My theory on GUS clicking
Ranseus (Jim Knepley)
speakers I normally use, so I've got in on a pair on Infinity RefEL's with
a Sony amp. Things are improved, a lot, holy cow, wow, much better.
Anyway...
With this setup, the notorious Gravis UltraClick (TM) made it's habbit of
clicking all too audible, but only on some things. Hmm, I thought, why not
all the time?
It clicked on .MOD files, .669 files, and in Star Control 2. It _DIDN'T_
click with Midi files or anything in SB emulation. Thus, my theory...
I'd be willing to bet that there are 2 calls that can be made to the
Ultrasound to have it produce sound (I'll call them PlayPatch and PlayDigi).
It seems that everything that uses PlayDigi (the .SND files, GUSMOD, etc)
will click. On the other hand, PlayPatch driven things (PLAYMIDI, SBOS)
won't (SBOS does use patches, type SBOS -debug if you don't believe me).
Anyone care to shed some light on this? If I'm right (I hope I am)
then it should be a fairly simple patch to eliminate clicking all together.
It's also possible that there's an attribute to the Gravis supplied patch
files that resist the clicking of other, less refined, patches.
Josh, when do we get the Renaissance (SP?) GUS developer's kit? :)
All the best,
Jim
--
----- kne...@cs.colostate.edu ----------------------------------------------
Jim "Ranseus" Knepley | Avoid rectal-cranial inversion.
Programmer, magician, | "Gun's don't kill, it's those bullet things"
computer geek. |
Gunn, C Stephen
it is also the chip used in the apple IIgs. The Ensoniq Sound Chip has a
bug when you pair oscilators to play via Swap Mode. (This is by far the
most popular method to play the digitized sounds.)
From what everone describes, when you are playing a patch, you are using
the FM synth.. However when playing digitized data, you have to set up
two oscillators to play different parts, and contine putting the rest
of the sound into the dedicated Sound memory. The only way to avoid this
is to use the One-Shot mode, but then you loose much of the flexibility
such as looping.
The bug consists of the ensoniq outputing a value of 128 through the D/A
converter when it swaps from oscillator to the other. Another work around
is to make sure that when the Ensoniq is going to swap, have a value in
the digitized sound that is close to 128. This means that you have to
somewhat engineer your instruments and be careful where they swap, and
what size of resolution you set for each of the oscillators. But if you
are writing a demo or something that wants awesome sound, it is worth
the time to avoid the clicks.
However, for just general sound files, there is not much that can be done
about the anomality. People in the Apple IIgs world have been living with
it for about 6 years now.. Ensoniq is aware of the problem, but last I knew
had no plans to fix the problem since it is minor (in their eyes)
--
| C. Stephen Gunn | GEnie A2PRO.STEVE | __ __
| Freshman Engineering | INET chu...@sage.cc.purdue.edu | (__X__)
| Purdue University | Ask About A2 University on GEnie! | II Infinitum!
Francois Dion
>
>With this setup, the notorious Gravis UltraClick (TM) made it's habbit of
>clicking all too audible, but only on some things. Hmm, I thought, why not
>all the time?
>It clicked on .MOD files, .669 files, and in Star Control 2. It _DIDN'T_
In all these, samples do not have a release enveloppe, and so the last value
on the DAC from an oscillator playing a sample is probably not near zero.
This causes two things. A click if the difference is huge between that value
and the zero which is now fed by the oscillator plus a transient (a high
pitched ping), and if the difference is medium, you could be hearing a
transient. This is the same thing for any DAC be it from an ADCOM or ROTEL
cd player or a Multisound card.
To correct that, you simply need to create a sample with a sampled release
(like 0.005 second drop to zero) or arrange the enveloppe so that the release
fades quickly (but it must fade, which is probably not done in SC2). Doing that
it should resolve the problem completely.
The only other possible source of clicking would be if the program uses for
example 32 channels, so by the time the DOC II goes back to channel 1, 1
zero sample got thru oscillator (corresponding to a lower error rate than
your typical CD player anyway but nastier, since it is a zero). This is
purely speculation on my part, and is based on the swap mode problem of the
DOC I. I do not have the information about that relatively to the DOC II, so
it may (and was supposed to) be corrected.
>I'd be willing to bet that there are 2 calls that can be made to the
>Ultrasound to have it produce sound (I'll call them PlayPatch and PlayDigi).
No, it's the same thing, only the enveloppes are taken from .pat or .sbs
files with midi and sbos, and in MOD files, none is done. The problem is
easy to correct while loading the sample to the GUS, or even before loading it.
>Anyone care to shed some light on this? If I'm right (I hope I am)
>then it should be a fairly simple patch to eliminate clicking all together.
