Author |
Message |
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Fri Jan 26, 2007 6:54 am Post subject:
dlyeight module Subject description: is setting '4' in 5ms mode 1 sample / tap?? |
|
|
While I am busy doing other things (looks over shoulder to check for suspicious colleagues), I wondered some things about FIR filters. I did some experiments with 'em before, but always used a clocked delay module with a 48kHz clock to get access to n-1, n-2, n-3, etc samples. I did not use the dlyeight module before, as the display and manual do not give any sample-accurate info, and an initial experiment didn't work out.
I ran some experiments on the demo software, using VU meters for monitoring and some simple s/h stuff to see how the dlyeight module behaves when clocked with 48kHz or 12kHz rates (see attached demo patch; don't turn up sound or activate outputs for your speakers sake).
When clocking at 48k and taking alternating outputs summed to mixers hardpanned left & right, I get 'high signal' / silence (left / right) when the delay is set to 0.02 ms (value 4). The same happens for 12k clocking when taking blocks of 4 consecutive taps mixed together and panned left / right. Can we safely assume that this delay setting is really a '1 sample per tap delay'? I noticed that the 48k experiment yields similar results at values 8, 12, 16 etc, but the 12k setting does not have this exact replication of results; am I correct to think this means it is really a 1 sample delay and not some multiple-sample delay setting that yields the same output?
My thinking is not too 'hi-fi' due to a nasty sinusitus+headache; is there anyone that understands my blabbering and can confirm or correct my thoughts?
Description: |
File to test exact 1 sample delay setting in dlyeight module. Needs some repatching to test both 12 and 48k clocks. DON'T turn up sound; it is for testing with VU meters (in the editor) only!! Don't hold me responsible for breaking any speakers if you do |
|
Download |
Filename: |
dl8test.pch2 |
Filesize: |
1.02 KB |
Downloaded: |
1480 Time(s) |
_________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
blue hell
Site Admin
Joined: Apr 03, 2004 Posts: 24079 Location: The Netherlands, Enschede
Audio files: 278
G2 patch files: 320
|
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Sat Jan 27, 2007 1:37 am Post subject:
|
|
|
Yes, I figured that out about the 0.01 ms per sample, but maybe it's just that the display of the module is not accurate at these low rates. I also noticed that for the lowest values, it appears that single value steps do not alter the delay setting, but this happens in blocks of 4 values.
I tried to make some experimental filters or chorus-like stuff (none of which sounded good btw) and noticed that in all cases the sound did not chang at these values: 0-3; 4-7; 8-11; 12-15; 16-19. This was similar in my clock experiments. I'm not sure about it not being at 96k, but I do think it implies there's no sample interpolation.
Will have a look at your patch, Blue Hell, and see if I can figure it out some more. Nice to hear that you find the same results.... _________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Sat Jan 27, 2007 2:56 am Post subject:
|
|
|
The clock and the flipflop operate both at 96k; the resulting squarewave has a frequency of 48k, which is brought down to 375Hz by the 128-divider (48k/128=375). I compared the frequency to an osc, and it is indeed 375Hz. So proves imo the 96k operation.
Now I guess there are two options: the display is wrong, or the delay works in blocks of two samples. These value-blocks that I mentioned earlier apply over the whole knob-range, leaving 32 functional steps that in your patch of which 16 steps result in a signal, and 16 result in silence. Isn't this just weird? For the display setting to be correct, value 127 or .66ms should be 64 samples delay.
Correction of the above statement: there is of course the third option of my brains making some crucial error here.... will report back... _________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Sat Jan 27, 2007 4:32 am Post subject:
|
|
|
Well, tested further a bit. I really think the display is off by a factor 2, so maybe this thread should be moved to the buglist.
In this testpatch, I use the 48k clock to clock a sequencer with 4 different values (lenght=4 and reset is at 1/4 of clock), patch the output of the seq into the delay, and readout the delay outputs with s/h's at the reset-moment of the sequencer.
The outputs show 4 pairs of the 4 sequencer values, which corresponds to a 96k reading of a 48k sequence.
