AxxSolder - JBC soldering controller

[c0d3z3r0]

As a guy from the semi industry, I often see engineers using schottky diodes such as BAT54 for protection. This is a bad design practice. They really leak current that varies with temperature. This leakage also differs from manufacturer to manufacturer. I've seen real pre-production designs failing specs because of it (analog front-ends and opamps).

What do you suggest as a replacement?

Usually, one would choose diodes with lower forward voltage than the internal protection diodes, so they start conducting before the internal ones. Unfortunately, TI doesn't tell us what diodes are used in the OPA2333, but only say there are diodes :/

The main question for me is: how much reverse current is too much in this specific case? I still don't believe the diodes get that hot on that board, at least not >30 - 40 °C and e.g. for Vishay BAT54s leak current is ~1-2 uA at 50°C.

@Warhawk can you share more details from your experience and a suitable alternative?

[drksy]

Well, voltages is all about references, right? The main question is: does the OpAmp measure a) voltages at +/- against GND or b) voltage at + with reference to -  and then compare them?

I have another idea. What if we don't isolate the 5V supply and don't make the controller board earth-referenced, the heater ground would be the same as controller ground, and to measure tip temperature, we use an opamp as a differential amplifier to measure the voltage difference between red and green. What are the downsides?

[miket6000]

My comment about the diodes getting hot is a somewhat generic one about the issues with Schottky diodes in general. That said you have a power switching controller in a plastic box with no vents, I wouldn't be surprised if the ambient was above 40°C.

The biggest issue in this case is that the leakage current will show up as a voltage on the thermocouple input. I'm not sure what thermocouples are in the handles in the JBC irons, but it's likely in the order of 40uV/°C and the input impedance is dominated by R21 and R22 which are 470k, so a small imbalance in leakage current could easily be enough of an issue to make the iron unusable.

[c0d3z3r0]

My comment about the diodes getting hot is a somewhat generic one about the issues with Schottky diodes in general. That said you have a power switching controller in a plastic box with no vents, I wouldn't be surprised if the ambient was above 40°C.

The biggest issue in this case is that the leakage current will show up as a voltage on the thermocouple input. I'm not sure what thermocouples are in the handles in the JBC irons, but it's likely in the order of 40uV/°C and the input impedance is dominated by R21 and R22 which are 470k, so a small imbalance in leakage current could easily be enough of an issue to make the iron unusable.

Thank you! JBC uses a custom theromcouple with ~40 uV/°C: https://www.eevblog.com/forum/beginners/jbc-t245c245-thermocouple-thermal-coefficient/

[c0d3z3r0]

Well, voltages is all about references, right? The main question is: does the OpAmp measure a) voltages at +/- against GND or b) voltage at + with reference to -  and then compare them?

I have another idea. What if we don't isolate the 5V supply and don't make the controller board earth-referenced, the heater ground would be the same as controller ground, and to measure tip temperature, we use an opamp as a differential amplifier to measure the voltage difference between red and green. What are the downsides?

That's basically what Aixun did, just with a differential amplifier without GND-reference and with SELV instead of PELV. Indeed, that might solve the problems of the T3A already.

Hmm, in either case, isolated or this, the stand sense won't work anymore. Aixun (AxxSolder as well?) uses a simple pulled-up GPIO to check if it's pulled down by the grounded tip's outer shell. This won't work with SELV as there is no connection to GND anymore. Since the outer shell has an internal connection to +24V, this problem can be solved by inverting the logic and using a voltage divider. Maybe that debounce circuit needs some adaption as well.



Ohhh, one more thing... having a undgrounded/floating setup, what will happen when connecting both mains/SELV and USB to your (grounded) computer? Dave's oscilloscope scenario again
AIUI, the SELV will be floating at a higher voltage with respect to earth (Y-Cap leakage). Connecting USB 1) creates a PELV szenario and 2) might kill the USB port on the computer.

[drksy]

Ohhh, one more thing... having a undgrounded/floating setup, what will happen when connecting both mains/SELV and USB to your (grounded) computer? Dave's oscilloscope scenario again
AIUI, the SELV will be floating at a higher voltage with respect to earth (Y-Cap leakage). Connecting USB 1) creates a PELV szenario and 2) might kill the USB port on the computer.

Yeah, this is a big problem isn't it? It's a compelling reason to make the controller board referenced to earth. If we have a floating supply feeding the controller and then connect a grounded USB cable, when the mosfet is on, the 24V goes from supply positive to mosfet to red to green (via tc) to earth (assuming tip is connected to earth), to the earth connection on your PC power supply, though the earth connection in the USB port, to the GND of the (no longer) floating controller, back to negative of the supply, essentially shorting it once again!

In the CD-2BC reverse engineered schematic, they solved this problem by isolating the UART bridge connected to USB. Perhaps you could use one of those popular ADUM3160's to isolate the USB ground (and hence earth) from the pcb ground. But after the USB ground is isolated, the "safe" 5V would no longer be available (unless going through a dc-dc isolator).

