Other Parts Discussed in Thread: ENERGYTRACE
Tool/software: Code Composer Studio
Could somebody point me to all the docs that specify the behavior of Energytrace for my setting? (details below) I've seen lots of juicy stuff to do with MSP432 and other targets and realize that with CC1310 I've got just the simplest sampling mechanism but I need some gory details.
I notice the sample times in the .csv file are irregular. Here's an example:
Time (ns),Current (nA),Voltage (mV),Energy (uJ) "0","12000","2800","0.2" "2000000","12000","2800","2.5" "2500000","12000","2800","3.3" "8500000","12000","2800","3.5" "12000000","12000","2800","5.8" "12500000","12000","2800","6.6" "17000000","12000","2800","6.9" "22000000","12000","2800","8.3" "22500000","12000","2800","9.9" "25500000","12000","2800","10.2" "31000000","12000","2800","10.4" "32000000","12000","2800","11.2"
So time starts with a sample at 12uA, then 2ms later it's still 12uA, then .5ms later it's the same, then 6ms later, then 3.5ms later, then 3.5ms, .5ms, and so on.
I'm starting to use custom tools to look at the data and I can make sense out of this. But I started some time ago jumping to the conclusion that the sampling was uniform and now that i'm really focused on it I see it constitutes a pseudo-random function. Is there anything that will influence the sampling to make it regular. No knobs apparent in the CCS interfaces.
But the thing that motivated me to write and ask about this is that when a long interval such as six milliseconds happens it potentially straddles one of the "radar blips" I've put into the execution stream to mark a spot as a first crude way to correlate my code locations with changes in energy consumption (e.g. "the consumption is jumping about X milliseconds after this blip"). So I now know I have to make these high current spike "blips" wider than 6 milliseconds to be sure I see something. Except I wanted to use a number of short spikes as separate, identifiable markers. But I can't unless I make every spike more than six milliseconds wide. So my parade has been rained on. If there is any way to eliminate the long delays between samples I want to use it. Another aspect of the unequal intervals is that I was expecting to cheap out on homebrew tool features by simply skipping the first X samples of an ET .csv file and that corresponding to a known amount of time where I could make graph X (time) axes zero-origin (e.g. relative to a significant "blip") but now I'm forced to preprocess the files so skipping (and truncating of trailing records) can be in terms of time.
So, where do I read up on the algorithm for these sample intervals, what might influence them, etc?
My target is a 32 pin (middle size) CC1310. I'm using an XDS110 that's a few weeks from Digikey TDSEMU110-U Serial 5158600160 also 51855-0001R Rev A and I've never refused it firmware updates.
I have Energytrace giving the target 2.8 volts. Would jacking it to the max allowed for the chip enable more even sample intervals?
Code composer Studio 8.3.0.00009, Simplelink 2.30, Ubuntu 18.04, I7-8700K 16GB RAM, M.2 250GB Toshiba SSD, Debug probe disconnected: only power and ground from the second XDS110 cable's "target supply Vout" and "Ground" connected to target during testing. Measurement duration set to five seconds. The XDS110 is connected to my computer through a powered USB hub. Would it change anything if it was directly connected?
For this particular tool on this target if I was running CCS on Windows would I get improved performance?
Are there any better tools out there for energy measurement?
Finally, are there plans to make EnergyTrace support for CC1310 on par with other chips such as MSP432?
Thanks,
