question for math wizards

[aajgss]

I am using a pressure sensor that gives me .2V to 4.8V from 0 to 6 KPA

I want to convert that firstly into mm H20 and then mm H2O to Knots. The first bit is easy. The equation for the mm H2O to knots is 7.7538*(mmH2O ^ .5).

That where I am having the trouble with no FP Math

All ideas welcome

aajgss

[mister_e]

SQR(mmH2O)*(504/65)

SQR(mmH2O)*(1543/199)

SQR(mmH2O)*(2551/329)

SQR(mmH2O)*(31116/4013)

Hint: Use MiscEl ;)


Careful with the math overflow

[aajgss]

Awesome stuff Steve

Thanks for your help

aajgss

ps not sure what you mean by MiscE?

[mackrackit]

Give this a try... :D
MiscEL (http://lmgtfy.com/?q=MiscEl)

[aajgss]

Got It!

Thought it was a ! and not an l - old eyes no glasses. Often get told I need them - maybe I do


Thanks
aajgss

[mackrackit]

Got It!

Thought it was a ! and not an l - old eyes no glasses. Often get told I need them - maybe I do


Thanks
aajgss
I know what you mean... without mine I walk into walls .. try stairs with a new pair of bifocals :eek:

[mister_e]

I'm sure in the same boat... without mine I can't even say if it's an elephant ... or a supermarket :eek:

[Archangel]

Got It!

Thought it was a ! and not an l - old eyes no glasses. Often get told I need them - maybe I do


Thanks
aajgss
You mean you tried to read that text and type it ? !!!

Thank Jobs / Gates / Linus for copy/paste !

[aajgss]

Ok,

Lets forget I said the first bits easy!!!

Instead of converting to mmH2O and then converting that to Knots, I decided to go straight from ADC counts to Knots.

So the equation comes out to be (SQR(ADC Counts))* ( 484/79).

Excel comes up with the correct answer. for 1 ADC count the result comes out at 6.1266 knots. Doing the same calculation in a PIC the result is 6 knots. It appears that the result is the integer part only.

The numbers I am trying to get is in the table

mmH2O ________Knots
15_____________ 30
26.5 ___________40
41.5 ___________50
60 _____________60
81.5 ___________70
106.5 __________80
134.5 __________90
166.5 _________100
201 ___________110
239.5 _________120


How do I calculate the correct value and still heed Steves advice to be careful of Math Overflow?


Thanks
aajgss

[Archangel]

Hello aajgss,
What level of resolution are wanting to display, I mean how many decimal places ?
You could multiply up by 10, 100, 1000 and display using DEC modifier and "." .
I didn't notice which PIC you are using but 18Fs can use long variables which if I am not mistaken are 32 bit vars.

[aajgss]

Hi Archangel

Resolution I would like is to 1 decimal place. The project I am working on is an electronic manometer to check calibration of the Air Speed Indicator for gliders.
The limits are critical at 30 knots. The ASI must be within +/- 1.5 knots at 30 knots.

I am using an 18F14K50. I did try a LONG variable and multiplied the counts by 10000, but still got the same results. I looked like I was getting a 16 bit math overflow. Most likely cause was me not knowing what I am doing. Is it just a case of using a LONG identifier or is there something else I need to do. There is an option in MCS to use PBPL. Is PBPL another part of PBP, a different program all together or some type of addin? I find the documentation for this kind of simple info a bit difficult to locate.


Thanks
aajgss

[Acetronics2]

Hi,

Your answer is here :


Excel comes up with the correct answer. for 1 ADC count the result comes out at 6.1266 knots


you have to use a signal amplifier to get more ADC counts for the same outputted voltage ...

A logarithmic amplifier could be a nice idea to get a linear ADC input voltage Vs Speed.

