Has anyone fooled with the last version of the speak model put up? I
have and am seeing some differences of PCT versus SD. I created the
following volume disturbance:
5*PULSE(20, 10)+10*PULSE(40, 10 )+15*PULSE(60,10)+30*PULSE(80,
10)+50*PULSE(100,10)
And then I looked at the graph of volume disturbance versus uttered
volume for each one.
Anyone want to discuss this?
From: David Goldstein
Subject: PCT versus SD
Date: 9/14/98
From [ Marc Abrams (980914.1116) ]
From: David Goldstein
Subject: PCT versus SD
Date: 9/14/98Has anyone fooled with the last version of the speak model
put up? I have and am seeing some differences of PCT
versus SD. I created the following volume disturbance:
As Bill pointed out so eloquently in his post this morning
Powers, 980913.0758 ) It's not so much SD _vs_ PCT as much
as it is how SD _has_ been used. You can virtually model
_any_ number of system types with SD _including_ PCT. It has
_not_ been done. I believe this exchange between Bill and
Bob is extremely important for both. SD _NEEDS_ PCT _badly_.
PCT can benefit greatly from using SD ( although it does not
_need_ to use it to model with it ). It is a tool that will
allow facilitation and exploration of PCT concepts and
ideas. I for one hope the interest continues for both Bob
and Bill, We all benefit from the relationship. I would like
to thank both for their patience. This is _not_ a PCT vs. SD
thing.
Now on to your question 
5*PULSE(20, 10)+10*PULSE(40,
10 )+15*PULSE(60,10)+30*PULSE(80,
10)+50*PULSE(100,10)
Seems like one helluva disturbance to me. .
OK, will do, but I need to clarify some stuff. This was done
in speak9.mdl? Attached to Powers (980911.0938)? Were the
variables "uttered sound intensity" or "perceived volume"?
and "volume disturbance" ?
And then I looked at the graph of volume disturbance versus
uttered volume for each one.
Anyone want to discuss this?
Sure would. just need to clarify. Thanks
Marc
Yes Marc it was speak9.mdl.
The variables which I graphed were: volume disturbance and uttered sound
intensity.
With the disturbance I posted the last time, I looked at the above graph
for the SD formulation and for the PCT formulation.
I am simple trying to understand why the results are different.
The reason I chose these variables was the idea that in a control
system, the output quantity (I assume uttered sound intensity) is a
mirror image of the disturbance quantity (I assume the volume
disturbance).
I like the idea that SD may allow a broader look at things. However, if
From: David Goldstein
Subject: Re: PCT versus SD [ Marc Abrams (980914.1116) ]
Date: 9/14/98
at the level of a single control system, one formulation leads to a different result than the other one, this diserves discussing. Marc Abrams wrote:
>From [ Marc Abrams (980914.1116) ]
>From: David Goldstein
>Subject: PCT versus SD
>Date: 9/14/98
>
>Has anyone fooled with the last version of the speak model
>put up? I have and am seeing some differences of PCT
>versus SD. I created the following volume disturbance:As Bill pointed out so eloquently in his post this morning
Powers, 980913.0758 ) It's not so much SD _vs_ PCT as much
as it is how SD _has_ been used. You can virtually model
_any_ number of system types with SD _including_ PCT. It has
_not_ been done. I believe this exchange between Bill and
Bob is extremely important for both. SD _NEEDS_ PCT _badly_.
PCT can benefit greatly from using SD ( although it does not
_need_ to use it to model with it ). It is a tool that will
allow facilitation and exploration of PCT concepts and
ideas. I for one hope the interest continues for both Bob
and Bill, We all benefit from the relationship. I would like
to thank both for their patience. This is _not_ a PCT vs. SD
thing.Now on to your question
> 5*PULSE(20, 10)+10*PULSE(40,
10 )+15*PULSE(60,10)+30*PULSE(80,
>10)+50*PULSE(100,10)Seems like one helluva disturbance to me.
.
OK, will do, but I need to clarify some stuff. This was done
in speak9.mdl? Attached to Powers (980911.0938)? Were the
variables "uttered sound intensity" or "perceived volume"?
and "volume disturbance" ?>And then I looked at the graph of volume disturbance versus
>uttered volume for each one.>Anyone want to discuss this?
Sure would. just need to clarify. Thanks
Marc
[From Bill Powers (980915.0713 MDT)]
From: David Goldstein
Subject: Re: PCT versus SD [ Marc Abrams (980914.1116) ]
Date: 9/14/98
With the disturbance I posted the last time, I looked at the above graph
for the SD formulation and for the PCT formulation.I am simple trying to understand why the results are different.
In this model, the disturbance MULTIPLIES the output, rather than adding to
it. Notice that you have left out a term of 1 in your disturbance. Since
each pulse generator has a value of zero before and after the pulse, when
they are all zero the multiplier becomes zero and there is no feedback. In
the original setup, the disturbance is defined as 1 + pulse(...), so its
value is either 1 or 2 -- but never zero.
