AFX vs TI89 different results! which is correct?
#1
Posted 28 April 2003 - 09:03 PM
try this...integrate
(1-i)/((x+(1-x)i)^2 between 1 and i
ti89 gives 4/5 + 2/3 i
casio crash!!!!
another big problem wich is worst!is that it gives WRONG RESULTS!!!
the result of integrating the expression....
(1-(1/2)cosx)/(1-cosx+(1/4)) between 0 and 2Pi is Pi with the AFX
AND THE REAL RESULT WITH TI89 AND WITH COMPLEX ANALISIS THEOREMS APPLIED IS 2Pi.
SO PLEASE IF ANYONE KNOWS WHAT IS WRONG HERE PLEASE TELL ME I?M BEGINING TO DESESPERATE!!!!
Since this I?ve lost my trust in AFX!!!!I don?t know if my results are right or not!!!
WHATS HAPPENING!!!!
#2
Posted 28 April 2003 - 10:20 PM
Ti has the right answer
The general integral will evaluate as:
x/2 + invTan(3 tan(x/2) )
as you can see, this is undefined at x = pi, or x = pi + 2k * pi
at x = pi, this will drop from undef (well, actually pi) to 0
quick evaluation of this graph shows that it's only valid for representing a fixed integral in the range pi < x < pi
since the AFX doesn't take this into consideration, but evaluates it normally (by substituting start and end points into the integral expression), it produces errors..
you see it yourself, if you were to substitute x with 2pi, the expression evaluates as pi (and for 0, it evals 0), so pi - 0 = pi
what you will need to do, is add pi (since it drops pi down) to the integral for every 2 pi, from and including x = pi
The AFX isn't too good at evaluating these things, so I would trust the TI more if I were you
I tried on the Classpad Emulator, and it seems to prodece the right answer, so they have improved the CAS a bit
As for the first one my Classpad emu gives the answer as:
4/5 + (2/5) i
(this answer is also obtained on the AFX, if you first cExpand the expression)
#3
Posted 29 April 2003 - 12:43 AM
I?ll take your advice..
And tell me where could I download the classpad 300 emulator?
byebye
epaloco
#4
Posted 29 April 2003 - 08:35 AM
For some reason they've taken the link away,
but the file is still hosted
So, here the direct link to the file: ClassPad Manager Setup
Note: It's a limited version of the Manager, in that it doesn't support all the features of the real counterpart. The final version, shipped with the calculator will.
The Manager will also be the communication tool between the computer and the calc (check out the manual)
#5
Posted 03 May 2003 - 06:39 AM
(At least that's what my calculus professor said, and I've verified the ti-92 answer)
#6
Posted 03 May 2003 - 02:30 PM
#7
Posted 03 May 2003 - 02:48 PM
It has nothing to do with mis-typed bracketsSo you have to verify that you type correctly ( check brackets, for example), because I, I don't have any pb with CAS !
The Casio CAS is simply not as advanced as the TI one
#8
Posted 03 May 2003 - 09:37 PM
If there's a bug in a piece of pc software its not so bad, a new version can always been downloaded easily. But it's a different story with embedded software, if the thing doesn't know that ln(-oo) isn't actually defined for real numbers it should just say it's simply not defined rather than give a false answer.
#9
Posted 03 May 2003 - 10:37 PM
same for 'oo +/ whatever*i = oo'
makes for some strange answers, like your examples shows
#10
Posted 03 May 2003 - 11:27 PM
oo+/- real should=oo
oo+/- imag=oo +/- imag
oo-oo, oo/oo not defined
etc...
AFX
oo+i=oo
ln,log(-oo)=oo But as we know the log function isn't defined for -ve numbers so must have a complex solution if any.
AS another example, taking lin(i)=(Pi.i)/2 is fine (this is done in radian mode), but for ln(oo.i) it spits out oo again where it should be oo+(Pi.i)/2........