I remember some really good work had been done here for Vu-meters ... ( a greek project ... ;) )

CA3089 or NE 604 could be some interesting keywords to try ...

have fun

Alain

[Demon]

...I am using an 18F14K50. I did try a LONG variable and multiplied the counts by 10000, but still got the same results. I looked like I was getting a 16 bit math overflow. Most likely cause was me not knowing what I am doing. Is it just a case of using a LONG identifier or is there something else I need to do. There is an option in MCS to use PBPL. Is PBPL another part of PBP, a different program all together or some type of addin? I find the documentation for this kind of simple info a bit difficult to locate.
...


I see it at least in section 3 (pages 17 and on), and in section 4 on Variables (page 23) and Division (page 37). It could be in other places; no fast way to use SEARCH in a book besides the table of contents (no index that I can see).

(I really need to reread the manual, they've added important information in there)

[aajgss]

Archangel

Thanks for the info. I think that may be the way to go. A 12 or 14 bit ADC may give me enough resolution.

Robert,
You hit the nail on the head. Exactly what I was meaning.

From the Manual
"PBP is the DOS version of the compiler. PBPW is the Windows version
of the compiler and can take advantage of all the PC’s memory. PBPL is
the long version of the compiler and can use 32-bit variables, as well as
all of the PC’s memory under Windows. PBPL can only be used with the
PIC18 devices.",

But when I use a LONG variable and get the same answer as not using it I am stuck between "Did I do it right" or " Do I have the right version. Do I have PBPL" Where does it say "The version you are using is PBPL or PBP? As far as I know I am using PicBasic Pro. The executable that MCS or MPASM runs is PBP.EXE

Remember, I am only new at this and finding this type of info that others take as commonplace is not so straight forward for a beginner like me. I also do a lot of searching this and other forums/ websites, but again I feel that I am expected to know this as its "simple stuff".

Please note, I am not trying to offend or anything like that, just trying to highlight some of the frustrations I have with finding / understanding the "simple stuff"

Regards
aajgss

[aajgss]

Sorry Acetronics, I did mean to address you, instead I put Archangel

aajgss

[Demon]

...But when I use a LONG variable and get the same answer as not using it I am stuck between "Did I do it right" or " Do I have the right version. Do I have PBPL" Where does it say "The version you are using is PBPL or PBP? As far as I know I am using PicBasic Pro. The executable that MCS or MPASM runs is PBP.EXE

Remember, I am only new at this and finding this type of info that others take as commonplace is not so straight forward for a beginner like me. I also do a lot of searching this and other forums/ websites, but again I feel that I am expected to know this as its "simple stuff".

Please note, I am not trying to offend or anything like that, just trying to highlight some of the frustrations I have with finding / understanding the "simple stuff"

Regards
aajgss


Can you give us more details on exactly what tools (software/hardware) you use?

I have yet to use long division either. Off the top of my head, I recall a setting in
MeLabs programmer to state what compiler/options to use.

EDIT: Correction, in MicroCode Studio.

- View.
- Compile & Program options

There is a toggle to Use PBPL. (I have MicroCode Studio Plus version, can't comment for sure on basic version)

I would start by enabling that if I were to use PBPL on a PIC 18F device.

The first thing I would do is a simple test program connected to a LCD.
That way I would make sure to learn how to use the syntax properly
before integrating it into a larger program.

[aajgss]

Thanks Robert,

I am using PBP V2.6 with a PicKit 2 programmer and a low pin count USB board and 18F14K50.

I have had some success by enabling PBPL in MCS.

The values I am after and the values I get are really close. Probably close enough for this particular project.

mmH2O _______Knots ________Pic Output to serial port
15_____________ 30______________ 30.09
26.5 ___________40_______________39.56
41.5 ___________50_______________50.16
60 _____________60______________60.19
81.5 ___________70______________69.61
106.5 __________80______________80.2
134.5 __________90______________89.72
166.5 _________100______________99.75
201 ___________110_____________109.79
239.5 _________120_____________119.8


To get these values I used




knots var long
knotsRem var long
counts var long
pause 100
MAINLOOP
for counts = 0 to 10000
knots = ((SQR(counts*100))*1563)/10000
knotsRem = (SQR(counts*100)*1563)//10000
serout2 portb.7,396,[ #Knots,".",dec3 knotsRem," ",#counts,","," Knots",13,10]
'pause 10
next
Goto Mainloop


But there are a few "oddities" I dont know how to deal with.One thing I did see was a printout like this. 1st coulmn is Knots and the second
the counts corresponding to that value.the question I have is why so many values the same, and then it jumps by about .5 knots. As I said, not critical in this instance, but may be in a future project and if I can get it sorted now all the better.

