Conference napalm::commusic_v1

< Note 939.0 by AKOV75::EATOND "Finally, a piano." >
>	1)  There's a lot of talk about the Nyquist theory.  I've read the
>formula, but how do I apply it in practice?  For one thing, how in the world
>do I detect what the overtone frequencies of my sampled sound really are, in
>order to apply the formula?  Or am I only interested in the fundamental
>frequencies?
Since you're asking the wrong questions, you can just ignore it.
Overtones count, too, BTW.
>	2)  Is the practice of looping simply a matter of trial and error?
>Any tricks to it?
I don't know what real implementations are on current machines, but the
wave should be crossing the same instantaneous pressure in the same direction
at the sample start and end points.
>	3)  On my machine, there is an envelope generator.  How does one use
>this on sampled sounds?
By amplitude-modulating the sample with  the eg.
>
>	Perhaps another approach to this topic might be to create an example
>(an egg sample? - no, no).  I've wanted to try sampling my voice doing some
>oohs or ahs for background in my recordings.  These, obviously will need to
>apply looping to be of any musical use to me.  How do I go about setting it
>up, sampling it, and then editing that sample?
Beware of sampled voice.  It's hollow played lower than the sampled
pitch and munchkiny played higher.  I have an SPX90 which can be played like
a cruddy mono sampler, and an SK1.  Voice should be sampled around the pitch it
will be used.

You can be sure this is oll korrect since I have no experience of the matter
and don't know anything about it.
Tom

Some of my favorite SK1 samples:

a toy triangle, looped; hold down the lowest f, f#,g#,a# for a long time,
you get big bells chiming.

Blow into the mike (the SK1 needs a separate mike to sound half-way OK
(oll korrect).  Play wind clusters, changing pitch to get a change in wind.

duck call .

lip pops.

acoustic pianos 88-note clusters.

Tom

>For one thing, how in the world
>do I detect what the overtone frequencies of my sampled sound really are, in
>order to apply the formula?

    What? you don't have a builtin frequency sensor ? Tom's right, the
    overtones are extremely important for realism.. so it's the top
    end that determines sampling rate. If you have a sweepable parametric 
    EQ or a graphic EQ, start high and drop the fader/sweep until the
    sound dulls out, then bring that back up. Look at the frequency
    you settled at, multiply by 2, and there's your desired sample rate.
    
>2)  Is the practice of looping simply a matter of trial and error?
>Any tricks to it?

    SOME samplers do the hard work for you. SOME samplers have software
    that will find the zero-crossing points and give a click-free loop.
    Some OTHER samplers are a pain in the ass. That's why there's such
    a number of 3rd party firms offering loaded diskettes. IF your 
    sampler isn't one of the 'good guys', you should A) get a Mac and
    visual editing S/W, or B) buy a library somewhere (or copy one from
    store you bought it from) or C) spend endless hours tryin' ta find
    the perfect loop.

>3)  On my machine, there is an envelope generator.  How does one use
>this on sampled sounds?

    The envelope generator is your friend. The EG will restore waveform
    characteristics such as the release shape, in the analog domain,
    allowing you to store LOTS more samples in memory. The sample might
    just be the initial attack and enough for a short loop.. everything
    else (shaping the waveform) can be done with AHDSR parameters.
    
    karl moeller

RE < Note 939.1 by PLDVAX::JANZEN "Tom LMO2/O23 2965421" >

	Thanks for the advice about voices.  I'll ignore the rest.

	Dan

    All of the following is pertinent to the Mirage and probably to
    others...
    
    Nyquist Theorem - says you can't sample a sound with frequencies
    any higher than .5 the rate. Yes you can, but they sound like dog-do.
    Higher frequencies will show up as low frequencies, often with no
    harmonic relationship to the source. This is why a steep input filter
    is needed.
    
    Looping - Trial and error works, but is maddening as you are constantly
    moving both Loop Start and Loop End points in search of that Holiest
    of Sampling Gods, the vaulted "Zero Crossing".
    
    Think of a wave...
    
    .           .          .          .          .          .      
     .         . .        . .        . .        . .        . .
      .       .   .      .   .      .   .      .   .      .   .
    -- .---- .---- .----.---- .----.---- .----.---- .----.---- .------0Point
        .   .       .  .       .  .       .  .       .  .       .
         . .         ..         ..         ..         ..         .
                  A           B                  C
    
    In order to get a smooth loop, both the start and end must be at
    a 0 crossing, like A&B. If you were to try to loop from A to C
    you would get a tick, or worse.
    
    On the Mirage, the sampling frequency must be a multiple of the
    frequency of the note you are sampling. Unfortunately, most samplers
    don't have an infinite variety of rates. So, you are often forced
    to retune your source in order to match it to an available rate.
    On the Mirage A220 must be sharp 25 cents in order to come close.
    This may not hold for all samplers, but the mirage can only loop
    between page boundaries. Consequently, in order to get a true sound
    out, you don't necessarily put a true sound in.
    
