Conference napalm::commusic_v1

    I made my closing statements in the MT-32 note that spawned this.
    
    To summarize, I think the only "problem" here is how companies
    advertise synths with these kinds of architectures.  Some are, IMHO,
    unacceptably deceptive, particularly Roland.

	I have found it to be no real problem.

	In sequencing, I use the ESQ-1 and the EPS a lot. Both have voice
	limitations. However, I have managed to use up to 5 instruments
	on the ESQ-1 without losing notes, because there's never more than
	8 notes playing at a time. As you say, the oscillators are software
	entities as opposed to hardware entities, so context switching 
	works. Just don't hit too many notes at the same time.

	The EPS isn't such a problem, usually. I've found that the "big" 
	patches, I don't use much anyway, they are cliche' and waste
	memory. (pure opinion).

	Having more voices available than MIDI channels, I'm more concerned
	with the MIDI limitations than the timbrality issues.

	/pjh

    Are osciallators really just "software entities"?  I thought that, when
    push came to shove, we were still dealing with a hardware limitation.
    No? 

    Maybe I don't understand how these things really work anymore.

-b

"Effective Polyphony."  I love it.  Or we could call it "Music Power," which
should get a smile from audio fans as old as I am.

	Eirikur

    
    Re: .4
    
    On an old EM cover, they showed an MT32 with the top off, and I saw a
    pair of custom ICs with Roland logos, some memories with rev stickers
    and some Burr-Brown DACs (like, two, one for each channel).  This,
    along with the *price* of an MT32, leads me to believe that the thing is
    a software engine right up to the point it spits out a data stream to
    the DACs.  
    
    I'd really like to see some input from anyone who has played with the
    circuits or seen a service manual.
    
    							Brian

    
    I suppose this is where a reliable reviewer would come in handy.
    You can't really blame Ensoniq for claiming 21-voice polyphony, given
    their architecture. It's up to the reviewer to balance statements about
    'amazing sounds' with comments on corresponding depletion of available 
    voices. If a synth really does sound great using mostly single voices,
    then it's a bonus that there are lots of voices available. But if
    someone produces a design with 32 but each patch or tone
    requires at least 4 of those voices, then there's not much to shout 
    about. 
    
    So, I don't think it's misleading to claim the 'available polyphony',
    but I think a reviewer who didn't point out a poor voice-to-sound
    relationship might justifiably be thought to be sponsored by the
    manufactirer. 
    
    Richard.

These days, with those software occillators, the limiting factor tends to be
CPU speed (with influence from software efficiency).

The other tricky spot is bandwidth into the D-to-A.  If you have 32 oscillators
delivering data at close to your full D-to-A rate, you have to do some fancy
mixing to get to something to send out.  That's 32 (probably 64) add operations
per D-to-A clock cycle, AFTER all the synthesis has been done.

	Eirikur

    Well, first, I wish you wouldn't call them "oscillators", because
    there's usually a whole lot more than just an oscillator involved
    (e.g., envelopes, filter, LFO, etc.).  Furthermore, some (most?
    at least all the ones I know about) multioscillator architectures
    (e.g., MKS-80, JX-10, Matrix-1000) don't allow you to reuse "unused"
    oscillators.  Note that this is not because these are "old" synths,
    but because even in "modern" synths it's not the "oscillators" that
    are dynamically assigned, but some larger construct that includes
    one or more oscillators (e.g., consider 4-op or 6-op FM).
    
    So what do we call them?  I call them voices, but that's a pretty 
    confusing usage these days.  The feature that Brian is describing
    is generally called "dynamic voice allocation", consistent with
    my use of the term "voice", which is why I chose it.  But there's
    another reason.
    
    I think we already had this discussion.  I remember proposing that
    we take a cue from traditional musical notation.
    
    A voice is an isolatable timbre.  Like a human voice, or a single
    acoustic instrument.  On a synth, this might be layered with other
    timbres to produce a richer sound.  I'm going to ignore for the
    moment "layers in time", where one timbre is responsible for the
    "attack" of the layered sound, and another is responsible for the
    "sustain", and neither is actually playing at the same time.
    
    Some number of voices may be playing in unison, i.e., following
    the same (monophonic) "line" in a score. This is called "doubling". 
    Doubling may also be done by the same timbre, in order to produce a
    "chorus" or "ensemble" effect, provided there are some (however subtle)
    differences among the constituent timbres.  Layered sounds constrain
    their constituent voices to *always* play in "unison" (in quotes,
    because they may actually be transposed from one another, say by
    an octave or a fifth).
    
