Author Topic: EEVblog #1035 - Flaming DIY Power Supply  (Read 4822 times)

0 Members and 1 Guest are viewing this topic.

Offline EEVblog

  • Administrator
  • *****
  • Posts: 24775
  • Country: au
    • EEVblog
 
The following users thanked this post: jonovid

Offline Brumby

  • Supporter
  • ****
  • Posts: 5693
  • Country: au
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #1 on: November 03, 2017, 12:12:08 PM »



I think I would like to see some follow up from the producer before I jump in.
 

Online wraper

  • Supporter
  • ****
  • Posts: 6415
  • Country: lv
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #2 on: November 03, 2017, 12:24:26 PM »
The way that cap was soldered in between of 2 terminals, means a lot of mechanical stress. Even more stress happened when wires were screwed to those terminals. MLCCs are very susceptible to mechanical stress and microcracks may happen causing leakage and shorts. As of lower voltage specified part, MLCCs are very tolerant to overvoltage and usually survive even 5x times above their spec. So this unlikely to be an issue.
 

Online wraper

  • Supporter
  • ****
  • Posts: 6415
  • Country: lv
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #3 on: November 03, 2017, 12:28:37 PM »
I actually as example can say about one MLCC on Raspberry pi 2 and 3 failing short or heavily leaking. It always the same C97 cap nearby to mounting hole even though there are multiple of the same type capacitors on the same power rail.
 

Online kalel

  • Frequent Contributor
  • **
  • Posts: 773
  • Country: 00
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #4 on: November 03, 2017, 12:52:16 PM »
This big issue aside, these units are interesting, especially the slightly cheaper lower current ones. But once you add the case, plus a powerful enough power supply input, you can probably get some entry level complete power supply units for the cost.

Of course, if you already have a powerful input supply and possibly a case, and just need a converter, it's probably a good deal. Well, if it's more reliable than this unit in the video.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 1774
  • Country: au
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #5 on: November 03, 2017, 01:17:18 PM »
The way that cap was soldered in between of 2 terminals, means a lot of mechanical stress. Even more stress happened when wires were screwed to those terminals. MLCCs are very susceptible to mechanical stress and microcracks may happen causing leakage and shorts. As of lower voltage specified part, MLCCs are very tolerant to overvoltage and usually survive even 5x times above their spec. So this unlikely to be an issue.
I'm not sure I'd agree with MLCCs being tolerant of sustained over voltage, they handle transients and high peak voltages for short periods (minutes etc) but the reliability of them plummets at higher voltage. In this example with Dave's power supply thats probably not coming into it as the device wasn't running for significant periods of time and your suggestion of mechanical stress is a strong contender for the failure mode. There are some good manufacturer application notes on how to align and stress relieve parts on PCBs using slots or other parts to control where the strain occurs.
 

Offline djQUAN

  • Regular Contributor
  • *
  • Posts: 237
  • Country: ph
    • My DIY website
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #6 on: November 03, 2017, 06:19:17 PM »
The way that cap was soldered in between of 2 terminals, means a lot of mechanical stress. Even more stress happened when wires were screwed to those terminals. MLCCs are very susceptible to mechanical stress and microcracks may happen causing leakage and shorts. As of lower voltage specified part, MLCCs are very tolerant to overvoltage and usually survive even 5x times above their spec. So this unlikely to be an issue.
I'm with you on that. On the previous company I worked for, they take great pains about mechanically stressing MLCC parts. Then for high reliability power lines, we always used two in series.
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 6256
  • Country: nz
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #7 on: November 03, 2017, 06:44:42 PM »
Anyone else notice this bodge. At least i hope it's an intentional bodge.

Greek letter 'Psi' (not Pounds per Square Inch)
 

Offline Towger

  • Super Contributor
  • ***
  • Posts: 1182
  • Country: ie
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #8 on: November 03, 2017, 08:32:38 PM »
Anyone else notice this bodge. At least i hope it's an intentional bodge.

Good catch.  Bodge or contamination?

It is concerning as to what else failed.  Any half decent bench type power supply should happily survive direct shorts on it's output.  Dave said there was a burning smell again, the thermal camera would have come in useful.  Is the charred FR4 conducting 11W of power?  Still the CPU should not be rebooting, unless the input PSU was current limiting.

Dave do you still have your 'PSU Killer',  it would be interesting to see a working example being out to the test.
 

Offline NANDBlog

  • Super Contributor
  • ***
  • Posts: 3626
  • Country: be
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #9 on: November 03, 2017, 09:06:44 PM »
I guess it went from "interesting, only 20 bucks" to "not recommended" then.
 

Online wraper

  • Supporter
  • ****
  • Posts: 6415
  • Country: lv
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #10 on: November 03, 2017, 09:22:11 PM »
I'm not sure I'd agree with MLCCs being tolerant of sustained over voltage, they handle transients and high peak voltages for short periods (minutes etc) but the reliability of them plummets at higher voltage.
Type 2 and 3 ceramic MLCCs extremely drop their capacitance at high voltage, so voltage ratings may take it into consideration. I haven't done extensive testing. But I've done some experiments, as running 16V X5R capacitors on 120V for like for 10 minutes. I don't say they remain reliable at this voltage, just that generally it should not catch fire, especially that fast.
 

Online wraper

  • Supporter
  • ****
  • Posts: 6415
  • Country: lv
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #11 on: November 03, 2017, 09:42:06 PM »
Anyone else notice this bodge. At least i hope it's an intentional bodge.
I don't see it on the other parts of the video like 10:25, so likely it's a piece of something which got there during video filming.
« Last Edit: November 03, 2017, 09:44:23 PM by wraper »
 

Offline bundy

  • Contributor
  • Posts: 7
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #12 on: November 03, 2017, 10:05:07 PM »
The way that cap was soldered in between of 2 terminals, means a lot of mechanical stress. Even more stress happened when wires were screwed to those terminals. MLCCs are very susceptible to mechanical stress and microcracks may happen causing leakage and shorts.

Totally agree, the orientation of this cap is a no-go. If there is absolutely no other place for this capacitor then I would have rotate it 90 degrees, so the capacitor terminals are in parallel to the connectors.

Bram
 

Offline DrChiron

  • Newbie
  • Posts: 2
  • Country: us
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #13 on: November 04, 2017, 04:58:34 AM »
Greetings:
1. Now that you have cleaned up the damaged area, I would check to see the voltage at the front output binding posts now that the USB monitoring is non-functional.
2. I notice that there are 3 major traces in the area of the damage:
a: output +
b: output -
c: the circuit that would have been connected to C33 if the cap was fitted (no info on what that signal is)
3. Do not discount the possibility that that last circuit could have been damaged as a result of the cap failure.
Why was it necessary to use the ceramic type of cap used?
Also, I think that some output hardware fuse is needed there. The failure seemed to last too long.
 

Offline jaromir

  • Regular Contributor
  • *
  • Posts: 121
  • Country: sk
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #14 on: November 04, 2017, 05:15:56 AM »
Quote from: Psi
Anyone else notice this bodge. At least i hope it's an intentional bodge.
The "bodge" wasn't there from beginning of the video
https://snag.gy/4HFhAc.jpg
My hobby projects: https://hackaday.io/jaromir ----------- http://jaromir.xf.cz/
 

Offline M4trix

  • Regular Contributor
  • *
  • Posts: 238
  • Country: hr
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #15 on: November 04, 2017, 05:18:41 AM »
Anyone else notice this bodge. At least i hope it's an intentional bodge.

Those are tin whiskers and it's a normal process while tin ages.  ;)

j/k
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1310
  • Country: de
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #16 on: November 04, 2017, 07:13:05 AM »
Dave, that's been a very instructive video!

In automotive applications, there were several design rules on MLCCs, for safety reasons, that the whole xx k$ car does no burn.
These rules were violated in this application:

- Don't place MLCCs at the edge of a PCB (biggest mechanical stress during PCB assembly / further manufacturing bending situations)
- Never place single MLCCs directly on the battery input (automotive clamps 15 & 30)
- Use relaxed PCN layout and handling on MLCCs, like enough space to connectors, not too much solder mass, smaller case size, (1206 gives higher mechanical stress, due to mismatch PCB vs. MLCC ceramic), no manual soldering, etc.
- Use enough headroom in voltage rating, for sufficient voltage compliance and ESD resistance, especially on capacitors, which have direct access to connectors to the outside of the application.
- use 100V types when the board supply system (or the application itself) might introduce inductive spikes on the supply line

If MLCCs are used at the battery input, there were several safety measures to be implemented:

- use two MLCCS in series, each from different supplier. If one fails fatally (shorting), the other will avoid complete short
- use safety types, with 'soft' or 'flexible' termination. These fail open, but do no short
- Use small case sizes (1206 not preferred), and avoid "on-the-edge" values, that's the highest capacitance / voltage type of a case size.
- Don't use too small case size, as these are more susceptible to ESD damage... 0603 and 0805 preferred.
- Generally, use MLCCs of known-good suppliers only (which also have a QA system). These also implement additional surge testing in their production line, for example.

Frank
« Last Edit: November 04, 2017, 07:54:17 AM by Dr. Frank »
 
The following users thanked this post: thm_w, Bud

Offline metrologist

  • Frequent Contributor
  • **
  • Posts: 881
  • Country: 00
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #17 on: November 04, 2017, 07:40:58 AM »
The 5015 version does not appear to have that cap
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 1774
  • Country: au
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #18 on: November 04, 2017, 09:33:59 AM »
I'm not sure I'd agree with MLCCs being tolerant of sustained over voltage, they handle transients and high peak voltages for short periods (minutes etc) but the reliability of them plummets at higher voltage.
Type 2 and 3 ceramic MLCCs extremely drop their capacitance at high voltage, so voltage ratings may take it into consideration. I haven't done extensive testing. But I've done some experiments, as running 16V X5R capacitors on 120V for like for 10 minutes. I don't say they remain reliable at this voltage, just that generally it should not catch fire, especially that fast.
I've plenty of data from products I've worked on, and failure near to or slightly above the rated voltage of an MLCC is greatly elevated above the estimates from HALT experiments, there is a short reference to the same within this comprehensive document:
https://nepp.nasa.gov/files/25843/2013-Teverovsky-pres-MLCCsBME-n263.pdf
Immediate failures on first power on are typically from damage by mechanical stress, but the failures into the lifespan of the product are strongly affected by operating voltage and many companies will have mandatory derating factors based on their extensive historical data.
 

Offline cdev

  • Super Contributor
  • ***
  • Posts: 2798
  • Country: 00
  • Friend
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #19 on: November 04, 2017, 10:08:24 AM »
When I first started reading up on this kind of DC-DC converter, I found something that made me drop the idea. Some race condition or potentially destructive circuit thing could happen suddenly and unexpectedly.

Something like that.

But, I don't remember the specifics, just that something I saw made me say "No" to myself and just drop the idea.

It was in a thread here. I don't remember the title of the post. I doubt if I bookmarked it. I just made a mental note to myself not to buy one.

--

Is this the kind of supply that seems to always need a small load for stability?
« Last Edit: November 05, 2017, 12:50:08 AM by cdev »
"What the large print giveth, the small print taketh away."
 

Offline smithnerd

  • Regular Contributor
  • *
  • Posts: 83
  • