Breaking the encryption on the Lerdge series mainboards so I can try porting Marlin 2.0 to it.
Patched bootloader that does not reset security bits on startup. Must be loaded directly with an ST-link, will not load correctly through the lerdge X boot updater
octet-stream - 64.00 kB - 01/10/2019 at 18:33
This appears to be the LCD controller - most of the commands match up to what we see in the code.
Adobe Portable Document Format - 3.40 MB - 01/09/2019 at 18:28
script to call and then flash. Be sure to adjust paths as appropriate.
x-sh - 288.00 bytes - 01/03/2019 at 16:16
openocd script for resetting.
cfg - 312.00 bytes - 01/03/2019 at 16:15
This is the base bootloader from a lerdge-K system.
macbinary - 64.00 kB - 12/19/2018 at 19:41
I've got a minimal marlin build up and running on both X and K, with the boards wired into a printer to test with. Doing so has revealed a few problems I would not have seen otherwise:
1. My LCD isn't filling the background correctly. The Logo during boot appears clipped, and the boarders are not filled in black.
2. Hotend temperatures read from PC1 are not working correctly. i can swap the pin definition to PC0, swap the connector, and the hotend behaves correct. This doesn't affect the MP MIni, so it's not clear to me what might be happening.
The solution for #2 is to write some diagnostic code to narrow down if pure, plain "analog read" works for the pins. Once basic pins are verified, I'll begin PRs to add this to Marlin main.
Just a note that adding the I2C EEPROM definition to the Lerdge pins and enabling #EEPROM in the configuration was all it took to get settings to save correctly. Next up, I want to take a swing at emulating the 3 button interface on the touch screen, since unlike Mickey, I don't currently have an encoder (it's on a boat from China).
We're getting there. This weekend I'll be wiring the Lerdge-X into a printer.
With the help from Jason, Dzenik, and jmz52's help I was able to adapt my custom version of Marlin4MPMD to work with the LerdgeX hardware. The majority of the work was spent adding in the STM32F4 HAL library instead of the STM32F0 HAL library, and then running through inconsistencies with the Lerdge hardware. As a proof of concept I now have full functionality including motion, auto bed leveling, SD support, a janky looking version of the Marlin GUI, and Octoprint support.
Because it is derived from Marlin4ST originally, the fork is based off of an older Marlin 1.x base and is only compilable through openstm32, which is a lot more complicated to generate and install binaries through. Given the current development state of STM32 arduino support, it made more sense for me to continue bare metal development to make this proof of concept. The right way to move forward however is to provide HAL support through Marlin 2.0, which is what future efforts will entail.
One of the key difficulties developing for Lerdge was centered around the issue of the STM security bits. By default, both the bootloader and the application binaries include a piece of code that will enable the security bits of the processor. This is intended as a copy-protection feature that will erase the entire flash memory if the device is reprogrammed over SWD. However, by side-loading our own custom binary that reads out the entire flash image we can extract the raw code for the bootloader and application binaries and modify them to disable this check. I've added a patched version of the 1.0.2 bootloader. This is import because it allows direct debugging and loading of application code on the processor without triggering a full erase.
The LCD code was another challenge in that we had an unknown panel with an unknown pinout connected via ribbon cable. Through some sleuthing we were able to determine that the communication was done over a 16-bit parallel data interface to the panel and the FSMC peripheral was used to send/receive commands. After deducing that standard 0x2A(row select), 0x2A(column select), 0x2C(RAM write) sequence was able to change the colors on the panel, I deduced it was similar model to the ILI3941. However, reading the device ID provides 0x7796, which indicates it is likely a generic ST7796 panel instead, which coincidentally shares the same command set as the ILI device. With this and some code provided by jmz52 to scale up the Marlin output to a larger resolution, I was able to provide the Marlin GUI to cover most of the Lerdge screen.
This was intended as a proof of concept more so something to be distributed, IMO the best course of action is to wait until full Marlin 2.0 support is provided.
With some help from Dzenik, I managed to get a working LCD on Marlin 2.x, though some of it is hacked together. Initialization isn't working the way I'd expect, so I'm manually calling the LCD init code.
After that, the LCD writes just work, though I don't currently have pixel tripling working.
Now, Mickey had already gotten the LCD working on his Marlin build - his is based off of Marlin4ST, mine is from 2.x, but we're both working toward the same goal. I'll get the scaling working soon, then hammer out the issues with LCD init so it works with U8G the way we'd expect.
Sometimes, it's less about what you do and more about what other people are doing around you. In this case, someone was already working on adding support for SDIO to Marlin.
This gave a nice way to abstract away the details of block reading, and THAT gave me a simple way to plug in code. I started using the STM32SD library (and still am using code derived from it) but hit a problem where no matter what, the block reads failed.
After a great deal of investigation, I found that the block read calls were passing in the block SIZE as the block number, meaning that in many cases, huge chunks of memory were getting trampled. A few fixes later, we have the following output from the ESP8266:
M20 Begin file list echo:Cannot open subdir JOURNA~1.COR echo:Cannot open subdir JOURNA~1.LIV echo:Cannot open subdir JOURNA~1.COR echo:Cannot open subdir JOURNA~1.SCA echo:Cannot open subdir JOURNA~1.LIV echo:Cannot open subdir JOURNA~1.COR echo:Cannot open subdir JOURNA~3.LIV TEST.GCO 304884 TES~1.GCO 304884 End file list ok
This code needs to be vastly cleaned up - I literally imported the bsp_sd h and c files directly, but they don't use DMA, there's the afore-mentioned "passing the wrong parameters" bug, and in general, a lot to do before I can proclaim it works. Also troubling is that if I re-init the SD card, it crashes, but that's a problem for another day.
