E-paper hacking: fastest possible refresh rate-----make money online

Add caption

today on Applied Science I'm going to

talk about these e-ink or ePaper

displays these things too are made

popular in the Kindle and other ear

readers and now they're available in the

hobbyist market you can play with

putting them in all kinds of different

projects of course one of their

signature features is that they retain

the image even when it's not connected

to anything at all so it takes literally

zero power to maintain the image and of

course the the look of it is much more

like paper than it is a liquid crystal

display they also have these interesting

red white and black displays which are

pretty cool it's not a color filter this

will talk about the tech that goes into

these but it's actually really unusual

how these work so the setup I have here

today is a TT microcontroller that we're

programming through the Arduino

interface and it's driving this display

over the spy bus you can see I bought

this one from WAV share and then we have

this glorious scope from Tektronix on

loan from Tektronix to monitor the spy

bus wave share is not a manufacturer of

these panels that they are a I guess a

distributor that also makes these

breakout for it so it's very convenient

to have a board that converts the flat

flex that's on the display to this pin

header and also they provided some

firmware which is a great starting point

so let's zoom in on the display and

check out some demos ok I'm going to run

the demo that comes direct from wave

share it's just 2 frames and so it's

compiling and downloading and you'll see

when the image transitions first you see

the inverse of it then all white then

all black then the image this is the

frame and then it's going to switch back

to the first frame and as you can see

it's a pretty long process so you can't

change any part of the display without

going through that entire multi second

refresh routine and the early Kindle and

other eReader type things had this same

characteristic to them

you can't just update the screen you

have to clear it entirely however if you

know how to tweak the firmware you can

make the screen do anything you want

actually interestingly the controller is

very programmable

and it's sort of security by obscurity

you can load any values you want into

the registers and cause it to do

anything you may hurt the display but

let's take a look and see what's going

on okay so I'm gonna load up the demo

reel that I made with my custom firmware

and we'll see how fast the update is so

the first one clears it that's just a

standard update and now you can see the

frame rate is about 3 Hertz so we don't

have to flash the whole display and we

can update any part of it within a third

of a second this one shows what an

animation would look like it at close to

3 Hertz and you can see that I'm drawing

text up here and I'm slowly moving the

text in the x-direction across the

screen to prevent burnin which we'll

talk about a little bit later the ball

going around in circle doesn't really

cause much burn in but you will see if

you look closely there's a ghost image

as it was going to a full white screen

so you can still see that ghost image

now we'll go back to wave share these

are all full updates as you can see

because it's going through this cycle it

that should look familiar you can see

when it's done cycling there's no more

ghosts left like the text is completely

clean when it's going through its image

update process you can see a little bit

of remnant so we'll talk about this a

little bit more but I think that the

pulse sequence I've come up with is more

or less safe but it's possible that if

you you know some would abuse the

display like this eventually there will

be burn in permanent maybe certainly

short term burnin for sure let's talk

about how these things actually work and

then it'll make it clear why we have

this problem with burnin so if we were

to cut one of these displays and look at

the cross-section from the side we would

have sort of a smooth top which is an

electrode and we'll call that common and

then within the display there's a whole

bunch of chambers and these are not

necessarily pixels they're just little

chambers within the display and within

each of those there are positive and

negative particles and they're submersed

or they're suspend

in a fluid oil of some kind and the

trick is that the positively charged

particles are black let's say and the

negatively charged particles are white

so if we put an electrode array on the

bottom and in this case each one of

these is a pixel and if we put a

positive voltage up here in a negative

voltage here the field will cause the

negative particle to migrate to the top

and the positive particle to migrate to

the bottom so what we'll end up with is

white particles if those are negative

going to the top of this pixel and then

as we scan across we can change the

fields for each one of these pixels

pretty cool in its most basic form you

could just put one voltage across this

thing and if you're standing through all

the pixels you could say okay you know

positive on top negative on bottom for

this one positive on top negative and

scan through it and the display would

work at least once or twice but the

problem is with electrostatics and

fields like this eventually the parts of

the display will become charged

themselves and so if you keep putting

this electric field on the display