    There are 2 kinds of loops. Short and long. Short loops are only
    one page long (256 samples) and produce very static tones. In order
    for these to work, special attention must be payed to the sampling
    rate and the frequencie of the source. Zero crossings MUST fall
    on subsequent page boundaries. Long loops cover many pages. The
    loop end point can be moved off the page boundary in order to find
    0 crossings. All loops must start on page boundaries.
    
    I recently started some heavy sampling on the Mirage and have been
    documenting the process and putting together all kinds of tables,
    graphs, etc. I had originally planned to start a note but you beat
    me to it. I'll be posting stuff here.
    
    Edd

    > ... A) get a Mac and visual editing S/W ...
    
    Mac Schmac.  Get an ST and visual editing software (write your own?).
    
    Just kidding.

    There are programs to do frequency analysis of samples.  Pick a
    range, and it will do a DFT over the range and give you a graph
    of amplitude versus frequency.  I have a home grown one which does
    that for me (when it is not crashing).
    
    Steph

RE < Note 939.3 by CANYON::MOELLER >

>    What? you don't have a builtin frequency sensor ?

	Since I didn't see any graphic facial expressions, can I assume that
	a lot of sampling instruments include this?

	I like that idea of using an EQ and gradually dropping out the high
	end.  That makes a lot of sense.  Of theten or twenty articles I've
	read so far on applying the Nyquist limit to sampling, nobody really 
	mentioned detecting what the highest overtones were.  Do most people 
	just inherantly know how to identify parts of the frequency spectrum?


>    SOME samplers do the hard work for you. SOME samplers have software
>    that will find the zero-crossing points and give a click-free loop.

	The MKS-100 does, by default, auto-looping.  But there have been times
	where it will choose a point to loop that makes it sound like someone
	threw in the sound of a car horn at the end of the attack.  It is
	at this point, I have assumed, that I'd want to go in and select a
	more appropriate loop-point.


	Thanks,

	Dan

    That car horn sound is probably caused by a short loop. Even though
    it's found a zero crossing, it's only looping on part of the entire
    waveform. You're right, find another loop point...
    
    Edd

RE < Note 939.5 by JAWS::COTE "Note stuck? Try Kawai..." >

	Pardon my ignorance, but all this talk of page boundaries and
such confuses me.  I don't remember the manual with my MKS speaking of such 
things.  Is that something described at length in a Mirage's documentation or
is that more of a generic concept as memory relates to any micro-processor?
Does the Mirage reveal to you where page breaks are or is that something you
aquire an intuitive feel for when you've been getting a lot of experience in
sampling?

	You mention the need of an input filter.  On a machine that doesn't 
appear to house one, would a normal graphic or parametric EQ work?

>    On the Mirage, the sampling frequency must be a multiple of the
>    frequency of the note you are sampling. 

	Can you give an example?

	Thanks,

	Dan

> What? you don't have a builtin frequency sensor ? 
    
    .. this was joke .. like, an organic sensor in your ears, see ?
    
    I'm just allergic to little winky faces. sorry.
    
    km

    re: Edd Cote's replies..
    
          Edd, have you actualy DONE any USEABLE samples on your mirage?
    With hours and hours of painful persistance I have not even been
    able to come close to obtaining a sample that I would consider
    using/storing on disk. I do not have any kind of visual editor 
    and get completely confused on the present settings of parameters
    with the single hex display. Also, in my opinion, in order for the
    mirage (8 bit) to sound realistic you really need to MULTI sample.
    This means doing up to 8 wavesamples (such as the accoustic piano).
    This may sound negative, but I LOVE the mirage but would not
    consider anything(right now) except for factory disks, or third
    party disks.

    
    Good Luck
    
             Frank

    Yep, I've done quite a few - mostly pure academia (sampling one
    of my synths to see how close I can get).
    
    I've only saved 2, but that's a function of forgetting to buy blank
    floppies rather than the quality of the samples. I've got a nice
    'flutey' type sample I stole from the MKS-30 and a 'live' sample
    of my Gibson bass guitar with the tape-wound strings. (Stevie K
    will pronounce this a 'garbage' sample. I'll take that as a plus);^)
    ...anyhow the sample NAILED the sound of my bass.
    
    I do my sampling by using the headphone PFL jack on the board, running
    this to the anti-aliasing filter.
    
    Frank, do you have the Mirage Advanced Sampling Guide? If not, Union
    music sells them for $20. Worth every penny just for the tables.
    
    Hint (probably usable for all samplers) - If your using a short
    loop and you get the "car horn" Dan mentioned, listen to the pitch
    of the 'horn' relative to the beginning of the sample. If it's right
    on pitch, great. If not, retune your source in the direction of
    the 'horn'. Try to match it and see what your results are.
    