143.708 8404, Knots
143.708 8405, Knots
143.708 8406, Knots
143.708 8407, Knots
143.708 8408, Knots
143.271 8409, Knots
143.271 8410, Knots
143.271 8411, Knots
143.271 8412, Knots
143.271 8413, Knots
143.271 8414, Knots
143.271 8415, Knots

I am guessing its the integer math thing again, just after some clarification


Acetronics,
Yes, its not me its Excel !!

I did recalculate it in Excel by using the "ROUND" function, but his rounds up or down. I would expect that integer math always rounds down. I am guessing correctly?
Thanks
aajgss

[Acetronics2]

Acetronics,

Yes, its not me its Excel !!



Are you really sure your Math is good ???

I'm not ...



for counts = 0 to 10000
knots = ((SQR(counts*100))*1563)/10000
knotsRem = (SQR(counts*100)*1563)//10000



1) due to your ADC ... "Counts" granularity is 1/1024 or 1/4096 ( let's dream ) ... just a detail ... but important.

2) knots = ((SQR(counts*100))*1563)/10000 = SQR ( counts ) * 1563 / 1000 ... YESSS, no need to multiply "Counts", as granularity will remain the same ... !!!

3) you can't use the remainder trick ... as it is a SQUARE function ... remember SQR ( A ) + SQR ( B ) != SQR ( A+B )

Ok, I don't have the easy solution at the time ... but' Im on the way !!!

Alain

[aajgss]

Hi Acetronics,

Here is some brief details of what I did and why I did it.

ADC CountsADC VoltsmmH2OKnots10.00030.0411.5620.00060.0812.2130. 00090.1222.71

[aajgss]

Hi Acetronics,

Here is some brief details of what I did and why I did it.
ADC Counts ADC Volts mmH2O Knots

1 0.0003 0.041 1.56

2 0.0006 0.081 2.21

3 0.0009 0.122 2.71



now using

knots = ((SQR(counts))*1563)

and looking at the integer value that I would expect would be returned

ADC Counts ADC Volts mmH2O SQR(counts) Integer valuegives me Knots
1 0.0003 0.041 1 1 1.56

2 0.0006 0.081 1.414 1 1.56

3 0.0009 0.122 1.73 1 1.56



now using

knots = ((SQR(counts*100))*1563)/10000



ADC Counts ADC Volts mmH2O SQR(counts*100) Integer valuegives me Knots

1 0.0003 0.041 10 10 1.56
2 0.0006 0.081 14.14 14 2.188
3 0.0009 0.122 17.32 17 2.657


So I guess what I am saying is that it was the only way I could see of getting close to the resolution I needed.

As for the remainder, that is the .56 at 1 count, .188 at 2 counts etc.

Anyway, that was my train of thought. I am sure there must be a simpler way, i'm just not able to see it. Well I'm off to find that section of the forum for ideas to improve PBP "Floating Point Math"

regards
aajgss

[Demon]

Yup, I tried three counts in Open Office Calc using 2 versions of the formula and got identical results:

((SQRT(cell))*1563)/1000
((SQRT(cell*100))*1563)/10000



1 1.563
2 2.210
3 2.707


Are those above the expected results?


Or are these?



1 1.560
2 2.188
3 2.657

[aajgss]

Hi Robert,

I think the difference comes from the multiply by 100. This gives the result of for example 3 counts 17 and not 1.732.
But as I said one decimal place is close enough for me.

aajgss

[aajgss]

Thanks to all for you help to this stage.

Now I'm off to create a new thread for the next bit. I have the hardware setup and producing data, just a bit of trouble reading it.

aajgss