CFW for KSGER/Quicko STM32 Soldering Stations

[AndyC]

What's up guys? Are you using the latest fw? ( 11-3-2021, "SW: git c612c32" )

Yes, on 3 different stations. JCD, KSGER 2.1 & 3.1.
I had time to scope the JCD during the week, and I found the cause of the oscillations (the opamp). With ADCDelay=35 it runs fine.
The KSGER 2.1 works fine with the latest build.
The KSGER 3.1 (with the black 3.0 board) needed some tweaking. Had to slow down the PID quite a bit (P25 I10 D0). Handle shake-wake is misbehaving, but this was a custom build, so I'll have to look into it again.
The temperature readings are quite low when doing calibrations, see attached photo for KSGER ADC numbers.
BTW, that's Swedish you see on the KSGER screen

[DavidAlfa]

Nice! Are you using EMA filtering? That should do the trick.
You are lucky that I added automatic widget sizing to the contents! (Unless it hits OLED boundary limits).
You had to try that when every little thing caused a hard fault crash... "English is ok"
With that words (I would too if using it in Spanish), better to start thinking on a 10" screen!

[AndyC]

Tested a few builds on the KSGER 3.0 today, and it works fine with the precompiled .bin and also the F101 & F103(!) build. No need for PID adjustments. Using EMA, factor 2. 
I must have messed up the clock settings on the previous build which made it rather nervous.
The one issue remains: shake wake. I see constant 'wave' flashing when the input is pulled high. Switching noise?

10" screen, yeah! Soon we'll have usb and wifi and watch TV on our soldering stations, only using the iron to heat our coffee
(I have been eyeing the T12X stations with the 128*128 screens though)

A question about the PWM and ADC functions. The documentation states:

Code: [Select]

|<----------------- PWM TIME -------------------->|
|_____________________________[ADC READ]|<-DELAY->| NEXT CYCLE
|<- POWER ->|_____________________________________|
Is this correct?
When troubleshooting the JCD, it led me to believe it was more like this:

Code: [Select]

|<----------------- PWM TIME -------------------->|
|___________|<-ADC_DELAY->|<-ADC READ->___________| NEXT CYCLE
|<- POWER ->|_____________________________________|


[DavidAlfa]

Yes, it's wrong, I will tell Mike to fix that (Edit: fixed), he very kindly made the user's guide 
The firmware works like in your correction.

Code: [Select]

|<----------------- PWM TIME -------------------->|
|___________|<-ADC_DELAY->|<-ADC READ->___________| NEXT CYCLE
|<- POWER ->|_____________________________________|


The one issue remains: shake wake. I see constant 'wave' flashing when the input is pulled high. Switching noise?

How are you pulling the input?
Some KSGER users had a similar issue in the stock fw, the station would wake up randomly, it got better with bigger capacitors.

If you connect the wake/shake input either to GND or 3.3V, it should only blink briefly.

If I touch the input, the icon it will stay on all time, due the 50/60HZ mains noise I'm injecting with my body capacitance.
So, it's detecting like the handle was moving at 100/120Hz (it detects rising and falling edges).
The shake icon has a 50mS timer before going off (I don't remember the exact value right now).
What happens is that it gets enabled again before turning off. Not a bug, just noise.

[AndyC]

How are you pulling the input?
Some KSGER users had a similar issue in the stock fw, the station would wake up randomly, it got better with bigger capacitors.
If you connect the wake/shake input either to GND or 3.3V, it should only blink briefly.

The input is pulled high by the onboard resistor and pulled low by the handle tilt switch like normal.
I changed the 3.3V output capacitor to a bigger one, but no improvement. Then I just inverted the switch in the handle
Which capacitors 'need' upgrading on the 3.0 board?

[DavidAlfa]

I have no idea, I don't have that board...I read that in a KSGER v2.1 thread.

How are you testing it when you say you are "pulling high"?

My handle had the switch reversed, it didn't detect anything while soldering, only when putting it upside down.

[AndyC]

How are you testing it when you say you are "pulling high"?
My handle had the switch reversed, it didn't detect anything while soldering, only when putting it upside down.

Maybe I'm using the wrong words, not a native speaker...
When the handle was in the stand, the tilt switch was open -> shake input is at 3.3V + lots of noise
When holding the handle with the tip pointed up, the tilt switch was closed -> shake input is ~0.3V + very little noise
So, open switch gave flickering shake indicator and the station never went to sleep. Inverting the tilt switch fixed this.