    Alot of sampling is trial and error and error and error...
    
    Edd

    Oh yeah, I don't have any VES either. I posted the hex->decimal
    conversion chart right next to the mirage though.
    
    Are you using the MASOS disks???? When I first tried, I was just
    writing on top of the last sound in memory, and getting all the
    filter, EG and other parameters for the old sound applied to the
    new one. Are you setting parameter 77 (user-multisampling) to "on"?
    If not, you're always getting the default values...
    
    I multisampled the bass using 2 samples. For the lower I played
    "A" on the low E string. For the upper I played "A" on the A string.
    I wanted to use 4 samples (one for each string), but that would
    have made them too short.
    
    Edd

    > Can you give an example?      Dan Eaton...
    
    Let's try to sample an A at 220 hz. What instrument doesn't matter
    a whole lot for this example...
    
    In order to get a reasonable analog of the source, a sound must
    be sampled at least twice, at the peak of the waveform and at the
    bottom. In order for a good loop, you'll have to have your zero
    crossing on page boundaries.
    
    With 256 samples per page, the highest number of waves you'll be
    able to accurately sample will be 128.
    
    220*128=28,160  - The optimum sample rate for A220. The Mirage does
    not support this rate however, the closest rate it does support
    is 28,571. In order to get a good sample, the pitch of the source
    must be increased in order to coincide with the available rate.
    The following is how I figure it out...
    
    A*128=28,160           (too flat)
    
    Bb*128=29,824          (1 semitone higher)
    
    29,834 - 28,160 = 1674 hz difference or approximately 16.74 hz for
    each cent between them. (Logs be damned, at 220 hz this is close
    enough.) 
    
    28,571 (closest available rate) - 28,160 = 411 hz delta between
    source A and available rate.
    
    411/16.74 = 24.551971 cents difference. This is the amount the A
    will have to be sharpened by in order to get it to line up with
    256 samples/page. 
    
    See? Nuthin to it...
    
    Edd

Karl.  Do you have to go thru all of this to sample on the EMAX or is
-.1 pertainant to a Mirage only?

Todd.

	I think I see what you're saying, Edd.



	Go out and get a lot of factory samples!

RE < Note 939.15 by JAWS::COTE "Note stuck? Try Kawai..." >

	First of all, you never explained pages and page boundaries.  This
doesn't have anything to do with Washington Senators now, does it?

>    In order to get a reasonable analog of the source, a sound must
>    be sampled at least twice, at the peak of the waveform and at the
>    bottom. In order for a good loop, you'll have to have your zero
>    crossing on page boundaries.

	I can understand this part.
    
>    With 256 samples per page, the highest number of waves you'll be
>    able to accurately sample will be 128.

	I have no idea where the 128 comes from, except that it's half the
256.  Are you saying that you need to have two cycles of the wave-form, 
therefore, since you can't fit 2*220 on a page, you have to take half the
value?  I'm lost.
    
>    220*128=28,160  - The optimum sample rate for A220. The Mirage does
>    not support this rate however, the closest rate it does support
>    is 28,571. In order to get a good sample, the pitch of the source
>    must be increased in order to coincide with the available rate.
>    The following is how I figure it out...

	Do you have a chart of values that the Mirage supports?  Is this where
the number 28,571 comes from?

	Why did you subtract 29,834 - 28,160, especially when Bb was 29,824?  Or
was that a typo?  

	Do you think you could write this out in a generic equation?

	Did you get all this stuff from the 'Advanced Sampling Guide'?  Was 
there none of it in the owner's manual?  I wonder, if that is so, how I can find
out the corresponding info for the MKS-100.  You mentioned the value of the
Advanced Sampling Guide Book for the tables alone.  It would seem that if this
kind of stuff has enabled you to improve your sampling, one such book would be
imperative for each sampler.

	How much of the success of your samples do attribute to the use of this
information and how much do you attribute to simple luck?

	Dan

>	First of all, you never explained pages and page boundaries.  This
>doesn't have anything to do with Washington Senators now, does it?

Sorry. The Mirage has 65,536 bytes of memory, broken up into 256 ($FF) 
units called pages. Each page is further broken down into 256 ($FF) units
called samples.


>    In order to get a reasonable analog of the source, a sound must
>    be sampled at least twice, at the peak of the waveform and at the
>    bottom. In order for a good loop, you'll have to have your zero
>    crossing on page boundaries.

>	I can understand this part.
    
>    With 256 samples per page, the highest number of waves you'll be
>    able to accurately sample will be 128.

>	I have no idea where the 128 comes from, except that it's half the
>256.  Are you saying that you need to have two cycles of the wave-form, 
>therefore, since you can't fit 2*220 on a page, you have to take half the
>value?  I'm lost.

Not quite. In order to determine the frequency of the note, the bare 
minimum number of samples *per wave* is 2. Anything less than 2 and
you get low frequency aliasing noise. This is where Nyquist fits in. 

