Power supply cable problem - GT2560 a+ nearly burned !

[exodius]

Hello

I bought a i3 pro W on november 2017, i'm very happy with it : kit is well done, mount it very carrefelly, place it on a wood board and screw it so X/Y/Z can't move, calibrate and you'r ready to print !

Happily i have upgrade marlin firmware long long time ago, because the GT2560 a+ board was using the marlin v1.0 or something like that.

So after nearly 1 year of intensive printing (7kg of PLA, 1kg of ABS) and some minor problems, here it is : hardware failure !
Heating bed not heating - Bed Power Led is not lighting. Printer is displaying a nice hardware failure, octoprint says thermal runaway from the bed.


Now look at what i have found when i got to my printer :




You can see that the +12V rail that power the heat bed is burned inside the connector, the plastic has melted so much that it is not more here, and the socket is burned too.

MORE !



The solder HAS MELT !!!!
Lucky me, is was not touching the other pad so the power supply was not short.


Now, let's have a look as why this has happend

Have a look here : https://forum.digikey.com/t/common-comp ... ectors/328
under the section 4 Pin ATX +12V Power Connector, you can find a lots of manufacturer codes for those connectors.
Go to molex web site, enter the part number and have a look :
https://www.molex.com/molex/products/da ... USINGS.xml

They clearly says : MAX current 13A with 45750 series terminal crimped with 16AWG wire.
I had a look to other manufacturers and models, and they all they the same.

Product specs :
https://www.molex.com/pdm_docs/ps/PS-45750-001.pdf

I had a deep look onto the supplyed power cable from the kit, wich i was using : no AWG gauge mentionned on it
I tryed to measure the diameter of the wire, counted the number of strands and all the like i can do.
I am now thinking that the wire is at most 18AWG, wich is OK for the connector, but not with the current usage. Please correct me if i'm wrong.

MK2 heat bed : https://reprap.org/wiki/PCB_Heatbed
Resistance is around 1 Ohm @ 12V

so it's 12 amps form the power supply, that must go to the power connector of the GT2560 and all the rest of circuit.


So here is the whole problem :

Wire is just sufficiant gauge to handle the current when crimped into the terminal,
Wire and terminal are used at the maximum rating, wich is NOT GOOD when you run a heat bed that is ON 80% of the time while printing abs, and more than 50% of time when printing PLA, and when the printing heat the bed when cold, is ON more than 15 minutes to go 100°C (ABS printing)

A LOT of heat is generated from the connection, because material a used to the limit, and after some hundreds of hours, the plastic housing (nylon ?) melt slowly, plastic particles go inside the connection, making things worse.

Heat is going SO HIGH, THAT EVEN THE SOLDER ON THE PAD HAS MELT !

The 15 Amps fuse is not in play at this time, because the circuit is ok after the fuse, and the problem is before the fuse.

If the crimping inside the connector is not done correctly, more heating can came from this.

When i tryed to removed the connector from the socket, i gently pressed the latch, it broke in my finger.
The socket was nearly glued to the connector, i had to pull more than usual to get it off.

SOLDER JOINT HAS MELT !
When i removed the socket from the board, i had trouble to melt this solder with my soldering station running at 280°C, had to go to 310°C and add PbSn solder to get it gone. (Yes i know your are using RoHs solder with has a higher melting point)

What was the temperature inside the connection ? so high that the solder has melt. NOT GOOD
However, this board is of high quality : the pads and tracks around haven't lift, even with a good heating from the desoldering operation. REAL GOOD POINT ! (i have had some crap board from samsung tv, pads and tracks lift up just from a 280°C iron)
Pads and tracks size seems ok to handle the current from the heated bed (i did not measure this one).
Even the fuse holder, wich is nearly touching the socket, is ok.
The wires of the power cables are ok when i cut them just after the connector and stripped them, but the copper clearly show signs of overheating.

More :
The Ground (0V) connection, have absolutly NO signs of overheating.

So, this power cable and this connector/socket are used to the MAX current capacity, ALL THE TIME !
This is not good at all, any electronic/electric engineer will tell you that.


