The Randomizer Laboratory
-
The Randomizer Laboratory is a lab made to test randomizers. This was not made by me, but rather one of the current* FP save holders, SandwichLizard. I am simply helping by giving him my randomizer to test. Not only that, but I am also showing this as a thread here for him.
(*since 3/2/2014 UTC+8)
Reliability is how well a randomizer responds to inputs; a reliability less than 1 means that it does not always listen to the input. A randomizer with reliability 1 is a perfect randomizer; it always listens to the input.
Low/High is a crude, but at least helpful calculation on the randomness of a randomizer: The closer the number is to 1, the better the distribution. This approximation is calculated by dividing the lowest output amount with the highest output amount. Proper shift registers or flip-flops must always register a value of 1. Randomizers must have a value of less than 1, but is far better as it gets closer.
The second algorithm is the total output count divided by the product of the largest output number and the amount of outputs, or ToC / (LoN * AoO)
Example stats:
A2 = Algorithm 2. Algorithm is number 1 unless told.
*Low/High is tested for 1000 pulses unless told.
SandwichLizard's PT-based 8-output true randomizer
-Reliability: 1.000
-Low/High: 0.974 (159076), 0.980 (159076, A2)
PTuniverse's 4-output true randomizer
-Reliability: 1.000
-Low/High: 0.942, 0.989 (21753, A2)
SandwichLizard's 4-output electron randomizer
-Reliability: 0.960
-Low/High: 0.715
SandwichLizard's 4-output HYGN randomizer
-Reliability: 0.803
-Low/High: 0.815
Save:
1474563View Save 1474563Created by: SandwichLizard
-----
Credit to:
mechaman: Counters
PTuniverse: Randomizers
Edited 4 times by PTuniverse. Last: 3rd Mar 2014 -
New Lab
1476438View Save 1476438PTuniverse:
Low/High is a crude, but at least helpful calculation on the randomness of a randomizer: The higher, the better. This approximation is calculated by dividing the highest output amount with the least output amount. A higher Low/High means that it is more random.
This should read:
Low/High is a crude, but at least helpful calculation of the distribution of output pulses amongst the different outputs. The higher the number, the more even the distribution. It is calculated by dividing the LOWEST ouput value BY the HIGHEST ouput value. A shift register or flip flop tested on this should have a distribution of 1.00. A good randomizer should have a number less than 1.00. this value is subjective based on how you want your randomizer to perform.
Edited once by Sandwichlizard. Last: 4th Mar 2014 -
reliability of 1.00 means there where an equal number of input pulses as total output pulses. if you have a number less than 1.oo the device missed pulses. If you have a number greater than 1.00 then the device is generating extra pulses. I want to build a parallel pulse detector to record whether extra pulses are parallel or serial.
Low/High is a crude measure, but it does give a value to quantify.
-
I wasn't wondering what that meant (if you are talking to me).
Just the bold statement there seems biased.
-
Oh thats OK. I put that data out there. I'd feel a little braggy also, that is a very innovative device. In fact here is some new data.
Sandwichlizard's (PT tech) 8 Output "true" Randomizer
Input Counts: 159076
Total Output Counts: 159076
Reliability ratio: 1.00Output A: 19774
Output B: 19936
Output C: 19848
Output D: 19766
Output E: 19799
Output F: 19773
Output G: 20294
Output H: 19886
Low/High = Out D/Out G = 19766/20294= 0.97398
-
Added info.
-
You should include the number of count tested for each data set
SandwichLizard's PT-based 8-output true randomizer
-Reliability: 1.000
-Low/High: 0.974
-Cycles tested: 159076
-
That's strange, I tested for up to 21500 pulses, and the low/high becomes 0.975. I have my custom low/high formula (Total Output Counts / (Greatest Output * Amount of Outputs), where my 4-output yields 0.989, and your 8-output 0.980.
-
if you want to get critical here it a link to a page that explains "STANDARD DEVIATION" and gives an example of how to calculate it. http://www.mathsisfun.com/data/standard-deviation-formulas.html
that would be a more precise and generally excepted metric of performance with real world comparison.