   
>    220*128=28,160  - The optimum sample rate for A220. The Mirage does
>    not support this rate however, the closest rate it does support
>    is 28,571. In order to get a good sample, the pitch of the source
>    must be increased in order to coincide with the available rate.
>    The following is how I figure it out...

>	Do you have a chart of values that the Mirage supports?  Is this where
>the number 28,571 comes from?

Yep, I got a chart.

>	Why did you subtract 29,834 - 28,160, especially when Bb was 29,824?  Or
>was that a typo?

Typo.   29,834 is the correct value.

>	Do you think you could write this out in a generic equation?

Argh...

Let n = a note.
Let n+1 = n raised a semitone. (If n=A, then n+1=Bb)
Let F= The frequency in Hz of n.
Let R= Sampling Rate

So try this...

        (F(n+1)*128)-(F(n)*128)
        -----------------------   = C    (Hz per cent)
                100      

          R-(F(n)*128)
          ------------  = Offset (in cents) to n
              C
      
Let's try it...

Bb= 233 hz, (n+1)
A=220       (n)

       (233*128)-(220*128)       29,834 - 28,160       1674
       -------------------    =  ---------------   =   ----  = 16.74
              100                    100                100

        

        28,571 - 28,160
        ---------------  =   24.551971 cents offset to A-220
            16.74


See??   ;^)    (I dunno who's loonier. You for asking or me for trying to
                explain.)


>	Did you get all this stuff from the 'Advanced Sampling Guide'?  Was 
>there none of it in the owner's manual?  I wonder, if that is so, how I can find
>out the corresponding info for the MKS-100.  You mentioned the value of the
>Advanced Sampling Guide Book for the tables alone.  It would seem that if this
>kind of stuff has enabled you to improve your sampling, one such book would be
>imperative for each sampler.

The rates came from the ASG. The math is mine. The owner's manual is trash.

>	How much of the success of your samples do attribute to the use of this
>information and how much do you attribute to simple luck?

Without the ASG, getting a good Mirage sample was 100% luck. With it, I'd
guess at 70/30 (Science/Luck).

>	Dan

Edd
                       

    Oh yeah, and this doesn't yet take into consideration the .5 octave
    offset (A to E) that you have to set up because the Mirage does
    'unity playback' on E (for the lower half).
    
    If you don't program this offset, no matter what pitch goes in,
    you get it back by playing the E key.
    
    Don't confuse this offset with the one discussed in the previous
    note. They're different....
    
            
    Edd

RE < Note 939.19 by JAWS::COTE "Note stuck? Try Kawai..." >

	Do all sampling instruments have an architechture with 'pages' and page 
boundaries as you described?  Is this differ acording to resolution?  (i.e., 
Mirage is 8 bit companded, MKS-100 is 12-bit linear)  Would one have to get
precise numbers from the manufacturer, or can I guestimate?

>Not quite. In order to determine the frequency of the note, the bare 
>minimum number of samples *per wave* is 2. Anything less than 2 and
>you get low frequency aliasing noise. This is where Nyquist fits in. 

	I think this is basically what I meant, I just said it in a reverse kind
of way.  But now another question comes up.  We said before that the Nyquist
theorem deals with the highest frequency of the overtone series.  Does that
mean that this example of sampling a note at A220 is irrelevant, unless 1) it
has no overtones, or 2) I'm confusing two completely different lines of
sampling terminology?

	Thanks for the equations.  It appears to me the next step I have is
to determine the specifics for my own machine's architecture.  Once I have
that I can adapt your math to the MKS.

	Dan

    Isn't sampling fun?????
    
    I've no idea about other samplers architectures.
    
    The example I provided was for a theoretical sine wave, with no
    overtones. It doesn't take into consideration any of the overtones,
    harmonics, etc, which make up the spectrum of an instrument. But
    the principles still apply, and it's a place to start.
    
    I can post the numbers for the Mirage if anyone wants 'em...
    
    Edd

    
    
    Edd, I have been reading some of this and a lot of doesn't make
    sense.
    
    Are you telling me that in order to get a sample you actually have
    to *detune* the pitch you are sampling in order to get perfect 
    waveforms?
    
    That is, if you want to have exactly 360 x N degrees of wave rather
    than 360 x N + Y where Y is a remainder angle?
    
    How do you then reproduce the pitch *in tune* when you play it back?
    
    No wonder Ensoniq and Roland are pushing commercially-made
    samples....it seems that for users to get a good library of samples
    on their own is pretty difficult.
    
    			Brian_who_can't_even_figure_out_the_SK-1

Hey Cote, 256(10) /= FF(16).  
256(10)=100(16)
Tom

    
     I take all morning trying to explain this to the best of my
     knowledge only to get jumped on...Perhaps Mr. Janzen will 
     enlighten us to the finer points of sampling? After all,
     we're talking an SK-1 owner... Please tell us how to make
     lip pops, Janzen...
    