[DavidAlfa]

Neither am I!
I was not sure if you were pulling high the handle itself or the pin by connecting it to vcc.
But even when leaving the handle in the stand caused the noise? With no movement?
Can you confirm the actual resistor value? I think it was 4k7, more than enough...
And there was another 4k7 resistor in series to the stm32 pin (as a protection method).
Shouldn't cause any trouble.

Given you already know how to build the de, open CubeMX (.ioc file) and enable pullup resistor for wake input. And check if it gets better.

[DavidAlfa]

My JBChinese tips are about to come after 2 long months... I hope I don't get the same crap like the last time!

I used the station a bit these days, and noticed that the detection was a mess, it jumped like crazy.
So nobody realised that iron detection was broken since long time ago?
I don't know why, but I was disabling the PWM completely when not detecting the cartridge. Fixed by leaving the PWM slighly on.
I remember doing exactly this few months ago to fix the detection! I might have had another brain stroke... 
I mostly spent the day in testing and debugging to enhace the filtering delay/response.

The iron detection works spot-on now (at least for me), no resistor needed in the amp input, even with very low detection delays (tried 100mS and perfect).

This is the last changelog:

Disable error detection during the first second, to let the readings get stable.
This is safe, as during this time only the boot screen is shown.
The PWM is almost disabled (10uS pulses to enable iron detection).

Disable PWM if supply voltage is lower than 12V to avoid the possibility of the mosfet going into ohmic region and overheating.

Enabled Shake input pullup also in KSGER v3.0 profile.

Setpoint widget inverted for better reading. Increased delay to 1 second before returning.

Enable Tip selection in sleep / No iron screens.
Changed Tip selection widget to smaller letters, also added title "Tip selection".

Deleted DEMA filtering since it was pretty much useless in this application.

Enhaced EMA filtering response by comparing to last average.
Typical reading noise is around 10-40, so that's what we want to filter.
Reset filter if delta>200, add half the difference if > 100.
This allows a much faster response, yet the filtering still works perfectly fine.


Fix iron detection by always leaving PWM to min value (10uS pulses).
(Critical failures will always disable PWM completely).

Clear the plot data when entering sleep/error screens, so it appears clean when it returns to run mode.
Don't collect plot data if not in run mode.

Added more errors messages.
If only iron detection is triggered, the classic "NO IRON" is shown.

If more (not critical) errors are active, a code-based error message is shown.
- No iron (No cartridge inserted)
- NTC reading too high, > 70ºC (Shorted, open-circuit or serious overheating)
- NTC reading too low, < -60ºC (Shorted or open-circuit)
- Supply voltage too low (Less than 12V).
- Internal failure. It means the PWM has been disabled for safety reasons.
   Actually the only reason why it could happen is if the profile is undefined after passing  the ID + checksum checks, very unlikely unless there's some kind of memory corruption.
   (Maybe I'll move this one to critical errors, not a good thing if it happens).

Error code is a 8 bit bitfield (Displayed in HEX):
x x x x x x x x
7 6 5 4 3 2 1 0
- Bit 7: Global flag (Any error is active)
- Bit 6: Unused
- Bit 5: Unused
- Bit 4: Internal failure
- Bit 3: Supply voltage low
- Bit 2: NTC reading low
- Bit 1: NTC reading high
- Bit 0: No iron detected

Multiple errors can be shown, however only 3 will fit in the screen.
The displayed error code can be used for determining the cause.

For example, a KSGER with no handle connected will trigger No iron+ NTC low,  showing "ERROR 85, No iron detected, NTC read low".

Facing an extreme case: Internal failure + voltage low + NTC low + No iron.
All these errors won't fit in the screen, but the code will be 9D.

And, of course, I updated the binaries.
Since everything seems OK, I made the first release!
Say Hello to v1.0 

[AndyC]

I used the station a bit these days, and noticed that the detection was a mess, it jumped like crazy.
So nobody realised that iron detection was broken since long time ago?

Aha! I thought that my KSGER 3.1 was misbehaving in more than one way. I'll have to reflash it and test.
Did not see any issues on the 2.1 version though.

This is the last changelog:

Wow, great work!
I really like the inverted setpoint display.
My OCD will miss the F103 builds since I have only those MCU's, and I really need the extra MHz so I can solder faster 
This FW and the Bourns rotary encoder is the upgrade of the year!

Edit: The KSGER 3.1 is running better than ever on the new FW.

[DavidAlfa]

Haha... since I optimized the code to not waste all the cpu power to redraw the screen contents every time, the speed is pretty fast even at 16MHz!
The slowest by far will be the v2.1, as the display must be updated by software. And i2c is slowwwww, even being overclocked.
Before, it was a big mess to make 3 builds (F101,F102,F103) for every KSGER board! Since it works the same... it saved a ton of time!
If it could be automated with a make script, then it would be much easier. Since I don't have that knowledge, I'm tied to do it manually...

