Topics created by wsanders

I got the sketches to work, They will be posted to the above repo shortly. The key changes from some of the existing code out there that does not work anymore:

After WiFi connects, the delay must be increased to at least 1 sec:

while ( !WiFi.isConnected() ) delay(100); delay(1000);

Preallocating the source buffer seems to be essential:

const int preallocateBufferSize = 128*1024; const int preallocateCodecSize = 85332; // AAC+SBR codec max mem needed void *preallocateBuffer = NULL; void *preallocateCodec = NULL; ... // Don't forget to malloc the buffers preallocateBuffer = malloc(preallocateBufferSize); preallocateCodec = malloc(preallocateCodecSize); ... srcbuf = new AudioFileSourceBuffer(file, preallocateBuffer, preallocateBufferSize); ... // Preallocating the MP3 generator doesn't seem to be essential //gen = new AudioGeneratorMP3(preallocateCodec, preallocateCodecSize);

You must call M5.begin with the 6th argument false to disable the M5.Speaker code. With this argument, the skeptch will produce audio, but at half speed and half pitch:

M5.begin(true, false, true, true, kMBusModeOutput, false);

My M5Core2 library version is 0.1.9. I have 2.0.9 of the M5 board. definitions.

@teastain I posted an issue in the library repo, perhaps the ambiguous comment will be clarified. M5Core2 was updated last Jan 29, latest update to M5Unified was 3 days ago.

@wsanders if you are programming in Micropython (the language behind UIFlow) use thonny, otherwise forget Arduino and use visual studio

I have a first edition Core2 and Core2 AWS and have no issues updating the firmware.

Hello, @wsanders

The app0 and app1 partitions are used for firmware update with rollback in case of failure.

It is also used by the OTA function.

https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/system/ota.html

@teastain said in Core2 buttons don't work after M5.Lcd.setRotation(3)?:

@themusicman This is "The Community" website supported by user volunteers.
You should raise an Issue over on the M5Stack Github page,
https://github.com/m5stack/M5Core2
-Terry

Hi Terry, sure will do. I simply followed the links on the official M5Stack page where 'Forum' is listed under the 'Explore' menu hence considered this an official forum of M5Stack. There's no indication on the M5Stack page that this is anything other than an official page of theirs, quite misleading many would agree I am sure, especially considering each of the other links on that page head to official M5Stack pages such as GitHub, News, YouTube etc.

In any event, many thanks for the clarification that this is a site run by the community, that's good to know.

Also, unfortunately, on GitHub I see infrequent replies to issues posted. It appears M5Stack are not particularly supportive to customers when it comes to official replies to issues. I've returned a device to the company I bought it from which was purchased only last week due to what appears to be a strange and possible design issue where the M5Core2 device is unable to be placed into BOOT mode in order to recover/restore factory settings. At £45... bricked after a week is not something I expected for what is essentially a very high spec ESP32 board!

The issue I have overall with M5Stack is that their devices are not cheap low cost £3.50 ESP32 boards, they are quite expensive items relative to what they are and as such M5Stack don't appear to run an acceptable official after-sales support service. This makes this community even more valuable.

Thanks again Terry.

@wsanders At least with respect to the LCD library: The only function implemented so far to manipulate whole images is drawBitmap. drawJpg and drawJpgFile are commented out in https://github.com/m5stack/M5StickC-Plus/blob/master/src/M5Display.h so I assume they aren't working yet.

So my workflow for now is:

Save the image from gimp in gimp's ".h" format. This is smaller than a xpm or bmp. You will get a static char *data structure of all the pixels in the image. The .h file includes a macro to extract the pixels:

#define HEADER_PIXEL(data,pixel) {
pixel[0] = (((data[0] - 33) << 2) | ((data[1] - 33) >> 4));
pixel[1] = ((((data[1] - 33) & 0xF) << 4) | ((data[2] - 33) >> 2));
pixel[2] = ((((data[2] - 33) & 0x3) << 6) | ((data[3] - 33)));
data += 4;
}

Write your own function rgb888to565 to compress the pixels into a uint16_t.

Draw a bitmap of the image as fast as you can:

#include <M5StickCPlus.h>
#include "1.h"
int pixel[3];
// pointer fu to preserve the start of .h data
char *datastart;
uint16_t *bitmap;

void setup() {
M5.begin();
M5.Lcd.setRotation(3);
bitmap = (uint16_t *)malloc(height * width * 2);
}

void loop() {
M5.Lcd.fillScreen(GREEN);
datastart = data;
for (int16_t y=0; y < height; y++) {
for (int16_t x=0; x < width; x++) {
HEADER_PIXEL(data, pixel);
bitmap[60*y + x] = rgb888to565(pixel[0], pixel[1], pixel[2]);
}
}
M5.Lcd.drawBitmap(0,0,width,height,bitmap);
data = datastart;
}

Or you can use the Sprite library, which works well.

@wsanders M5Stack hit the space limit for the MP firmware and had to remove several fonts. V2 firmware should fix this issue.

Thanks for the hint. That code won't compile: 'MLX90640_BadPixelsCorrection' was not declared in this scope. If you comment out the reference to MLX90640_BadPixelsCorrection it compiles and works OK. In the meantime I got a UIFlow version working more or less. Download from https://github.com/wsanders/WSThermalCore2Cam/blob/main/thermtest1.m5f. It's pretty crude but it autoranges OK if the temperature differences aren't too bad. Using ASCII art pixels because it's faster than drawing two rectangle (you STILL can't draw a borderless rectangle in UIFlow!)

Sorry - I deleted my previous post accidentally. It was full of errors anyway.

@wsanders
i think so, but i have trouble lighting LEDs with m5stick

Unfortunately not. Arduino work on a lower level than MicroPython and so runs directly on the hardware. Micropython is an interpreted language and so need to install a boot loader/ interpreter to run.
It can be inconvenient to have to install the MP boot loader when you want to switch back but its now a simple process.

@wsanders I got the program to run with lightsleep, but only in "Internet Mode" and not "App Mode". Battery life was no different with lightsleep than with wait, about 2 hr. (Wait is probably is just a call to lightsleep, just in a more organized manner.)

It appears that in 1.9.6 Wifi doesn't come back up after lightsleep, I seem to remember the above callback function worked in 1.9.5 even after lightsleep. Now it hangs.

Back to square one. I think the best way is to have two separate programs, one to sync the watch to NTP, and one to run the clock with the time from the RTC (so I can still use the watch outside of my home wifi network.)