G600 Microscope dead keys / sporadic inputs

[Rooster Cogburn]

I have this dirt cheap G600 microscope:



I think they're pretty awesome. Image quality is just fine, you can reverse the stand and have the microscope 'overhang' in the other direction, and if I put it on the little monitor stand my scope stands on it's in a perfect position. I think it's very usable for SMD soldering or reading of small component labels, checking for solder bridges or other defects. Input lag is acceptable and you can put a cheap UV filter over the opening so you have a piece of glass & metal at the bottom, no risk of heat / flux fumes breaking it.

I originally bought one for ~25EUR shipped. After a year or so of perfect operation the buttons below the screen started randomly triggering. Really annoying when the zoom level changes or the menu pops up while you're soldering something. I could never figure out what the issue was. There was no visible damage to anything, I reset it with the little pinhole button, removed the battery, let it sit for a day, made sure the flex cables etc. were all in a good spot. When I look inside the microscope it seems like the mechanism that adjusts focus (big knob) might brush against the flex cable connecting the buttons to the main board. Hard to tell. No idea.

I simply bought a replacement, just buying the main unit was even cheaper. This one looked slightly different (bigger 'Microscope' font) and somewhat different picture/colors. After a year or so this one also just started to have sporadic ghost inputs on the buttons out of nowhere. Then a minute later the keyboard completely died. For some reason the power button still works fine, even though it's on the same input board connected through the same flex cable. I'm actually mostly fine with this, I don't need to use the menus or the record function or anything, I just want the thing to display an image. I reseated the flex cable connecting the input and main board, no change. Power button works, nothing else.

Anybody else has one of these and experienced this issue? Any idea how to fix it? I'd be perfectly happy with this thing if it kept working

Also, giant warning inside:



 

[ygi]

I have its big brother, the G1200. When I press and hold a button, it won't register another button press until I release the first one. Basically, all it takes is one of the buttons failing in closed position and the rest won't work (power button still works, though). I guess they used the cheapest microswitches they could find for the G600.
Pretty sure those buttons go straight to the main IC with nothing in between, so if they work fine, somehow the main IC input pins got zapped.

[Rooster Cogburn]

So maybe that huge warning label has some merit :-)

I guess I should've done my homework and checked if the buttons are actually working. Mechanically they feel fine, though.

It's interesting that one of them failed with sporadic inputs while the other one had sporadic inputs for like 10s and then the entire control panel went dead. I didn't check, but I assume the power button is connected differently then.

[DavidAlfa]

These tactile switches fail a lot that way.
Not a direct short, but can develop resistance (few Kohms) between their contacts when open.
As the pullups are usually weaker, (ex. 100K), they'll pull down the signal.
Get some new switches and replace them, don't bother checking which ones are bad, sometimes they fail intermittently.

[Rooster Cogburn]

I had no idea, interesting. I bought one of those 20 types of surface mount switches sets from AliExpress a while ago, I might even have something matching around.

Why do you suspect the pullups are so weak? Wouldn't you normally use something like 5-10K?

[DavidAlfa]

It depends, but often I find pretty high values, since MCUs have very high input impedance these days, so 100K is already a lot more than needed.
A cheap hack is to lower the pullup resistor to ex. 1K, 470ohms. The switch will live some more, or not... they can also fail shorted.

Also, some devices use the ADC to read the switches, there's a pullup resistor, and another for each switch (connected in parallel), making a voltage divider when pressed down. The mcu recognices which button is pressed based on the adc reading.
If any switch develops resistance between contacts, the reading be wrong, causing ghost presses or not working at all.

[Rooster Cogburn]

Well, seems like these dumb few switches are more interesting than I thought! Since this unit has so few switches I would've just assumed a one pin per switch arrangement.




(are there no spoiler tags on this forum to hide large images that break up the flow?)

Seems like you were spot on. On the six pin flex cable there are 2 ground, 2 dedicated to the power switch and two lines going through the switches. Switches connect with a 0-67K resistor between VCC and KEY.

The pictures of the board I posted are from the old microscope with sporadic random inputs. I can't measure any fault on these switches.

The newer microscope where all buttons but power went dead seems to have a slightly different board where VCC and GND are merged. Guess that works, too. But here when I measure between VCC and KEY I get ~80K resistance. Something clearly has failed. I used hot air to remove all switches. They all tested fine in isolation. I still get 80K. I noticed something that looks like flux residue on the board. I cleaned it off with IPA. Now there's no connection between VCC and KEY. I started soldering switches back in, occasionally checking resistance, a few MO. After all is done I clean the board again and resistance seems to fluctuate a bit, 0.5-2MO. I clean the board again and measure 5MO. Maybe I should've just replaced the buttons since I have suitable replacements, but well, the old ones seemed to be no issue.