You have to take into account that most users who print ABS put the printer in an enclosure, wich make ambiant temperature much higher, and making this situation worse.


Risks :

When done correctly, all this overheating just make a hardware failure, wich is detected by marlin firmware, IF YOU USE LASTS VERSIONS AND ENABLE THERMAL RUNAWAY DETECTION.

If there is just a little more problem in the connection, like wire a bit too tiny (AWG gauge not respected) or the crimping is bad or the soldering joint is bad, then the heat problem could become a FIRE PROBLEM : PCB or wire could burn.

I don't think you would to hear that a customer had his house burnt because the power connexion is wrong, and thing can go worse (dead people...)



So, i really hope you will take those into consideration :
You geetech guys MUST use upgraded firmware WITH termal runout function enabled !
USE ANOTHER CONNECTION TYPE TO SUPPLY POWER TO THE BOARD

Why to use a molex connector on the board side, where on the other side there are ring terminals, wich are no problems ?
Is if it done so the user can't plug power in reverse, no, they can plug in reverse on the power supply bloc side.

I really think that you should use ring crew terminal on the board side too, like the power supply bloc, and nothing detachable that use those 1/2" square pins, and use a 14 AWG wire to definitly solve this problem.

Thanks for your attention and reading of this long post.

Duwelz Loïc
Electrical Engineer

[_kaktus_]

Hello.


You did not write about the most important.
The direct cause of this failure was the poor quality of the contact in your connector.
In this case, a resistance is generated at the contact of the pins, and when a high current flows ... and so on,
and so on.

Recently, someone presented a similar case.

This is a very serious problem

[exodius]

Hello _kaktus_

I did not talk about the poor quality of the contact in the connector, because i suspect there are of good quality.

Of course, if the mating of the pin not following the specifications from the constructors (amp or tyco for exemple), this is another problem.


thanks for your feedback

[_kaktus_]

One more comment.

Your consideration concerns a single cable and connection.
And everything is ok.

But...

The wires and connections are 4.

The fact is that in your case everything has failed.

Perhaps it would be necessary to increase the number of pins in the main power plug, or consider using screw terminals.
This would improve the reliability of this sensitive connection.

[Dilbert0815]

Another one yesterday:

P1150866.JPG (318.96 KiB) Viewed 28748 times

P1150870.JPG (335.07 KiB) Viewed 28748 times

If you watch carefull you may realise the green connector block for the heat bed.
This was replaced some days ago to a 24A rated genuine brand. The noname removable connector showed already burn marks when removed.

I had also replaced allready most of the original 2x2x1.5mm2 cable with a single 2.5mm2 speaker wire pair and XT60 connector on the printer frame. Just a 10cm receptable with the ATX connector were staying at the board.
I noted already that the (suposed) 1.5mm2 cable got warm to the touch with the load, now the 2.5mm2 stays cool.

I did a small comparisation of the orginal cable with a general 1.5mm2 speaker cable. To me the speaker cable appeares to be beefier:

P1150873.JPG (204.29 KiB) Viewed 28748 times

The MOSFETS on the board were upgraded with the bed connector to Nexperia PSMN2R0-30PL, these are 2mOhms, vs. 15mOhm for the original STP55nf06. The later have also a quite worse logic level. At 5V they just about usable, but poor Rds_on. The Nexperia are way better there, but cost about 4x at a german electronic supply shop (~1,70€ each). But now they are pretty cold, the heatsink serves now just a mechanical purpose.

Acually the GT2560 is not a bad board. There are way worse boards. Its just that for 10$ extra the board could be made realy good.
Just use better FETs and better connectors.
A little more space for more relaxed (and safer) routing and lastly please: add the EXT3 and genuinly free ports of the Mega 2560 for better extention.

[_kaktus_]

Hello.



It's good that you noticed the fault in time, behind it something was damaged.

I'm wondering what has been changed, since such faults have been occurring for some time.