Did you modify the code to set a so low setpoint, right? Because it's 180ºC minimum by default.
Under ~100ºC the TC reading is pretty much dead. If it reads something, can't be trusted. See this:
https://hackaday.io/project/94905-hakko-revenge/log/144548-hakko-t12-thermocouple-is-not-type-k

The setpoint should be centered, but that widget is stretched so much to fit 5 characters and fill the entire screen, that it's actually a little outside the display.
If the code detects that, it disables the alignment calculation, and draws at x=0.
Since the widget is designed to have always 5 digits (xxxºC), that would never be an issue.
But you had to touch, didn't you? 


Did the shake input get better?
Check that the errors are correctly detected when removing the handle connector (KSGER should complain about NTC low and iron detection).

[AndyC]

Did you modify the code to set a so low setpoint, right? Because it's 180ºC minimum by default.

Yes, of course I had to touch it
The original 3.1 FW has that setting, and while I don't trust it, I find it useful when testing handle assemblies. I can get a reading and insert and remove the tips without burning my hands

Did the shake input get better?
Check that the errors are correctly shown when you remove the handle connector.

The shake input problems were bad cap and soldering.
The error display seems to work as it should on my KSGER's and on the JCD.

[DavidAlfa]

I see. I do it with some pliers, haha. If I have to wait until the tip cools down... would need to shave/eat/sleep few times!
Anyways, the PID doesn't work below 100ºC. It just sets PWM off.
You can modify iron.c line 151 from 99 to 100, and set min temp to 100 in the widget.
That way the setpoint widget still shows nice, and setting it to 100ºC disables the PWM.

You replaced some caps? Can you show which ones?
It would help others with your same problem.

Aaaand a litle bug, I forget to ensure that PWM was never 0 after PID calculation, that would cause detection issues if the tip was removed when the PID was off.
Fixed and updated everything.

[AndyC]

You replaced some caps? Can you show which ones?
It would help others with your same problem.

I replaced C9 on the black 3.0 board and redid some solder joints with leaded solder. Overall poor solder joints on this board.

[mastershake]

they are sending out a new psu with the ksger now. v 2.05 one major thing is the one trace doesnt run under the heatsink anymore otherwise its pretty close to the same thing. the one heatsink in the one that just got here is a mess though im going to remove it and straighten it out and re mount it. the thing was supppper warped and leaning way off to one side. i was gently able to straighten it upright wise but the thing is mangled warped wise. also they did not use any thermal paste or material at all on either one in this version.  i thought i mentioned the sleep issue a while back. i did have that but i dont remember which build it was now. i may have forgot to mention it i cant remember.

[DavidAlfa]

As you see the sleep issue is not fw fault (Also happens in original fw), it occurs in few boards, but due a hw proble.

Btw I'd like to add support for the older ksger v2.0 board, what the original PTDreamer fw used.
However nobody asked ever, strange.

[DavidAlfa]

Finally, the C245 tips arrived.

Edit: Same crap with fancy steel cover.
The tips loosen when hot, they are badly pressed.
I'm done with cheap JBC copies.

They're pretty good, but all of then are too small for soldering heavy stuff, only suitable for smd parts.
The overall finish is very good.
Couldn't calibrate them, I lost the DMM temp sensor weeks ago

[cosmin1]

My chinese 245 behaves well. But the 210 tip is some crappy material. Oxidizes quickly.

[DavidAlfa]

Define "well".
Have you used genuine JBC before to compare experiences?
I worked with one for 6 years, used almost every day. I changed one tip only once, because it was worn out.
And I can say the heat transfer was amazing even with the small tip.
This is absolute crap... Will buy some genuines. I have a friend where I bought all the parts for the repairs, he still makes me the discount...
I can get them for 22 euro. With 2 tips I'm more than OK. A "knife" tip for the heavy stuff and a 1mm round type for the rest.

Anyways I'm barely using it lately, so I'll leave it for the future.

Talking about soldering. Someone send me a private message in other forum asking for help.
Dammit, he used a gas soldering iron for the fine electronics, I felt in my soul all the pain the board had suffer, all the lifted/burned pads ....
Told him to inmediately stop using that thing coming from the hell and get at least a cheap temp controlled iron.
He's doing the soldering training with the $$$ board, skipping the practice with some cheap damaged one 

[cosmin1]

Someone send me a private message in other forum asking for help.
Dammit, he used a gas soldering iron for the fine electronics, I felt in my soul all the pain the board had suffer, all the lifted/burned pads ....

