The power saving is a complex stuff and we are sometimes facing some strange behaviors which are far from the expected results. That's why it's always interesting to measure the real benefit of a new development or to validate the theory used for this development. Nevertheless, it's sometime a bit difficult for extracting some results with a not optimized software what is often the case when you are working with latest kernel release. If you are able to measure a small part of the complete platform power consumption, it becomes easier to extract the improvement linked to your development.
This wiki page explains How to measure some specific power domains like the ARM part on a snowball in order to validate the benefit of your development.
Start by removing the EMI filters FB3, FB4, FB5, FB6, FB8 and FB9 which are not so necessary when using a stabilized external power supply. These free places will be used to connect the ammeter and measure some power domains.
You can see below the components (red circle) which must be removed
- component 1 : Vsmps1
- component 2 : Vsmps3
- component 3 : Varm
- component 4 : Vape
- component 5 : Vmod
- component 6 : Vsmps2
Once modified, your board should look like the next picture
I advise you to deport the connectors on an expansion card to prevent damage to the tracks of your snowball by manipulating the connectors. For normal use, you only need short-circuit the terminals of each connector as the picture below
Otherwise you just need to remove the short-circuit and insert an ammeter between the 2 pads of connectors
The modified board has been tested with following Arm core frequency : 400Mhz/800Mhz/1Ghz. Different CPU idle states have also been tested.
WorkingGroups/PowerManagement/Doc/Power_snowball (last modified 2012-01-09 22:53:33)