Kenji's K5000 Message Board Digest - Overview K5000 Resources - Overview
The Eat at Joe's Kawai K5000 Message Board Digest
Things that Should Be in the Manual (Details!)


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


FF dB Scale?
 Thursday, 18-Dec-97 11:52:52

      Message:
      192.28.2.19 writes:

      Anyone know the dB scale of the formant filter? What would a 24dB/octave filter look
      like? How steep is the low pass template in SoundDiver?

      leiter@skypoint.com 

------------------------------------------------------------------------------------------

Re: FF dB Scale?
Tuesday, 23-Dec-97 16:35:12 

     195.232.60.79 writes:

     Hi Leiter - hi K5000 phreax!
     Yes, the FF amplitude scale is what I need so much since I used the formant filter
     for the first time. (Remember my remark in the PWM posting?) 
     Well - now I have an answer. I have done measurements, and they are just finished: 
     It's 0.75 dB per step.


     Jens Groh 

------------------------------------------------------------------------------------------

Re: FF dB Scale?
Friday, 26-Dec-97 11:48:32 

     195.232.52.125 writes:

     I forgot to say:
     The amplitude scale in the "Harmonic Level" menu seems to be the same.

     Jens Groh 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Analysis Of The Synthesis
Tuesday, 23-Dec-97 17:16:31

      Message:
      195.232.39.207 writes:

      I just replied Leiter's question concerning the formant filter amplitude scale.
      But this was only the first of a whole bunch of questions that come now into my
      mind: 

      What about the frequency scaling of the FF envelope? Milliseconds for envelope
      time values? Decibels for envelope level values? The LFO period in
      seconds? The vibrato amplitude in halftones? And so on... - all unanswered.
      Why doesn't Kawai tell us in the manual what this synthesizer exactly does?
      (I'm afraid no synth manufacturer likes to explain the details. My bad
      experiences: I had to reverse-engineer the "LA" synthesis principle of my
      Roland D-10 to understand what it actually does - it's different from _any_
      published explanation I've read.) 

      Back to Kawai: How can we avoid reverse-engineering the K5k (partially, of
      course)? And if we can't avoid it, how can we do it together? Anyone
      interested in participating? 


      Jens Groh 

------------------------------------------------------------------------------------------

Re: Analysis Of The Synthesis
Tuesday, 23-Dec-97 20:59:38 

     199.86.47.85 writes:

     I'm up for it. Let's try e-mail to Kawai 1st. Their Australian site seems to be
     the best one in English.

     Thanks for the ff info.

     leiter@skypoint.com 

------------------------------------------------------------------------------------------

Re: Analysis Of The Synthesis
Wednesday, 24-Dec-97 04:43:38 

     153.37.15.18 writes:

     I've already mailed kawaius about this on Dec. 21st after reading Leiter's original
     post. I asked them these questions: 1-the volume scale for all volume
     controls (not just the formant filter). 2-At what point the Key Scale to cutoff
     frequency follows the note being played exactly (so you can have consistent
     cutoff of harmonics relative to the fundamental via the DCF). 3-The scale that the
     frequency cutoff dial follows before any modulation or scaling.

     Still no reply, although I don't expect one soon due to its being the holidays and
     all.

     A lot of these new questions are good too. I think the best approach would be to
     compile a list of questions the names of people who want to know so we
     can send them a well organized petition for large amounts of technical information
     (hopefully with a large number of user's names attached).

     Additive synthesis is such a mathematical approach to sound creation that I think
     these specifics would be very helpful....

     Let's do this.


     Kenji 

------------------------------------------------------------------------------------------

Re: Re: Analysis Of The Synthesis
Wednesday, 24-Dec-97 10:36:39 

     192.28.2.19 writes:

     Also, we could let them know we'll be posting the info (or Kenji will!), which
     should help. It's all user information, not competitor information, so it's in their
     interest to be supportive. We might want to try Kawai Australia, they seem better
     organized on the web than Kawai US.