    The bottom line, though, is, doubling, whether by the same or different
    timbre, uses up "voices".
    
    Multiple "lines" make up a "part".  I.e., a part, played by some
    particular aggregate timbre, may be "polyphonic".
    
    Multiple parts may play simultaneously.  The total number of lines
    is the current total degree of polyphony, and must be less than
    the number of voices available, or voices must be "stolen".
    
    You wanna make music, you need to understand this.  You can't make
    more notes than you got voices.  You need to know how many "voices"
    your lines use up.  You wanna do barbershop quartets, you need four
    singers.  You wanna do "Spem in Alium", you need 40 singers.
    You wanna sound like the Morman Tabernacle Choir, you need 100s
    of singers.  But the music probably only requires 4 singers.
    
    What's the problem, other than some "truth in advertising" to argue about?
    
    No "n-voice" synth ever fooled me.  I usually assume "n-voice" means
    "n/2 notes with interesting sounds".  For some fancy synths it might
    mean "n/k notes with sounds of interest level k".
                                                                
    len.
    

	A bit on a tangent...

	It would be interesting to poll the COMMUSIC readership to ask what
your current studio's "base polyphony" (the number of basic voices you have 
available without layering/doubling) and its "effective polyphony" (to use
Brad's coined phrase).

	I have often spoken of my studio as being the equivalent of x number
of musicians (in a practical sense, based on the type of vocing I tend to
use).  Currently, it would look something like this:

	Base Polyphony:

	TX81Z:		8
	Matrix6R:	6
	Piano:		8/16 (preset, depending on patch)
	YS200:		8

	Effective Polyphony:

	TX81Z:		8 (I have some *good* patches)
	Matrix-6R	6 (no layering required)
	Piano:		8/16 (I use both types of patches)
	YS200:		4 (used mostly for layered sounds with internal effects)

	For my style of playing/arranging/sequencing, number of musicians
I'd need for any given song to equate:	16 to 18

	Dan

    	RE -.whatever
    
    	The new Peavey workstation is supposed to be based on that idea. 
    It's (according to a dude [not Jeff] at Union) got 3 custom DSP's that
    just do whatever.  So the synth driver tells them what to do when and
    they do it, meaning more voices could spring up (fall down?) from
    upcoming OS changes.  I like that idea, I wonder if Peavey can do it,
    though...  I much would rather see Ensoniq do that.
    
    --mikie--

    Maybe we shouldn't call them voices.  if the DSP routines are
    re-entrant, we should call a voice a process, or an entry, or an event
    list entry, or an event list entry pointer, or an event list entry
    allocation pointer, or a dynamic event list entry allocation pointer,
    or something.
    
    Maybe voice is better.
    Tom

	In an off-line discussion an interesting point came up...

	How many voices do you really *need*, practically speaking.  Upon
analysis, I don't think I've ever used much more than four different timbres
at once and no more than a maximum of 16 or 17 "voices" at a given time.

	Dan

    I think we had thus discussion once, 'cuz Tom went and looked up
    some hairy scores to see how many voices and notes were going at
    the same time.
    
    My pop arrangements don't typically use more than 4 or 5 timbres
    (e.g., bass, piano, pad, lead, drums), but "classical" stuff tends
    to use a lot more, though only in the big tuttis does everything get
    used all together.
    
    The most complex chords that get used in pop arrangements might
    typically be 7ths or 9ths, and even with some doubling you'd be
    hard pressed to need more than 6 notes per timbre (if you used 6
    notes on all timbres, you'd have an awfully dense arrangement).
    
    More realistically, a pop (in my case usually oldies) arrangement
    might "voice out" as:
    
    	bass	1
    	piano	4
    	pad	4
    	lead	1
    	       ---
    	total  10
    
    
    len.
      

    I propose that polyphony be defined according to the maximum number of 
    notes you can get out of a synth at one time.  Many synths have a
    minumum as low as 1.  Consider this, the TX81Z is an 8-voice synth.  BUT, 
    if you set up chorusing right (or wrong, depending) all eight voices sound 
    on one note.  Effectively, it becomes a one-voice instrument.  Any
    claims between 1 and 7 have little meaning when describing the TX81Z.
    Another confusing instrument is the CZ-101.  Most folks think of it as
    a 4-voice instrument.  Then, they find out it's possible to get 8 notes
    at a time out of it.  To avoid confusion, the CZ-101 should be
    consistently advertized as an 8-voice instrument.  
    
