Random Numbers

ChedderCheeser · Aug 3, 2011

Hi guys. I am trying to figure out how I can create RANDOM numbers within a certain range. The first one (w0) needs to be between 10 and 50. The second (w1) needs to be between 25 and 175. Any ideas?

eclectic · Aug 3, 2011

ChedderCheeser said:
Hi guys. I am trying to figure out how I can create RANDOM numbers within a certain range. The first one (w0) needs to be between 10 and 50. The second (w1) needs to be between 25 and 175. Any ideas?

Try a quick search.
http://www.picaxeforum.co.uk/showthread.php?7799-Random-Numbers&highlight=Random

e

Jamster · Aug 3, 2011

Inacurate but here is how i do it

Random creates a number between 0 and 65535 so:

65535/1638		Creates a number between 0 and 40
Ans+10		Creates a number between 10 and 50

65535/436.9		Creates a number between 0 and 150
Ans+25		Creates a number between 25 and 175

BillyGreen1973 · Aug 3, 2011

Jamster said:
Inacurate but here is how i do it

Random creates a number between 0 and 65535 so:

65535/436.9

Creates a number between 0 and 150

Ans+25

Creates a number between 25 and 175

Not possible directly due to integer maths only.

You could look at the 'min' and 'max' commands to limit the results within a certail range

nick12ab · Aug 3, 2011

possibly a bit inefficient, but

Code:

main:
   random w0
   if w0 < 25 or w0 > 75 then main

Exchange w0 and main for any other variable/label. Code doesn't advance until the number is between 25 and 75 inclusive.

eclectic · Aug 3, 2011

Based on hippy's thread in post #2

Works on a real 08M2

Code:

#picaxe 08M2
pause 1000

main:
For b10 = 0 to 255
RANDOM w0
 w1 = w0 // 150 + 25
 
 sertxd (#W1,", ")
 
next

e

srnet · Aug 3, 2011

ChedderCheeser said:
Hi guys. I am trying to figure out how I can create RANDOM numbers within a certain range. The first one (w0) needs to be between 10 and 50. The second (w1) needs to be between 25 and 175. Any ideas?

You cant.

PICAXE has a 'pseudo' random generator, which may be good enough for you. Its important to seed the number correctly, or each time you start it you may (should really) get the same sequence. Easiest way would be to check the status of an external event, such as a button press, before starting the routine.

Jamster · Aug 3, 2011

BillyGreen1973 said:
Not possible directly due to integer maths only.

You could look at the 'min' and 'max' commands to limit the results within a certail range

The "." was not meant to be there, oops...

nick12ab · Aug 3, 2011

I decided to run a test on a 40X2 to see what the sequence was. This was the sequence after running it for a while:
With word variable:

Code:

Ran it a second time and the exact same sequence was reproduced so seeding is very important.

With byte variable:

Code:

Just a shorter sequence.

inglewoodpete · Aug 4, 2011

When working with random numbers, never use a byte variable with the RANDOM command. Never use MIN and MAX as they polarise the values. Preserve the value of the random register.

If you have an ADC input in use, use it to create the seed. If you don't then use a spare ADC input pin (left open circuit) as a seed. If need be, add a length of insulated wire to the input. One of the rare times that noise (inconsistent readings) is good.

inglewoodpete · Aug 4, 2011

Double double post post

graynomad · Aug 4, 2011

It's almost impossible to get random numbers with a processor. Any given RND-style function will produce exactly the same sequence for a given seed. So you need a random seed.

Back to square one.

You need what is called an "entropy source" ie a source of truly random events and normally there are none on an embedded controller. If you have buttons or something that a human is pressing or some other "random" event then that can be used by timing the event, but if not you are out of luck.

Re the unconnected ADC input, you can't just read the ADC, that's decidedly non-random. If you take say 100 readings and accumulate bit 0 into an array then "whiten" the array (basically dedupe the bits) and grab a couple of bytes from the array to use as your seed that works OK. But there is still definite grouping (at least on an AVR). Here's an example of the distribution using this technique (about 11000 samples, each reading the ADC up to 1000 times)

Note the distinct banding. I'm still not sure what causes it but general discussion on the Arduino forum concluded that all sorts of things influence the reading, including internal signals. It may also have been the whitening algorithm.

Having said all that I find the ADC method to be "random enough" for my purposes, ie that of generating a unique serial number for a PCB at run time. If you just want to shuffle your MP3 songs it's overkill, and if you need full on encryption it's nowhere near enough.

fernando_g · Aug 4, 2011

inglewoodpete said:
If you have an ADC input in use, use it to create the seed. If you don't then use a spare ADC input pin (left open circuit) as a seed. If need be, add a length of insulated wire to the input. One of the rare times that noise (inconsistent readings) is good.

Of course, if your noise level is too low, you may want to precede your input with a "noise amplifier"

Now seriously, a back-biased base-emitter junction is a surprisingly good noise generator. Couple that tho your ADC input and voila! a quite good random number.

Random Numbers

ChedderCheeser

Senior Member

eclectic

Moderator

Jamster

Senior Member

BillyGreen1973

Senior Member

nick12ab

Senior Member

eclectic

Moderator

srnet

Senior Member

Jamster

Senior Member

nick12ab

Senior Member

inglewoodpete

Senior Member

inglewoodpete

Senior Member

graynomad

Senior Member

fernando_g

Senior Member