eventually this positive charge will

accumulate in the display and then your

display won't work anymore because when

you put a voltage on there you're going

to be fighting against all the charge

that's in there so one of the tricks to

driving these things is you actually put

an AC voltage on this and rapidly cycle

it to prevent it from becoming polarized

there's actually a similar problem with

LCDs you can't run an LCD with just DC

voltage because it will polarize and

stop working if you search around on a

paper distributor sites you'll see that

some of the newer displays are actually

rated for what they call partial refresh

so basically doing this without any

flashing or flickering get a new image

on the screen without any of that and I

think the trick is they're basically

just loosens the sort of the usage

restrictions and so they're saying okay

you can bend the rules a little bit by

putting just a DC voltage on there to

get a quick update but then you better

do a full refresh at some point near in

the future

I think you know the reason that it's

hard to find instructions on how to do

this is because they know that you're

going to degrade the panel's performance

if you don't have the right routine so

it takes a little bit of experimentation

to get all this working additionally

this display that I'm using this 4.2

inch actually has two complete frame

buffers inside of it and so it knows it

keeps one frame buffer is like the

currently displayed image and then the

other frame buffer is the image that it

wants to display next and what it does

it knows that if there's if this pixel

or if this pixel is currently white then

it knows that all the positive charges

in the top or whatever and if it knows

that it wants to convert that pixel to

black in the next frame it actually

gives it a different sequence of

positive and negative voltages than if

it's going to stay white in the next

frame so the the display has this thing

internally called the lookup table and

it basically has a set of waveforms that

we're going to put on these pixels and

the waveform is selected based on what

the pixel was in the last frame and what

the pixel is in the new frame and it

seems kind of overkill but the obviously

they wouldn't go through all this

trouble if it wasn't necessary so coming

up with this correct drive sequence is

basically the secret all these are the

all of these a paper displays so what

about these red ones you might think

well it's probably just a color filter I

mean it's probably just white pixels

underneath a red thing but that's

actually not what's going on it's a true

three color pixel it can select whether

it's going to be red black or white and

I searched drown that took a long time

to find this out but I think the way it

works is kind of like this imagine that

you have black pixels that are small and

positively charged and then you have

really large red ones that are also

positively charged and then you've got

small white ones that are negatively

charged so then if you put this thing

between the two electrodes have the

large ones and the small black ones will

have the same motive force applied to

them if we put it this field here minus

on the bottom and positive on the top

both of the red and the black will be

attracted to the top let's say

the red ones are going to move more

slowly because they are bigger even

though the charge is about the same and

it's in this viscous sort of fluid so if

you put a voltage on here quickly the

black ones will make it to the top

faster than the red ones and the pixel

will turn black if you want to turn the

pixel red what you do is you first make

it black so that you're in this

configuration with the white ones in the

bottom the red ones kind of in the

middle and the black ones at the top

then you switch the field causing the

black ones to move down and the white

ones to move up but before they have a

chance to switch places the red ones

will actually end up being the ones that

are closest to the top and the black

ones will be kind of below and the white

ones too haven't started moving just yet

I think each manufacturer has their own

special trick of how this works and even

searching patents and things it's

actually very difficult to figure out

what's going on here but when we start

looking at the code you'll see there's

special pulse sequences to get the red

ones to the top it's amazingly effective

I mean it's a very high contrast sort of

display and there's no I mean you can

see the background is actually a little

bit white compared to a a black and

white only ePaper you can see the

background on this one is much paler

it's much truer to white this one's a

little bit reddish probably because

those red particles aren't fully at the

bottom but nonetheless it's a very cool

look okay let's take a look at the code

that runs this thing and the changes

that I've made to do this fast update so

the code from wave share is free and

public domain and so I've made some

changes to it and then redistributed it

on my blog check the description for

links to all this stuff and what it is

is in our Dino library and an Arduino

demo sketch so you can get the thing

running and so in libraries it's going

to be this EP D 42 directory and almost

all of the work is done in EP d 4 inch

to cpp the sketch itself is pretty

self-explanatory I'll also include a

demo basically the same demo reel that

you saw there and it's relatively

