(Now with pictures!) I built Sergey's 80W metcal-compatible soldering station.

[rfmerrill]

I found your experiments with the NTC quite informative. It seems their use case is really fairly specific.

I usually see them in use on mains AC inputs, where the voltage is much higher and current much lower, so they can be rated for a much lower max current and the resistance isn't as big of a deal. The voltage drop produced by the NTC will be on the order of normal variation in mains supply that you'd expect anyway, and at 120VAC the potential inrush current is really high: an NTC with a cold resistance of 8 ohms could cut the immediate current spike from 50-200A to <15A, which may be way more than what the system will draw normally, but brings it below the 15A rating of household wiring.

NTCs' main advantage is simplicity and cost--it's a single component and they're relatively cheap. It has other downsides like not handling quick power cycles.

[rfmerrill]

Thanks t0m for the inspiration: I discovered that the EastRising/BuyDisplay OLED ties its screw holes to its internal ground. Neither the winstar OLED nor the adafruit LCD do this. Bleh!

Edit: it's easy to fix, the screw holes are only attached to the ground plane on the back layer and the thermal reliefs are really easy to cut

[rfmerrill]

I got rid of most of the glitchiness on unit #2. Things I did:

• Ditch the NTC
• use a shorter and thicker gauge wire for the ground from the output jack (I'm just using two separate wires because I mounted the jack right by the footprint on the board
• Cut the thermal reliefs on the display module, confirm that ground is not shorted to housing when it's installed
• Mod in the connection between tip detect and the enable on the main supply
• Switch the main power transistor back to the "good" silpad from the cheap amazon one (which I used briefly because I couldn't find the good ones)
• Change the screw grounding at Q2: For this unit I initially tried a fairly thin wire directly soldered to the source pin of the transistor. I switched back to how I did it on the first unit which is to run a thicker wire from the ground leg of D19, but I kept the small wire and just connected it to the thick wire across the shortest distance I could

I still had significant glitchiness after step 4, but some of the steps 1-4 did change the behavior.

After step 6 I now have no display garbling, no freezes or blanks, and only a rare clean reset when I pull the iron tip out under load.

[rfmerrill]

Since I'm looking at it now, I thought I'd give this info:

2x of the BuyDisplay displays (PN ER-OLEDM1602-4Y-4BIT6800) cost US $31.02 plus $25.97 shipping to California and they arrived 7 days after I placed the order

3x of the Winstar displays (PN WEH001602AGPP5N00001) cost US $9.79 each but with $59 shipping to California (and the shipping is the same for 1 unit) and arrived 11 days after I placed the order.

The Winstar displays seem clearly superior, but depending on your situation the BuyDisplay/EastRising one may be easier/cheaper. I bought those directly from buydisplay.com and the Winstar ones from Unisystem

[t0m]

That is one hell of a shipping charge. Holy crap.

Mind you, I'm a total cheapskate. My build is going inside a box that I got for $1 in the local junk shop.

[rfmerrill]

That is one hell of a shipping charge. Holy crap.

Yeah as far as I can tell they're the only English-language website where you can order them shipped to a US address, and the only shipping option is FedEx. I'm sure if you're ordering dozens it ends up being totally reasonable...

[rfmerrill]

Confirming if anyone is curious: the station absolutely freaks out if you don't have the ground wire connected. I wonder if the issues I'm seeing might just be too high impedance to mains earth?

[rfmerrill]

Unit 2 is struggling with microcontroller resets and false tip detects now. I wonder if it's "broken in" or if the colder weather is affecting it (like seemed to be the case with the first unit). This is actually somewhat good news as it means that these problems were not the result of the ferrite cores/inductors I was using.

I kind of want to try tracing back whatever is causing the microcontroller resets, but there are a few obstacles:

• The bad behavior tends to happen much less if the housing is open. I don't understand RF enough to explain why, but I can only assume the cause is either due to signal taking a different return path (through the lid of the housing) or the radiated noise bouncing around the housing instead of escaping.
• The pins I would want to probe, near the microcontroller, don't have a close enough ground to avoid just picking up the radiated noise on the scope probe.

That being said, I do have a few hypotheses:

• The low-pass filter in the tip detect circuit is located really close to the programming header. About 10-50 mA AC current flows through C76 and that might be enough to make the ground plane nearby on the bottom quite noisy. The ground return path for that current passes very close to the reset pin on the programming header
• The 13.56MHz reference signal coming from the microcontroller passes right past the reset pin as well as directly under Q3

[quadtech]

Perhaps you could try an external 13.56MHz oscillator mounted close to the MOSFET driver IC,
and change the firmware to generate an Output Enable signal to turn on the oscillator -

https://www.digikey.com/en/products/detail/sitime/SIT8008BC-13-33S-13-560000G/11473743

[rfmerrill]

It was the goddamn display again.

Oh my god I spent hours trying *everything*. Moving grounds around, even added a second DC-DC module to power 5V directly off the input

Then I disconnect the display ground and it works.

I connect the display ground to one of its screws and it works even better

what the fuck

[rfmerrill]

So it turns out sometimes you get the best result if the display is grounded *only* to the housing and *not* to the mainboard?!?!

[rfmerrill]

OK, Unit #2 is now working perfectly after making the following changes from its previous fairly usable state:

• F connector connected to the board with a short length of RG174
• F connector wrapped in copper tape on the inside, soldered to coax shield
• 5V linear regulator replaced with second buck converter module, connected as close to power input as possible
• 47nF C0G cap from microcontroller reset pin to ground (same ground as the crystal load capacitors
• Both buck converter modules grounded only on the output side--5V grounded near D13 footprint, 10V grounded near C36
• C76 relocated to top side of board--connected between L12 lead and top ground plane
• Display grounded by one of its screws on the side closer to the RF section

Resets and freezes on tip removal or heavy load appear to be totally gone... fingers crossed.

[rfmerrill]

In case anyone is following this thread and is curious--I've been using one of the units basically exclusively since March, and it hasn't failed me yet. The occasional resets and display blanking when you pull out a tip has not been solved, but it also doesn't really bother me.

I have had a few tip cartridges fail (lose their inductance/base station won't drive them) but I can't tell if it's a higher than normal failure rate. I have been using it pretty heavily.

[WaveletSea]

From minute 15:08 to min 16:45 the host of the video shows you the spots where Sergey has done it's mistakes & also how to fix them all.
Good luck.

[rfmerrill]

Oh... it would be very nice if there were a translation. I don't speak russian xD

It might be simplest to just make a daughterboard with the microcontroller circuit on it with better layout for the microcontroller problems.

Also his heatsink is sideways... might not matter, but it's always best to have fins oriented vertically.