Author |
Message |
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
|
Back to top
|
|
|
mosc
Site Admin
Joined: Jan 31, 2003 Posts: 18197 Location: Durham, NC
Audio files: 212
G2 patch files: 60
|
Posted: Tue May 03, 2005 12:16 pm Post subject:
Re: Bug in delay lines? |
|
|
Chet wrote: |
I'm using v1.30. Could someone with 1.32 try it and see if it's fixed? |
I'm running 1.32 and it does exactly as you describe, but the noise is very low level. _________________ --Howard
my music and other stuff |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Wed May 04, 2005 1:51 am Post subject:
|
|
|
I've noticed these artifacts for quite some time too.
They are, unfortunately, not the only ones. I've also often discovered a strange buildup of subtle high-freq noise within plucked waveguides with long decay times. When playing polyphonically with lots of sustain pedal, it really gets intrusive.
I ascribe most of these phenomena to the "14-bit-ness" of the delays and the inescapable up- and downsampling needed.
Equipping a 24bit modular system with 14-bit delays is -sorry to say this, Clavia- just STUPID!
As a dedicated physical modelling student, these issues have been a big disappointment for me personally. (I've found myself considering the detour into Reaktor quite often lately...)
I think, a "HQ" ("high quality") button on all delay modules, which halves the available delay time while doubling the bit depth, should have high priority in the wishlist! |
|
Back to top
|
|
|
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
Posted: Wed May 04, 2005 5:42 am Post subject:
|
|
|
I agree completely, Tim. It gripes me that my Nord Modular can still make better-sounding models than my G2. Especially plucked ones.
I've tried PM stuff using Reaktor. It can do all the PM modeling that the G2 can do, and it has some advantages: greater precision, and the ability to create packaged "macros". But it took me a lot longer to build stuff in Reaktor than the G2. I lost patience, and now rarely use Reaktor anymore. For me, the G2 editor is just more enjoyable. I can experiment and try "what-if" ideas more quickly.
P.S. I think the HQ button on the delay lines is a great idea. I hope we see it. |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Wed May 04, 2005 3:55 pm Post subject:
|
|
|
Interesting thoughts, Chet.
I've dabbled a bit with Reaktor already (building some simple stereo processing tools) and agree: it's somehow not as "hands on" as the Clavia modulars. Although I wonder if it isn't just about getting more used to the workflow of it -in other words: just a matter of time. Can't judge that myself (yet).
I just can clearly tell that PhM patches on the Reaktor have a far better sound quality than on the G2. The sonic resolution is just audibly higher -logically so, as the delay lines are bound to be better than 14 bits.
So I've already been considering getting Reaktor and a dedicated hardware standalone computer platform. However, as a live performer, stability is paramount, and I've heard many stories about Reaktor not being the most reliable. I guess I'll have to ask around a bit.
However, if the NM Classic performs better for PhM stuff, that also might be worth looking into. Hmm...
...or, of course, the "HQ" button! That would help tremendously. Although I have the impression that G2 updates are more about consolidation rather than innovation, and my gut feeling is that not much more new stuff is going to happen for the G2. Dunno, but it's my gut feeling. I hope I'm wrong. |
|
Back to top
|
|
|
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
Posted: Wed May 04, 2005 5:53 pm Post subject:
|
|
|
I think Reaktor 4 is a good deal. For a modest sum, you get an amazing synthesizer construction kit. Regarding reliability, my experience with Reaktor 3 was awful. But Reaktor 4 is reportedly better, and I haven't spent enough time with it to form my own opinion.
In my experience, Reaktor needs a lot of computer horsepower. It was rarely an issue for me, since most of my patches are monophonic. But if you'll be programming polyphonic pads, you'll want the fastest cpu you can afford. A typical PC still can't match the 8 dsp chips in a G2, imo. A 5-rank pipe organ I made for the G2 couldn't be duplicated in my 1.5GHz laptop.
Actually, your comments have stimulated my interest in Reaktor again! Maybe I'll put some patches together this weekend! |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Thu May 05, 2005 2:32 am Post subject:
|
|
|
Chet wrote: | Regarding reliability, my experience with Reaktor 3 was awful. But Reaktor 4 is reportedly better, and I haven't spent enough time with it to form my own opinion. |
I've dabbled a bit with R4, and it hung up on me a few times. But I was patching while using it as an insert effect in Cubase, which (according to a friend who is an VST expert and R user) is asking for trouble. If you use it as a standalone app, it reportedly is much more reliable.
Quote: | A typical PC still can't match the 8 dsp chips in a G2, imo. A 5-rank pipe organ I made for the G2 couldn't be duplicated in my 1.5GHz laptop. |
...okay.
I was thinking about your comment concerning the NM Classic and the quality of its delay lines. I only know that they can only be very short, but it seems to be sufficient for making waveguides.
I was wondering though: what is, roughly, the polyphony you can achieve on the Classic? Let's say, for a single-waveguide plucked model, no fancy frills ...? Can you give me a rough figure?