And yes, detecting whether the iron is the the stand is more convenient if the controller is earth referenced. So maybe it's just best practice as Marco Reps said to make the controller referenced to earth. It solves all of these weird problems at once.

[c0d3z3r0]

Ohhh, one more thing... having a undgrounded/floating setup, what will happen when connecting both mains/SELV and USB to your (grounded) computer? Dave's oscilloscope scenario again
AIUI, the SELV will be floating at a higher voltage with respect to earth (Y-Cap leakage). Connecting USB 1) creates a PELV szenario and 2) might kill the USB port on the computer.

Yeah, this is a big problem isn't it? It's a compelling reason to make the controller board referenced to earth. If we have a floating supply feeding the controller and then connect a grounded USB cable, when the mosfet is on, the 24V goes from supply positive to mosfet to red to green (via tc) to earth (assuming tip is connected to earth), to the earth connection on your PC power supply, though the earth connection in the USB port, to the GND of the (no longer) floating controller, back to negative of the supply, essentially shorting it once again!

In the CD-2BC reverse engineered schematic, they solved this problem by isolating the UART bridge connected to USB. Perhaps you could use one of those popular ADUM3160's to isolate the USB ground (and hence earth) from the pcb ground. But after the USB ground is isolated, the "safe" 5V would no longer be available (unless going through a dc-dc isolator).

Yes, the ADUM* would work perfectly here but the CC pins need isolation, too. Probably a simple digital isolator works here (there is some application example from AD, have checked roughly only, yet).

And yes, detecting whether the iron is the the stand is more convenient if the controller is earth referenced. So maybe it's just best practice as Marco Reps said to make the controller referenced to earth. It solves all of these weird problems at once.

No, it creates all the weird problems Aixun has, because this is exactly what they did. It would be just a copy of the T3A then

The stand detect should be fine with that circuit I posted. Needs testing ofc

[c0d3z3r0]

BAV199 could be a good candidate for OpAmp protection

[Warhawk]

@Warhawk can you share more details from your experience and a suitable alternative?
...
TI doesn't tell us what diodes are used in the OPA2333
...

(writing the post the second time as i accidentally closed the tab )

I am sorry it took me longer to reply.
Using Schottky diodes is relevant, they just need to be carefully selected. I checked some of my designs and  the 1PS79SB70 device from nexperia is frequently used. Check the DS. Alternatively, I used BAS40 (still not perfect). I am with the team red, therefore esd protections such as TPD1E series find often place on my boards. Note the maximum 10nA leakage. They are intended for digital signals but work for ADC protections too.

[Thinking aloud]
I also wonder, if protections such as D12, D1, D3 are really necessary. I would maybe add only 22pF ESD caps on the input and call it good to go. The current to the input pin of the ADC is limited by series resistors. This reminds me that it is important to find out if S/H time is long enough.

Reverse polarity protection is maybe something I would like to see. D9 could be only unidirectional. Reverse polarity on the input would blow the fuse. Nevertheless, it would be likely more elegant to use an N mosfet + PWM charge pump from the MCU. That's often used for in automotive.

I also wonder if using the OPA2333 makes sense here. I see it is configured as a difference amplifier here. This gives us ~60dB CMRR top. There are maybe parts for this. I like that the author used the charge bucket filter C15, C16. Not all opamps like transients when S/H circuit opens. There is often ringing.

Regarding the diodes and TI.

OPA333 datasheet, the chapter 7.3.2, describes the protection. Normally, we do not specify (characterize) ESD cells on the input as they are not intended to be used for operation. (there are few exceptions). What matters is the maximum input current to the pin. This is defined in the Absolute Maximum Ratings, chapter 6.1. I believe there's a typo 1 vs 10 mA (see the note 2).

[Warhawk]

To all - I am genuinely interested in contributing to this project. Could someone give me a brief tl;dr; what the problem with grounding and Aisun is?

I work nearly daily with small isolated power supplies (sn6501, 6505, 6507, ucc25800, LM5180, integrated modules, etc. etc.). It should not be a deal to do any kind of isolation necessary (including isolating signals). However, I need the problem statement.

[c0d3z3r0]

Well, there were different reports of weird problems the Aixun has, which are probably based on the fact that they use a PELV supply (secondary GND earthed).

Only now I understood what might be the problem from this post: https://www.eevblog.com/forum/reviews/aixun-t3a-misbehaving-on-grounded-pcbs/

Current flows not only from to the middle connector back to GND but also through the earthed thermocouple, the PCB, the earth connection.... that could be reason for the weird behaviour there and it's definitely not good soldering a grounded PCB with fine traces because there will be 4A (1/2 * 8 A) flowing through the PCB  See red arrows. I think this finally is the best argument for not using a PELV supply.

[c0d3z3r0]

[Thinking aloud]
I also wonder, if protections such as D12, D1, D3 are really necessary.

Unsure about D12 but D1 and D3 are crucial for ESD protecting the GPIOs. Many T3A GPIOs were killed already. Check the T3A thread.