     Also, re the envelope parameters: it helped when I figured out that the time
     parameter is a rate, not a duration. If you change the level and you want the
     duration to stay the same, you have to change the rate to match. I don't know
     which way would be better, but there it is.


     leiter@skypoint.com 

------------------------------------------------------------------------------------------

Re: Analysis Of The Synthesis
Wednesday, 24-Dec-97 04:55:06 

     153.37.15.18 writes:

     I forgot to mention the one good thing the manual has along this vein - it does
     chart out the frequencies of the various bands of the formant filter (with bias set
     to 0). It's on pg. 45 in my manual. Basically, each band is a semi-tone. Maybe this
     means that somewhere there is a designer at Kawai who knows that we
     want to know this kind of stuff....


     Kenji 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Rate parameters of harmonics envelopes
Monday, 26-Jan-98 05:00:07 

     193.96.226.61 writes:

     Here is a hint for you brave ones who dare to edit the K5000 harmonics envelopes:
     When I wanted to give each harmonic an individual "note-on delay", I tried to set
     the attack level to zero so that the attack phase becomes a delay phase. The
     remaining parameters are then still sufficient to form a simple ADSR curve. But,
     I noticed that the attack phase was always skipped. Obviously, it's not
     possible to do it that way, and that's because the K5000 uses a rate value rather
     than a time value for each phase:

     If the start level and the target level of a phase are equal, the target level
     is reached at once, regardless of the rate value!

     The start value for the attack phase is always zero, thus it was the same as the
     target value that I had entered. Now it was no more difficult to find an simple
     solution: I set the attack level to 1 (a barely audible level) instead of 0, then
     the rate parameter worked fine.
     Have fun!


     Jens Groh 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Rates in the guise of time values
Monday, 02-Feb-98 04:01:54 

     130.67.1.177 writes:

     leiter@skypoint.com wrote:

     QUOTE

     Also, re the envelope parameters: it helped when I figured out that the time
     parameter is a rate, not a duration. If you change the level and you want the
     duration to stay the same, you have to change the rate to match. I don't know
     which way would be better, but there it is.

     UNQUOTE

     The time parameters seem to be some sort of inverse rates. Here's my latest
     experiment with the DCA Envelope.

     Set Decay1 Time and Decay1 Level to the values below and set the other parameters
     to 0. Trying to measure time by counting beats for the duration of the
     tone, I get the following results:

     D1T D1L Time Proportion

     127 124 11 1/2
     126 124 7 1/3
     125 124 5 1/4
     124 124 4 1/5

     The resulting time function is probably more sophisticated than the simple fractions
     I have listed above, but it does look as if Time is proportional to 1/(129 -
     D1T). Anyone got a better idea or more solid information?

     In any case, Time appears to be proportianal to (127 - D1L), so "rate" is probably
     the right word, as leiter says.

     Tore

     tl001@online.no 

------------------------------------------------------------------------------------------

Re: Rates in the guise of time values
Monday, 02-Feb-98 11:25:13 

     193.96.226.61 writes:

     Hmm... this throws a different light on the DCA envelopes. Is it the same "rate
     algorithm" as in the ADD/DHL envelopes after all? How can they keep the
     envelope on a constant value then? Can they? - I didn't test it. (See "Rate
     parameters of harmonics envelopes" thread.)

     There is another similarity of these envelope generators: I observed that both
     the DCA_Key_Scale_To_Envelope_Level and the ADD_DHL_Key_Scale_To_Gain parameter
     seemingly do not function (at least in some cases). 

     Is this a bug? - What do you think? 


     Jens Groh 

------------------------------------------------------------------------------------------

Re: Re: Rates in the guise of time values
Monday, 02-Feb-98 12:46:15 

     130.67.64.84 writes:

     Yes, I have also noticed that some of the Key Scale parameters apparently do not
     function or have a function different from what is intended. I am trying to
     test it all systematically.

     Meanwhile, have any of you found a way to translate DCA Envelope settings into
     Harmonic Envelope settings? Not to mention a similar translation for the
     DCF Envelope. Figuring out this machine will take some time...

     Tore

     tl001@online.no 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Re: Re: Re: Re: Another bug ...
Wednesday, 04-Feb-98 12:16:54 

     194.172.230.108 writes:

     Yes!
     - and no. Well, I think I found out what happened, and it looks no more like a real
     software bug but still very odd:
     The neutral value (the MIDI documentation says: "default") of the user macro
     controllers (GC#5...8) is zero, not 64. No offset is added to this value before it
     is scaled with the depth value and added to the destination parameter. Furthermore,
     with the maximum scale (depth = +31 or -31) you can reach a range of
     only about 63(!) of 100 units (percent or whatever it is). 
     So, I couldn't turn the effect completely down with the controller if the effect
     DRY/WET value was greater than 0 (for positive depth values) or greater than
     63 (for negative depth values). 
     This is in contrast to the behavior of the "user" knobs in the "common" settings.
     There, you can use them for both positive and negative changes of the assigned
     parameters, and their neutral value is 64! 
     Anyway, thanks for your hints! 

