![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]()
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
SG100 is delivered
with driver for Windows 95/98 and Windows NT/2000. The device is connected
to the computer through the 9-pin serial port. Power is taken from the
port. Supports all bit rates up to 115,200 bit/s. Throughput is about
10,000 bytes/s for the 115,200-bit/s rate. Output is processed using statistical
and cryptographic methods, and passes any statistical test. Resistant
to external electromagnetic fields and high resistance against power fluctuations. Features
Prices and AvailabilityThe SG100 is available in two basic packages; Developer and Runtime. Developer pack
Runtime Packages
Prices and specification subject to change without prior notice.Important Tech Note!The current version
of the SG100 do not work on some multi-function Statistical TestsThe simplest statistical test is to check if the SG100 noise strings has about the same number of ones and zeroes. A test program (N1_TEST.EXE, included in Developer Package) is written that counts bytes and bits. The output is given in absolute and relative frequency. To make comparison easy the difference between a relative frequency of 50% and observed frequency is computed relative to the standard deviation. These values are seldom higher than three, for random output. Note, that as the program outputs a large number of sigma values, it sometimes happens that a sigma value higher than three is found. This is normal for random strings. If in doubt, accuracy can be increased by counting a longer noise string. If we, as an example, count 6,400,000 bytes and find 25,603,990 "one" bits then we have a relative frequency of 0.50007793 Sigma = 1.1 That is 50.008% one bits. To increase accuracy we count 441,600,000 bytes. We find 1,766,378,269 "one" bits yielding a frequency of 0.49999385 ( Sigma = -0.7) That was very close to 50% "one" bits and 50% "zero" bits. Desperately we can read 1,651,200,000 bytes and count to 6,604,734,712 "one" bits and the frequency is 0.49999506 ( Sigma = -1.1). Download complete test results (25K) pLab load test of the SG100The pLAB Research
Group of the Institut für Mathematik, Universität Salzburg has
conducted a load test of the SG100. The report contains commented
simulation results for SG100. Each page contains the plot of the
truncated Kolmogorov-Smirnov-values and the according uppeirtail-probabilities
for the Load Test'(LT). DiehardThe SG100 also passes the Diehard test. The Diehard test, by George Marsaglia, consists of several statistical counts that should have a specified distribution if the input string is random. By comparing observed counts to a theoretical count we can see if a string is random or not. For a sample of size 500: mean SG100.DAT using bits 6 to 29 1.942
Chisquare with 6 d.o.f. = 2.61 p-value= .143850 The observations above
are to few to give high accuracy. This problem originates in that the
Diehard program do not adjust the sample sizes to a larger test file.
The forthcoming revision of diehard may correct this problem. Robert Davies test of SG100Robert Davies have tested hardware random number generators, including the SG100, for a lottery application. Link to Robert Davies lottery page Electrical & RFI/EMI MeasurementsEMC Test ReportsEMC Test Report: Emission of electromagnetic disturbances EMC Test Report: Immunity to electromagnetic disturbances Electrical Characteristics and Measurements SG100 eBookA schematic diagram
of the SG100 circuit is displayed below. To the left we find the You can also choose
to download the test in pdf format
|
Home | ![]() |
Company | ![]() |
Shop | ![]() |
Products | ![]() |
What´s New | ![]() |
Site Map | ![]() |
Contact Protego |