MagTag screen washes out with predominantly black images

[CuGat]

• MagTag
• Arduino IDE
• Adafruit GFX
• USB power source
• Good USB cable (power & data)
• Refresh every 10 mins
• No peripheral devices attached
• Multiple MagTags incl. brand new ones test the same way

Small amounts of black to screen -> normal operation:

MG1.jpg (112.5 KiB) Viewed 189 times

Large amounts of black to screen -> defective operation (grey washout left side):

MG2.jpg (152.97 KiB) Viewed 189 times

Using color other than black -> display works fine:

MG3.jpg (134.17 KiB) Viewed 189 times

Display is being operated well outside minimum refresh time requirements
https://www.good-display.com/news/80.html - The refresh intervals should be at least 180s (except partial refresh mode)

I note this - https://www.good-display.com/news/79.html - 42 / What are the possible reasons for E-paper displaying greyish but not full black image? / If a greyish not full black image appears it's usually the boosting circuit does not have enough voltage to drive the black particles onto the surface and that led to a greyish display result.

During full display.display() refresh, screen goes through normal annoying flashing etc and reverse video image normally happening in the middle of that sequence (black background, white lines) appears fine. Then screen reverses to show regular image, and the washout occurs. Washout is severe enough to obscure 95% of image in many cases. Attachment filesize limit too small to attach video of behavior. Repeating screen refresh several times does nothing to fix. Purchased manufacturer's devkit and screens to test same issue on different drive circuit -> several weeks before shipment arrives.

Appreciate any insight you can give me - thank you in advance.

Relevant code, vanilla setup:

Code: Select all

//In decl:
 
#define EPD_DC 7 
#define EPD_CS 8 
#define EPD_BUSY -1 
#define SRAM_CS -1 
#define EPD_RESET 6
ThinkInk_290_Grayscale4_T5 display(EPD_DC, EPD_RESET, EPD_CS, SRAM_CS, EPD_BUSY);
 
#define COLOR1 EPD_BLACK
#define COLOR2 EPD_LIGHT
#define COLOR3 EPD_DARK
#define COLOR4 EPD_WHITE
 
//In setup:
 
display.begin(THINKINK_GRAYSCALE4); 
.
display.clearBuffer();
display.setTextColor(COLOR1);
display.setRotation(4);
display.setTextSize(2);
display.setCursor(25, 15);
.
In loop, with barCount a counter to 36 bars, and various scaling factors which are constants:
 
for (uint8_t i = 0; i < 5; i++) {  // draws 6-pixel thick histogram bars
// (X_offset * 1.5) is the starting offset to Y-axis; add 5 to exactly offset the first bar from axis; draw 6 lines at 1++ pixel off first for fat bars; then draw next bar 7 pixels away
      display.drawLine(X_offset * 1.5 + 5 + i + barCount * 6, height * 1.5 + Y_offset, X_offset * 1.5 + 5 + i + barCount * 6, displayBar, COLOR1);  // 7 pixel shift to right per bar
    }
display.display();  
.

[CuGat]

Update:

Changing to mono solves the problem of screen washout, at the expense of losing the 4 grayscales. Which kinda sucks given the grayscale capability of the MagTag screen. So screen boost voltage etc. is fine, but there might be a GFX driver issue.

From:
display.begin(THINKINK_GRAYSCALE4);

To:
display.begin(THINKINK_MONO);

[adafruit_support_carter]

Interestingly, both variants are used in the guide. The shipping demo uses mono:
https://learn.adafruit.com/adafruit-mag ... pping-demo
The quotes example uses grayscale:
https://learn.adafruit.com/adafruit-mag ... es-example

Can you create a short simple sketch that demonstrates this issue and post full code.

[CuGat]

Shortest sketch that shows the problem, result is the same whether use 5 second e-paper update or mnfr recommended 180 second update. Uncomment whichever time preferred. Uncomment whichever driver instance preferred.