>It's also possible that there's an attribute to the Gravis supplied patch
>files that resist the clicking of other, less refined, patches.
Enveloppe 8)
Ciao,
--
Francois Dion
' _ _ _
CISM (_) (_) _) FM Montreal , Canada Email: CI...@ERE.UMontreal.CA
(_) / . _) 10000 Watts Telephone no: (514) 343-7511
_______________________________________________________________________________
Audio-C-DJ-Fractals-Future-Label-Multimedia-Music-Radio-Rave-Video-VR-Volvo-...
Christopher Daniello
>From what everone describes, when you are playing a patch, you are using
>the FM synth.. However when playing digitized data, you have to set up
Uh, the Gravis Ultrasound does not use FM synthesis. The patches it plays
are digital samples. You can convert a wav file into a patch if you want
even. Are you sure this is the same chip?
:1
--
/// The Fierce Pancake /// "...because you have another eyelid."
/// pan...@ccwf.cc.utexas.edu /// -Galaxie 500
Ranseus (Jim Knepley)
>the FM synth.. However when playing digitized data, you have to set up
>two oscillators to play different parts, and contine putting the rest
>of the sound into the dedicated Sound memory. The only way to avoid this
>is to use the One-Shot mode, but then you loose much of the flexibility
>such as looping.
The problem here is that the GUS has _NO_ FM synth to it. A patch is
a collection if digitized samples.
Also, I've heard that the GUS chip (what's it called now anyway?) is only
based loosely on the Equisonix chip and is an improvement.
Francois Dion
>it is also the chip used in the apple IIgs. The Ensoniq Sound Chip has a
>bug when you pair oscilators to play via Swap Mode. (This is by far the
>most popular method to play the digitized sounds.)
This is what i tought at first when i got the technical note, but this mode
is not needed for the DOC II and so it is not the problem. There was also
another problem (that i talked about in my other post) but i just got email
saying it was corrected. So the problem is the enveloppe.
The DOC II is much more sophisticated than the original DOC chip as found
on the Apple II gs. In it's oem form, it is labelled:
ICS1614 9225 204409 which is a better version than what can be found on the
audiomaster. (Remember, i did bring this up in may 92 and everybody said i
was an heretic, that it couldn't be an Ensoniq DOC II chip as found in
EPS16+ keyboards?)
>From what everone describes, when you are playing a patch, you are using
>the FM synth.. However when playing digitized data, you have to set up
>two oscillators to play different parts, and contine putting the rest
>of the sound into the dedicated Sound memory. The only way to avoid this
No FM synth here.
>converter when it swaps from oscillator to the other. Another work around
>is to make sure that when the Ensoniq is going to swap, have a value in
>the digitized sound that is close to 128. This means that you have to
You are close to the answer, but it's not for the good reason. In mod files,
you load all the samples in the GUS memory. You playback each sample on
a different channel (4 used for mod files), and the problem arise at the
end of the sample. See my other post.
>somewhat engineer your instruments and be careful where they swap, and
>what size of resolution you set for each of the oscillators. But if you
>are writing a demo or something that wants awesome sound, it is worth
>the time to avoid the clicks.
Just do a release on the 0.005 second at the end.
>However, for just general sound files, there is not much that can be done
>about the anomality. People in the Apple IIgs world have been living with
>it for about 6 years now.. Ensoniq is aware of the problem, but last I knew
>had no plans to fix the problem since it is minor (in their eyes)
I dont seem to get that on my EPS-16+ 8). The DOC II is much improved over the
old DOC I.
Francois Dion
>Uh, the Gravis Ultrasound does not use FM synthesis. The patches it plays
>are digital samples. You can convert a wav file into a patch if you want
>even. Are you sure this is the same chip?
It's not the same chip. It's not even the exact same chip as for the
audiomaster card (which is limited to 768K).
David Seah
>it is also the chip used in the apple IIgs. The Ensoniq Sound Chip has a
>bug when you pair oscilators to play via Swap Mode. (This is by far the
>most popular method to play the digitized sounds.)
According to Paul Travers at Forte, the GUS doesn't use a straight
Ensoniq 5503. Apparently, Forte licensed an Ensoniq product and
had their own ASICs made. Whether or not they fixed the swap-mode bug
was unclear.
The clicking I've heard from my GUS seems to occur at the beginning
of playing samples. The swap-mode bug in the IIGS occurs fairly
regularly due to swapping from the odd-to-even oscillator. The limited
sound RAM makes these swaps occur more frequently, so the clicking is
more frequent. Anyone have any low-level stuff on the GUS's GF1
registers?
--
Dave Seah (se...@ee.rochester.edu, AFCD...@aol.com)
That's Rochester, New York!
