Is auto brightness a battery killer?

[Guest nrev]

Edit: I no longer believe auto brightness is such a battery drain, please see second post.

I know this has been mentioned on the forum before but as it is buried deep in other threads I thought it would be worth highlighting again.

Setting screen brightness to auto from the power control widget consumes more power all of the time than leaving it on maximum brightness permanently (and a lot more than on minimum). This is according to PowerTutor as previously reported in this post.

Interestingly, setting auto brightness through settings->display->brightness->automatic (rather than cycling through the power control widget brightness settings) does not show the same high power consumption in PowerTutor. This was reported in this post as a bug report on FLB (but it wasn't taken further as development on FLB stopped around that time). I see the same thing in CM7, so it doesn't seem to be ROM-dependent.

I am a little suspicious of the latter, though, as PowerTutor reports the same low power consumption even when the phone has switched to maximum brightness in bright light. If PowerTutor is to be believed then the obvious thing to do is to have auto brightness configured through settings and make sure you never touch the brightness icon in the power control widget.

Edited October 3, 2011 by nrev

[Guest nrev]

I am a little suspicious...

If PowerTutor is to be believed...

OK, I've now read the paper published by the developers of PowerTutor:

L. Zhang, B. Tiwana, Z. Qian, Z. Wang, R. P. Dick, Z. Mao and L. Yang, “Accurate Online Power Estimation and Automatic Battery Behavior Based Power Model Generation for Smartphones,” in Proc. Int. Conf. Hardware/Software Codesign and System Synthesis, Oct. 2010.

As I understand it they built a model based on measurements of HTC Dream and HTC Magic phones that used knowledge of lithium-ion battery discharge rates so they could use battery voltage readings to predict power consumption. Otherwise they have no way of measuring power usage in real time. However, in their method the power consumption of the LCD screen is estimated purely from brightness settings - no doubt it was accurate on the HTC phones they used to derive the estimation models, but it is meaningless for our Blade. So, all PowerTutor is doing is reading the brightness value and multiplying this by some factor modelled from measurements on the developers' HTC sample phones. In effect it is just a plot of the brightness level that PowerTutor thinks the phone is set to. I reckon it doesn't read the actual value of brightness either, which could explain why it is constant when set to auto brightness regardless of the state of the backlight: when auto brightness is turned on through the power control widget it pessimistically assumes a constant high value and when set through the settings menu it optimistically assumes the lowest value.

In summary I think PowerTutor is useless for estimating the power consumption of the Blade's LCD screen. I have uninstalled it, and I suggest everyone else does the same.

However, it still seems possible that auto brightness may consume additional power to continually sense ambient brightness, BUT I have no idea how much power this might be or whether it is in any way significant.

Edited October 3, 2011 by nrev