You likely have 4MB Flash storage, but 520KB SRAM. Of that, I believe only like 380KB of SRAM is usable, but you can get them with 4 or 8MB PSRAM chips on board that can extend the memory
I have no idea if FluidNC would need changes to address that extra RAM or if the extension is seamless, or how much more it would cost, but I’m willing to do some research down that path to figure it out, especially if that makes it all less failure prone .
Just make sure if you share finances with anyone you tell them I did not suggest this, I suggested just the opposite. Now I see this is an opportunity for you to dig into that network a bit more.
Sounds like you have the skill set to prove that antenna is bad and the swap proved it, I do have external antenna versions I can swap you out with or I can just send you out another internal antenna version.
There are settings if you compile your own to change up the memory/versions. I thought we were on the largest one. Adding memory (cost) to me is not as beneficial as figuring out exactly how to test it and optimizing what is there. I am confident the fluid guys have squeezed out ever drop of memory they possibly could. That is not to say I don’t think things can get better, Jamie proved that already, there are things to be fixed still. I will say if we have a test that everyone can replicate we then have an easy issue to present to the fluid team and they will surely be more interested in fixing it, or one of the smart people around here can more easily fix it for them.
As an example why I am not getting memory issues, I should swap out with one of you that is to see if it is hardware or not. We are using $3 boards, not $9 boards. It is starting to sound like I need to start using $9 boards just to eliminate a variable. My first instinct is never to throw more money at a problem, I throw time at it…probably even worse but that means it is cheaper for everyone after me.
The idea with the jackpot using the esp32 is price. I am not trying to make the best board in the entire world that works in every circumstance. I am trying to make a board as inexpensively as possible that works for 99% of us. This all started with the Ramps board and some drv8825 drivers, and every single time we have switched boards it just keeps going up in price. Ramps were $7 and the drivers were like $1.5 each. So to me it feels very weird to be at a $140 board almost 9 years later for honestly not all that much gained. When an SKR has an issue I spend a lot of time trying to fix it. Back in the day if a ramps had an issue I just threw it away if I did not see the issue in two minutes or less.
Mike I think we should move this over here, Genuine vs Knock off esp32's - #35 by bitingmidge Not entirely the same but might actually be an issue.
I have read more about ESP32 wifi in the last few weeks than I care to admit.
Ryans advice for something that just works is correct:
But if that is a non-starter for you then this advice is also good.
And required disclaimer:
I can understand not giving home networking advice, because every network is different and changing things can break your network and at worst brick your router.
I did do some digging, so here’s what I found. The ESP32 does indeed like its own quiet little network and a strong signal.
I move my network access point as close to the CNC table as I could.
I made a guest network/vLAN just for the ESP32
I set the network for 2.4 GHz only, WPA2-PSK.
Here’s biggest bang for the buck and simplest test for this. If your router has a built in guest network turn it on. Get as close to your router or router assess point as you can. Connect the ESP32 by itself to the guest network.
Since making these changes I have a -47db average connection with no network drops or connection errors. Testing was done with 4 ESP32s, 2 mounted on Jackpots and 2 free standing, 1 of each had an external antennas. External antennas provided a consistent -3db gain.