On 07/16/2014 07:36 PM, Hans de Goede wrote:
Hi,
On 07/16/2014 04:54 PM, Olliver Schinagl wrote:
Hey guys,
I've been doing some initial temperature measurements and calibration and want
to share my initial results. I left most of the information on the wiki, but
feel free to ask me anything here.
http://linux-sunxi.org/Temperature_Calibration
Interesting, but the ods is a bit hard to read without a
more detailed legenda, esp as you call every line in the graphs
foo (delta), can you explain what is what in a bit more detail ?
Ok let me go over each column first then :) I agree it was mostly
personal at first.
Freq, obvious, the SoC frequency of scaling_setspeed
Ambient temperature, the room temperature measured by a thermometer next
to the board, as can be seen in the pics
Ambient Humidity, equal for humidity I recorded it, not sure if it
matters for anything
die temperature, the temperature what the die says it is, e.g. the
thermal sensor inside the chip
package temperature, the temperature measured at the top center of the
chip package. (either thermocouple or IR camera
die temperature delta, die temperature - ambient (this is important
because the ambient temperature changes through out the day, so the
delta should be consistent.
package temperature delta, package temperature - ambient (as above)
delta, horrible name, temperature difference between die and package.
This is really the most important datum
current, current as reported by the PSU
voltage, voltage as reported by the PSU
power, current * voltage
I've also been doing some testing with the temp sensor on A10
boards myself, as the current linear curve we have works
ok-ish for the A20 but seems to be way of for the A10.
I haven't looked at the A20 at all, and for the A10 I used 2 different
Lime's so far ( should have noted that). On the 3.4 kernel at that.
While I have access to several Lime's (a few thousand I suppose in a few
months :p) I don't have the time to perform that many tests, as it takes
a minute or two for the temperature to settle. I can see that testing
anything under say 300 MHz may not be hugely interesting, but it does
help with the overall graph. Many datapoints are tedious but good.
I guess doing a full run takes 2 or so hours, time which I cannot easily
justify sparing. But I do agree we need to test a few more boards at the
very least, accurately.
I've noticed that there is quite a large spread between different
A10 SoC-s. So it would be good if instead of measuring one board
very accurately you could measure multiple boards.
I will see what I can do. I did notice the quality differs a lot between
SoC's. I have one board that reached 97C on the outside of the package!
(the 14-07 table). While a nother board had trouble reaching 85C even
according to the internal sensor. (so 65-70 ish on the outside). None of
the boards even can run stabily at 1008 with the ambient temperatures
where seeing here. I guess at 20 C, they work, above that, it gets tricky.
All that said and done, we may go for A20 due to the lower power
requirements/heat generation so can test that then :)
Hopefully other people have access to thermocouples too! :)
Olliver
Regards,
Hans
--
You received this message because you are subscribed to the Google Groups
"linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
