Was I supposed to know delay lines are only 16 bits wide?

[LLR3]

Does it say somewhere in the manual that the signal resolution for the delay modules (on G2) should be less than the normal 24 bits?

I was trying to make a signal packer in order to turn one 96 kHz 24-bit audio delay line into two 48 kHz 16-bit audio and sixteen 6 kHz 8-bit control signals (just to waste some spare DSP cycles of course!), but unfortunately I ran into the fact that the delay lines on G2 are only 16 bits wide...

Well, maybe you already knew all this, but in case you didn´t, you can try it yourself by attenuating a clock pulse signal so that it´s just barely strong enough to trigger something (ANYTHING above zero equals to condition true on G2 logic operations). Next try putting a delay module in between the attenuated clock signal and the trigger destination. Now, try re-amplifying the clock signal until it becomes visible on the trigger led again. And then count the bits!

I´ve yet to try the same on G1. (Though I´m quite sure it couldn´t possibly have gone unnoticed this long!)

*Llare

[Chet]

That's an interesting observation.

I've noticed on percussive Karplus-Strong patches (which use delay lines as tuned resonators) that there's some really nasty harmonic distortion audible as the sound dies away. I've never been able to figure out why. 16-bit resolution would explain it.

I think the G1 has 24-bit resolution on the delays, because I don't hear that distortion on my G1 patches. It was one of the first things I noticed when I began trying to copy some G1 patches to the G2.

Bummer.

[mosc]

If this is indeed the case, then maybe that is a good item for the Wish List, a button on the delay modules to put them into 24 bit mode. It will give you less time delay, but in some cases it may be worth the tradeoff, eh?

[cebec]

oh yeah, let's make this a wish list item... i tend to use short delays on the G2, anyway, and leave any longer delay/reverb duties for native or outboard processing...

hopefully it's possible!

[LLR3]

I think the delay lines should have 24-, 16-, 8-, and 1-bit modes available for highest quality audio, "long play" audio, control and logic signals accordingly.

There should also be an option to use lower sampling rates AND a switch for turning the interpolation on or off.

*Llare

[Chet]

Perhaps the external RAM is only 16 bits wide.

Even so, it's solvable. The delay module could split the 24-bit signal into two paths, send them through two parallel delays, and then combine them back together again. It would require twice as much RAM, but it could be done.

I'd bet the comb filters work this way, too. I've heard the same distortion on them when the signal level gets low.

[cebec]

geez, what else is or could be running with 16 bit depth and why?

[Kassen]

[cebec]

the digitizer only goes to 12 bits or is 24 bits when OFF

[jksuperstar]

I also support the idea of smaller bit width delays, such as 8b and 1b. That opens up the system for logic and midi delays, which would help alot with moving things from the audio signal chain to the control signal chains to improve DSP usage.


Just to do the math, how much RAM is hanging off each DSP? 96k * 3 = 288k samples/sec, * 2.7sec = 778kB of RAM. We might be able to calculate if Clavia is intending to use 24bit data in the RAMs, and there is simply a bug in how they juggle 3bytes at a time efficiently through the 2byte wide RAM, or if they really are only using 16bits at a time.

[Kassen]

[Rob]

[cebec]

What else is 16 bit, then? The G2 manual and Clavia website says the audio signal path is 24 bits... Are they incorrect, then? Apparently, the signal is getting degraded at various points that we are unaware of.

[Rob]

[Chet]

Would the delay lines in the comb filters be 16 bits, too? I think I hear the same artifacts on them when I use low signal levels.

[Kassen]

[ian-s]

The missing 8 bits are the LSB's, so a signal with a peak of +-64 (unity) is represented by only 14 bits. Scale the signal by 4 and all 16 bits are used but you are on the threshold of clipping.

This is a bit disappointing but I can't say I would have known if it hadn't been pointed out.

[Kassen]

Ah! check, that´s not as bad as I imagined it could be then.

Thanks for that. I probably wouldn´t have noticed either but I hear you can get realy clever if you ask enough questions. "curiocity makes the cat" as the old Russian saying goes.

;¬)

[mosc]

[jksuperstar]

I'm still not convinced that because the memory is 16bit wide, that the delay *has* to be 16bits wide.

Rob, can you confirm this was Clavia's intent (or mistake)?

[Kassen]

[davep]

That must be founded on a translation error. When you think about it it makes no sense that way. (well, perhaps with the dmca it does in the u.s.....)

I love funny translations! When I was in China recently I was in a park that had about 20 large signs - not painted on wood, but actually CARVED in STONE that said: "We can't stand the sight of mattress fragrant grass".

I assume it was supposed to be telling me to stay off the grass. I think the funniest part is the fact that someone went to the trouble of carving 20 signs into large rocks without checking the translation.

BTW, my favorite 'colorful' saying is from Romania and translates into: "A dog could walk between her legs with a stick in his mouth".

[Rob]

[Rob]