    Deceptive advertising?  Only if you are an uneducated consumer.  This
    is much like benchmarks for computers.  If you buy a computer based on
    which computer can solve an FFT the quickest, what you have is a
    computer that can solve an FFT the quickest.  You have not necessarily
    bought the machine that handles *your* problems the quickest.  My
    understanding is that the Digital sales force has run into this very
    type of problem and have had to educate those who would buy based only
    on a particular set of benchmarks.
    
    Similarly, 32-voices is an appropriate feature to advertize with regard
    to the MT-32 or the Proteus.  8-voices is an appropriate feature to
    advertize with regard to a CZ-101 or a Fairlight (it's 8-voice, isn't
    it?).  Whether you can do anything useful with those voices is a whole 
    'nuther matter.  Caveat emptor.
    
    
    Steve

    
    I think the idea about being an educated consumer hits the nail on the
    head.
    
    I've had a couple of demos of new gear recently where between myself,
    the salesman and the manual we couldn't scope out how the hell the box
    worked.  It certainly makes it harder to make decisions on what to buy
    when you have to spend many hours poring through documentation, press
    reveiews, talking to guys who actually sprung for a unit, etc. just to
    find out if it will do what you want.  
    
    Sure was a lot easier when I was just buying guitars  8^)  8^)  8^)
    
    						Brian

    Synthesizers are now about as complex as personal computers, or more so
    depending on measures of complexity.
    Therefore, manufacturers should provide sales training and complete
    well-written technical documentation, examples, demos, videos, classes,
    seminars, and help lines.
    If manufacturer does not provide these things, I don't buy their gear.
    Well, OK, I don't buy any gear, but you get the point.
    Tom

    
    But they don't sell as many 2000 dollar synths as they do 2000 dollar
    PCs.
    
    Tom's not buying any don't help!
    
    Richard.

    > I think the idea about being an educated consumer hits the nail on the
    > head.
    
    I disagree completely.
    
    I don't think part of being an "educated consumer" is knowing when
    a particular manufacturer has abused terminology.  The issue here
    is not education - it's standardized terminology. 
    
    When I buy a car billed as a 12 cylinders, I expected them to only 
    count the kind of cylinders that we all think of, and NOT anything
    that happens to be cylindrical.
    
    When it says "comes with air conditioning", I expect it to mean a
    system that cools the passenger and NOT something like a carbuerator.
    
    Now, in deference to Chad, it may not be a "lie" to describe a
    carbuerator as an "air conditioner" - it does "condition" air - but
    I don't think we can invoke "El Caveat Emptor" to excuse such usage.
    Do you?
    
    Prior to Roland's D-series, "32 voice synth" meant something.  It
    implies a level of polyphony in practical applications that the 
    MT-32 does not deliver.
    
    	db

    A brand new baby acronym - 'EP'.  
    
    The Emax has 8 actual, physically identifiable DACs.  Using 'dual'
    mode, I can actually drive 16 samples simultaneously thru them.. BUT
    each dual sample pair is only playing one note.  SO - IMO the Emax has
    an EP of between 4 and 8, depending on loaded presets.  My favorite
    drum kit uses a stereo snare and kick - thus eating 2 voices per note..
    except that they're only used for about 1/10 second.
    
    The Kurzweil, with 24 advertised voices, CAN be driven to the point
    where it truncates voices, but I have to work at it - the only way
    it's ever run out of "EP" is when the patches used have heavy layering.
    
    Luckily my fave Ksounds are mostly 1-layer.  So I figure the Kurzweil
    averages out between 16-24 voice EP, and the Emax about 6 EP.  And when
    I DO run out of polyphony (or more likely sample space in the Emax) I
    can always record a track on the tape recorder, synced with FSK v1.0.
    
    karl
    

    
    Re: .19
    
    Dave, I'm not advocating that a manufacturer can "lie" to me just
    because I'm willing to spend hours of research to uncover the truth,
    but in the end it does come down to that.  Manufacturers always hype
    their products, and you just have to be filter it out.  Knowing what we
    know now about machines like the MT-32, I doubt any of us are going to
    take polyphony claims on face value anymore, OK?
    
    Re:  .17
    
    Tom made a good point about training.  Seems most of the "seminars" I
    go to at music stores tend to be flashy demos, although if you wait
    around long enough you can chat with the manufacturer's rep a bit after
    everyone else has left.  Why can't manufacturers and stores hold
    sessions that support the users *after* they buy?   I mean real
    hands-on sessions showing how to use some of the arcane features of the
    hardware, showing neat patch approaches, etc.  
    
