How do you do a battery bench?

[dreamlayers]

I just did a battery bench on my 5th generation 30GB iPod to see if a patch I submitted saves power.  I am not happy with the results because the initial voltages were different and the log stops at 15%. 

I guess the log stopped at 15% because the iPod was shut down by hardware, because the battery is old and has increased internal resistance which caused an excessive voltage drop when the hard drive tried to spin up.  Of course this is just a guess.

I wonder if the different initial conditions are different because the first test (unmodified Rockbox) was done after a few topping up charges, while the second test (Rockbox with patch) was done after a charge from a discharged state.  At the start there's also the initial browsing to the directory I wanted.

I wonder whether to redo the test while obsessively trying to ensure that conditions are as close as possible.  Does anyone have any particular hints for this?

Is a battery bench pointless if I can't get a log that goes down to 0%?  Do I really need a battery bench when i have current readings (including FireWire input current with two different DMMs).  Can the differences in discharge slope tell anything about power consumption?

[soap]

This is why I typically pay attention to the 90%-10% part of the curve.  (Perhaps you should pay attention to 85%-15%).
Not sure how to read your FS entry - but if I am reading it right the discharge curves you got were the same slope, just different offsets (due to the differing start voltages?)?

PS - your battery is getting old.

[saratoga]

Reading your comments on the tracker, perhaps fitting a curve or spline to the shape of the discharge slope and using the derivatives would give you a more accurate reading?  From what I understand, the discharge curve of a battery should be the same no matter how you arrive at a given level of discharge.

[dreamlayers]

If I simply use percentages, I can be misled by local variations caused by hard drive activity.

Yes, fitting a curve or spline might be a good idea.  But how about just aligning the curves by horizontally moving one?

[saratoga]

If I simply use percentages, I can be misled by local variations caused by hard drive activity.

Yes, fitting a curve or spline might be a good idea.  But how about just aligning the curves by horizontally moving one?

Thats basically the same idea as fitting a curve and extracting the slope right?  I think it should work, though its not very quantitative.