| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1157.1 | My 2� (what the IRS will leave me, if I'm lucky) | ECADSR::SHERMAN | No, Rodney. That's *old* science! ... | Tue Jan 19 1988 12:08 | 16 |
| Seems to me that the CZ-101 is a machine that has relatively low
bandwidth and low resolution (8 bits). The more expensive machines
have higher resolution and about the same bandwidth. So, the aliasing
tends to disappear in the low resolution of the CZ-101. To get
rid of it at higher resolutions (more bits), it would be necessary to
increase the bandwidth (increase the sampling frequencies or digital
synthesis frequencies, for example). For this reason, the TZ is
probably really quiet because while the resolution went to 12 bits,
the internals have more accurate synthesis (reflected by the many
frequencies available in the TZ that are not available in other
4 op machines). My guess is that if they had added 12-bit resolution
to the TZ, but left the frequency adjustments the same as in other
machines, they would have wound up with more chance of aliasing.
Just a hunch ...
Steve
|
| 1157.2 | The problem is more difficult algorithms. | BOLT::BAILEY | Steph Bailey | Wed Jan 20 1988 13:50 | 26 |
| I also feel the premise that the CZ-101 is quiet because it is simple
is accurate. It probably doesn't have a large frequency bandwidth,
so the can put an aggressive filter on the output.
Another source of noise in most of my digital synths is the poorly
shielded analog components in the output stage. My K5m has a buzz
which is audible both through the output jacks, and through the
air, when the output is turned off. This is probably a switching
power-supply introducing noise in the output. All this high-speed
digital stuff is starting to consume power, and radiate all sorts
of junk.
The TX81Z has synthesis comparable to the DX7 (12-bit), and the
DX7 sound is noisy, to my ear. I believe the aliasing noise in
the DX7 comes from not being able to ``calculate'' the numbers fast
enough, since the nominal frequency of the D/A converter is 50KHz,
which could provide a very quiet output (listen to a CD player).
The DX voices definitely have frequency limitations which are far
below the 25KHz ceiling imposed by the converter.
The problem is generating the values for 16 (or 32!) simultaneous
voices from some obscure synthesis algorithm. Tom Janzen's
note about synthesis on an 11/23 give a good sense of perspective.
Steph
|
| 1157.3 | Not the same! | JON::ROSS | we is wockin'.... | Wed Jan 20 1988 14:14 | 33 |
| no no no no..... TWO different things here!
Two different noises:
resolution: you have so many bits to represent a wave with.
The fewer number of bits with a large (enough) signal(and
you want to get the peak to be all bits 'on') leaves a gap
for the signal between any ONE bit changing. Um, so, if you
had a wave at low volume (what happens on the tail, eh?) it
can be as bad as having a resolution of 1 bit bouncing between
0 and 1.....BUT THATS A SQUARE WAVE, AND YOUR REAL WAVE IS NO
DOUBT MORE COMPLEX.
Thats the grit that you hear at low volumes. More bits/lower grit.
Also, using non-linear encoding changes everything...
Aliasing: With the FM units, what seems to happen is that the
MULTIPLIED or env. gen. modulated (or both) waves (even if sine)
can result in frequencies that violate the nyquist rule. The
filter isnt perfect, so some UNWANTED FREQUENCIES leak thru. This
sounds like 'ring modulation'.
One left: (see? I lied.) Its also true that at high levels, the
operator sine waves are again only represented by so many bits,
so theres a lack of resolution, and some grundge....
You too can set up your FM units to generate the above and HEAR
what Im talking about.
have fun.
ron
|
| 1157.4 | Grunge As Standard Feature???? | AQUA::ROST | Lizard King or Bozo Dionysius? | Thu Jun 02 1988 15:45 | 28 |
|
Re: reply to 1300.8 which discusses noise in the Kawai K1
Since I first entered this note, all I seem to hear about with new
digital synths is "noise". Every under $2000 machine seems to be
the target of complaints about noise.
What's going on here?
Is noise now acceptable to the masses of musicians?
Maybe I'm being alarmist, but a low-cost synth that is hobbled with
bad noise is not of much interest to me. I heard a few patches
on the Yamaha FB-01 when I borrowed one that shocked me as how
*blatant* the noise was. OK, for $300 what do I expect.
In particular, the fact that so many people seem to have found these
low-priced machines noisy means that the noise has to be pretty
obvious or else you wouldn't hear about it so much.
Maybe in this headlong digital revolution we're taking a step
backwards?? Cheap analog machines never had these kind of problems....
Geez, now I'm becoming an old fart 8^) 8^) 8^) 8^) 8^) 8^)
|
| 1157.5 | Noise is a factor of price. | DYO780::SCHAFER | Brad - DTN 433-2408 | Thu Jun 02 1988 16:39 | 18 |
| It seems that noise is more of a low-midrange machines problem. I've
listened to lots of the new units - everything from the D-110 to the
K1m to the TX81z to the D-50 ... and they ALL have some level of noise
evident.
(Strangely enough, the digital grunge that everyone dinged Yamaha for
with the old DX7 architecture machines is NOTHING compared to this new
stuff. Compared to a D-50, my TX7s are dead quiet.)
I think this boils down to another case of "you get what you pay for".
I listened to an Emu III a couple months ago. No noise whatsoever. And
it only cost $13k. Bottom line - don't expect an $800 machine to act
like an $8000 machine.
-b
PS I know I sound like a broken record, but I have yet to hear ANY digital
synth as quiet as my OB-Xa.
|
| 1157.6 | Would oversampling help? | BARDIC::RAVAN | | Tue Jun 14 1988 14:37 | 5 |
| Could the digital synth manufacturers learn a lesson from the
CD folks and oversample the 8 bits to 12, then put a long slope
filter on the output to clean up the quantization noise?
-jim
|
| 1157.7 | How About 4 Times Oversampling Filter Plus Dithering | DRUMS::FEHSKENS | | Tue Jun 14 1988 18:16 | 14 |
| Oversampling 8 bits to 12 requires 16-times oversampling (4 bits
worth), which may have some heavy consequences for the design.
Most CD oversampling is 2 bits worth, 4-times oversampling.
Making any part of a design go an order of magnitude plus a bit faster
is something of a job. But it would mean easier to design and implement
filters, with less phase shift and ringing. Sounds like a tradeoff
they ought to explore.
Another option to decrease quantization noise is dithering, which
is a filter independent solution. Maybe they're already using it,
synths designs seem top be pretty proprietary.
len.
|