Also, since your disturbance can have several pulse generators on at the
same time, the total magnitude of the multiplier can be greater than 2. The
system is optimized for a loop gain somewhat greater than 2, but if your
gain gets too large, the system will blow up.
I should have called attention to the fact that the disturbance in this
model is not our usual additive disturbance, but a multiplicative one.
Just to be sure we're on the same page, how about posting your model with
the modified disturbance?
Best,
Bill P.
Attached is the modified model with the disturbance I spoke about.
Why multiply rather than add like the usual PCT model?
Bill Powers wrote:
[From Bill Powers (980915.0713 MDT)]
>From: David Goldstein
>Subject: Re: PCT versus SD [ Marc Abrams (980914.1116) ]
>Date: 9/14/98>With the disturbance I posted the last time, I looked at the above graph
>for the SD formulation and for the PCT formulation.
>
>I am simple trying to understand why the results are different.In this model, the disturbance MULTIPLIES the output, rather than adding to
it. Notice that you have left out a term of 1 in your disturbance. Since
each pulse generator has a value of zero before and after the pulse, when
they are all zero the multiplier becomes zero and there is no feedback. In
the original setup, the disturbance is defined as 1 + pulse(...), so its
value is either 1 or 2 -- but never zero.Also, since your disturbance can have several pulse generators on at the
same time, the total magnitude of the multiplier can be greater than 2. The
system is optimized for a loop gain somewhat greater than 2, but if your
gain gets too large, the system will blow up.I should have called attention to the fact that the disturbance in this
model is not our usual additive disturbance, but a multiplicative one.Just to be sure we're on the same page, how about posting your model with
the modified disturbance?Best,
Bill P.
Speaking Effort= INTEG (
effort adjustment,
INTENDED PERCEIVED VOLUME/PERCEIVED OUTPUT SENSITIVITY)
~ NSU
~ |
effort adjustment=
IF THEN ELSE(formulation switch = 0,volume error*GAIN,volume error / PERCEIVED OUTPUT SENSITIVITY\
/time to adjust effort)
~ NSU/Second
~ |
PERCEIVED OUTPUT SENSITIVITY=
5
~ PWpSqM/NSU
~ |
time to adjust effort=
100
~ Second
~ |
volume error=
IF THEN ELSE(error test=1,0,
INTENDED PERCEIVED VOLUME-delayed perceived volume)
~ PWpSqM
~ |
error test=
0
~ Dmnl
~ Switch to force the error to zero so that the resulting behavior can \
be observed.
>
formulation switch=
0
~ Dmnl
~ This is a contants set it to 1 for a standard SD formulation and 0 \
for the formulation used in speak5 by Bill
>
delayed perceived volume=
DELAY3(perceived volume,PERCEPTUAL DELAY)
~ PWpSqM
~ |
OUTPUT SENSITIVITY=
5
~ (watt/SqMeter)/NSU
~ |
INPUT SENSITIVITY=
1
~ PWpSqM/(watt/SqMeter)
~ |
INTENDED PERCEIVED VOLUME=
30
~ PWpSqM
~ |
received sound intensity=
Uttered sound intensity*volume disturbance
~ watt/SqMeter
~ |
perceived volume=
received sound intensity*INPUT SENSITIVITY
~ PWpSqM
~ |
PERCEPTUAL DELAY=
0.25
~ Second
~ |
Uttered sound intensity=
Speaking Effort*OUTPUT SENSITIVITY
~ watt/SqMeter
~ |
volume disturbance=
PULSE(20, 10)+5*PULSE(40, 10 )+15*PULSE(60,10)+50*PULSE(80, 10)+50*PULSE(100,10\
)
~ Dimensionless
~ |
GAIN=
0.125
~ NSU/PWpSqM
~ |
From: David Goldstein
Subject: Re: PCT versus SD [From Bill Powers (980915.0713 MDT)]
Date: 9/15/98
********************************************************
.Control
********************************************************~
Simulation Control Paramaters
>
FINAL TIME = 120
~ Second
~ The final time for the simulation.
>
INITIAL TIME = 0
~ Second
~ The initial time for the simulation.
>
SAVEPER = 0.25
~ Second
~ The frequency with which output is stored.
>
TIME STEP = 0.025
~ Second
~ The time step for the simulation.
