Casio FX991EX

[Tugdual]

I was shopping in my local super market when I found a HP300s+ for a pretty modest price. I picked it and had mixed feelings. The concept is interesting, the screen very dark, the keys are also not very good. I was quick to find that this is actually a Casio (fx92 if I'm not mistaken) and hence got interested in the Casio fx991EX. It has been a long time I was looking for a decent non programmable calculator covering binary/hex, measurement unit conversions, quick functions calculation (like f(x)=...). I ordered a fx991EX and received it today but was not expecting much after the hp300s+. What a nice surprise, this Casio is really good! First impression is the casing is less bulky, lighter. Screen is much brighter, color choice on keyboard pleasant and easy to read. Keyboard is very responsive and enjoyable (much better than the hp).
I'm very impressed with this little gem which is modest but does everything very well. It covers basic matrices, complex numbers, function solver, binary/hex/octal/decimal, integration, derivative and many other features. It has a history buffer, Ans, error management (that let you fix the issueetc...) All very simple, very natural and pretty quick.
I compared speed and some features with the Hp50g which remains the queen, my best calc all time but the casio is pretty quick and is a very decent calc for the price. I would definitely recommend it over a 35s... I'm surprised to no see more praise and thought I would share my impression and I'm curious to know yours for those who know the product.

[Csaba Tizedes]

Is it easy to disassembly? I think I will buy one 991ex and one 82ex and I install the 82ex dark number keypad keys into the 991ex.

[lrdheat]

The 991EX is a nice calculator. My only wishes would be that it could retain the work when turned off. Turning it off clears the history. Also, when creating a table of x and f(x), I wish that that "f(x)" would still be available to utilize in the calculation mode. Instead, one must retype the equation of interest.

A neat item, at least on the 991EX version, is that one can solve a derivative to =0, so as to find a maximum or minimum of a function! Likewise, one can solve an integral to equal some value of interest, say, "3", where one of the interval points over which the integral is evaluated, is defined.

[Tugdual]

(07-12-2018 08:28 PM)Csaba Tizedes Wrote:  Is it easy to disassembly? I think I will buy one 991ex and one 82ex and I install the 82ex dark number keypad keys into the 991ex.

There are screws in the back so it should be pretty easy to open.
Not sure what it looks like inside though.... also I wouldn't recommend opening it since it may loose the keyboard which would be a pity considering that it is really good.

[Albert Chan]

Just wondering about Casio REPLAY key.

Instead of grabbing the calculation FORMULA from history,
I want to grab the RESULT instead. Can this be done ?

(Similar to RPN ENTER, to push what is shown to ANS)

Retyping is tedious and error prone, and the value is not the same.
The stored result has extra precisions not shown.

I owned Casio fx-3650p, and numbers cannot be "grabbed".
Gone thru its manual, but no luck.

Same with the Casio fx-991ex ?

[ijabbott]

(07-14-2018 12:21 AM)Albert Chan Wrote:  Just wondering about Casio REPLAY key.

Instead of grabbing the calculation FORMULA from history,
I want to grab the RESULT instead. Can this be done ?

It seems not. You can do it on the TI-36X Pro / TI-30X Pro MultiView, which is the calculator I'm most likely to grab to do simple calculations with.

[brickviking]

(07-12-2018 11:08 PM)Tugdual Wrote:  

(07-12-2018 08:28 PM)Csaba Tizedes Wrote:  Is it easy to disassembly? I think I will buy one 991ex and one 82ex and I install the 82ex dark number keypad keys into the 991ex.

There are screws in the back so it should be pretty easy to open.
Not sure what it looks like inside though.... also I wouldn't recommend opening it since it may loose the keyboard which would be a pity considering that it is really good.

You won't lose the keyboard, as the circuit board is plastic-riveted to the keyboard. It makes it a pain to service a dicky switch, but at least you won't lose any keys. There's a couple of EEVBlog Youtube videos about the 991EX, this one here shows the insides.

(Post 256)

[Albert Chan]

Hi, Ian Abbott

Sorry if I go off-topic a bit ...

You mean TI can do BOTH ?

Grabbing formula from history, like the Casio
Grabbing result from history, like RPN ENTER

And, not just "shown" result, but it's full precision values ?

What about in equation mode, showing x1, x2 roots.
Can it grab those too ?

[ijabbott]

(07-14-2018 03:32 AM)Albert Chan Wrote:  Hi, Ian Abbott

Sorry if I go off-topic a bit ...

You mean TI can do BOTH ?

