PRS-500 PRS-500 power consumption measurements

[rothe957]

I charged my PRS-500 today using a lab power supply.
The battery was empty (reader came up, said "LOW BATTERY", then "NO BATTERY", then shut off immediately).
After I found out that USB charging won't work without driver,
I used plan B: a power plug from a universal adapter set and a lab power supply. I set it to 5.20V, turned output on and the reader started charging right away at about 560mA. Now, about 2 hours later, current is down to 140mA.
I then switched the reader on, it showed the "starting up..." screen.
During startup it consumed about 140mA more than when it was off (and only charging).
As long as the unit is just sitting there and no buttons get pressed, there is little or no additional current consumption (less than 5mA, maybe zero; the accuracy of the ampmeter in the lab power supply does not allow for more precise measurements, sorry).
A page turn causes a short (<2s) increase in current consumption, with a peak of about 220mA.

So much for the raw data.

Now a few ballpark figure calculations just for fun:
Lets assume a page turn consumes 180mA for 2 seconds. This means 360mAs or 0.1mAh per page turn. Judging from the physical battery cell size, it has a capacity somewhere in the 800mAh range. At 0.1mAh per page turn, you would get 8000 page turns out of a full charge in a perfect world. Well, pretty close to the 7500 claimed by Sony, considering that these are just ballpark figures..

^_^

[Alexander Turcic]

VERY interesting work!

It would be cool if you could test power consumption with books stored in internal memory vs. books stored in external memory such as an SD card. We have numerous reports that the latter uses considerably more power per page turn.

[rothe957]

Originally Posted by Alexander Turcic

It would be cool if you could test power consumption with books stored in internal memory vs. books stored in external memory such as an SD card. We have numerous reports that the latter uses considerably more power per page turn.

The measurements were taken with books in internal memory, no memory card inserted.
Hmm, unfortunately, I don't have a SD card or MemoryStick here now. I'll be able to test it next week thursday or friday.
From what I know about the usual current consumption of SD cards, I estimate a memory card would raise the page turn current by about 40-60mA.

Some more info: When charging was complete, current consumption from the lab power supply with a book page on the display was shown as zero (lab ps ampmeter, not sure if it was really zero, lets say: below 1mA). Page turn current peaked at about 220mA (drawn from external power supply, no mem card).
Varying the external supply voltage from 4.8V...5.2V didn't make a difference in current consumption.

^_^

[NatCh]

Very nice analysis, rothe957, very nice, indeed.

[TadW]

Originally Posted by rothe957

Some more info: When charging was complete, current consumption from the lab power supply with a book page on the display was shown as zero (lab ps ampmeter, not sure if it was really zero, lets say: below 1mA). Page turn current peaked at about 220mA (drawn from external power supply, no mem card).

I guess page turn also varies depending on the complexity of the content.

Nice work, rothe!

[Azayzel]

This is a cool experiment, especially to give real-world numbers to corporate marketing claims. I too, would be interested in the power consumption using and SD card; hopefully you can find locally for ~$10 or so to test it out (could always pull one out of your digital camera or old-school MP3 player ;-) )

Thanks for informing us, 8k pages is friggin cool!

[kovidgoyal]

You can charge it from USB without the driver. Just use linux and libprs500.

[rothe957]

Originally Posted by kovidgoyal

You can charge it from USB without the driver. Just use linux and libprs500.

Well, I admit that I consider installing the driver the easier way.
Compared to installing Linux and then getting and installing libprs500.
I have a few linux machines in my office/workshop, but these are all embedded non-x86 devices without USB.
No, as tempting as it sounds - I think I stick to the lab power supply, and try to remember to bring the Sony driver CD next time.

^_^

[rothe957]

Well, I added a current sense resistor (1 Ohm) to the setup, and connected a scope.
X axis is time, 1 second per grid. Y axis is current, 100mA per grid.
The small "1->" marks the zero current line.
Screenshots were done the lotech way, with a camera.

