PaperBack: A Desktop EPaper Monitor

Here's a picture of my current status, with 4 working PCBs:

Unfortunately, I destroyed some outputs of 2 of those ESP32s in the process. One drives the displays but skips every other line (probably the data lines). One... drives the power supplies but doesn't update the displays (same).

But the new one - my third - drives an EPD perfectly! Isn't experimentation grand?

Next step is waiting on the PCBs for the conversion board and a components shipment. Here's a teaser:

Ready to build your own? I just posted a wealth of IP to Github, ready for you to purchase some PCBs and match my results. I build the PCBs in Kicad, but I've also included Gerbers and a PDF Schematic for you non-Kicaders.

At this point the project matches the results of some of the projects I mentioned in this log. The board in the picture is an ESP32, which has enough pins to drive the 6" ePaper display without shift registers.

I will continue to improve the software on the ESP32 and make it do some funky web tricks - so even if the rest of the project is a dud we'll have a product.

In parallel, I sent the first draft of the conversion board off to @oshpark, who generously sent me a coupon. With some luck and a lot of VHDL and C, I'll be able to take VGA in at 800x600. I'll then have the board grab a frame and convert it to something the ePaper Display can read, then pump it out to the screen. If all goes well, we'll perform that symphony up to (it'll be an option) roughly 1x a second or so.

Let me clean up a bit and I'll toss them up on Github. Here's a preview:

On the left is the new run, on the right is the old one. I fixed the mistake that caused me to bodge, moved some components, and made more parts surface mount.

The only problem? It takes me about 90 minutes to put one together! Stay tuned, I'll post again when everything is uploaded.

Just a snap - I'm still waiting on a parts order to put this batch together.

Fingers crossed for no bodges!

I'm working on the VGA graphics conversion boards in parallel. Expect an update soon.

The next step is taking some monitor input and getting it to display on the ePaper Display.

I'm going with VGA - I'll try to use a MST9883 (that's what was on eBay). The canonical part is the Analog Devices AD9883; essentially it will save a lot of steps with this project. It takes RGB input and recovers a pixel clock from HSync and dumps all of the pixel colors in parallel.

Also, there is an I²C component - EDID on VGA and settings on the MST9883. I'm going to just use a microcontroller to do these.

Add up memory plus the MST9883 plus a couple SPI busses plus the ePD signals and you're looking at a minimum of 68 to 83 I/Os on the FPGA/BIG_CPLD (depending on color depth), probably 71 minimum to get 4-bit 'color'.

Any suggestions on FPGAs here? I'm reasonably agnostic, I just want something reasonable with plenty of LEs.

I may have missed that IoT deadline, but eye on the Prize - I'm targeting Best Product anyway.

Above is what I've got working so far.

I sent out for a second round of PCBs, Green/ENIG. I added a few creature comforts, but sorry to say I kept it safe and stayed with the LT1945 for high voltage generation. The TI all-in-ones still had a lot of parts and I'm not super confident I can solder unexposed leads.

Anyway, hopefully this one doesn't need a bodge and I'll be publishing it. Here are the differences:

• 10 Control signals (down from 10, I'm tying off RL and SHR, keeping the 3 voltage on/offs)
• 8 Data lines
• Switching to DPAKs on the 15V/-15V Regulators (nice to sit flat this time)
• Adding a resistor divider to one of the ADCs on the ESP32
  • (Even though it isn't going to be part of the final build, I might keep a few around with the '32 onboard)
• Switched some TH components to SMD
  

... and that's about it. Again, I'll publish it if I got it right this time - I don't want people to have to read all my entries to know not to build a version, haha.

And in the meantime I'll work on the second board: converting video inputs into something that can be displayed onePaper. From what I've seen, we can make this work - probably at 1.1 frames per second maximum, heh.

I've put what I've got up in Github; get it while it's hot! I've got firmware for the ESP32 and my Python desktop image conversion utility so far. I'll work on something internet-ty next [Ed: see below].

I'll go back and fill in some more comments when I get some time; hopefully it's understandable as is.

Here are the numbers on screen clears and screen draws after my cleanup:

# Clear the screen

Clear cycle.
Took 1075 ms to redraw the screen.
Going active again.

# Next to draw an image

Draw cycle.
Took 1531 ms to redraw the screen.

There are probably some bugs in there, but... I'll plow ahead with that IoT (I-o-Useful-T) deadline looming [Editor: Never did get it running before Monday morning...]. Call it 1 second to clear and 1.5 seconds to draw.

I'm not sure how useful it'll be to you all, but I'm cleaning up and commenting my code before I post.

I'm using the Arduino Core for the ESP32 and the standard signal names for these ePaper displays - so even though the boards aren't working, maybe you can borrow my timings or look at how I'm driving the screen. As I promised, I'll spin up another board - but since I have two more days and an ESP32 I'll be working mainly on Internet features this weekend.

Also, I'm adding greyscale support to the display, using a similar method to NekoCal's. I use progressively greater number of refresh cycles to change the brightness of an area. I'm experimenting with 16-grays or 4-bit color and it doesn't look so bad right now.

Once I lock that in I'll post.

Here are my current timings so you'll know if I made progress when the code goes up:

4 output cycles coming!

// The next one is a clearscreen
Took 1327 ms to redraw the screen.
Did anything happen?  Adding one and trying again.
4 output cycles coming!

// The next one is drawing a picture.
Took 6351 ms to redraw the screen.
Did anything happen?  Adding one and trying again.

And what am I drawing in those 6.3 seconds? E-Paperception, of course!

Okay, more soon - and more speed. Stay tuned.

A picture is worth 1,000 words.... or in this case, 120,000 unsigned 8 bit integers.

I'll have a better write-up soon and publish some code (c'mon, it's 1:00 AM now here), but here are the details:

• 6" ePaper Display (this model is a LB060S01-RD02)
• Bodged PCB (see earlier article)
• ESP32

Pretty excited about this - now I can turn this into an Internet of Thing while I work on the "standard monitor" part.

A series of (both unfortunate and scheduled) events kept me busy in the last few weeks, but I did have some time to send out v1 of the PCBs.

Unfortunately, in my haste my schematic was (very) wrong - I messed up the negative feedback stage on the LT1945. Pretty badly too - the one I powered up is no longer with us.

Umm, don't build this. Just shake your head with/at/near me:

I was able to bodge a few by removing C1 and building some nonsense on protoboards, but I'll spin a new board. @Johnny Karamello suggested I take a look at TI's selections and specifically the TI TPS65185x, so I'll gather my courage and think about how to hand solder a QFP with a bottom pad (Suggestions? Big hole on the bottom, flood with flux and paste?).

Here's a bodge in all its glory!

Captain Steven Hiller: What do you say we try that one again?
David Levinson: Yes, yes. Yes. Without the "oops".

Send me your funniest/most impressive/most hacky bodges for motivation? Haha - stay tuned, I'll sort it out.