Code: Select all

#include <Wire.h>
#include "Adafruit_ThinkInk.h"

#define EPD_DC 7
#define EPD_CS 8
#define EPD_BUSY -1
#define SRAM_CS -1
#define EPD_RESET 6
#define COLOR1 EPD_BLACK

ThinkInk_290_Grayscale4_T5 display(EPD_DC, EPD_RESET, EPD_CS, SRAM_CS, EPD_BUSY);

void setup() {
  display.begin(THINKINK_GRAYSCALE4);  // <-------------- switch these to see problem
//display.begin(THINKINK_MONO);
  barDraw();
}

void loop() {
}

void barDraw() {
  display.clearBuffer();
  // data bars
  int32_t millisPrev = millis();
  uint32_t barTime = 5000;      // 5 sec (demo speed)		<--------- update time
  //uint32_t barTime = 180000;  // 180 sec (mnfr rec)	<--------- update time
  for (uint8_t barCount = 0; barCount < 36; barCount++) {
    while (((millis() - millisPrev) < barTime) && (barCount > 0)) {
      yield();
    }
    for (uint8_t i = 0; i < 5; i++) {
      display.drawLine(38 + i + barCount * 6, 102, 38 + i + barCount * 6, 10, COLOR1);
    }
    millisPrev = millis();
    display.display();
  }
}

MONO (starts & ends fine):

Forum 1.jpg (158.3 KiB) Viewed 139 times

GRAYSCALE (starts fine, progressively gets washed out, eventually -> almost no image)

Forum 2.jpg (170.39 KiB) Viewed 139 times

[adafruit_support_carter]

Thanks for test sketch. Not seeing that behavior here. GRAYSCALE4 looks OK (same as MONO) testing on a MagTag here.

Is that a new MagTag? It looks like one of the NeoPixels has been knocked off:

np.jpg (24.71 KiB) Viewed 133 times

[CuGat]

Yeah that MagTag's been through a few international flights jammed into a bag :) Unfortunately same behavior with brand new MagTags.

Given that I show the same behavior, with this code, on multiple Magtags, our experiences differ! I guess I'll post a video showing the behavior. Are you allowing your MagTag to update sequentially in time, per the code, i.e. 36 sequential display refreshes, or just digitizing a final picture and placing that entire image on the screen with a single refresh?

[adafruit_support_carter]

Just running your sketch above as is. It takes a bit for the whole sketch to complete since it draws them one at a time with a 5 second pause for each. But when done, looks like below.

And just to be clear, this is with:

Code: Select all

  display.begin(THINKINK_GRAYSCALE4);  // <-------------- switch these to see problem

magtag_test.jpg (94.5 KiB) Viewed 130 times

[CuGat]

Forum doesn't allow me to upload an mp4 of the behavior. Compressed video as the image transitions from updating fine to updating washed-out available at https://www.dropbox.com/s/5gkgjvkmr7kd3 ... d.mp4?dl=0
Do you have libraries loaded I don't? Newer version of MagTags or different batches than the 5 units I purchased Jan '22? Could you post your entire code if there are any variances? Should I send you my MagTags for testing at Adafruit? Any other ideas?

[adafruit_support_carter]

Running your sketch above as is. No changes.

Latest libraries for everything.

Newer version of MagTags or different batches than the 5 units I purchased Jan '22?

Unfortunately same behavior with brand new MagTags.

Is the "brand new" MagTag from Jan '22 also?

[CuGat]

5 from Jan 2022. I'd buy more but they've been OOS for a long time. I have the mnfr's break out board and devkit for the EPS, I guess I could start probing voltage levels on that flex connector. The thing that is so strange is it works perfectly for me, using Grayscale, if the bars are short. It craps out only with long bars and only when there are a lot of them (like 75% + of the 36 bars drawn). Feels like insufficient screen drive for the electrophoretics when a lot of black required. Yet you don't show the defect. It's my identical code running in both places, you're pretty Ada-mant. COVID parts? Earth's core reversed direction in 2009 and the Pacific Northwest hasn't caught up yet to NYC?

[adafruit_support_carter]

The MagTag's I'm using for testing are from early 2021.

For the MagTag in the photos above with the damaged NeoPixel and history of international flights - the behavior could just be from usage.

So it's more interesting for the one that is "brand new" but is showing this behavior. It's from Jan '22 and has never been used until now?

[CuGat]

Yes, Jan 22 purchase and used only for the forum post. Back in a week and will unbox and try all 3 additional NIB Magtags I have - update then - thanks for the inputs and patience