EEVblog 1769 - REPAIR by REMOVING a Component?

[EEVblog]

Can you repair a PSU by REMOVING a component? Well, in this case, yeah, kinda.

00:00 - My NerdQaxe++ Bitcoin miner has failed. Guru Meditation!
02:13 - Bitmain 1370 chipset thermal paste
03:54 - The flipchip die under the microscope
04:48 - Thermal paste application triggering
06:55 - Time for a new strategy R2. The Vreg temp sensor
07:53 - A look at the open source hardware schematic in KiCAD
https://github.com/shufps/qaxe
09:42 - The TPS53647 4 Phase ASIC Buck PSU controller
https://www.ti.com/lit/ds/symlink/tps53647.pdf
10:23 - The CSD95472Q5MC Synchronous Buck NexFET Smart Power Stage
https://www.ti.com/lit/ds/symlink/csd95472q5mc.pdf
11:50 - We can try wacking a heatsink on + Jaycar field trip
12:41 - Did that fix it?
14:10 - It's becoming obvious what the fault is now
16:24 - Thermal camera testing
16:56 - Oscilloscope time on the 3 phase PSU
18:21 - Desoldering time, maybe
19:42 - Will removing this component work?
23:52 - AutoBalance multi-phase buck conversion for the win

[Hydron]

Does self destruction count as removal? Once fixed a bench PSU which would blow the mains fuse by ramping it up on a Variac. The lower voltage destructively cleared the shorted ceramic cap across the bridge rectifier without popping the fuse, and it worked happily ever after.

[EEVblog]

Does self destruction count as removal? Once fixed a bench PSU which would blow the mains fuse by ramping it up on a Variac. The lower voltage destructively cleared the shorted ceramic cap across the bridge rectifier without popping the fuse, and it worked happily ever after.

Putting a high current supply on a rail to "blow out" a shorted part or a hairline short is a time honored method of repair.

[thm_w]

Not that it was the issue but 80C seems quite conservative for a thermal cutout.
Its directly measuring the die temp, so 100 C should be fine, recommended in CSD95472 datasheet is up to 125 C.
Each IC can power 2 ASICs easily, I guess the third is added for overclocking headroom and maybe alternate ASICs.

There are some improvements mentioned in a 5.1.1: https://github.com/shufps/qaxe/pull/31 don't see anything obviously mentioned as causing failures though.

[EEVblog]

Not that it was the issue but 80C seems quite conservative for a thermal cutout.
Its directly measuring the die temp, so 100 C should be fine, recommended in CSD95472 datasheet is up to 125 C.
Each IC can power 2 ASICs easily, I guess the third is added for overclocking headroom and maybe alternate ASICs.

In this case the shorted MOSFET was dragging the Vreg down to 1.0V which isn't enough to run the ASIC's, so it had to be removed.

[Psi]

You really don't know what country the device will be used in, and if the office will have AC or if it will be a hot warehouse somewhere.
So I get why they added the extra phase when it can obviously run at an ok temp on 2 phases in an office with AC.

[ballsystemlord]

Does self destruction count as removal? Once fixed a bench PSU which would blow the mains fuse by ramping it up on a Variac. The lower voltage destructively cleared the shorted ceramic cap across the bridge rectifier without popping the fuse, and it worked happily ever after.

Putting a high current supply on a rail to "blow out" a shorted part or a hairline short is a time honored method of repair.

Yikes! I'd have thought you'd break several parts in the process of doing that.

[ballsystemlord]

@EEVblog Dave, try attaching the heatsinks to the back of the board with thermal paste (you'll also need a way to stop them from moving over time). The backside is doing the heatsinking.

Another option, if you have to heatsink on the front, in thermal putty. It's made to fill larger gaps than thermal paste.

[Psi]

Does self destruction count as removal? Once fixed a bench PSU which would blow the mains fuse by ramping it up on a Variac. The lower voltage destructively cleared the shorted ceramic cap across the bridge rectifier without popping the fuse, and it worked happily ever after.

Putting a high current supply on a rail to "blow out" a shorted part or a hairline short is a time honored method of repair.

Yikes! I'd have thought you'd break several parts in the process of doing that.

na, when a rail is supposed to handle X volts, but it has a short, you can usually put up to X volts on it safely.  All that happens is whatever is faulty gets hot or explodes, depending on your current limit.   
There are exceptions of course, some FPGA stuff requires rails to come up in specific order, and there's probably other exceptions too.
Normally you only use enough power so the heat shows you where the fault is on the thermal camera. Blowing stuff up does carry some risk of PCB damage.

[Hydron]

Thermal cameras feel like cheating sometimes - a good one can see milliwatts of dissipation so you can be very caution with your short-finding current. I used one on the weekend to confirm location of a short on a R6581T analogue board (sadly as expected it was in the unobtainable gate array ASIC).