straightforward especially if you use

the LCDs from Sparkfun or Adafruit

before basically the ideas that you can

stir

- frame buffer and edit the frame buffer

by using these you know draw a filled

circle routines or put text in there and

then you send it to the display and tell

the display to update and some time will

go by about a third of a second wait

until idle and then you can do something

else with the display so pretty

straightforward so let's take a look at

the CPP file from wave share if you're

new to editing C code at this level it

may look a little gnarly but just stick

with it it's not too bad I would

recommend starting from the sketch and

seeing what functions are being called

here and then tracking them down in EPD

4-inch to CPP so you can see we've got

EPD set partial window so if we scroll

through here in fine set partial window

there it is

and I've modified the function slightly

from the original wave share function

and I've added one new parameter called

DTM which is this data transmission mode

so I mentioned that the display has two

frame buffers one for the old image and

one for the new image and you can

actually edit each flame frame buffer

completely independently so you can tell

it whether you want to insert the new

image of the old image at first it

doesn't seem to make much sense but

remember that the waveform that is being

given to each pixel is determined by the

difference between these two frame

buffers so if you power off your device

you can actually restore the old frame

buffer from non-volatile memory so that

your display always has the correct

waveform it's kind of a little bit

unnecessarily complicated you can

basically just always assume that you're

starting from a white display or you can

even write your firmware to always leave

it in a white display state you know

it's kind of up to you anyway the next

thing that happens is display frame

quick so let's take a look at that

here's display frame quick it's pretty

similar to the original display frame

the main difference being I've got set

LUT quick here and I've gotten rid of

the just the delay commands so let's

take a look at the difference between

set LUT in the original and set LUT

quick will scroll up the command happens

to be above it in the file

so the original set LUT

uses these variables lu TV comm zero and

Lu TW and as you can see they have some

descriptive names here white to white

black to white so this is the difference

between frame buffers for each pixel so

it picks out a waveform from this lookup

table for each of these pixel

transmitted transitions so the original

set LUT just uses these variables and I

have like a parallel function that uses

quick quick variable functions or quick

variables and at the end of this file I

have all of the stuff written out so

let's start digging into what this is

actually doing okay so here we've got

the lookup table on the left and the

data sheet on the right that explains

what all these values do and it was very

difficult to find this data sheet this

is actually for the IL 0 3 7 3

controller which is not the one used in

the 4.2 inch ePaper display if you start

off with the correct data sheet the 4.2

inch ePaper and track down the

controller that it's using it's the IL 0

3 8 9 and all of the lookup table

information has been removed from this

data sheet because the industry thrives

on secrecy or something but luckily

there aren't that many ePaper display

controller manufacturers and they happen

to use the same lookup table format in a

lot of different displays however it's

not the same in all of them and so you

have to start poking around and lots of

trial and error and whatnot the thing

that was the tip-off here is that these

look-up tables are 42 bytes long hey

magic 42 and the this datasheet Specht

42 byte look-up tables so I knew I was

on the right track oh I should point out

that the wave share files came with this

lookup table that's how I knew it was 42

bytes because there was in fact an

original lookup table in the way of

sheriff files however funny enough if

you look through the code or written the

original code carefully it doesn't use

them in fact it even has a function cut

called set LUT but it never calls the

function or I think it maybe it did call

the function but it didn't set up the

display to use it properly I'm not going

to get into the super low-level details

but if you use my code it works just

fine this is stick with that there

is actually a couple of really nasty

typos in the original wave share code

okay let's take a look at what this

thing's actually doing you can read this

thing left to right top to bottom just

like a book and the very first bite out

of the sequence tells it what voltages

you want to put on that specific thing

and if we look you actually can control

V comm which is the what we're calling

the common electrode the one that's

shared among all pixels so you can

actually control that voltage and if we

go to the datasheet for V comm the

lookup table for V comm that first byte

selects four different voltages and it

can be v c md c VD h VD l or floating

the subsequent four bytes tell this

thing how how long you want those

voltages to be applied for this part of

the waveform and it says number of

frames and so in this example over here

we've got hex 40 gives it the voltage

values that we want so we can come to

our calculator here and say hex 40

that's zero one zero two zero zero zero