I would be very keen to hear what your lecturers/teachers have to say about this. Granted, we don't often need to take the log of -oo but when we do it should give a logical answer based on (for instance) Euler's exponential law (I forget exactly what it's called) which says that r.exp(i)=cis(theta)
#11
Posted 03 May 2003 - 11:32 PM
#12
Posted 04 May 2003 - 12:12 AM
I tried that integration as well (AFX2.0+) and got a totally bonkers answer.
integral((1-i)/(x+(1-x)i)^2,from 1 to i)
gives back the question (ie can't integrate) but changes the sign of the lone x in the denominator!! So it's answer is:
integral((1 - i) /(-x+(1-x)i)^2,from 1 to i)
If I do this:
integral((1 - i)/ (x+i(1-x))^2,from 1 to i)
answer is zero...
If I first expand (1-x)i:
integral((1 - i))/(x+i-xi)^2,from 1 to i)
the answer is 1/10-i/10
Nomatter if I expand the (1-x)i first or not, my ti-92 consistently gives 7/5+i/5+(1/5+3i/5).i which expands to 4/5+2i/5.....
As for integral((1-(1/2)cosx)/(1-cosx+(1/4))) from 0 to 2Pi my afx also gives pi in CAS, but numerical integration gives the correct answer, 2pi.
I am beginning to think there is a rather sleepy Homer Simpson inside there pressing random buttons!!
#13
Posted 04 May 2003 - 02:03 AM
You have to use cExpand to make the integral work
I also explained the seccond integral, from 0 to 2pi
The CAS is limited, but doesn't have all that many failures as you claim.
It's when you analyze the answers you see that the errors come from it's limitations, and not random assumptions
-- EDIT --
Just sent an email to Saltire about some of these issues
#14
Posted 04 May 2003 - 02:57 AM
#15
Posted 04 May 2003 - 05:57 AM
By the distributive law, (1-x)i = i - ix and there is no reason why CAS should treat the two quantities any different (they are the same are they not?), whether it works with cExpand or not, the CAS IS flawed. I'm following up this argument because I think it is important - I've paid a lot of money for something that hasn't had all of its creases ironed out yet. This is akin to buying a new car whose headlights only work when it's raining/on the main road/parked in your garage. It's fine to use the calc for just playing games on, but I actually bought it to USE.
Good on you for contacting Saltire, I'm keen to hear their reply.
PS:
Casto Productions - I have a 2.0+ which is supposed to come with algebra2 and the revised CAS pre installed...
#16
Posted 04 May 2003 - 11:40 AM
I don't know how you managed this, but it does NOT change the sign of the lone x....integral((1-i)/(x+(1-x)i)^2,from 1 to i)
gives back the question (ie can't integrate) but changes the sign of the lone x in the denominator!! So it's answer is:
integral((1 - i) /(-x+(1-x)i)^2,from 1 to i)
again, I don't know how you got that answer.. manual substitution doesn't work unless you first cExpand your expressionIf I do this:
integral((1 - i)/ (x+i(1-x))^2,from 1 to i)
answer is zero...
Same with thisIf I first expand (1-x)i:
integral((1 - i))/(x+i-xi)^2,from 1 to i)
the answer is 1/10-i/10
I believe I explained this quite thorough in my previous reply..As for integral((1-(1/2)cosx)/(1-cosx+(1/4))) from 0 to 2Pi my afx also gives pi in CAS, but numerical integration gives the correct answer, 2pi.
The AFX can do e^i...My fx100w scientific calculator cannot raise e^i and gives an Ma ERROR. This is a limitation - it doesn't pretend to be able to do something when it can't, consequently giving a wrong answer.
Anyways, there's a big difference in how numerical software evaluates expressions, and how algebraic software does it. If you still are reffering to your logarithmic errors you recieve when dealing with oo, those are all root in the same error..
We can all agree the CAS is limited, and has a few fatal assumptions. These errors are really not acceptable for a calculator, but once you know about them you can easilly avoid the bogus answers (cause it does not generate wrong answers at will)
If not for anything else, at leaste the Casio CAS makes you think
#17
Posted 04 May 2003 - 05:47 PM
My point is this just isn't acceptable...