I decided to clean the board of the first microscope as well, now it measures a few dozen MO resistance between VCC and KEY. Guess my brushes have been so contaminated with various types of noclean and Amtech flux that wiping a clean board with the brush and IPA lowers the resistance of the solder resist

The new microscope with the dead buttons still has issues. When pressing the MODE button it sometimes registers as MENU or LEFT. Sometimes pressing LEFT is missed. I should've measured the resistance of the closed MODE switch, hm. The old microscope with the sporadic inputs didn't seem to show any in a few ~10min tests.

[DavidAlfa]

That's the analog method I said... Very sensitive to unexpected impedance changes.
People abuse flux a lot. Once you have the pads clean, use ak almost invisible layer on each pad, do each pin pin in a single operation, quick and efficient.
Avoid those huge blobs I see in most yt videos to compensate their poor soldering skills, reworking the same pin 15 times.
I usually start with no flux, apply solder to a single pad, put the part and solder the pin, doing all the alignment.
Then I put a very small quantity and do all pins in a single pass (when possible).
Don't flood the thing with alcohol, ot the flux will enter the internals. Take a rag, drop some alcohol into it and rub the board. Repeat few times with clean parts of the rag.
It will be pretty clean, and you'll avoid contamination.
Not a problem with most parts, but in switches, connectors, potentiometers... definitely!
Once the flux enter the insides, no amount of alcohol will fully remove it.
For ICs, sure, I also like to flood it and use a brush for the best result.

[Rooster Cogburn]

Yeah, pretty much like you said. I think the newer microscope was misbehaving due to the flux and I assume the MODE button has some sporadic resistance when close. I didn't check for that because I was assuming the problem was when the switch was open as there were 80KO on the line with no switch pressed. Would you say this is in generally a pretty poor method of hooking up a bunch of switches? It also seems on the microcontroller side it would be hard to interrupt for that and you'd have to poll them. Maybe have a comperator interrupt or something when any voltage at all makes it through...

I tend to avoid the gel flux. It's so sticky and hard to remove. It's on your parts, your tweezers and when you wipe the PCB with IPA you'll just get a sticky brush. It's very effective but a PITA. So I work with noclean flux with <5% actual flux whenever I can. I only use the gel flux when I use hot air or when I need to work on something for longer periods of time, like removing solder bridges from a QFP IC or so.

[DavidAlfa]

That's the problem.
There're no resistors pulling to gnd, so if no switch is being pressed, you shouldn't measure 80K, if anything, in the Mohm range.
Assuming it's disconnected from the main board, or it'll affect the reading.
Not really a poor method. It's simple, fast, you only need one mcu pin for up to 8-10 buttons or so (depends on the ADC resolution).
I've had the same problem with certain displays, having dedicated pins for each button, but they had a weak 100K pullup.
The switches developed 10-20K resistance, and the display bevahed completely crazy, mainly (actually funny) trolling people by lowering the brightness when they weren't looking at the screen. It was completely random.
You can make it more tought to that issue by lowering the resistors to few Kohms, but using the same ratios.
Ex:
100K.. put 1K.
82K... Put 820r.
And so on. As long as the voltage dividers drop the same voltage, it'll work.
But if it has very specific precision values, it might not be possible.

[Rooster Cogburn]

It was in the MO range after I cleaned the flux residue next to the test pads. And I'm sure if I measure the MODE button it would sporadically close with some non-zero resistance. I didn't do that because I was investigating why there's a connection even when no switch is pressed, I only tested the switches open.

If you look at the PCB next to the capacitor, there's an unpopulated pad for a resistor to ground...

In any case, I'm fine with the way it is now. I learned something (thanks!!) and my microscope is working good enough now. The MODE button switches between live view, picture & video modes, and I never need to press it as I only use the live view. I have the exact replacement switch I need, so I could just swap it if I ever have any issues.

[DavidAlfa]

I guess the unpopulated resistor it's a ”better safe than sorry" thing, you never know if something will need a last minute change after fabrication, for whatever reason.

[Rooster Cogburn]

I think so. Like adding a pad for a decoupling capacitor and not populating it. Doesn't cost you anything and if you find you need it you don't have to do another spin on the PCB.