     Put up or...
     
       * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
                 
    Meanwhile...
    
       
> Edd, I have been reading some of this and a lot of doesn't make
> sense.

	Welcome to the club.... ;^)
                        
> Are you telling me that in order to get a sample you actually have
> to *detune* the pitch you are sampling in order to get perfect 
> waveforms?
    
> That is, if you want to have exactly 360 x N degrees of wave rather
> than 360 x N + Y where Y is a remainder angle?
 
Yep, that's exactly it. Y would cause you to have "Zero-crossing" problems,
since it would be X% through the wavecycle...
   
> How do you then reproduce the pitch *in tune* when you play it back?

     Remember, you're not doing either of 2 things. You're not sampling
     the note "A" or using a sampling rate that produces "A". All you're
     doing is trying to stuff *complete* wavecycles into a given area
     of memory. Transposition is done on playback by reading the sample
     out at the speed which will cause 220 hz.      
     
 
Edd
	

    ass I see it....
    
    You are TRYING to get some multiple of waves to 'fit' into your sample
    interval. That way you maximize storage, simplify looping, and Im
    not sure what else, BUT it has NOTHING to do with sampling theory.

    It has to do with limitations of memory in current samplers.

    Thank you.
    
    ron
    

    Oh and by the way, the 256 is the number of discrete values from
    $00 to $FF. I do use the chart. May I send you a copy?
    
    I can prove myself a fool, I don't need you to do it. 
    
    Edd

    Re: Rost....
    
    One way to look at it is to imagine to different rates, the input
    rate and the output rate. You detune to match the input rate. Then
    on playback, it's output at a different speed. That's how 220.25
    gets back down to 220...
    
    Re: Walkin' Won
    
        Right. Because the input has been quantized to discrete values
        you often can't pick the 'perfect' spot to loop from/to. You
        have to pick something close. The idea is to get the 'close
        enough' point (0-Xing) on a page boundary. Maybe other samplers
        let you move the loop start point at individual sample resolution.
        The Mirage doesn't. Only the loop end point allows this fine
        of an adjustment.
    
        Maybe by reversing the waveform, inserting a loop end marker
        and then re-reversing???
    
    Edd

RE < Note 939.28 by JAWS::COTE "Note stuck? Try Kawai..." >

>        Maybe other samplers
>        let you move the loop start point at individual sample resolution.
    
	Hey, that sounds familiar!  The MKS has loop start *and* end points!
Maybe there's hope for me and mine yet!

	Dan

    The Mirage also lets you move both the loop start and end points.
    The start point however, can only be moved by pages (256 sample)
    increments. If it's not a ZC yer outta luck...
    
    The end point can be moved by pages OR samples...
    
    Can't figure why they did that....
    
    Edd

>Karl.  Do you have to go thru all of this to sample on the EMAX or is
>-.1 pertainant to a Mirage only?
>Todd.
    
    You MUST be joking.
    
    The Emax, in common with the Emulator II(+), has two methods of
    automating the loop process, Crossfade Looping(tm) and AutoLoop(tm).
    
    There are no 'page boundaries' or funny hex readouts.. the sample
    is moved thru quickly and coarsely by a slider control, or down
    to the individual sample using a 10-key numeric keypad.
    
    Although I admire the price of the Mirage, esp. the stereo Mirage, 
    quite a good buy, the hoops for loops that Edd's been describing
    might accurately be described as a HIDDEN cost for the Mirage.

    Regarding 'cycles' and 'waves' in sampling.. the length of a waveform
    cycle becomes smaller as the frequency increases. This creates an
    interesting situation when trying to single-phase loop (a single
    cycle) a high sound: the smallest loop length the E-II can arrive
    at manually is 64 samples.. the Emax can loop smaller yet. Generally,
    higher pitched sounds are harder to loop within a single period.
    they should be looped somewhat closer to the attack than lower freq.
    sounds.
    
    How do you know the exact length of a single period? Determine the
    freq. of the sample in Hz., divide into 1 second, or use the
    AutoLoop(tm) function.
    
    karl
    

                                                       
                                                       
    For those of you who have been shocked frightened and stunned into
    never considering sampling by -1 to -26. Let me add some light releif.
                                                       
    Sampling on an s700 is a doddle (English for Easy Peasy (or is that
    English too?)). All you do is stick a cord with the sound source
    in the "line in" socket, press a button called "New Sample". this
    allows you to set 3 parameters. The trigger input sensitivity, to
    help you there is a graphic LED display of the sound source a bit
    like on a cassette deck. You set the Note number to sample at, this
    can also be retuned later. And the sampling rate (I always go for
    the highest unless I am sampling a phrase of a given time). Then
    you press New Sample again and the machine stores the sample as 
    you play it.                                                    
                                                                    