Believe it or not the sign sign did change for the integration! I'll tell you what software version I have soon. It may be a bug introduced to the 2.0+ CAS. My point about having to cExpand first was that CAS doesn't seem to have heard about order of operations. Actually I don't see how even that could affect it but I was as surprised as you about the answers I got!
#18
Posted 08 May 2003 - 07:53 AM
Quoting some key parts:
I'm happy at least they seem to care.. this is more of an answer than I ever got from CasioThanks for the input. Saltire developed the original CAS for Casio, but Casio has taken over development of the CAS for the past few years.
I've responded with my opinion to each issue below. I've also forwarded your email to the developers at Casio.
About infinity and imaginary numbers
An expected explanationYou are correct, and Casio is correct. infinity is not a number, but a limit. As the real part goes towards infinity, the angle between the real and imaginary parts tends towards 0, thus the limit goes to infinity.
limit(ln(-x),x=infinity) --> infinity
Because infinity is actually being treated like an extended number though, I believe you may be correct. infinity+i*pi can be considered to be the same as infinity MOST of the time, but not all of the time. Any time the pure imaginary component is used exclusively, I believe Casio will give the wrong answer.
limit(im(ln(-x)),x=infinity) --> i*pi
...but Casio returns...
limit(im(ln(-x)),x=infinity) --> 0
I do not think this is easy to correct through, because it is an underlying assumption used by the CAS. If it is changed, many integrations and limits may break.
If Casio does make this change it will take a large amount of re-testing and corrections to other parts of the CAS.
The rest was all about how the other errors were fixed on the Classpad, but they didn't know if they could do the same on the AFX due to memory constraints..
i.e, the integral of (1-(1/2)cosx)/(1-cosx+(1/4)) , in the range 0 to 2pi:
the fact that it doesn't manipulate or rearrange expressions automatically to solve themThis problem has been fixed on the Classpad. Due to memory constraints is may take time for this fix to work its way into the FX.
This has been improved on the Classpad. It is possible that for now, this improvement might not be added to the FX due to memory constraints.
#19
Posted 08 May 2003 - 02:47 PM
I read what was on the official casio site about the newest calc but I was wondering if anyone had any real dealing with an actual unit. It says something about 500k RAM and 4MB flash availible. But did Casio break the flash memory into chunks again instead of simply having the OS handle the memory allocation? And half the buttons of a normal calc aren't on the main body, so are most commands through the touch screen and drop down menus?
#20
Posted 08 May 2003 - 02:56 PM
I'll get a TI-89...
#21
Posted 08 May 2003 - 06:42 PM
#22
Posted 08 May 2003 - 08:01 PM
#23
Posted 09 May 2003 - 06:49 AM
Large screen is good too.
The whole 92/89 OS is really intuitive, its great. I wouldn't mind trying out a 49g too. One day when prices eventually fall...
#24
Posted 09 May 2003 - 02:32 PM
#25
Posted 10 May 2003 - 02:02 PM
#26
Posted 11 May 2003 - 02:40 AM
#27
Posted 11 May 2003 - 04:15 AM
#28
Posted 11 May 2003 - 06:44 AM
I use cas for checking whether my solution to things like f'''(x) is correct late at night when I'm dropping off to sleep and can't read my hand writing any more!
Peace
#29
Posted 11 May 2003 - 07:54 AM
-- EDIT --
@Crimson:
Ok, it's a two word post
But it's a reply to xyz, and a statement to show that my opinions is the same as that of his
I can agree it's not particularly contributing to the subject, but it is contributing to any coming debate obout the purpose of CAS and my stand point in such a debate
However, you have point, and I'll try to get my act together
#30
Posted 11 May 2003 - 01:32 PM
I'll leave it to you to edit or delete.
On topic: you have to remember that even though a person may be able to do something without a CAS it is sometimes faster to do it with, that I can understand.
--EDIT--
good, we have to set a good example. Also, I really don't want to see Mohamed explode and kill you .
#31
Posted 11 May 2003 - 03:03 PM
#32
Posted 11 May 2003 - 03:19 PM
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users