[Grbl_Esp32 Issue#475] Send hello message on connect instead of init()

#1 – MitchBradley 于 2020-07-12

I am a developer on both the CNCjs and Grbl_ESP32 projects, so, given enough time, I should be able to give a definitive answer and solution to this one. I will dig deeper.

#2 – atlaste 于 2020-07-13

> I am a developer on both the CNCjs and Grbl_ESP32 projects, so, given enough time, I should be able to give a definitive answer and solution to this one. I will dig deeper.

I can reproduce this every time and have been a professional software developer for many years, so if you need any more details (with or without code locations), just give me a shout.

#3 – atlaste 于 2020-07-13

> I am a developer on both the CNCjs and Grbl_ESP32 projects, so, given enough time, I should be able to give a definitive answer and solution to this one. I will dig deeper.

I can reproduce this every time and have been a professional software developer for many years, so if you need any more details (with or without code locations), just give me a shout.

#4 – MitchBradley 于 2020-07-15

Lacking the following information, I cannot duplicate your setup:
* What computer are you running CNCjs on
* How did you install CNCjs? (electron app, docker, direct npm install, etc)
* On what flavor of ESP32 are you running Grbl_ESP32?
* What kind of USB serial chip does it have – CP2102, CH340, etc

On my setup, I sometimes see inconsistent startup behavior but most of the time it works. I have tracked the problem down, at least in part, to the following interesting hardware behavior. First, it is important to note that nearly all ESP modules connect the SoC’s EN (active high to enable, low to reset) line and IO0 “download mode” line to the serial chip’s DTR and RTS lines, so that DTR high, RTS low will reset the CPU.

When a program is connected to the serial port, the EN signal remains high as long as the connection persists. But when there is no connected program, I am seeing periodic low-going pulses on EN. The pattern is one pulse that lasts for 3 us, then a 750 ms delay and another pulse. The two-pulse pattern repeats every 5 seconds until a connection is opened. Furthermore, there is also a pulse when a connection is closed, but not when a connection is opened.

This suggests that something in the USB driver or stack is periodically probing with the result that the ESP32 chip is getting doubly-reset every 5 seconds – at least on this Windows 10 system with a CP2102 USB serial chip.

I tried bypassing the 4-port USB hub that was between the computer and the CP2102; no change.

#5 – atlaste 于 2020-07-15

> What computer are you running CNCjs on

Windows 10 Pro x64. The pc I’m running it on is a Dell XPS 15, 9550.

> How did you install CNCjs? (electron app, docker, direct npm install, etc)

Windows installer (https://github.com/cncjs/cncjs/releases), desktop app, x64, 1.9.22.

> On what flavor of ESP32 are you running Grbl_ESP32?

ESP32 DevkitC rev 4 WROOM-32
This one to be exact: AZDelivery ESP-32 Dev Kit C V4 WLAN WiFi Development Board .
Data sheet: https://cdn.shopify.com/s/files/1/1509/1638/files/ESP–32NodeMCUDevelopmentboardDatenblatta3bf98d8-6a53-4d26-8f1c-c61b1c82af39.pdf?76837 .

> What kind of USB serial chip does it have – CP2102, CH340, etc

CP2102 according to the data sheet.

> First, it is important to note that nearly all ESP modules connect the SoC’s EN (active high to enable, low to reset) line and IO0 “download mode” line to the serial chip’s DTR and RTS lines, so that DTR high, RTS low will reset the CPU.

I don’t see a reset happening when reconnecting. I believe this is exactly the issue. When i reconnect, I can do GRBL commands like $$ to get the current settings.

> I tried bypassing the 4-port USB hub that was between the computer and the CP2102; no change.

I’m not using a hub, it’s directly connected to the PC.

> [EN signal] / double reset

Hmm that sounds feasible. I don’t have access to a scope, but can put a multimeter between EN and GND to see if it goes low. If it remains high, there’s a different problem. Will have to get back to you on that.

#6 – MitchBradley 于 2020-07-15

EN only goes low for 3 microseconds per pulse and the pulses are widely spaced relative to their widths, so it is not likely that you will see them clearly with a multimeter. With my multimeter, I say the voltage dip ever so slightly from time to time, from 3.215 to 3.212 or something of that order. What you will see, if anything, is highly dependent on your meter.