zero zero so zero one is VD h plus vc

MDC and then the next one is hex 17 so

we'll try that in decimal its twenty

three and as it happens a frame the

frame rate is 50 Hertz and you can also

change that through the things but let's

just say it's 50 Hertz so then 23 times

you know 53 hertz gives you how the

length of time that that voltage will be

applied in some of the other data sheets

they actually had a better graphical

representation of this so you can see

this is kind of a little bit easier to

see what's going on here so out of all

these little bits and bytes that are put

into that lookup table you'll eventually

end up with a voltage waveform that's

applied to the pixels and that's what

makes this whole thing tick so let's

take a look at the original lookup table

versus my lookup table

so here's white to white original and

white to white quick so you can see the

original one is actually pretty

complicated it's a hex 40 which is the

voltages we're going to apply then at 17

which is the time

the voltage is going to be on and then

this fifth bite is actually our sixth

bite is actually the number of times to

repeat that whole part of the sequence

so this is voltage forty times seventeen

repeat it twice then it does voltage

ninety times seventeen times seventeen

repeat that twice yada-yada

so the original pulse sequence quite

long but here's what I did I took the

final voltage that was going to be

applied to it a zero and then the final

amount of time zero E and I'm only going

to repeat it once so basically I

truncated the look-up tables just using

the last value from each waveform this

makes sense because the last thing that

the waveform does is sets the pixel to

the proper value or if this is the

proper color so what this thing is doing

is let's say you want to go white to

white it's actually first converting it

to black then converting it to white

then to black then to white again but

that's very messy looking and if we're

trying to do a quick update that's no

good so we just go straight to white and

if the pixel was already white to begin

with we just make it white again no

problem and so you'll see that that's

true for all of these things the

original black to white LUT looks like

this and again I just cut the last bit

out of it let's see if we can actually

probe those voltages to bring this thing

full circle and see if our code is

actually working the way we think it is

so you'll see here I've I've cut away

the encapsulants it was actually a

polymer that was covering up this whole

area on here and this long bit of

silicon is actually the driver chip that

is doing the the work of converting our

serial commands into this giant you know

matrix scanned pixel array of that is

the display itself you'll see the

abbreviation Co g which is chip on glass

which is that long piece of silicon the

actual driver chip another little gotcha

is that the glass itself has a lot of

wiring on here but the wiring is all

covered with a very hard insulating

layer some kind of oxide so if you just

take a even if you take a relatively

sharp steel probe and just touch down on

the surface here you can't access any of

the electrical

signals in there because the steel isn't

hard enough to cut through it instead

what we have to do is use a tungsten

carbide scribe err and cut away the top

oxide layer and then get a probe down on

there so I'm gonna zoom in even further

and see if you can see this here's a

closer view and if you look carefully

you can even see they were nice enough

to leave us some test pads on here and

that one even says comm so that's

probably be calm and then a little bit

lower you can see here where I scratched

away the coating and believe it or not

this actually works pretty well you

would think and you scratched away the

top oxide layer it would blow away the

conductive layer underneath but as it

turns out you can put a probe here and

it actually has pretty good conductivity

so let's move over to the side of the

display and you can see where I'm going

to get the individual pixel lines this

is the edge of the display and if you

look closely you can see this honeycomb

structure those are the ended those are

actually not pixels those are just the

cups that contain these colored

particles and so the actual pixels can

be independent from this sort of cup

structure and then here's a little bit

where I've cut away the encapsulants and

scratch it with this scriber and then

try to get a probe on there

I forgot her I wasn't thinking this is

actually an active-matrix of transistors

in there so the row and column inputs

are scanned and so seeing any sort of

meaningful waveform from there is very

difficult because it's spread out across

all the pixels and so I guess we could

change the entire display that I was

having trouble probing so just based on

this diagram here it's much easier to

measure V comm so let's take a look at

the scope now and measure V comm since

it's much easier and then will alter the

let the lookup table and see the effect

it has one of the neat features of this

new oscilloscope is that it has eight

channels that can be configured as

digital or analog on each one so when

it's digitally you get eight digital

inputs you could actually have 8 by 8

digital inputs or you could have all

eight analog inputs or some combination

thereof so today I have eight digital

inputs here and we're keeping track of

the spy bus Plus this data command thing

