SHARP Memory Display Booster Pack

[chicken 630]

@@greeeg a minor detail on not driving DONE: DRVn goes low 50ms before the interval ends. So in our scenario the pulse length would be ~850ms. Still worth a try though.

[chicken 630]

Toggling VCOM in hardware (part 2b)

 

A quick belated update: @@greeeg was right that omitting the DONE signal saves a lot of power as there's no charge current for the RC circuit.

 

I tested this by removing R3. I left C4 in place, as a bit of decoupling won't hurt and might even be a good idea. (edit: actually, tying DONE to GND might be the better idea)

 

Here's the modified circuit:

 

As expected, the wave form is now 860ms high, 50ms low. I haven't tested yet what the display thinks about the inverted pulse width, but as observed by greeeg, if you follow the datasheet by the letter it shouldn't matter.

 

Idle current measures exactly the same 37.4nA as above. The peak seems much shallower, the ammeter briefly jumps by about 0.6nA to 38nA or so. As suggested above, I will have to do some measurements with Energy Trace to find the actual peak profile.

 

I think it's save to say that with this approach the average power consumption of the VCOM circuit is now below 38nA.

 

With this we'd be at 2% of the display's static current (2uA), i.e. 5x lower current than the original goal of 10%. And it's less than 10% of a G2553 in LPM3 (500nA). I'd call that negligible in pretty much any real world scenario.

 

Edit: added schematic

[chicken 630]

Toggling VCOM in hardware (part 3)

As suggested by @@greeeg I tested the VCOM circuit with EnergyTrace.

I compared two BoosterPacks. One with the TPL5100 based VCOM circuit, and one without. To maximize the impact, I display a static image while the MSP430G2553 is in LPM3. There's no software VCOM update in both scenarios.

Theoretically that should give us something in the range of 2.5uA (2uA for the display, .5uA for the MSP430). EnergyTrace shows a mean current of 2.9uA (0.0029mA) for both scenarios, which seems close enough. The impact of the TPL5111 (37.4nA as measured previously) is smaller than the least significant digit that EnergyTrace displays (100nA resolution).

EnergyTrace can compare a measurement to a previously saved profile. Here Reference was recorded with VCOM circuit, and Live is without:


Results varied a bit over several runs, but the differences for total energy and battery life was always a negative impact of the VCOM circuit of 1.5% to 2.5%. Again pretty close to the theoretical impact of 2%.

Displaying a static image for more than 8 years on a coin cell battery sounds pretty cool

One last question remains: How do we know the hardware VCOM actually works?

An interesting property of the SHARP displays is, that its circuit is visible on the glass and changes from light to dark gray based on voltage. Here three video i recorded with a USB microscope:

1) VCOM toggling in software

2) VCOM toggling in hardware

3) No VCOM toggling

I think we can declare this a win.

[yosh 121]

Brilliant writeup @@chicken

[greeeg 460]

This is one reason I love working with the sharp memory LCD. You can see the circuits physically running, I'm assuming its because of how they're made inside the liquid crystal display.

@@chicken Can you see a visual contrast difference when using VCOM vs not?

 

 

Off topic, But an interesting read discussing battery life from CR2032 cells. http://www.embedded.com/electronics-blogs/break-points/4429960/How-much-energy-can-you-really-get-from-a-coin-cell-

[chicken 630]

@@chicken Can you see a visual contrast difference when using VCOM vs not?

 

Only in the traces. The pixels on my display are reflective and I can't see a contrast difference.

 

Off topic, But an interesting read discussing battery life from CR2032 cells. http://www.embedded.com/electronics-blogs/break-points/4429960/How-much-energy-can-you-really-get-from-a-coin-cell-

Interesting but one shouldn't draw 10mA+ from a button cell anyways, as one of the commenters was pointing out. I was hoping for a true long term analysis, self discharge and such.

[greeeg 460]

Interesting but one shouldn't draw 10mA+ from a button cell anyways, as one of the commenters was pointing out. I was hoping for a true long term analysis, self discharge and such.

Sorry, I actually thought I was linking to the much longer article the Jack wrote. That one must have been an earlier article.

 

http://www.ganssle.com/reports/ultra-low-power-design.html

This goes much further in depth, with some actual long term results.

 

Some good stuff, including some hidden traps, like current draw from cheap decoupling caps.

[chicken 630]

Oh yes, I remember that article. In the nA range even cleaning up your flux can make a difference. 38nA is a 100M resistor at 3.3V.