Also: According to Clavias comparison chart, the Classic has an internal resolution of 18 bits (16 bits plus 2 headroom, I presume). If this also applies to the delay lines, then that is a 24dB better S/N ratio compared to the G2. That, imo, is a considerable difference. |
|
Back to top
|
|
|
ian-s
Joined: Apr 01, 2004 Posts: 2669 Location: Auckland, New Zealand
Audio files: 42
G2 patch files: 626
|
Posted: Thu May 05, 2005 3:28 am Post subject:
|
|
|
I guess for the purpose of PM, where your feedback gain needs pretty tight control, the G2 delays only use 14 bits but technically they would be 16 bit due to the 12dB headroom implied by Clavia Units. So probably the same as the classic.
Actually, has anyone definitively confirmed the 16bit delay word size? I know Rob said the memory chips are 16bit types, but has anyone done the memsize/samplerate == 2.7 seconds calculation, or just got Clavia to confess?
I’m still hoping that the odd behaviours reported could just be down to coding bugs.
|
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Thu May 05, 2005 4:21 am Post subject:
|
|
|
A revealing phenomenon was described at:
http://www.electro-music.com/forum/topic-4838.html
If the Classic has the same delay line bit depth as the G2, I wonder why -according to Chet- they sound better. Maybe there is something else involved too.
Anyway, I've been wading through the NI Reaktor forum a bit, and the magnitude of search results when using the keywords "reaktor crashes" is rather disconcerting. After all, I'm a live performer. |
|
Back to top
|
|
|
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
Posted: Thu May 05, 2005 4:32 am Post subject:
|
|
|
tim wrote: |
I was thinking about your comment concerning the NM Classic and the quality of its delay lines. I only know that they can only be very short, but it seems to be sufficient for making waveguides.
I was wondering though: what is, roughly, the polyphony you can achieve on the Classic? Let's say, for a single-waveguide plucked model, no fancy frills ...? Can you give me a rough figure?
Also: According to Clavias comparison chart, the Classic has an internal resolution of 18 bits (16 bits plus 2 headroom, I presume). If this also applies to the delay lines, then that is a 24dB better S/N ratio compared to the G2. That, imo, is a considerable difference. |
In the Classic, I could get 2 notes per dsp with the very simplest plucked string and blown pipe models. But a more realistic expectation is one note per dsp. A couple of models couldn't fit into a single dsp, so I moved the final eq and post-processing into another slot, and patched one of the Classic outputs back into its input. Yuck. Bass instruments, like bass guitars or tubas, pretty much can't be done at all, because there isn't enough memory in the Classic to create delays that long.
The Classic has 24-bit internal resolution for everything, since the only memory in the system is the 24-bit-wide memory within the DSP itself. This is trimmed back to 18 bits in the DACs. The internal headroom is 2 bits, so 22 bits is available for most processing, including the delays. I notice a big difference in plucked models. Plucked Classic strings have a clean decay, while plucked G2 strings have terrible artifacts in them. Interestingly, the G2 String Oscillator doesn't have these artifacts. Maybe it has a built-in "HQ" algorithm, or some other kind of internal compensation. |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Thu May 05, 2005 6:45 am Post subject:
|
|
|
Quote: | Bass instruments, like bass guitars or tubas, pretty much can't be done at all, because there isn't enough memory in the Classic to create delays that long. |
Even when you invert the feedback to kick things an octave lower? What's the lowest you can get?
I'm asking this because I just spotted a mint G1 rack... hmm....
Quote: | Interestingly, the G2 String Oscillator doesn't have these artifacts. Maybe it has a built-in "HQ" algorithm, or some other kind of internal compensation. |
Yes -now if only it had blue mod inputs for damping and decay. That would solve many problems.
Interesting observation though: the string osc uses only half the memory of a comb filter. Strange.
The whole thing doesn't make sense to me at all anymore. |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
|
Back to top
|
|
|
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
Posted: Thu May 05, 2005 9:43 am Post subject:
|
|
|
Thanks for the patch! I'll give it a try this evening when I get home. |
|
Back to top
|
|
|
jksuperstar
Joined: Aug 20, 2004 Posts: 2503 Location: Denver
Audio files: 1
G2 patch files: 18
|
Posted: Thu May 05, 2005 9:53 am Post subject:
|
|
|
I don't have my G2 in front of me, and will try to look at these also. But have you repeated these experiments on the G2 Demo? Here, there presumably would be NO memory limitations in size or width (other than what you have installed).
