I did a little work on figuring out the relationship of switch settings to bus speed settings on the XLR8 ZIF Carrier the other night. Here's what I came up with:
Bus Speed = 25 MHz + 3.2 MHz * [5:8] + .2 MHz * [1:4]
where [5:8] is a four digit binary number with 5 the most significant bit and 8 the least significant bit. Similarly for [1:4].
So, for example, if you set switch 5 On and all other switches Off, you have [5:8] = 1000 binary = 8 decimal and [1:4] = 0000 binary = 0 decimal. So your bus speed setting is:
25 + 3.2 * 8 + .2 * 0 = 50.6 MHz.
Turn switch 4 On and you add .2 MHz:
25 + 3.2 * 8 + .2 * 1 = 50.8 MHz
Want to subtract .6 MHz from the first example (only 5 On)?
Turn 5 Off and 6, 7, 8 on so that [5:8] goes from 8 to 7. That subtracts 3.2 MHz. Then set 1, 2 and 4 on (1101) which is 13 X .2 which adds back 2.6 MHz.
Turn on just switches 5, 8, 1, and 3 and you have:
[5:8] = 1001 binary = 9 decimal
[1:4] = 1010 binary = 10 decimal and:
25 + 3.2 * 9 + .2 * 10 = 55.8 MHz.
I suspect that the reason why speed measuring software such as Clockometer is always a little off these readings (and off of the XLR8 chart readings) is for one or both of two reasons. First the actual mechanism for setting the speed is truly bizarre. It depends on the ICD2051 Clock Generator chip, which arrives at the speed through a quotient of two 7 bit numbers supplied to the chip. These quotients probably don't result in the exact numbers on the chart or from the equation above. Second, the speed also depends on the Clock Generator multiplying a 13.xxxx MHz clock signal and that clock signal might be slightly off from oscillator to oscillator.
This also probably explains why there is a PIC microcontroller on the XLR8 cards. It may do other things, but it is almost certainly running the program that translates the straightforward DIP switch settings into the bizarre serial input that the ICD2051 requires.
