OMAP-L138 - power calculator - capacitive load

Hi Milan

For these power models, we have assumed the trace lengths to be more dominant compared to input/output capacitance. There were different school of thoughts on low power / low MHz vs high power/ high MHz devices. OMAPL138 family tends to fall somewhere in between.So for end user modeling, we have essentially given trace length as the variable.

On the power characterization board the EMIFA has a NOR, NAND and an FPGA hooked up. Usually on parallel busses we tend to use the trace length defaults for the fastest switching IOs as a proxy for the entire bus.

If for whatever reason you think this is not giving you reasonable estimates, you can also choose to do a simple nCV^2F calculator with trying to guessitmate the switching frequency of the address/control vs data busses and using "C" values from the connected devices on your board.

Regards
Mukul

Hi Mukul.
Thank you for your thorough reply.

I disagree that trace length capacitance may be considered dominant. Also, I do not think that each user should make their calculations at their end. Thus, I suggest the power calculator be improved in the following aspects.

1. Please add "EMIFA 8/16 data bus width" option.

2. Please add "Sum of input pins capacitance connected to a single EMIFA trace" option, separately for Data/Address/Control.

3. Consider adding input pin capacitance option for other interfaces. (Although I do not use it, the UPP clearly calls for it due to number of traces and typical operation freq I think.)

Also, please move "Nu. of PRU pins" label to the right or put it to the relevant cell's comment  Current localtion is really confusing for EMIFA users.

Many thanks in advance,
Milan

Hi Milan

Thanks for the inputs. I understand your position on the trace length concept. We have had differences in opinion on this approach even internally at times.

Unfortunately there is no plan to update the power estimation spreadsheet. The spreadsheet is meeting most of its purpose and is consistent with the power estimation spreadsheets for other C6000 devices.We have extremely good success with this model with several customer bench correlation exercises.

I have not had any other customer request such enhancements or different approach to the model, so it is not feasible for us to make such changes based on a single request. 

I recommend following my previously stated approach if you wanted further analysis on this or alternatively see if running your use-case on a prototype board can give you more accurate estimates for your EMIFA use-case.

Regards

Mukul