I'd be interested to hear the difference, if any. |
|
Back to top
|
|
|
cebec
Joined: Apr 19, 2004 Posts: 1098 Location: Virginia
Audio files: 3
G2 patch files: 31
|
Posted: Fri May 06, 2005 7:16 am Post subject:
|
|
|
i'm not sure how to test this but i have the G2 demo here. could you tell me, again, what i'm listening for with each module selected? |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Fri May 06, 2005 7:55 am Post subject:
|
|
|
The signal you are hearing is the lowest 8 bits of an audio signal (in this case, an oscillator), passed through various different delay lines. There should be no sonic difference between them (a delay should only delay and not warp), but there obviously is (as described above). It would be interesting to know if the demo version exhibits the same behaviours. |
|
Back to top
|
|
|
cebec
Joined: Apr 19, 2004 Posts: 1098 Location: Virginia
Audio files: 3
G2 patch files: 31
|
Posted: Fri May 06, 2005 8:52 am Post subject:
|
|
|
ok, i'm using demo 1.32.
the orig. signal is very very quiet.
delaysingle attenuates it but it sounds the same otherwise.
delayA is silent
Comb sounds the same as orig. moving the freq. knob causes some weird warping or artifacting at lower freqs
StringOsc sounds the same as orig. moving the freq. knob here, has the same effect as above (see Comb) |
|
Back to top
|
|
|
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
Posted: Fri May 06, 2005 8:04 pm Post subject:
|
|
|
tim wrote: | OK, this issue isn't letting me rest! So I've set up a comprehensive delayline testing patch... |
That's an interesting test.
The fx delay does seem to be a 16-bit delay. That doesn't bother me, actually.
The delayA, the same one I included in my patch above, is the worst-sounding of the bunch. For fun, I modulated the delay length with a constant module, and there were all kinds of artifacts as the delay length was changed to various positions. That's pretty much my original complaint. So although it seems to be a 24-bit delay, there's something squirrelly going on.
I take back what I said earlier about the String Oscillator not having artifacts. It does, although they're not loud. I made a small patch with two KS variations: one with a comb filter, and the other with a string oscillator. I gave them both very long decays. They both sounded pretty similar, and both began buzzing as the decay became very quiet. But they both seem much better than the DelayA module.
So it seems like if I want to make a PM patch, a string oscillator or comb filter is a better choice than the straight-up delay modules. That's not always possible, unfortunately, but it seems like the preferred way to go.
Thanks for the patch, Tim. That was very helpful. |
|
Back to top
|
|
|
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Sat May 07, 2005 12:14 am Post subject:
|
|
|
I mostly use the comb filter (too lazy to patch delay line keyboard tracking ). It's not perfect either. Comb filter-based KS models with long decays have a distinct background noise buildup as the sound fades away. Especially noticable at high polyphony.
Checking out the comb filter on our "LSB delay line test bench" reveals the following strangeness: When you wiggle the filter freq knob (=affect delay length), you should hear some pitch smearing happen -but nothing does. Nada! It's as if the LSB isn't processed but just passed through. This most probably is the cause of this high freq artefacts.
I think its time to hip Clavia to this thread. |
|
Back to top
|
|
|
Chet
Joined: Nov 19, 2004 Posts: 231 Location: Lititz,PA,USA
Audio files: 7
G2 patch files: 35
|
Posted: Mon May 09, 2005 5:00 am Post subject:
|
|
|
I sent the thread to Clavia. Hopefully they'll take a look at it. |
|
Back to top
|
|
|
ian-s
Joined: Apr 01, 2004 Posts: 2669 Location: Auckland, New Zealand
Audio files: 42
G2 patch files: 626
|
Posted: Wed May 25, 2005 7:51 pm Post subject:
|
|
|
I just had my G2 open for the first time, trying to see if I could see anything obvious wrong with the pitch stick which has mysteriously failed 2 months out of warrantee. Anyway, the memory chips, like Rob said are 16 bit, 256K so 262,144/96,000 = 2.7306 seconds. I guess the smaller delays might use the on chip 24 bit memory but I doubt it.
I was thinking of using the delays to sync the G2 output with electro mechanical elements and near real time video effects in installations. But that would be reducing the bit depth by 8 bits, I feel ‘short changed’. I hope Clavia can come up with a full quality version of a couple of the delay modules, for instances where quality matters more than total delay time. |
|
Back to top
|
|
|
jksuperstar
Joined: Aug 20, 2004 Posts: 2503 Location: Denver
Audio files: 1
G2 patch files: 18
|
Posted: Wed May 25, 2005 9:35 pm Post subject:
|
|
|
I think the on-chip memory is the "zero-page" memory used for inter-module (cable) communication. Hence the limit on modules/module outputs. |
|
Back to top
|
|
|
cebec
Joined: Apr 19, 2004 Posts: 1098 Location: Virginia
Audio files: 3
G2 patch files: 31
|
Posted: Thu May 26, 2005 12:50 pm Post subject:
|
|
|
yeah, i hope there's some kind of compromise, as well. all over their site and in the manual, the G2 is touted as having a 24 bit/96 kHz signal path and processing depth/rate. 16 bit delays don't really fit into that equation. so the reverbs and probably a few other modules are 16 bit, as well? i think we need to nail down which ones so users can decide whether or not they want to use them in a patch. |
|
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: Thu May 26, 2005 1:09 pm Post subject:
|
|
|
Could it maybe be the case that the delay laines are 16 bits indeed but that some compression / expansion takes place on in / out puts ?
Or maybe a sample rate conversion takes place on in / out puts ? Maybe that could account for the buzzing (as demonstrated in the first patch in this thread) ?
Jan. |
|
Back to top
|
|
|
|