    Once stored you play back the sample. with a bit of experience you
    can usually tell whether the sample will work or not. The machine
    has an autoloop feature which you can overide if you get the    
    aforementioned "honk". You overide the loop point by moving the 
    sample part parts are 1 - 32767 (magic number?).   You can move
    either the start or end in 1 or 100 part sections.             
                                                                    
    Let me say the Akai sample library is naff therefore I have always
    produced my own samples. I sample from records, blowing down mike
    stands, my son's activity centre makes some great percussion sounds
    with the ratchets and rattles. I even have a Korg Compact Sample
    Disk from which I have taken a number of samples.               
                                                                    
    Don't be put off by the technology. This is not the best sampling
    machine available by any means. But it's great fun.             
                                                                    
        			Paul.                                                            
                                                                    
                   

    I think I understand what you're saying, Paul, but I need a little
    more clarification on the 'activity centre'. Most of the activity
    centres I've had demo'd are fine on the bell and the squashy
    button bit, but the rattles are rubbish. I thought for a long time
    it might be down to poor documentation, but Ive since heard that
    other people get on fine with them. Could it be your son has had
    specific mods made to his activity centre? And if he has, would
    he be willing to divulge them to noters who wish to harness their
    own activity centres to the latest in musical tehnology?
    
    (Between you and me, I reckon what the guys in -1 to -26 know about
    activity centres you could inscribe on the back of a postage stamp).
    
    Richard.

    What with all this sampling talk being bandied about, I ran home
    last night and booted up the Mirage...
    
    ...and found a method for determining the amount of detuning necessary
    for perfect short loops. Is it any good? Well, let's just say I
    was getting *perfect* loops in less than 5 minutes...
    
    First, set the sampling rate/time parameter (73) to the closest
    value which corresponds to the note you intend to use as a source.
    (This info is in the ASG, pg. 73). Turn multisampling on (parameter
    77). Now, take your sample.
    
    When done, turn the loop switch (parameter 65) on, and press the
    key. You'll probably hear a distinct change in pitch when it starts
    looping. If you don't, you're all done (at least with the looping
    part). If the pitch is different, simply tune your source to match
    the pitch of the loop! No kiddin', it's that easy. The playback
    rate will take care of bringing you back to correct pitch.
    
    Of course, after you get your perfect loop, you then get into the
    next phase, modifying the envelope, filter and filter envelope,
    LFO, keyboard scaling and other various goodies.
    
    ...like falling of a log.
    
    Edd 
    

RE < Note 939.32 by MINDER::KENT >
                                                       

	I don't know, maybe I'm looking for an extremely complicated process
when, in reality, it is quite simple.  The MKS-100 has almost the exact same
procedure as you described with your S700.  There are four buttons;

	1) Record button - When selected, it sets the machine up to record the 
sample.  It confirms the memory location to which you are about to sample to.

	2) Mode button - Allows you to selct note to which sound will be 
recorded (i.e. C4, G2, ...)  This can be altered in the editing process as well.
Allows you to set Auto-trigger, and sampling clock.

	3) Standby Button - Creates a level meter in the display and allows you
to set the trigger level.

	4)  Start Button - Start sampling.

	Come to think of it, having heard non-mirage sampling examples, I 
wonder if the reason I don't have a lot of charts and tables available to me
to maximize the sampling process is because I don't need them.

	Still, I appreciate hearing the particulars of sampling technology that
Edd has been writing up here.  It's obvious that he's done his homework.

	Anyway, for those who may be interested, I brought in my manual today 
and have printed below the editing parameters available on the S-10/MKS-100.

	Recording Key Number
	Bank Tune
	Loop Tune
	Scanning mode
	Loop type (one shot, manual, auto)
	Start point (1-32767)
	End Point (Manual)
	Loop length (Manual)
	End Point (Auto)
	Loop length (Auto)
	Key Follow
	Pitch Bend (on/off)
	Vibrato (on/off)
	Envelope Velocity Sensitivity
	4-stage envelope rates and levels
	Dynamics Sensitivity
	Auto Bend Rate
	Auto bend Depth

    
    
    Re .33
    
    Actually Steven, my son, is really hacked off beacuse the activity
    centre he has, cost us 13 pounds only 4 months ago. Because of the
    advance in technology it's been superceded by a model from LEGO.
    I hear that "Mommy's Dead Good Toy Shop" has them knocked down to
    4.50. 
    
    He should have waited for the new model. He's only now getting prepared
    for commusic 14. At the current rate he'll be 42.
    
    				Paul.
                         

    What's the sampling rate on the S-10/MKS-100???
    
    Variable? I didn't see it in the list.
    
    What about input filtering?
    
    Edd

RE Note 939.37

>    What's the sampling rate on the S-10/MKS-100???

	Selectable: 30KHZ or 15KHZ.
    
>    What about input filtering?

