dynamic power consumption

Hi all,

I was wondering if anyone has experience measuring the dynamic power
consumption of the OMAP chip on the BeagleBoard. I plan on probing the
appropriate Vdd pins with a scope and calculating the dynamic power
via the equation P = CfV^2. Obviously, this requires knowing the
switching capacitance of the package, and I was unable to find this
information in the datasheet. Is there a way to find this information?
Can it be calculated from values in the datasheet? I've poured all
over the Internet for related information, but can't seem to find
anything. Any help would be much appreciated. Thanks.

Beagle was not designed for doing dynamic power measurements on all of the voltage rails. The TI EVM has that capability and as adding the series resistors required to implement this is very disruptive, it was not implemented. There is one point on the board where you can do measurements for current consumption on the board but not on individual rails. . Measure voltage drops is the way to go although that can be challenging as you need to keep the drops small so as not to cause issue in the operation of the device.

Knowing the capacitance of each pin is tough as it is still connected to the PCB and the bypass and filter caps are on the board. This is not a reliable way to do it. Not to mention that are a bunch of pins on each rail spread out over the part and alot of these pins are no een accesible. It is a .4mm pitch BGA.

Gerald

Hi,

You can use a powermeter. If you can't put your hands on one, just an
amperimeter can do the trick, since you know the voltage applied to
the board. Then you can just multiply the current you are reading by
input voltage (5V). This is what a powermeter does. Note that the
current will fluctuate, you will read an average power value. And the
amperimeter is connected in series between the board and DC power
source

Regards,

Thiago

2011/2/23 Gerald Coley <gerald@beagleboard.org>e:

Power meters will cause voltage drops and introduce noise into the voltage rails. Too much voltage drop and you can have issues.Plus these meters are anything but dynamic in nature.

Gerald

Thanks for the reply Gerald. I agree that the powermeter is really not
the way to go. Plus, I am interested in the consumption of the OMAP
chip itself and not the entire platform. I know I can measure on R6,
but will that give me accurate measurements as far as the OMAP goes,
or is that a more platform-level power consumption? For the dynamic
power, I figured I could probe the VOCORE_1V3 rail and if I knew the
capacitance I would be good to go. Really my goal is to produce a
power vs. time plot. Do you think I could just probe that rail and
treat the capacitance as a variable, thus giving multiple plots for
different capacitance values? Or should I just scrap the idea of
measuring dynamic power?

But VOCORE is only the core voltage rail. There are several rails you would need to probe to get the power of the chip itself. Take a look at the TRM for the processor and look at the power system. Make sure you understand all the different rails. Then see if you think you can measure all three of the main rails and most likely several of the other rails all at the same time. Use this to see if you still think it is feasible.

Gerald

Also, there is some good info on the TI Wiki site. They have a spread
sheet with a break down of the individual rails taken in different
sleep/suspend/running modes.
Most were taken with the EVM but they sould be very close if not
exactally the same, at least for the chip rails.

Does R6 indicate the power of the platform or the OMAP chip?

I think measuring the 3 main rails is feasible. The schematic
indicates that there are test points on the board, so I should be able
to just use an oscilloscope, right? Three probes shouldn't load the
rails so much to the point that the chip stops working.

Also, I was looking at the IBIS models for the OMAP chip, but don't
know much about them. Would it be reasonable to use the package
capacitance given in the model as my C for dynamic power? I don't need
an insanely accurate measurement. The relative changes during
processing of different workloads is really what I'm interested in.