     Jens Groh 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


KS to Gain once again
Sunday, 15-Feb-98 08:25:32 

     130.67.1.73 writes:

     Experimenting with a piano sound, I just could not understand why the tone was
     so mellow. After careful comparison with other sources I finally found the
     culprit: KS to Gain in the DHL Common screen was set to +45. Reducing it to 0
     or some negative value cleared up the sound and brought back the high
     harmonics.

     This has been mentioned before, but some explanation for things like these is on
     the very TOP of the wishlist as far as I am concerned. Is there anything in the
     words "KS to Gain" to tell any of you that this is really a brightness control?

     Whether or not this is a bug is not my prime concern. What I want is
     DOCUMENTATION so that I don't have to waste time on things like that. 
     Besides info on "hidden features" like this one, we sorely need some documentation
     on the finer details of rates and levels. It is probably too late to ask for
     times instead of rates, but if there is some free program memory left for extra
     features, I wish Kawai would let us have parallel screens with times (possibly
     logaritmic ones) in them.

     Tore

     tl001@online.no 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


KS to Cutoff Frequency
Wednesday, 18-Feb-98 05:01:53 

     130.67.2.236 writes:

     Kenji wrote on Friday, 05-Dec-97: QUOTE

     Example: How do you set the DCF to filter the harmonics the same way for each
     note you play, instead of filtering certain frequencies?

     Could it be around 31 or 32? I wish the manual was more specific about the
     exact values that these stupid binary numbers represent...

     UNQUOTE

     I have just received my "final" K5000R, so I peeked through the German manual
     that comes with it. On the DCF page, KS TO CUT, it says that a value of
     +63 makes the Cutoff frequency follow the keyboard exactly. ("Bei einem Wert
     von +63 läuft die Cutoff-Frequenz exakt mit der Tastatur.")

     I am not sure whether this question has already been answered - or indeed
     whether the statement above describes what actually happens - but I thought it
     might be worth while to mention this extra piece of info. 

     The German manual also says that the recommended normal value for TOTAL GAIN
     in the DHL COMMON page is 50. For the rest, it is just as tight-lipped as the
     English manual, unfortunately...

     Tore

     tl001@online.no 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Re: Formant n' Macro Q
Tuesday, 17-Feb-98 16:58:59 

     192.86.155.91 writes:

     1. To save macro settings you enter "edit" mode and then tweak the macro knobs.
     If you save them from there (preferably to a different memory slot/patch
     name) it should record the changes the macro knobs have made to the sound. I know
     this works because it has been an endless source of frustration when I
     am trying to edit the properties of the macro knobs for my patches. I have to
     make the settings, save the patch, test them in play mode, and then go back into
     edit mode each time I want to change them. If I test them in edit mode it screws
     up my patch permanently. It can REALLY screw up your harmonic and
     formant filter levels (if you raise them they might "clip" and then they won't
     return to their original shape when you bring them back down - a little complicated
     to explain easily but try it if you really want to know and you'll see what I mean).
     There's something for the wishlist - a system setting that enables or disables
     this feature, although it's my guess this might not be feasible with the hardware.


<<SECOND PART CLIPPED AND MOVED TO "ADDITIVE SYNTHESIS AND FORMANT FILTER THEORY DIGEST">>


     Hope all of my little parenthesis and ramblings didn't confuse anyone, but I'm
     easily sidetracked.


     -Kenji

------------------------------------------------------------------------------------------

Re: Re: Formant n' Macro Q
Wednesday, 18-Feb-98 04:08:45 

     194.172.230.108 writes:

     Does someone know whether the "permanent macro clipping" occurs in play mode, too?
     This would mean that you should be careful not to turn the knobs too far when
     you stand on the stage and play your K5000. If you turned too far, and you
     want to return to the original sound, you have to reload the patch from memory
     which causes a little pause. Not so good...


