delay memory. you have to read out the value at r1 and copy it to a new register e.g. macs new, r1, 0, 0. then you reuse the r1 adress for a new write process w2. this saves 1 memory place (of thousands), but its not really usefull and unsave. i hope it works that way max.

JoshuaChang · May 7, 2011, 12:37 PM

thanks for the technical reply, now i understand the delayline workflow!

Max M. · May 7, 2011, 12:45 PM

Quote:

i hope it works that way max.

hehe, typically it works (or at least it worked at times we tested it - when E. had implemented those hidden tram alignment flags in kX) - but you never know if it will work under every possible condition :)
So I actually (well, depending on my mood :) prefer to have even 2 sample gaps in between lines (including the end of plugin lines). Well, I guess you don't really need to bother with those things in context of C simulation (there're different problems anyway) - but in case of an abstract "fx8010" machine - yes, the {0,1,1,2}->2 thing should be valid code.

stylus02 · May 7, 2011, 12:50 PM

now, let me ask: does anyone work on a fx8010 simulator (function set) in c/c++?

May 7, 2011, 08:54 AM	#1
JoshuaChang HardwareHeaven Lover Join Date: Mar 2003 Posts: 127 Rep Power: 0	a small question about emu10k's delay line~ does the code below means 2 x 1sample delay? Code: idelay write w1 at 0x0; idelay read r1 at 0x1; idelay write w2 at 0x2; idelay read r2 at 0x3; __________________ lumetools.blogspot.com

May 7, 2011, 11:23 AM	#3
Russ HardwareHeaven Extreme Member Join Date: Jan 2005 Posts: 5,563 Rep Power: 62	Re: a small question about emu10k's delay line~ Yes (assuming that those addresses are not changed later by DSP instructions used by the plugin).

May 7, 2011, 11:49 AM	Thread Starter #4
JoshuaChang HardwareHeaven Lover Join Date: Mar 2003 Posts: 127 Rep Power: 0	Re: a small question about emu10k's delay line~ but it use 4sample tram size? each delay line will use (actual delay size+1) tram size? __________________ lumetools.blogspot.com

May 7, 2011, 12:35 PM	#6
stylus02 HardwareHeaven Addict Join Date: Jan 2008 Location: germany, sb0090 Posts: 259 Rep Power: 0	Re: a small question about emu10k's delay line~ what max showed is a "trick" to save delay memory. you have to read out the value at r1 and copy it to a new register e.g. macs new, r1, 0, 0. then you reuse the r1 adress for a new write process w2. this saves 1 memory place (of thousands), but its not really usefull and unsave. i hope it works that way max. __________________ http://kxm.dyndns.org

May 7, 2011, 12:37 PM	Thread Starter #7
JoshuaChang HardwareHeaven Lover Join Date: Mar 2003 Posts: 127 Rep Power: 0	Re: Îòâåò: a small question about emu10k's delay line~ thanks for the technical reply, now i understand the delayline workflow! __________________ lumetools.blogspot.com

May 7, 2011, 12:50 PM	#9
stylus02 HardwareHeaven Addict Join Date: Jan 2008 Location: germany, sb0090 Posts: 259 Rep Power: 0	Re: a small question about emu10k's delay line~ now, let me ask: does anyone work on a fx8010 simulator (function set) in c/c++? __________________ http://kxm.dyndns.org