I say this is all a bit weird, but my conclusion is that values 0-3 correspond to no delay (all taps reflect the input value), 4-7 to 1 sample, 8-11 to 2 samples, etc.
Hence, the max delay is 32 samples per tap, or 2.6ms total delay if I'm correct.
Description: |
Another visual test patch for the delightful dlyeight module at it's lowest settings |
|
Download |
Filename: |
dl8test_02.pch2 |
Filesize: |
1.53 KB |
Downloaded: |
1420 Time(s) |
_________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
|
Back to top
|
|
|
blue hell
Site Admin
Joined: Apr 03, 2004 Posts: 24079 Location: The Netherlands, Enschede
Audio files: 278
G2 patch files: 320
|
Posted: Sat Jan 27, 2007 6:04 am Post subject:
|
|
|
Ah, this is starting to get interesting
Maybe the delay does work with 16 bit samples then and not 8 ? Here we go again ....
Will try the patches later. _________________ Jan
also .. could someone please turn down the thermostat a bit.
|
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Sat Jan 27, 2007 6:38 am Post subject:
|
|
|
Blue Hell wrote: | Ah, this is starting to get interesting
Maybe the delay does work with 16 bit samples then and not 8 ? Here we go again ....
|
You mean 32 bit instead of 16, right? Else _________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
blue hell
Site Admin
Joined: Apr 03, 2004 Posts: 24079 Location: The Netherlands, Enschede
Audio files: 278
G2 patch files: 320
|
Posted: Sat Jan 27, 2007 6:40 am Post subject:
|
|
|
Fozzie wrote: | Blue Hell wrote: | Ah, this is starting to get interesting
Maybe the delay does work with 16 bit samples then and not 8 ? Here we go again ....
|
You mean 32 bit instead of 16, right? Else |
still waking up _________________ Jan
also .. could someone please turn down the thermostat a bit.
|
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Sat Jan 27, 2007 7:12 am Post subject:
|
|
|
I've tried the precision of the dlyeight outputs by feeding it an very low attenuated logic high (attenuated to the amount that it *just* triggers a trigger in on a drum osc).
The attenuation threshold for direct triggering is 0.01*0.01*0.01*0.5 (gains on levelamp modules). If the dlyeight module is used, the threshold is twice as high; the last levelamp should be >=1.0.
Here's where it starts to get weird: for a mod-delay the threshold is only 0.01*0.01*0.29 (!), as for the other standard delay modules. The clocked delay however is the same as direct triggering. For the comb filter, it is 0.01*0.01*0.01*1.5
I thought there was just one difference; either 24 bit or 16 bit. Seems there are other bitdepths as well
Edit: just realized, that probably the dlyeight module might be the only long-delay module at 24 bit minus one (headroom?) bit, right?
edit 2: that would be kinda ironic if it is so, as this is a non-modulatible delay, so perfectly unsuited for any PM _________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
Rob
Joined: Mar 29, 2004 Posts: 580 Location: The Hague/Netherlands/EC
G2 patch files: 109
|
Posted: Mon Feb 05, 2007 6:12 pm Post subject:
|
|
|
The DlyEight module is off by 4 ticks on the scale, the knob positions 0 to 3 simply "do not work" and the input is passed unaltered on all eight outputs, so all delays are zero. At position 4 to 7 the eight delay outputs give delays of 1 to 8 samples. Then at knob position 8 to 11 the delays are from 2 to 16 samples with gaps of a sample. At knob position 12 to 15 they are 3 to 24 samples with gaps of 3 samples. Etc.
So, the display is utterly incorrect, as there are actually only 32 ranges on the knob and not 128 as could be expected from the display.
