Thread: Touch Excessive Battery Use?
View Single Post
Old 01-19-2012, 12:00 PM   #55
azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.azazel1024 ought to be getting tired of karma fortunes by now.
Posts: 182
Karma: 346596
Join Date: Jan 2012
Location: Ellicott City, MD
Device: Nook simple touch, iPad 2
Backlit devices will also still use more power when turning pages than with not as the processor and display processor and probably GPU have to render the next page and display it on the screen. However, the backlit and idle CPU usage and main board power consumption, let alone wifi, etc on a backlit device is going to consume orders of magnitude more power than the page turn operation is likely to agregate out to.

Reading 10 pages an hour on a tablet is probably only going to use in the single digits of percent less power than reading at 50 pages per minute.

If the nook is awake it is using markedly more power than if it is asleep. The CPU has to be powered on, if in "idle" state, the IR emitter(s) for the touch screen have to be running the main memory likely has to be in a higher power state, etc. That said, idle power consumption is pretty low.

Actually we could come out to a rough number of actual power draw rate. 30 minutes per day for 2 months is an average power draw from a 5.66wh battery of 94.3mw of power. From what I know of DDR memory, the ARM8 CPU and some other stuff, and idle power draw of roughly 50mw actually seems probably close to spot on. That would leave about 44.3mw left over for page turns, averaged. That gets you 2.65wh used by page turns in the 60hrs/2 months of usage. Lets say 2,000 pages turns in that time (seems resonable that they used that measuring stick, that is 33.3 pages per hour) you get 1.325mwh of energy per page turn, or about 10w used for half a second, which is WAY too high for what the CPU, screen rendering engine and the eink screen must use to redraw the screen, even with a full refresh. A more likely "target" power consumption is probably in the 2-4w realm for "system" power consumption during a page refresh before it falls back to idle power consumption. To get even on the high end of 4w for a 500ms page redraw you come out to about 15mw average power consumption over those 2 months of reading, leaving you with an idle power consumption of about 80mw.

Of course that doesn't take in to account power lost during sleep, but in sleep all you have to do is maintain a watch on the power/n button to wake the system and low power mode for the DDR memory to keep things in memory. 80mw is already pretty gosh darn low, but I'd bet (if nothing untoward is going on) sleep power use is probably on the order of 1-3mw, otherwise there would be absolutely no reason for the nook to go to sleep if it didn't save power.

So going back to the usage model for a minute, if we said the nook used 4w during page turns, if you read at 66 pages an hour instead of 33 pages an hour, you'd up your power envelope from ~95mw average to 110mw average, reducing battery life by 15% by doubling your page turn rate.

Now one other thing to consider that I didn't take in to account is that partial page redraws likely use less power than a full refresh. Also just because a page turn (under 1.1) supposedly take 450ms (lets round to half a second), that doesn't mean that system power actually returns to idle istantly at the end of that time. I'd bet that the system agent keeps the CPU above idle for a period of time after a page turn to facilitate fast page turns by keeping CPU clocks up, maybe pre-rendering a page or so in advance at that time as well keeping the screen renderer power up and whatever that needs to do to remain primed, etc, all waiting to see if there is going to be another screen touch event or button page turn.

Now, I'd imagine the system really does return to an idle monitoring state pretty quickly, but it probably takes anything from an additional 250-2000ms to return to this state after the screen redraw occurs. So the idle power consumption could well be below 80mw if you, say, assume 2-4w for the half second to redraw a page for 500ms, dropping to maybe 500-2000mw for anything from 250-2000ms after the page draw.

If something like that were occuring, instead of 15% faster battery usage by doubling your page turn rate, it might be more like 20-50% faster battery usage.

With my limited experience with the nook ST, but based on what I went through and my knowledge of tablets, computers and low power processing I'd suspect (at least with a baseline 33pph reading rate) doubling your page turn rate while reading is going to use the battery probably closer to a 25-35% faster rate than either 15% or 50%.

Of course total number of pages on a battery charge will go up as you read faster as less power is used for idle/sleep, though with diminishing returns. You might get 60hrs of use at 33pph and 2000 total pages. You might get 48hrs of use at 66pph and roughly 3,200 total page turns. At 132pph you might get 40hrs of use and roughly 5,300 total page turns and so on.

So really, idle uses quite a bit of power as a function of overall battery usage. Page turns use orders of magnitude more power than what idle does, but energy wise, you are at idle for dozens of times longer than it takes to turn the page so the amount of energy consumed by turning pages is relatively little (at 33pph you spend .46% of your time turning the pages and 99.54% of your time at idle or doing something other than "turning the page").

Now a tablet at idle with its active screen uses a boat load of power. An idle ARM9 CPU (which I think has a higher idle power than an ARM8 does, though maybe it is the other way around) uses roughly 10mw (absolutely lowest power ARM9, most use somewhat more than that, but lower than 50mw in general). However, other systems resources, such as the wifi chipset, memory subsystem and especially the backlight/screen use substantially more. Even with wifi off, you are looking at a total system idle draw of roughly 1.5w for some of the better tablets with the screen brightness settings in the lower range. Compared to almost absolutely no more than 100mw of power consumption averaged for something like the nook simple touch, that is more than an order of magnitude more idle power consumption than a nook ST uses on average!

Sorry, long post, but I thought I'd go through some of the math and at least my madness to my method of coming up with why idle power consumption on an ebook reader is actually resonably substantial (though by no means exhorbitant) in comparison to how much power you are using while turning pages.

This is of course not taking in to account that BN, Amazon, etc may actually be taking in to account sleep power usage in those 2 months, 1/2hr a day and whatever hidden standard of pages per hour they are using to get those 2 months. And sleep power consumption is SOMETHING, even if really stupid low (it might well be low single digit miliwatts). Heck 5mw over 23hrs is the same as 1hr of reading might be assuming ~95mw of average on/reading power consumption (course sleep power consumption is possibly/probably lower than that). So sleep power use is probably more like 1-2mw just to keep the DDR powered up enough not to lose memory and to monitor those two buttons.

Of course at 1mw power usage numbers, we start getting in to self discharge rates of lithium ion batteries. IIRC a lithium ion battery (depending on the chemistry) will self discharge at roughly 5-10% per month. Voltage monitoring circuit uses some power, they don't self discharge in a classical sense, and dumb lithium ion batteries (which is not what is in most devices) don't really self discharge, it is actual capacity loss...but they "lose this" much slower, takes about 7 years depending on storage conditions for a lithium ion battery to go dead in that it has no rechargable capacity left (or effectively none). At 1mw in sleep (if we say it is that high), you get about 7-8% of battery capacity used per month by sleep power consumption plus 5-10% from self discharge of the battery. Doubling to 2mw gets you 14-16% per month from sleep.

Sorry, long post, I tend to fixate.
azazel1024 is offline   Reply With Quote