Getting closer and closer to that first test print.
Also, another member has successfully initialized and written to the LCD. I haven't done this yet, but they have. So, bit by bit, it's coming together.
I'm uploading the cfg and scripts I use to reset the lerdge board. THe paths on these are probably all wrong, but it should make your life a little easier if you're hacking.
First off, some clarifications from discoveries since the last log:
1. The USB B /CH340 and ESP8266 share the same USART, Serial1. This is why plugging in USB kills the ESP - they didn't want conflicts.
2. I've had no success getting USB working through the female connector, despite having a spare A->A cable. The serial that works is the output of the CH340, and it should be zero surprise it works since it's basically a dedicated FTDI clone.
Secondly...never accidentally leave test code in.
I removed a "blink forever" line from Marlin's main, re-enabled watchdog, and boom, I can issue commands to Marlin. After a few more adjustments it will be time to plug it into the printer.
Did I mention you can do WIFI printing? Yeah. You connect to Marlin by telnetting to Port 23 on the ESP and using the Serial bridge. Soon Windows, Com0Com will give you a virtual com port for Marlin. Or ESP3D could be run.
What does not work:
SD. Marlin uses SPI. This board uses SDIO. I need to test if SDIO via Arduino libraries works.
LCD. Nope, not even a little bit, yet. I'll need to figure out FSMC initialization before it will work.
All changes are pushed to my github, so you can enjoy if broken builds of Marlin are your thing.
start echo: External Reset Marlin bugfix-2.0.x echo: Last Updated: 2018-01-20 | Author: (xC0000005, Lerdge-X Config) echo:Compiled: Dec 29 2018 echo: Free Memory: 51035 PlannerBufferBytes: 1344 echo:Hardcoded Default Settings Loaded echo: G21 ; Units in mm (mm) echo:Filament settings: Disabled echo: M200 D3.00 echo: M200 D0 echo:Steps per unit: echo: M92 X80.00 Y80.00 Z4000.00 E500.00 echo:Maximum feedrates (units/s): echo: M203 X300.00 Y300.00 Z5.00 E25.00 echo:Maximum Acceleration (units/s2): echo: M201 X3000.00 Y3000.00 Z100.00 E10000.00 echo:Acceleration (units/s2): P<print_accel> R<retract_accel> T<travel_accel> echo: M204 P3000.00 R3000.00 T3000.00 echo:Advanced: B<min_segment_time_us> S<min_feedrate> T<min_travel_feedrate> X<max_x_jerk> Y<max_y_jerk> Z<max_z_jerk> E<max_e_jerk> echo: M205 B20000.00 S0.00 T0.00 X10.00 Y10.00 Z0.30 E5.00 echo:Home offset: echo: M206 X0.00 Y0.00 Z0.00 echo:PID settings: echo: M301 P22.20 I1.08 D114.00 echo:SD init fail G1 X ok M503 echo: G21 ; Units in mm (mm) echo:Filament settings: Disabled echo: M200 D3.00 echo: M200 D0 echo:Steps per unit: echo: M92 X80.00 Y80.00 Z4000.00 E500.00
After my last post, I went back and studied how code that actually can initialize the board correctly worked. With a little borrowed code from micropython sample, I was able to initialze the board successfully, and communicate outbound at much faster rates. Inbound data works in a test sketch, but not in Marlin yet. Now I can use USB instead of having to wire up to the serial1 via a FTDI clone.
Now I have an ESP8266 hooked to Serial1 and output via USB, so in general, this is very good.
Notes on the ESP module - For some reason, if I plug into USB, it cuts power to the ESP8266 module. These two do not share the same pins, so there's no logical reason I can find for this. Marlin supports two serial ports in 2.x, so it would be nice to enable both. Perhaps there's a cutoff circuit on the expansion module that can be disabled - I'm not sure yet.
I'm just happy to be able to send and receive data via USB.
I noted in the previous log that the failure to run at faster baud was probably clock related. Today I built a binary that dumps the oscillator and clock configs to see how the bootloader and reboot works, and here's what I found - "Letting the bootloader configure the clock" is bad.
Here's the clock config from both:
OscillatorConfig: OscillatorType:15 (all of them) HSEState:0 LSEState:0 HSIState:1 HSICalibrationValue:16 LSIState:0 PLLConfig: PLL.PLLState:1 PLL.PLLSource:0 PLL.PLLM:16 PLL.PLLN:192 PLL.PLLP:2 PLL.PLLQ:4 Obtaining Clock Config: Obtained Clock Config ClockType:15 SYSCLKSource:0 AHBCLKDivider:0 APB1CLKDivider:0 APB2CLKDivider:0 Flash Latency:5
So, that's what is wrong. If I attempt to initialize the oscillators, the board hangs, so I need to debug to this and figure it out.
If you own the lerdge USB/Wifi printing module, you've probably looked at that 8 pin header and thought "It would be nice to do serial printing." And it probably would. However, in the lerdge store's description, they basically say other ESP8266-01 modules don't work.
It's sort of a lie, sort of not. Lerdge does not release their firmware (though anyone with a module could dump it), and they don't document how the connection works. But that doesn't mean much to me, so I ordered a $2 ESP8266-01 from China, and when it arrived, I flashed ESP-LINk on it and plugged it into the lerdge. After adjusting for my hideously slow baud rate (again, probably a clock problem)
I got the following output:
So...can we enable standard Wifi Serial printing? Yes.
Is it easy/Working? Not yet. But it's not far off.
There's no custom code other than my marlin build at work here, so it's almost certain we can use the ESP8266 wifi printing code directly on the board once we get it running. I'll be checking the oscillator config soon to see why we're having problems, and I'm hoping that's directly related to why USB and SDIO don't work yet.