     Jens Groh 

------------------------------------------------------------------------------------------

Re: Re: Re: Formant n' Macro Q
Friday, 20-Feb-98 10:20:10 

     194.172.230.108 writes:

     Hi all! 
     O.K., I tried it out myself, but maybe some of you are still interested in the
     answer:  No, "permanent macro clipping" does not occur in play mode, just in edit
     mode. (Candidates are: Lo Harmonics Level, Hi Harmonics Level, Even/Odd Balance.)

     Jens Groh 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Those nonlinear rates and levels
Friday, 27-Feb-98 03:48:44 

     130.67.0.118 writes:

     At one time I believe I saw one of you - maybe it was Kenji - post information
     about the nonlinear nature of harmonics levels and how you had taken this into
     account in making a sawtooth wave. I cannot find that in the digest, so maybe
     the poster could point me to the right place or repeat the explanation, please?

     The thing is, the envelopes we get from books or PC sound tools need to be
     translated into K5000 terms, and for this purpose we need to know the precise
     nature of the rates and levels. The rates of the harmonic envelopes appear
     to be logarithmic. For a given level, the time needed to reach it seems to be
     proportional to 2^(-Rate/8). (Anyone know this for sure?) However, the level
     also needs to be part of this formula for the time, and it is obviously nonlinear.
     That's why I ask for that formula posted previously.

     Tore

     tl001@online.no 

------------------------------------------------------------------------------------------

Re: Those nonlinear rates and levels
Friday, 27-Feb-98 04:36:09 

     194.172.230.108 writes:

     Hi Tore!
     It's here: (Let's see if the following becomes a link; otherwise, you will find
     it "by hand".)   Things that Should be in the Manual (Details!) 

     0.75 dB per step means, as formula:
     amplitude = max_amplitude * 2^((level_value-127)/8) 

     Or:
     level_value = 127 + 8 * log2(amplitude/max_amplitude) 

     To create a sawtooth wave, the amplitude spectrum must be inversely proportional
     to the harmonic number, that is: 
     level_value = 127 - 8 * log2(harmonic_number)

     For those who are missing the base-2 logarithm on their pocket calculator - use this:
     log2(...) = ln(...) / ln(2)

     Have fun with your additive sawtooth!

     Jens Groh 

------------------------------------------------------------------------------------------

Re: Re: Those nonlinear rates and levels
Friday, 27-Feb-98 08:42:41 

     194.172.230.108 writes:

     (Sorry Tore, I should have read your post twice before answering - just a
     "levels" discussion is in that digest section, not the "sawtooth" discussion.)

     Jens Groh 

------------------------------------------------------------------------------------------

Re: Re: Those nonlinear rates and levels
Friday, 27-Feb-98 08:56:38 

     130.67.1.157 writes:

     Thank you, this was a very clear presentation. Unfortunately, these formulas
     don't seem to explain how the time varies with the level, when keeping the rate
     constant - there appears to be an inverse logarithmic relationship here. I'll
     do some more experiments before I quote numbers.

     Tore

     tl001@online.no 

------------------------------------------------------------------------------------------

Re: Re: Re: Those nonlinear rates and levels
Friday, 27-Feb-98 10:23:17 

     194.172.230.108 writes:

     Interesting! - I have made the same experience with the rate/time behaviour. For
     the rate influence, I can confirm your 2^(-Rate/8) formula. What may be the
     algorithm that generates the level ramp? (Now, that's "reverse engineering",
     isn't it? ;-) 

     I'm planning to publish a patch of mine here (Check out soon ADDpegio) that uses
     precice rate values to create an arpeggio just from the harmonics. It's fun! 

     Ha det bra! (Nej Tore, jeg snakker ikke norsk!)

     Jens Groh 

------------------------------------------------------------------------------------------

Re: Re: Re: Re: Those nonlinear rates and levels
Sunday, 01-Mar-98 07:45:40 

     130.67.65.183 writes:

     One thing that is certain about the time needed to change a level is that this
     movement is accelerated. And it appears that the acceleration is greater for low
     levels than for higher ones. Going from level 0 to 10 in the Harmonics Envelopes
     takes around 10 secs at a rate of 42, going from 50 to 60 takes 16 secs. The
     differential equation for this movement ought to be something like l'' = a0 - kl,
     where l'' is the second derivative of the level, a0 the initial acceleration, and k is
     a constant. a0 would vary depending on where we begin the movement, and going
     downward would give us a plus sign in front of kl. My calculus is VERY
     rusty, so correct me if I am wrong.