	I don't see any input filtering in my manual, but I didn't mention that
there is digital filtering in the Wave Modification section.  Two hi-Pass and
two low-pass filters.
    
	Dan

JAWS::COTE "AUTOLOOP??? Ha! Wimps..."                 7 lines  11-SEP-1987 10:34

    ... let's explore this extremely uninformed PERSONAL_NAME. 
    
    If I understand it, anyone with a sampler with decent loop editing
    S/W is something less than a man.
    
    My contention is that it takes bigger cojones to spend the money
    (and the time to make it) to get a capable unit than to 'economize'
    by getting a less capable sampler.
    
    I'd rather spend my music time making MUSIC, not poring over hex
    tables.
    
    best. karl
        
    
    

    Oh my! Looky what I done gone and did now, Mr. Wizard...
    
    Actually, it's compensation. Since I can't play keyboards to save
    my soul, I compensate by learning lotsa hex codes....
    
    REAL men don't use autolooperdoopers. They reset pointers, *in binary*,
    after mapping out wavetables on graph paper.
    
    Edd
    

    I'm afraid Edd is only half right.  Not only do you have to do looping
    by hand, but you also have to have to do it on a VAX 8800, with
    64 Meg of memory.  Keeping Ultrix pared to a minimum, that leaves
    you with 60 Meg of sample.
    
    And you have to figure out where zero crossings are by running adb
    on /dev/mem (the running system).
    
    Lets hear it for megabyte sonorities.  (Now that's fat?)
    
    Steph

	
    These are probably the same folks who save patches in RAM instead
    of punching cards full of sys-ex data.
    
    Edd

    Boy, I can't wait to get one too.  In the mean time I'll just have
    to make due with all those crummy analog sounds in my JX-10 and
    MKS-80.
    
    Actually, maybe I can convince myself that the 38 is really a low
    functionality sampler - it takes one sample with 20KHz bandwidth
    and a duration of 32 minutes, but you can only play it back at one
    pitch - the same one it was recorded at...
    
    len.
    

    Seems to me you could transpose by holding your hand against the
    source reel...
    
    Of course, your samples might truncate early...
    
    Edd

    re: a few back
    Ummm.  Pardon my obvious (perhaps naive and innocent) lack of
    understanding, but what exactly are cojones?  Like, do I have to
    wash my mouth out with soap if I mention them in front of my mother?
    Would "Would you please pass the cojones?" at the dinner table result
    in the same gasps and giggles that would result if I called All-Fruit
    a jelly?  If I mentioned them in church, would I stop all conversation?
    
    Steve

    Mirage owners...
    
    I spent the best part of a rainy Sunday doing a 16 part multisampled
    organ. (The how-to details of which I may post in another reply.)
    For the source I used a patch I made on the DX21 called "Soft Organ".
    It's kinda of a quiet, chapel type organ, kinda like you'd hear
    in a funeral home. Basically it sounds damn good, although the 3-4
    highest notes gave me a few problems. Chorusing is built in via
    detuned oscillators. The original, unmodified sample is in Upper
    and Lower program 4 if you want to hear that. Program 3 is a heavily
    filtered version.
    
    I'll be glad to give a copy to anyone who sends me a disc, ON ONE
    PROVISION. Review it here. I want some feedback.
    
    Send mail if interested.
    
    Edd

    What's the provision?
    	
    First born child?  :-)
    

    Uh, thanks, but kids and disc drives don't mix. ("Edd!! Lookit where
    I stuffed my marshmallow!!!")
    
    ...just some comments will be fine, tank you.
    
    Edd

	Tell me what y'all think about this one.

	I sampled a toy cymbal the other day in the hopes of adding a particular
sound to the percussion portion of a song.

	First of all, I noticed that I had to really crank the mike and the 
recording level to get it to show any kind of reasonable input level on the
sampler's meter.  This, of course, caused it to pick up additional 'atmospheric'
noise from the mike.

	Then, when playing back the auto-looped sample, I noticed a sound not
heretofore heard from the sampler, coming a bit after the decay of the cymbal.
It was something like a sine wave sweeping from a very high frequency on down
until it faded away.

	Was that aliasing?

	An inquiring mind wants to know.

    RE: levels....
    
    Cymbals are a bitch, all high frequencies. Len made me a tape of
    his Zildjians that is almost gratyingly loud, yet the meters on
    my tape deck hardly budge. Perhaps a noise gate between the sampler
    and the source will help. Dunno, don't own one...
    
    As to the other noise at the tail of the sample, I get something
    very similar on the Mirage, but without hearing yours I'd be hard
    pressed to say it's the same. Anyhow, all the symptoms sound exactly
    alike. Assuming that it's the same problem, I'd say it was the result
    of your output filter EG. It's open for the duration of the sample
    and slowly closes as the sample releases. Try tweaking the filter
    release rate. Works like a charm for me... Do you have some sort
    of generic parameters your machine is set to when you sample, or
    do you boot up your zither sample and then sample over it in memory?
    If so, you're probably getting all the filter, EG, etc., settings
    for the zither on your new sample. 
    