    I noticed that the local Ensoniq dealer gave up on a User's Group after
    the meetings degenerated into patch swaps (sigh).
    
    Re: .20
    
    I also like the term EP.  I hope people start using it in discussions
    of SGUs in the future.  That way we can further confuse noters new to
    the wonders of MIDI acronyms  8^)  8^)  8^)  8^)
       
    
    							Brian

    They don't need to give real instruction is using the
    advanced features of a synth because most people use only the presets
    (cf. another note!)
    Tom

>    Dave, I'm not advocating that a manufacturer can "lie" to me just
>    because I'm willing to spend hours of research to uncover the truth,
>    but in the end it does come down to that.  Manufacturers always hype
>    their products, and you just have to be filter it out.  
    
    There is a difference between "hype" and "deception".
    
    And deception doesn't have to be a lie.
    
    In a previous note I mentioned a "Satellite Antenna Dish" that wastried
    it's best to lead you to the conclusion that this $19.95 wonder would
    do everything the "big" dishes did without actually telling any lies.
    But actually all the thing is is a "rabbit ear" type antenna with a
    plastic parabolic "dish".  
    
    Does anybody remember that ad.  It is by far the best example how
    a deception needn't be a lie.  In fact, I thought it was so incredible
    that I tore it out and saved it (if anyone wants a copy - it's truly
    incredible).
    
    For example, it claims to "bring you movies, sports and special
    events just like an ordinary pair of rabbit ears.  No cable box
    or special attachments needed".  And it's "legal in all 50 states."
    
    It's also "guaranteed not to interfere with any satellite signal".
    
    
    > I doubt any of us are going to take polyphony claims on face value 
    > anymore, OK?
    
    I don't see this as progress.  But I agree that I will certainly not
    take any polyphony claims from Roland at face value.  At lesat they
    could have avoided the term "voice".  Perhaps they should call it
    a "32-partial" synth.
    
    	db

	How about if instead of saying "32 voice synth" or "32 partial
	SGU" we just call it a box capable of making 32 noises at the
	same time? *8') 

	SWAV1::STEWART, vacationing in AZ while my home node is updated

	P.S. in which case Roland seriously understated the MT32's capabilities!
	P.P.S. before anyone gets excited, I own one and mostly like it... 
	P.P.P.S. wonder how long db's been saving that ad up for a "discussion"?

re .23

If you had 32 fingers, or a sequencer that could send 32 note-ons to
a machine, and there is at least 1 program in that machine which would
sound all 32 notes you just played, then the machine is UP TO 32 voice
polyphonic.  Period.  How useful the sound is is not material the
machine's polyphony.

Do us a favor, Dave, and don't apply for a job  marketing DEC
multiprocessor systems, okay? 

8^)

    Hi from munich!
    
    I don't need any deference.  I would call Dave's examples from 
    the car deceptions too.  But the MT32 isn't a deception -- just
    because you don't care for the patches that allow such polyphony
    doesn't make it a lie.  Steve sherman's definition is the best.
    
    Is it a lie when a car manufacturer ays a car can dn 0-60 in 6.2 secs
    but fail to mention that you have to be a profi driver to get that
    kind of performance?  I don't think so.
    
    Dave, I've seen the ad you refer to.
    
    Things that are measrable are what are put in ads -- things like
    the ability to play 32 independent instances of a patch sounding
    (voices) or that the car is able to do it.
    
    The ads don't claim that you will be able to get 0-60 in 6.2 secs or
    that you will like the patches that allow 32 voices.  Only maximum
    functionality can be advertised, as that is objective or provable.
    
    The companies don't know how fast you the individual cn drive or what
    patches and associated polyphony the individual will achieve.
    
    We live in an imperfect world.
    
    More comments later when I am back in the USA.  These DECnet
    connections make this note-entering stuff to much of a pain in the
    backside.
    
    gruesse
    
    chad

    Chad,
    
    It boils down to the FACT that describing the MT-32 as "32 voice
    polyphony" omits a MAJOR detail in a way that deceives the buyer.
    
    IMO it is immoral to lead a customer to a false impression even if
    you manage to do so without telling a true "lie".
    
    Whether or not that constitutes "fraud" varies with your idea of
    fraud.   However, it seems silly to deny that it is "deceptive".
    
    	db