Gareth Barnard
if the sample starts and ends around 0 then the volume change is minimal.
Some poor samples start and end at a high sample value thus exhibiting the
annoying click. I dont have a gravis so i could just be talking crap.
C'ya
Roo.
Gunn, C Stephen
the other night on the #appleiigs channel on IRC. We decided that although
the chip may still have the anomoality in swap mode.. It had to have had
quite a bit of its architecture changed. (It can access 1Mb now instead of
64K bytes) this alone means that there are 4 bits more allocated to address
where the samples are located in memory. I tend to remember a couple being
free in the IIGS, but Im not sure..
Speaking of that ... could someone post some information about the actuall
control registers for this new Ensoniq Chip. Id like to compare them to the
One that makes my IIgs sound great.
Sorry about the confusion on the FM with the GUS. Somewhere I was lead to
believe that the GUS had a dedicated Frequency Modulation chip. If it
just uses ADSR Envelopes with the digitized playback. That means that it
could easily have support for asif instrument files.. or perhaps even a port
of MIDISynth from the IIgs.
sl...@cc.usu.edu
when the pointer in DRAM is changed and when the volume is moved in a large
jump from one area to another. Solution? There is a technique called volume
ramping that will slide the volume. Very fast.
If anyone wants more information, just email this account with a JOSH: in the
header.
Joshua Jensen
CyberStrike of Renaissance
Temporarily at sl...@cc.usu.edu
Ching Te Tai
>I felt like listening to my GUS on a better set of speakers than the walkman
>speakers I normally use, so I've got in on a pair on Infinity RefEL's with
>a Sony amp. Things are improved, a lot, holy cow, wow, much better.
>Anyway...
>With this setup, the notorious Gravis UltraClick (TM) made it's habbit of
>clicking all too audible, but only on some things. Hmm, I thought, why not
>all the time?
>It clicked on .MOD files, .669 files, and in Star Control 2. It _DIDN'T_
>click with Midi files or anything in SB emulation. Thus, my theory...
>I'd be willing to bet that there are 2 calls that can be made to the
>Ultrasound to have it produce sound (I'll call them PlayPatch and PlayDigi).
>It seems that everything that uses PlayDigi (the .SND files, GUSMOD, etc)
>will click. On the other hand, PlayPatch driven things (PLAYMIDI, SBOS)
>won't (SBOS does use patches, type SBOS -debug if you don't believe me).
>Anyone care to shed some light on this? If I'm right (I hope I am)
From my digital signal processing class... The clicking you described
sounds like the situation when you try to sample at a lower frequency
than necessary. Usually the theory says that you're to sample at 2
times the highest frequence in your input waveform (Nyquest (sp?)), but
it is hard to determine your highest frequency (e.g. there're
other sound sources in the room, like your monitor's high buzz, etc.)
and even if you do, due to timing errors, you need to sample at a somewhat
higher freq. than 2* the max. freq in input.
>then it should be a fairly simple patch to eliminate clicking all together.
>It's also possible that there's an attribute to the Gravis supplied patch
>files that resist the clicking of other, less refined, patches.
Here you are complaining about clicking and boy do I envy you. Ever since I
sold my Sound Blaster card, I'm w/o a sound card at all! (But, I'm
saving up money for the GUS... Can't wait 'til I get one!)
Ian Schmidt
manifested in the DOC and DOC-II chips. (disclaimer: I work with the DOC
on my Apple IIgs, haven't really had a lot of experience with the DOC-II).
The Problem
-----------
When the 550x is used in swap mode, there is a hardware bug where when swapping
from the odd oscillator of a pair back down to the even one, the output sample
briefly falls to the centerline. This (on the original DOC anyhow) doesn't
happen when swapping the other way.
The Fix, as sucessfully used on several Apple IIgs programs
-----------------------------------------------------------
Simply design your swapper such that the last 5 bytes or so of each swap page
are 'de-amplified' (linearly brought closer to $80). Thus, the distance from
the centerline will be minimized, and the click will be inaudible. For the
first 5 bytes of the next page, amplify up in the opposite and equal manner.
This insures that almost no distortion of the sound takes place, and allows
very clean playback.
If you have any questions, etc, I'll answer e-mail from programmers who
need more detail. I'm all for destroying the SoundBlaster before I buy my
PC sometime soon :)
--
Ian Schmidt - Two Meg Software: irsman@{iastate.edu | aol.com}
"I will choose a path that's | Author of AudioZap, MODZap, and more.
clear: I will choose free will" | Co-author of SoundConvert and DeskTracker.
- N. Peart, 1980 | Presenting: the compressed Amiga check _/