reset and the busy flag from the display

and then I have channel one reading V

comm so I was tack on here and then run

the program and hopefully you'll be able

to see the display updating back there

and we should see V comm change into

something else let's see if it shows up

okay and there it is so V comm is 10

volts or 11 volts for a while and then

at 0 and then it's back to 10 zero and

10 again and if we look at the LUT we

can see that makes sense because the

final line of the LUT has starts with

hex 40 and ends with hex - which means

repeat twice and 40 corresponds to this

higher V comm voltage so I'll put this

back to zero and save the LUT and then

run the demo again

you can see interesting the display

still works which is interesting you'd

think just making a random change like

that would completely destroy it but

actually it's it's still okay and then

it went away so since we updated the

what table to be zeroed at the beginning

this thing just stays at ground the

whole time let's finish off by going

back to the digital and seeing how fast

the framerate is on my hacked firmware

version so I'm going to turn off channel

1 since we don't need the analogue and

I'm going to change the trigger to

instead of triggering on channel 1 we're

going to trigger on channel 6 and since

6 is a digital it breaks out all the

different pins in there so d0 is the

serial clock it'll just start triggering

right away the record length is super

long 125 million points and a sample

rate is 62 and a half million samples

per second so we should get 2 seconds

worth of data here which would be plenty

I'm gonna run the code and I'm going to

stop it and when the code gets to the

part where it's doing the rapid frame

update I'm going to do a single

acquisition so I'm watching the display

and now is a good time so I'll hit it it

triggers very quickly and then 2 seconds

later we should be able to see all the

data there

okay cool so this is each frame being

updated so if we wanted to know what the

frame rate was we could use the cursors

and just get a real quick idea so from

the end of that one to the end of this

guy is 374 milliseconds 2.7 Hertz and so

that's that's about the frame rate that

I was getting now he can make it even

faster since we have control over the

waveform we can make it you know almost

nothing the problem is the display will

have very poor contrast and so it takes

time for those charged particles to rise

to the surface we can also take a look

at you can see that so the this

chunkiness here this is how long it

takes the microcontroller to actually

send the frame to the display and then

this is the time it takes the display to

actually show it so if we look at this

thing a little more carefully here we

can zoom in on it so then looking at the

zoomed in section I'll use the cursors

to measure the time it takes to actually

send the stuff to do the display roll in

here so 24 milliseconds out of the what

did I say 2 or 300 or whatever it was is

actually taken up just by the spy bus

and in fact we're cranking really fast

it's doing about 14 megahertz is what

this thing is measuring here I think

it's actually a little bit higher it's

closer to about 18 megahertz see if you

can zoom in on that zoom way in even at

62 and a half million samples per second

it's actually hard to see exactly what's

going on we'll bring in the cursors

again and

this clock edge it's like 20 megahertz

yeah it's somewhere between you know 14

and 18 megahertz clock if you look at

this the the data being sent has a lot

of dead time in between and it's

partially because the chip select line

has to toggle anyway it's getting kind

of detailed that this scope does make it

quite easy to see what's going on in

addition it's decoded the entire set of

data and has it available in sort of a

bus form like this another cool feature

is that if you know what you're looking

for you can search through all the data

so for example I have this search

already set up and it's looking for hex

20 hex 20 happens to be the code that

sets the V com LUT so we can use the

navigate buttons and then jump through

the data and find all of those 20 hexes

so if we zoom out a little bit we can

see what it's setting the V comm table

to be so 20 and even has this this drag

zoom which is pretty cool too ok see you

next time bye earn money online by typing pages, earn money online daily, online work at home and earn money, online earning, earn money online free, online money earning sites, earn real money online, e commerce ideas to make money, easiest way to earn money online, best way to earn money online, make money from home, make money online free, money making sites, earn from home, real ways to make money from home, online income ideas, make real money online, top online earning websites, earn money online fast, top 10 money earning websites, earn money without investment, earn money online for students, best sites to earn money, make money online 2019, earn easy money online, earn money online instantly, real money earning sites, make money from home online, make money online fast, earn dollar online free, online earning ways, best online money earning, earn money online with surejob, make money online, online money making sites, best sites to make money online, earn money online paypal, online earn money by typing, best survey sites, ways to earn money online, ways to earn money from home, make money online without investment, make money online surveys, 

s