HG324 Hardware Random Bit Generator

Users Manual

Testing random bits	Contents	Technical specifications

Benchmarks

We have performed a series of tests of the HG 324 maximum download bitrate using the program bitwr on a different hardware and on the different operating systems, supported by our software.

We found that both serial and parallel ports have always permitted the same downloading bitrate, that is our software was able to read parallel and serial port(s) equally fast on the same machine. This is clear due to the fact that both ports are governed by the same (special) hardware clock. However, due to very different linear front-end of these ports (parallel port has LOW logical level at 0V and HIGH level at +5V whilst serial ports use -7V...-12V as LOW and +7V..+12V as HIGH level) serial ports sometimes have worse performance. It seems that some serial ports have difficulties to handle cleanly such a big voltage swings. The most sensitive variable to measure this problem is the bias (you can check it with programs bitwr and zero) which gets bigger if there is a problem. In such case you can try to improve the bias by lowering the downloading bitrate. Ports of newer computers tend to be good, but one can never be sure without a check.

Below is the table of bitrates (in bits/sec) we got on the several computers we had at hand.

Computer	Operating system
	Win98	MS-DOS	Linux
Laptop w/ Celeron 600 MHz	722 k	873 k	1.24 M
Pentium II 400 MHz	510 k	475 k	1.19 M
Pentium MMX 200 MHz	481 k	748 k	1.24 M
Pentium 133 MHz	N/A	504 k	1.03 M
PC i486 DX2 66 MHz	N/A	218 k	644 k

Here MS-DOS means either a machine restarted from Win98 into the "DOS mode" or booted from the DOS 6.22 floppy. The two gave exactly the same performance in all cases.

It is interesting to note that Microsoft products always get a lower bitrates than Linux on the same machine. On one particular machine "Pentium 200 MMX" we had significant deviation of bias on the COM1 port under MS-DOS, while we got excellent results on the same machine and the same port under Linux ! Being only a simple users of what is given to us, we do not understand why such things happen, but somehow this does not came to as a big surprise. After all, we are attempting to use ports at much higher speeds they were intended to be used.

On the other hand, parallel ports on all machines performed well.

We also noticed that another program, bitplot, performed on some machines as much as 31 percent faster when the machine was booted by an MS-DOS 6.22 boot diskette with respect to the case when it was booted to the "DOS mode" starting from Windows 98.

Testing random bits	Contents	Technical specifications