    Aliasing often sounds like a 'buzzing' noise in your sample.
    
    Edd
    
                                       

    Aliasing would cause a noise moving *downward* in pitch only if
    there was some component in the source moving *upward*....
    
    Do you have such a beast?
    
    Edd 

re < Note 939.51 by JAWS::COTE >

	Yes, the envelope is wide open by default.  I believe I tried to
adjust the envelope so as to cut off the 'siren'.  Unfortunately, it also
had a detrimental effect to the sound of the cymbal itself; cut it off too
soon.

	Oh, well...

	Thanks,

	Dan

    Does the MKS-100 have a filter EG????
    
    When you say "the envelope is wide open by default" I assume you
    meant the *filter* is wide open. VERY rarely does any sound need
    a wide open filter. (Admittedly, something like cymbals might be
    the exception to the rule.)
    
    The Filter EG on the Mirage allows you to dynamically control the
    filter; wide open on the attack transient, down 50% on the sustain,
    close down at a rate of n for the release. Note that this is indep-
    endent of the familiar amplitude envelope.
    
    Edd

For cymbal recording, I would recommend the use of a compressor.  I've
noticed that on most vinyl recordings the cymbals are compressed a bit.

The mic also makes an incredible difference.  Anyone try one of the PZM's
on cymbals?  What do they sound like in the higher range of the audio spectrum?

Todd.

RE < Note 939.54 by JAWS::COTE >

	I delayed answering this until I had manual in hand...

	I searched the manual regarding the envelope generator and it gave no
explanation as to what component it was modulating.  My guess, from using it a
bit, is that it is either simply for amplitude, or for amplitude *and* filter.
I have yet to really experiment with its four filter parameters (LP1, HP1, LP2,
and HP2) to be able to say what effect envelope has on them.

	Just as an aside, it has become very frustrating to me that Roland says
so very little in their documentation regarding parameters.  The section on EG
in the book simply talk about rates and levels; absolutely NO MENTION as to 
what the rates and levels are modifying!  The manual with my JX8P is even worse.
Each parameter has a one or two sentence blurb about its editing parameters and
that's it!  If it weren't for a great article in Keyboard magazine about 
programming the JX8P and the JX10, I'd be totally in the dark about some of the
great programming options on this instrument!

	Meanwhile, back at the ranch...


	When I said "the envelope is wide open by default", I meant that the
levels and rates are set at their highest values. so the envelope would, by
default, look like this:

		_________________                          .
		|               |  as opposed to          . .
		|               | something like this:   .   .
		|               |		        .        .
 					               .             .
 					              .                  .

	Oh, in case it's not clear in this explanation, only one EG on the 
machine.

	Dan

    Dan,
    
    Could it be that you get a "short loop" in your sample, like 
    somebody mentioned before. I think I had something like that
    when I tried to do a sample on my S-10. Suggest you try the
    one shot mode on your MKS-100 and see if you still have it
    then. The "one shoot" mode only scans through the sample once,
    with no looping. At least that way you see if the problem is
    in the loop.
    
    Svein
    

RE < Note 939.57 by PILOU::MULELID >

	I was using one-shot mode.  That's why I was so surprised to hear it.
Thanks anyway.

    Dan

    This is in reply to some of the short loop problems Edd discussed
    earlier, as well as a clarification of what Emulator/Emax 'Autoloop'
    is..
    
        from 'The Art of Looping' Music Technology Sept 87
    
    "..A common problem with VERY short loops is that they end up playing
    a different pitch from the rest of the sound. this annoying shift
    is caused by a rather technical brawl between the sampling rate
    and the pitch of the instrument being sampled.
    
    For example you have used a sample rate of 44kHz to sample a flute
    playing an A440 note. Performing a little math shows that one cycle
    of this sound will  be 100 'samples' (however many bits wide that
    may be) long. If the loop is a little shorter or alittle longer,
    the cycle length is altered, and the resultant pitch is different
    fro the rest of the sample. To get the loop 'in tune', you will
    need a loop length of 100, 200 or 300 (and so on) sample words.
    This sounds great, but who wants to use an oscilloscope and calculator?
    Let the machine do the work..
    
    One technique for eliminating pitch shifts is called 'magnitude
    differencing' and appears in the Emax and Emulator II. this technique
    looks for matching portions of the waveform around the start and
    end points, and then adjusts the end point so that its position
    in the cycle corresponds to the start point's position. This renders
    a loop length that IS AN INTEGER MULTIPLE of the CYCLE LENGTH (emphasis
    mine- km) so that the loop is IN TUNE, and also minimizes loop clicks
    by keeping the waveform running in the same direction when it jumps
    fro the end point to the start point. "

    used emphatically without permission
    
    karl moeller