>
\\\---/// Sketch information - do not modify anything except names
V300 Do not put anything below this section - it will be ignored
*View 1
$0,0,Arial|10||0-0-0|0-0-0|0-0-0|-1--1--1|-1--1--1
10,1,Speaking Effort,517,152,40,20,3,3,0,0,0,0,0,0
10,2,received sound intensity,86,300,72,9,0,3,0,0,0,0,0,0
10,3,volume error,186,115,37,9,0,3,0,0,0,0,0,0
10,4,INTENDED PERCEIVED VOLUME,171,32,40,20,8,3,0,0,0,0,0,0
1,5,4,3,1,0,0,0,0,0,0,-1--1--1,1|(171,75)|
10,6,volume disturbance,274,330,58,9,0,3,0,0,0,0,0,0
1,7,6,2,1,0,0,0,0,64,0,-1--1--1,1|(164,325)|
10,8,delayed perceived volume,69,124,40,20,3,3,0,0,0,0,0,0
1,9,8,3,0,0,0,0,0,64,0,-1--1--1,1|(122,119)|
10,10,PERCEPTUAL DELAY,170,178,49,20,8,3,0,0,0,0,0,0
1,11,10,8,1,0,0,0,0,64,0,-1--1--1,1|(138,144)|
10,12,Uttered sound intensity,453,301,69,9,0,3,0,0,0,0,0,0
1,13,1,12,0,0,0,0,0,64,0,-1--1--1,1|(485,225)|
1,14,12,2,0,0,0,0,0,64,0,-1--1--1,1|(277,300)|
10,15,OUTPUT SENSITIVITY,568,257,47,21,8,3,0,0,0,0,0,0
1,16,15,12,1,0,0,0,0,64,0,-1--1--1,1|(530,309)|
10,17,perceived volume,82,206,51,9,0,3,0,0,0,0,0,0
1,18,2,17,0,0,0,0,0,64,0,-1--1--1,1|(84,259)|
1,19,17,8,0,0,0,0,0,64,0,-1--1--1,1|(77,177)|
10,20,INPUT SENSITIVITY,144,256,46,22,8,3,0,0,0,0,0,0
1,21,20,17,1,0,0,0,0,64,0,-1--1--1,1|(141,223)|
12,22,0,307,254,286,9,0,4,0,2,-1,0,0,0,0-0-0,0-0-0,|10||192-192-192
----------------------------------------------------------------------------------------------------------------------------------------------
12,23,0,304,228,40,20,8,4,0,16,-1,0,0,0,0-0-0,0-0-0,|10|B|0-0-0
NERVOUS SYSTEM UNITS
12,24,0,277,274,56,9,0,4,0,16,-1,0,0,0,0-0-0,0-0-0,|10|B|0-0-0
PHYSICAL UNITS
1,25,3,31,1,0,0,0,0,64,0,-1--1--1,1|(216,150)|
10,26,GAIN,421,33,18,9,0,3,0,0,0,0,0,0
12,27,48,308,151,8,8,0,3,0,0,-1,0,0,0
1,28,30,1,4,0,0,22,0,0,0,-1--1--1,1|(435,151)|
1,29,30,27,68,0,0,22,0,0,0,-1--1--1,1|(348,151)|
11,30,48,387,151,6,8,34,3,0,0,1,0,0,0
10,31,effort adjustment,387,168,50,9,32,3,0,2,0,0,0,0,0-0-0,0-0-0,|10||0-0-255
10,32,formulation switch,517,48,40,40,11,3,0,4,0,0,0,0,0-0-0,255-255-0,|10||128-0-0
10,33,error test,58,45,30,30,11,3,0,4,0,0,0,0,0-0-0,255-255-0,|10||255-255-0
1,34,33,3,0,0,0,0,0,64,0,-1--1--1,1|(122,80)|
10,35,time to adjust effort,423,200,58,9,0,3,0,0,0,0,0,0
1,36,35,31,0,0,0,0,0,64,0,-1--1--1,1|(410,188)|
12,37,0,216,227,40,20,8,4,0,16,-1,0,0,0,0-0-0,0-0-0,|10|B|0-0-0
PERCEIVED PHYSICAL UNITS
1,38,40,30,1,0,0,0,0,64,0,-1--1--1,1|(347,102)|
12,39,0,256,49,1,205,3,0,0,3,-1,0,0,0,192-192-192,0-0-0,|10||192-192-192
10,40,PERCEIVED OUTPUT SENSITIVITY,297,40,55,29,8,3,0,4,0,0,0,0,0-0-0,-1--1--1,|10||0-0-0
1,41,32,30,1,0,0,0,0,64,0,-1--1--1,1|(439,108)|
1,42,26,30,1,0,0,0,0,64,0,-1--1--1,1|(397,82)|
///---\\\
:GRAPH GRAPH
:TITLE GRAPH
:SCALE
:VAR delayed perceived volume
:SCALE
:VAR volume error
:GRAPH DIST_VERSUS_OUTPUT
:TITLE dist versus output
:SCALE
:VAR volume disturbance
:LINE-WIDTH 3
:SCALE
:VAR Uttered sound intensity
:LINE-WIDTH 5
:L <%^E!@
1:Current.vdf
9:Current
15:0,0,0,0
19:100,0
5:volume disturbance
[From Bill Powers (980916.0643 MDT)]
From: David Goldstein
Subject: Re: PCT versus SD [From Bill Powers (980915.0713 MDT)]
Date: 9/15/98
Why multiply rather than add like the usual PCT model?
I was getting ready to show that -- well, never mind, it will be done
another way.
1. Change the formula for "Received sound intensity" so your disturbance is
added rather than multiplied.
2. Make sure "Time to adjust effort" is 2.
3. Set Gain to 0.1
Now your model will work the same way for Bill and Bob.
Best,
Bill P.