Grabbing formula from history, like the Casio
Grabbing result from history, like RPN ENTER

Yes, as you scroll through the history with the up and down buttons, the selection cursor alternates between selecting a result or an input in a very intuitive fashion. Pressing enter copies and pastes the selected item to the cursor position on the input line for use in further calculations.

Quote:And, not just "shown" result, but it's full precision values ?

Just the shown result. It's basically just a textual copy and paste operation.

Quote:What about in equation mode, showing x1, x2 roots.
Can it grab those too ?

Not directly, but they can be saved to variables. The "poly-solv" mode (for solving quadratic or cubic equations with real coefficients) lets you copy each root to a choice of four variables (x, y, z, or t). It also uses the a, b, c, and d variables as the input coefficients and the r.h.s. of the equation. The "num-solv" mode stores the solution in x. It also remembers the previous equation. The "sys-solv" mode (linear equations of 2 or 3 variables) stores the solutions in the x, y, and z variables. It also remembers the previous set of equations unless you select a different number of variables. The variables can be used in the normal calculator mode.

There are eight normal variables in total: x, y, z, t, a, b, c, and d, plus the "ans" variable. (There are also three vector variables plus vector answer, and three matrix variables plus matrix answer. Vectors can be up to three dimensions. Matrices can be rectangular up to three rows and three columns.) These are retained on power off, as is the calculation history.

[Albert Chan]

Quote:Just the shown result. It's basically just a textual copy and paste operation.

Too bad. TI calculator sounded almost perfect.

Have you tried to prove it is really a textual copy ?
The "tape" might had the full-precision results memorized.

On my iPad, I use sci:pro calc.
It also have a "paper tape" you can grab previous result.

Although the grabbed result look like a textual copy, it is actually in full precision:

Quote:Sci:Pro Calculator History

9 SIN COS TAN =
0.0174549999

0.0174549999 ARCTAN ARCCOS ARCSIN =
9

9 - 9 =
3.7358621062e-33

All intermediates were grabbed from the tape, then minus 9, for the calculator forensic test.

[ijabbott]

(07-14-2018 12:00 PM)Albert Chan Wrote:  

Quote:Just the shown result. It's basically just a textual copy and paste operation.

Too bad. TI calculator sounded almost perfect.

Have you tried to prove it is really a textual copy ?

Yes, by doing 5/9 in "fixed 2" format, giving a displayed result of 0.56. The "ans" variable is full precision, so ans*9 gives 5.00. But copy and pasting the 0.56 result and multiplying by 9 gives 5.04.

You can copy and paste the expression that produced the result instead of copying the result, but if the expression is part of a chain using the "ans" variable, then it will use the current "ans" instead of the original "ans", so that might not work.

Quote:The "tape" might had the full-precision results memorized.

Only the last result in "ans" (and values in other variables) are full precision. The "tape" is just nicely marked up text.

Quote:On my iPad, I use sci:pro calc.
It also have a "paper tape" you can grab previous result.

Although the grabbed result look like a textual copy, it is actually in full precision:

Quote:Sci:Pro Calculator History

9 SIN COS TAN =
0.0174549999

0.0174549999 ARCTAN ARCCOS ARCSIN =
9

9 - 9 =
3.7358621062e-33

All intermediates were grabbed from the tape, then minus 9, for the calculator forensic test.

On the TI (and also Casio calculators), you'd just use the ans (or Ans) variable to do that.

The TI's result of the forensic test using "ans" isn't as precise as your Sci-Pro (which I guess uses 64-bit IEEE-754 format internally?). The displayed 10-digit result is 9.000001077 and the internal 13-digit result is 9.000001077372. That's a separate issue though.

For comparison, the fx-991EX does better on the forensic test with a displayed 10-digit answer of 9.000000012 and an internal 15-digit answer of 9.00000001178352. (I'm not sure the calculator really has an internal 15-digit precision answer, as simple calculations such as 1e-12+1-1 = 1e-12 versus 1e-13+1-1 = 0 suggests it only has 13 digits.)

On the TI, if you copy and paste the 10-digit displayed results from the history after each individual operation, the TI gives a final answer of 8.989942773, which is obviously less accurate than using the "ans" variable.

[Albert Chan]

Quote:I'm not sure the calculator really has an internal 15-digit precision answer, as simple calculations
such as 1e-12+1-1 = 1e-12 versus 1e-13+1-1 = 0 suggests it only has 13 digits.