First some measurements with files in internal memory:

Turning the reader on then off again, Baen RTF file, internal memory:


Page turn, Baen RTF file, internal memory:


Page turn, PDF file ('1984'), internal memory:



Next with files on SDcard (Sandisk 1GB):

Page turn, Baen RTF file, SDcard:


Another page turn, Baen RTF file, SDcard:


Page turn, PDF file (Atmel datasheet, with graphics), SDcard:



And another test set with a Sony MSmicro (512MB) with MS adapter:

Page turn, Baen RTF file, MSmicro:


Another page turn, Baen RTF file, MSmicro:


yet another page turn, Baen RTF file, MSmicro:


Page turn, PDF file (Atmel datasheet, with graphics), MSmicro:


RTF formatting + update of wait symbol, Baen RTF file, MSmicro:


Conclusions, anyone?

^_^

[dstampe]

The big spike followed by high consumption and noise is what I would expect from the eInk circuitry. This probably has a high-voltage DC-to-DC converter, followed by a matrix of drivers for row and column lines in the display.

This shows a big spike of current draw as it starts up and charges all its capacitors. Then the variable load producing the noise is a combination of this supply and the variable capacitive load as different lines in the eInk display are charged and discharged, Obviously this is a big user of power.

There seem to be 3 load levels. One is fairly low, probably the CPU coming out of sleep mode. The second appears to correspond to display preparation, as it's much longer for the PDF file display. Not sure if this is CPU use, display buffer being powered up, or flash memory access. The highest level seems to be when the eInk power supply is being used.

When used with external flash, the second level of current draw takes longer before the page display, so this second level probably does indicate internal or external flash memory access. (External flash access also appears to be much slower). Also, the CPU stays active for several seconds longer after a page update. So using external flash memory will shorten battery life by up to 40%, I would expect. Of cource, it looks like viewing complex PDF files is even more expensive.

[TadW]

So it's generally true to say that RTF is better than PDF in terms of battery life?

[Azayzel]

Hmm, wonder about the the BBeB format; especially for say a PDF that was converted using tools fromt this website? I'll post something you can test with here after it finishes transcoding. Didn't have any small PDF's already converted that I could u/l so you could test actual PDF vs. (hopefully) improved LRF. That is, of course, if you don't mind running the test with the samples I've provided.

Note: It's a freely available eZine converted from the distributed PDF to LRF using PDFRasterfarian v2.1.4.

Edit: Here are the files, thanks for posting your data!

[rothe957]

Originally Posted by Azayzel

That is, of course, if you don't mind running the test with the samples I've provided.

I'll see what i can do. It might be a while until I have access to the scope again, though. Please be patient.

Originally Posted by TadW

So it's generally true to say that RTF is better than PDF in terms of battery life?

I think it depends more on complexity. RTF is formatted text only. Formatted text in a PDF (see '1984' example) doesn't seem much different in terms of rendering time. A complex PDF with graphics etc. takes longer to render and thus needs more power per page turn.


^_^

[scotty1024]

Originally Posted by dstampe

This shows a big spike of current draw as it starts up and charges all its capacitors. Then the variable load producing the noise is a combination of this supply and the variable capacitive load as different lines in the eInk display are charged and discharged, Obviously this is a big user of power.

I agree on all points but one: that spike is the inductor in the eink DC-DC convertor being charged.

[dstampe]

Originally Posted by scotty1024

I agree on all points but one: that spike is the inductor in the eink DC-DC convertor being charged.

I guess it depends on the power supply design. When I'm designing circuitry like this, I tend to go overboard on capacitors for the output of the power supply to reduce EMI and improve requlation, so it can take a LOT of current at startup to charge these. On the other hand, most DC-DC boost converters for applications like this where size and weight are an issue would use very high frequencies (>500 KHz) and extremely small inductors, so these might not store much energy.

Anyway, I presume Sony took some care to get high efficiency from the converter, so most of the load can be assumed to be unavoidable given the voltage requirements and capacitance of the eInk display. Hopefully the evolution of these displays will reduce the voltage and increase the contrast as well.

It also looks like the display uses less than 40% of the total for most updates, so this may already be at the point of diminishing returns, Less time spend running the CPU and reading the flash would seem more likel;y to pay dividends.