I guess one has to be Swedish and be used to lock oneself up during long dark winters to come up with the actual function of this particular module, and then, when spring finally comes, forget how it actually works and put total nonsense in the display. |
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
Posted: Tue Feb 06, 2007 9:13 am Post subject:
|
|
|
Rob wrote: | I guess one has to be Swedish and be used to lock oneself up during long dark winters to come up with the actual function of this particular module, and then, when spring finally comes, forget how it actually works and put total nonsense in the display. |
Thanks for confirming my suspicions. Do you also have some inside intel on the bit-depth? Regarding measured delay lenght, memory usage and low signal testing, the results really suggest this is a kind of double-precision delay module (24bit?). _________________ Spinning at ~0.0000115740740741 Hz |
|
Back to top
|
|
|
dasz
Joined: Oct 16, 2004 Posts: 1644 Location: victoria, canada
Audio files: 29
G2 patch files: 56
|
Posted: Tue Feb 06, 2007 9:59 am Post subject:
|
|
|
This delay display is indeed disconcerting. A tick on a knob is meant to reflect some change otherwise it should be a multistage switch (or click-knob with n settings).
I really dislike the delay (& LFO) time knob when it is in clk mode. You turn it for a while before a change happens. A way better implementation on Clavia's part would have been something along these lines -- a 1/n or n/1 button, a multiplier knob and then the delay knob would delay in terms of 1/n or n/1 (where n = ticks * multiplier). The max clk setting on LFO's is waay to short.
Btw Rob, your website link is not working. this one is http://www.xs4all.nl/~rhordijk
And nice big g2 gif! where did you get it from?
/Dasz |
|
Back to top
|
|
|
Rob
Joined: Mar 29, 2004 Posts: 580 Location: The Hague/Netherlands/EC
G2 patch files: 109
|
Posted: Tue Feb 06, 2007 1:40 pm Post subject:
|
|
|
Fozzie wrote: | Thanks for confirming my suspicions. Do you also have some inside intel on the bit-depth? Regarding measured delay lenght, memory usage and low signal testing, the results really suggest this is a kind of double-precision delay module (24bit?). |
As far as I know it uses 16 bit and 512 memory locations in the 5msec mode. The DlyEight does not use interpolation, while the modulatable delays do use interpolation in the 5 msec mode.
Note that the code for the different max delaytime settings differs. So for other max delay settings the display might actually be correct. |
|
Back to top
|
|
|
Rob
Joined: Mar 29, 2004 Posts: 580 Location: The Hague/Netherlands/EC
G2 patch files: 109
|
Posted: Tue Feb 06, 2007 1:53 pm Post subject:
|
|
|
Perhaps you used an old link that leads to nowhere. This one should work:
Rob Hordijk's G2 Tutorial
dasz wrote: |
And nice big g2 gif! where did you get it from?
/Dasz |
Did you ever take a good look at the front of your G2?
/Rob |
|
Back to top
|
|
|
Fozzie
Joined: Jun 04, 2004 Posts: 875 Location: Near Wageningen, the Netherlands
Audio files: 8
G2 patch files: 49
|
|
Back to top
|
|
|
dasz
Joined: Oct 16, 2004 Posts: 1644 Location: victoria, canada
Audio files: 29
G2 patch files: 56
|
Posted: Tue Feb 06, 2007 2:24 pm Post subject:
|
|
|
Rob, by website link I mean the e-m profile link (the www button on every post of yours -- that's not working).
/Dasz |
|
Back to top
|
|
|
Rob
Joined: Mar 29, 2004 Posts: 580 Location: The Hague/Netherlands/EC
G2 patch files: 109
|
Posted: Tue Feb 06, 2007 2:47 pm Post subject:
|
|
|
dasz wrote: | Rob, by website link I mean the e-m profile link (the www button on every post of yours -- that's not working).
/Dasz |
You're right, my mistake! Forgot to update that old link.
Fixed it!
Thanks! |
|
Back to top
|
|
|
dasz
Joined: Oct 16, 2004 Posts: 1644 Location: victoria, canada
Audio files: 29
G2 patch files: 56
|
Posted: Wed Feb 07, 2007 11:35 am Post subject:
|
|
|
No worries, Rob. I was on a clicking spree when I found it didn't work.
/Dasz |
|
Back to top
|
|
|
|