[Warhawk]

[Thinking aloud]
I also wonder, if protections such as D12, D1, D3 are really necessary.

Unsure about D12 but D1 and D3 are crucial for ESD protecting the GPIOs. Many T3A GPIOs were killed already. Check the T3A thread.

Is there a circuit diagram somewhere? Do they have any ESD caps and current limiting resistors? I don't dispute the problem but there might be a different solution for that.

[c0d3z3r0]

Well, surely, there are always multiple options for one problem. What's the problem with just using (better chosen) diodes?

Schematics:
https://github.com/c0d3z3r0/aixun_t3a_rev/blob/master/schematics/aixun_t3a.pdf

[AxxAxx]

First of all, really nice to so much engagement and discussion about AxxSolder. I am at the moment in the djungle in southern india, turns out they have limited access to kicad and internet here. Getting home in about two weeks.

Some thoughts:
I will break out the AxxSolder hardware to a separate github repository (AxxSolder_hw) when I get home from my vacation. That make sense as hardware and software can then be released and versioned independetly. Git submodules could also work but it also makes it a bit more complicated for someone not used to git and git submodules.


@Pooya, yes, please list all issues you find.
If you list them here I'll transfer them to "Github Issues" so that they are collected in one place and it is easy to work on them.

At the moment the Temperature measurement is "all over" when no thermocouple (Cartridge) is present, I know. This is already documented as a Github Issue. One way to detect if no Cartridge is  present is to apply a very fast heating cycle and read the current response. That way also faulty cartridges could be detected.
There is no implemented current measurement at the moment. I also need to implement current measurement in order to do proper PID control with feedforward model based control.
If a temperature model is created for the tip it can be calculated how much energy is required to raise the tip to a final temperature. By integrating voltage*current for a specific amount of time that final temperature should be reached. A automatic calibration routine could be run to measure the thermal mass and heat dissipation of each tip.

Next to a thermal properties detection routine I think there should be a temperature calibration routine. By using this routine the user should be able to go through a couple temperature points and input the real measured temperature, measured by an external termometer. This in order to account for variations in components and tip geometry. The original JBC stations I have access to do only have a global temperature offset value as far as I know.

Neither of the above routines should be mandatory to use though. Just be there to increase accuracy and performance of thermal regulation if needed.

There was also a question about why not the built in USB-C PD support of stm32g4. Basically because I have never used it before.. But after reading about it, it looks like a nice solution! I have to look more into it. A possible downside is that I'm running out of pins on the current MCU and an external USB-C PD controller does use fewer pins from the MCU, only two I2C lines.
I will investigate this further. I will also look into the fact that 240W USB-C PD is coming and what that brings. That would be quite cool. Having a 240W soldering iron powered by usb! Can it really be possible? Downsides?

As for the grounding/isolation discussions: I believe that you guys have really valuable ideas and experience in this subject. This is not my main expertise so I'm glad to see that you want to contribute in this area.

@eliocor will sponsor the project with T115 handle/cartridge. This will enable me to measure the thermocouple characteristics, tune a initial PID as well as determine how this handle is detected (which pins are shorted in the handle). I believe that there is another pin that are used to sense if a T115 handle is detected than T210. As the current hardware does not have that pin automatic detection of T115 will most probably not work on the current hardware. But I'll implement that for next hw release. Highly aporeciated!

With this said. I am Happy to invite everyone wanting to contribute to the project. Fork the AxxSolder repo and submit a pull request.

[AxxAxx]

Ok, back to my home lab in Sweden!

I have since I came back done some firmware upgrades on AxxSolder. There is now a settings menue where you can set a number of parameters which are then saved to non-volatile flash memory, for example start-up temperature and "sleep time".
I have also started to implement current measurement and which will be used in temperature control and tip detection.

I have also updated my original JBC cartridge layout with the cross-sections. This because I realized that there are problably not only one TC in the C245 cartridges but 3 TC's.. Here is the updated post: https://www.eevblog.com/forum/projects/axxsolder-jbc-soldering-controller/msg5124267/#msg5124267)

Regarding the hardware.
I think the next steps are to implement USB-C PD power and DFU and take a look into the tip grounding topic.
I can drive and implement the USB-C features but I feel like I could get some help with a proper grounding solution.
@Warhawk you mentioned you were interested in contributing and had working experience in this topic. I would be happy to get your support (as well as anyone else's). I created a discord channel for a more chat-like discussion forum about AxxSolder hardware. https://discord.com/invite/WEFE46QuNv

[neverendingstudent]

Hey c0d3z3r0, I noticed in your post signature you link to the T3A reverse engineering project on Github.  Do you know of a similar project for the T420D?  Is anyone reverse engineering that model / making open source firmware?

[c0d3z3r0]

Hey c0d3z3r0, I noticed in your post signature you link to the T3A reverse engineering project on Github.  Do you know of a similar project for the T420D?  Is anyone reverse engineering that model / making open source firmware?

No idea, sorry