[Dilbert0815]

Actually the printer was used mostly unwatched in the back. This could have turened bad.
I just had a print job started and got surprised by the thermal runnaway protection in Marlin.
That was supprising as I spend almost all of yesterday calibration the PID of the hotend.
I had noted a little unusual thermal odor, but placed this to the PID calibration work with quite lot of hotend thermal cycles.

By the way, my cables allredy use crimped ferules on all screw terminals.
Obviously we can't overstate some general fire safety precautions with these printers, for the novice electrician these are fire hazzards in the making.

Next step is bridging the fuses on the board massivly and use an external fuse block from car, as this seems to be the next bottle neck.

P.S. To be fair, the event happened with updated ~14V on the power supply, which was the reason for the PID retuning.
So the current that was already component borderline was pushed may be a little over the edge. The replaced better components may have recovered a few other 100mV from the previous cable and connector losses am made the current even worse.
But it shows there is no real safety margin in the used components at 12V version. For 24V version the current situation is a little more relaxed.

P.P.S. Like the OP I have engineering degree and more than 20 years experience in automotive electronics development, even while last years mostly on the firmware side.
The firmware was already Marlin 2.x with runnaway protection, and a few other goodies like WS2812 LED control for caselight and event signaling.

[exodius]

Hello

I think the connector plating is not good/incorrect/whatever, and this cause the bad connection.
Or maybe the crimping of the wire is not correct : done with incorrect crimping force (maybe a tool that has gone bad ?)
Or it could simply be the wire that has gone bad

[Dilbert0815]

Hello

I think the connector plating is not good/incorrect/whatever, and this cause the bad connection.
Or maybe the crimping of the wire is not correct : done with incorrect crimping force (maybe a tool that has gone bad ?)
Or it could simply be the wire that has gone bad

Nope, I checked the connector.
The wire was still propperly crimped and any attempt to pull it from the (burned) terminal by brute force failed.
Its not a bad one, but it is working at max absolute rating.

Another problem is that the connector already seperates heat bed and remaining current at the poser supply. If the wire an therminal would be used realy doubled the stress from the heat bed would be distributed over both wire/connector pairs. This way the sole pair supplying the heat bed sees all the max current. Even with dual extruder working at max the second pair sees may be less than 50% power of the heat bed connector pair.

If someone whats to slightly do some precaution I would recommend shorten out both + and - pins repectivly with solder on the ATX connector so the sum current would distribute over both connector pairs. Thats not a perfect match, but at least would reduce the single point of failure in the heat bed pair.

P.S. My older A+ board came with these green removable plugs on all thermal outputs. They are RUICHENG 2ERCK, rated for 15A/300V. But these are just clones from Phoenix contact MSTB, rated at just 12A. And the Phoenix comes with massive metal terminals, not tin like the Ruicheng.

[exodius]

Hello

Thanks for your replys Dilbert

I have a nice milliohm meter somewhere in that horible pile of stuff in my basement. I will try to find it and mesure the resistance of the heatbed, let's says hummm in 20+ more hours (because it's printing atm).

Maybe we have a problem of very low resistance with the heat bed ? Really i don't know.

From the pics i send, i "seems", but i can be wrong, that the connector already got a nice solder bridge between the two +12V pins. Maybe the solder joint was not good ?

Got an answer email from geeetech :

Hello,
Thanks for contacting us!
Hope you have a great day!
Thanks for your feedback!
1.We think the issue lead by the poor contact between PSU and motherboard
2.There is a problem with the PSU input soldering of the motherboard
3.The 180W psu is not suit for your printer(We have upgrade it to 230W)
4.If it allows, we suggest you upgrade your PSU to 230W.
5.You can contact the sales to send you the new motherboard.

Regards

2.There is a problem with the PSU input soldering of the motherboard
This one !

Bad solder joint can result in lost of contact inside the joint (cold solder anyone ?)

I still think that the board desing should joint with a PCB trace the 2 +12V rails, and the schematics clearly not showing this joint. AND a better connector + more gauge on the power supply cable could definitly solve the problem.