[Coco] 6809/6309 Power Analysis

[Boisy G. Pitre]

> 
>> From what I gather the idea is to mesure the dynamic change in current from instruction to 
> instruction. What is intended is a current profile of the chip as it executes instruction. The 
> theory is that certain intructions will have more effect on the current than other instructions. 
> 
> What is the real problem is that the instructions are  not of equal time frame. Take the STA 
> insturction depending on mode will be three to five machine cycles. The MUL command 
> takes 11 cycles. If the processor is ran at say 2 MHz, you have a machine cycle time of 
> 500nS. The variation from one instruction to another will be short in duration. Now if you slow 
> the processor down to say 100KHz you can get a beter profile and then exptrapolate that 
> upward as the clock speed is increased.  

james,

This is an observation I also made today. It would be necessary to compensate for different cycle counts between instructions.

One would expect that the more cycles an instruction requires, the more power it would consume, but I expect that depends on exactly what the instruction is doing in those cycles. 

Perhaps it is possible to estimate the power profile of each instruction by looking at the cycle count?  I'm not saying that I should forego hard measurements, but with detailed data sheets, it would be possible to construct a relative difference between instructions. Your thoughts?

> I tend to think that at least on the 6809 there could be some change. The internal logic may 
> have several levels of gates that when swiched for various instructions will impact current 
> when some intructions are being performed. If the 6309 is indeed a microcoded processor 
> there may not be the change in current as the microcode is often stored in a microrom that is 
> say 2K by 56 bits. Then the microoperations would be more constant from instruction to 
> instruction and not have as great a variation as the MC6809.

It also occurred to me that given the 6309's lower power draw, that it may not be the ideal processor to measure.  I've got a 6809 in reserve and may just base my measurements on that processor alone.

Boisy