That was Casio trying to be "smart", by flushing close-to-zero to zero

http://people.ku.edu/~wzhuang/CAMseminar...ecrets.pdf

I was unable to figure out what the algorithm is:

Casio fx-3650p:
1e10 + 0.94 - 1e10 ==> 0
1e10 + 0.95 - 1e10 ==> 1

Casio fx-260 Solar
1e10 + 0.79 - 1e10 ==> 0
1e10 + 0.80 - 1e10 ==> 0.8
1e10 + 0.89 - 1e10 ==> 0.8
1e10 + 0.90 - 1e10 ==> 0.9

I really hate this feature. (above look like a bug to me)

With it on, 1e-13+1-1 = 0, can fx-991EX still claim 15-digits precision ?

[lrdheat]

Page 3 of my HP 42S manual says that it calculates with 15 digits internally, displays up to 12 digits.

If I input 1E-11 enter 1 + 1 -, I get 1E-11

If I input 1E-12 enter 1 + 1 -, I get 0

Similar behavior to the CASIO.

[lrdheat]

(Similar behavior in principle)

[lrdheat]

For that matter, in home, same exact behavior on the Prime.

[Tugdual]

Test 1
Enter this and equal
0.333 333 333 333
Ans=3.33...

Test 2
Just one more 3
0.333 333 333 333 3
Ans=\( \frac {1}{3} \)

And then
0.333 333 333 333 3 - 0.333 333 333 333
Ans=\( 3.00x10^{-13}\)

Test 3
One more 3
0.333 333 333 333 33
Ans=\( \frac {1}{3} \)

0.333 333 333 333 33 - 0.333 333 333 333 3
Ans=0

All answers in red are wrong. The calculator seems to be using more than 12 digits but essentially for rounding purposes.
If you also enter manually 3.14159265359 and press enter Ans=\(\pi\) which shows there is no symbolic calculation but pattern detection during a final rendering phase, during which fraction simplification also occurs.

[Tugdual]

The solar cell is a disappointment. If you remove the LR44 battery the calculator will power up and perform basic calculations but a sum will shut it down. That makes me think there is possibly a boost on the processor which is switched on for more complex calculations. As a result I regret I didn't pick the fx-570EX that has a solid battery only.

[Albert Chan]

(07-15-2018 09:17 AM)Tugdual Wrote:  The solar cell is a disappointment. If you remove the LR44 battery the calculator will power up
and perform basic calculations but a sum will shut it down.

Even under good light ?
Is it possible it were still doing sum but don't have the juice to show it ? :-D

Anyway, solar + battery is better than battery alone.

My Casio FX-115MS had a dead battery, but still work under good light.
The back is somehow stuck even with all 3 screws removed (glued stuck ?)
Cannot replace battery without breaking the calculator ...

FX-115MS also had some weird precedence, had FX-991EX fixed it ?

Keys: 6 / 2 ( 1 + 2 ) ==> 1 ?
Keys: SQRT 6 RCL Y ==> sqrt(6 * y) instead of sqrt(6) * y

Keys: SIN ( 30 DEGREE ) SQR ==> 0 ?

Above actually does sin(30 ^ 2) = sin(900) = sin(0) = 0
To do sin(30) ^ 2, had to do ( SIN 30 DEGREE ) SQR ==> 0.25

[John Keith]

(07-15-2018 01:09 AM)lrdheat Wrote:  For that matter, in home, same exact behavior on the Prime.

In fact, the same on all Saturn-based calculators since the 71B, including 32, 42, 28/48/49/50, etc.

[ijabbott]

(07-15-2018 11:51 AM)Albert Chan Wrote:  FX-115MS also had some weird precedence, had FX-991EX fixed it ?

Keys: 6 / 2 ( 1 + 2 ) ==> 1 ?

The EX gives the same answer, but automatically adds parentheses around the '2(1+2)' after you press '='. In other words, implicit multiplication is given higher precedence than '×' or '÷', but at least it visually removes any ambiguity.

Quote:Keys: SQRT 6 RCL Y ==> sqrt(6 * y) instead of sqrt(6) * y

Keys: SIN ( 30 DEGREE ) SQR ==> 0 ?

Above actually does sin(30 ^ 2) = sin(900) = sin(0) = 0
To do sin(30) ^ 2, had to do ( SIN 30 DEGREE ) SQR ==> 0.25

The EX behaves the same as the ES models here, not like the MS models. Pressing the '√' key or the 'sin' key automatically adds an opening parenthesis and the function argument extends to the matching closing parenthesis. (In math input mode, '√' uses a radical bar instead of parentheses, and you have to "cursor right" out of the radical bar instead of typing a closing parenthesis.)