 

I used these tips several times and they did the job. I have no genuine JBC tips to compare. Can only tell is ok considering my experience with other types of tips.
It's very possible to find big quality differences between different sellers.

[mastershake]

yeah the copies are super hit an miss. i have a few okay ones and mostly horrible ones. i only bother with genuine tips now. it sucks they are much more pricey but compared to the clones worth the cost difference imo unlike the t12 ones where the copies are usually pretty decent

[AndyC]

So, @DavidAlfa, after having used v1.0 for a while on different stations, here are some thoughts/questions:

• When the buzzer is set to ON, there are two beeps per edge detection of WAKE input. Is there a way to disable this? (Sounds like a radiation meter!)
• The new error handling is great!
• For a newly flashed KSGER, I sometimes get failed calibration (the temps are way off). How is successful / failed calibration determined?
• The original Quicko firmware is indeed terrible! I managed to save another station from the closed-source-crap. I saw it had support for 1321/1322 heaters, but who will use those when there's a T12 option?
• The standard calibration values for the Quicko almost made my tips glow, I saw 580°C at the 450°C calibration point. See attached photo.
• The temperature was not very stable on the Quicko, increasing EMA filter factor 2 -> 3 helped a bit though.
• Found an SSD1309 2.42" display for the KSGER 3.1, worked fine with offset = 0.

[DavidAlfa]

there are two beeps per edge detection of WAKE input.
Currently I have no buzzer. Are you sure it's that? It will beep when waking but also when reaching the temp.
So if you go to sleep and wake up quickly, both things will happen very closely in time.

How is successful / failed calibration determined?
Basically, it checks that T450>T350>T250. And than none of the new calibrated values exceed the ADC range.
So for example if your T450 was ADC 3900 and you measured 300ºC.
The calibration (Not exactly done this way) would be something like: 3900/300*450 = 5850.
The max ADC value is 4095, so the calibration will fail.
Otherwise it could store a number higher than the ADC, that would push all the power into it, while never reaching the setpoint, making a T12 candle.
This was another security check added some weeks ago.
Also, if you are getting unstable readings, the calibration will likely fail.


(the temps are way off)
That's the problem with so many tips and boards.
First, I set it a lot lower to avoid exactly that, then people complained that it was too cold on default calibration.
Then I put it normal, now they are getting candles 
As you see, almost nobody post their calibration results.
That way I can't make a proper table for each board and set a default calibration.

increasing EMA filter factor 2 -> 3 helped a bit though.
If it's unstable, increase the delay, not the EMA factor. The ema is designed to filter small spikes.
Also a bad power supply/noise will cause a lot of trouble.
Are the tips new? You know they do that for maybe 10 minutes, right?
Don't calibrate a new tip before lthe burn-in time!

The original Quicko firmware is indeed terrible!
Terrrible is not enough. There are no words in the human language to describe it.

The standard calibration values for the Quicko almost made my tips glow,
The Quicko board has adjustable gain, there's a pontentiometer in the board behind the display that can change the gain between 160 and 300.
You can guess what happens if the gain is too low.
Either set the gain higher or change the default T12 values than can be found in Core/Inc/settings.h
Lower the values by a 30% or so, and try again.

Found an SSD1309 2.42"
Yeah they're basically the same thing with bigger pixels.
Not willing to pay 15$ for them just to see the temp 

[AndyC]

there are two beeps per edge detection of WAKE input.
Currently I have no buzzer. Are you sure it's that? It will beep when waking but also when reaching the temp.
So if you go to sleep and wake up quickly, both things will happen very closely in time.

Well, it behaves like this:
https://youtu.be/Vbdyr-n5YeY
I did not see this behavior in earlier builds.

[mastershake]

So, @DavidAlfa, after having used v1.0 for a while on different stations, here are some thoughts/questions:

• When the buzzer is set to ON, there are two beeps per edge detection of WAKE input. Is there a way to disable this? (Sounds like a radiation meter!)
• The new error handling is great!
• For a newly flashed KSGER, I sometimes get failed calibration (the temps are way off). How is successful / failed calibration determined?
• The original Quicko firmware is indeed terrible! I managed to save another station from the closed-source-crap. I saw it had support for 1321/1322 heaters, but who will use those when there's a T12 option?
• The standard calibration values for the Quicko almost made my tips glow, I saw 580°C at the 450°C calibration point. See attached photo.
• The temperature was not very stable on the Quicko, increasing EMA filter factor 2 -> 3 helped a bit though.
• Found an SSD1309 2.42" display for the KSGER 3.1, worked fine with offset = 0.

wow sweet can you tell me where to get that display? id love to get a few of them for mounting these into a combo type unit?