     Whatever is the solution of this equation, I don't think it will be practicable
     to calculate such time functions whenever one wants to e.g. change a level while
     keeping the time constant. It will be simpler to make tables where one can see at
     a glance how to adjust the rate in the different cases. Hopefully, Kawai will
     come forward with some explanation for why the K5000 was designed this way and
     maybe tell us some simple ways of doing things.

     Tore

     tl001@online.no 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


MORF as built-in envelope calculator
Thursday, 05-Mar-98 07:24:07 

     194.172.230.108 writes:

     Hi all! 


     What about using the MORF feature to calculate DHE rates from given times? 

     It seems that you can set "anchor points" in MORF mode on a time scale, execute
     the morf, change back to DHE mode and see the results on a "rate scale",
     but it's not clear to me if this really solves the rate/time conversion problem. 

     Is the MORF time scale a "real" time scale? 

     Another problem seems to be that you have individual level values, taken from the
     selected morf sources, for each harmonic, but only one common time value
     for all harmonics. So you may have to make many morf calculations and combine the
     results somehow if the time values are harmonic-dependent. 

     Anyone with experiences? 


     Jens.

------------------------------------------------------------------------------------------

Re: MORF as built-in envelope calculator
Thursday, 05-Mar-98 10:25:20 

     130.67.2.216 writes:

     It seems to me that the MORF time scale is a "real" time scale, and that it DOES
     solve the rate/time conversion problem. To the extent that I have been able
     to test it, the rates are set so that the implicit "time load" of changing levels
     is taken into account. In fact, if you let one harmonic have the same value from one
     time slice to the next, it will even take care to set the corresponding levels in
     the HLE to two different values in order to avoid the skipping of stages that
     occurs when subsequent levels have the same value.

     But there appear to be some bugs in the MORF function. In particular, the Harmonics
     Levels of the MORFing source are sometimes set correctly, sometimes
     not. This will be corrected in the next version of the OS, I am sure...

     And it's a hassle that you have to save the sources before you execute the MORF.
     I wish the K5000 would read the sources being edited rather than the old
     version in flash memory.

     At least, the rate/time thing seems to be handled satisfactorily by the MORF function.

     Tore

     tl001@online.no 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Just talked to Troy at Kawai and here's the inside scoop
Friday, 03-Apr-98 19:35:59 

     208.250.189.192 writes:

     I talked to Troy just now and gained some info.

     1. There is no system 4.0 in the works right now, so I guess we just have to deal
        with the system as it is now.

     2. The K5000X has not been cancelled. It is still in the works. He says it is
        basically the K5000W and the K5000S combined. It will have 76 keys, all of the
        memory banks of the K5000S and the K5000W, a sequencer and tone generator, the
        arpeggiator, and the macro knobs - all in one machine. I didn't get a
        due date.

     3. Troy is requesting the "real life" values of all of the parameter settings and
        the algorithms used for these processes from the technicians in Japan. He thinks
        that most of it wouldn't be proprietary, so he should be able to get some of
        the info if it exists. The problem might be that it's not documented anywhere -
         although this seems highly unlikely to me...

     4. He is also looking into some of our bug reports I sent to him.


     Well, cross your fingers everybody.


     -Kenji 


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------


Attack and Release Times
 Saturday, 23-Jan-99 12:42:24 

      199.199.157.24 writes:

      Some info for the archive: I tried setting attack and release to 0 and looked at
      the .wav, and it looks like the minimum attack time is about 0.2 ms. An attack
      setting of 2 is about 1 ms and 3 is about 5 ms. The minimum release time is
      pretty long, about 20 ms. 

      In some places the waveform seems to try to start at a peak instead of a zero, or
      maybe it's starting a few ticks into the waveform, but that must account for that
      pop you get at short attack settings.


      leiter 

------------------------------------------------------------------------------------------

Re: Attack and Release Times
 Saturday, 23-Jan-99 17:02:54 

      128.8.201.10 writes:

      Interesting. Is this a common problem among 
      synthesizers? It seemed like the synth
      was never *EXACTLY* precise. Maybe those
      times are what give it that feeling.

      Ivan.


------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------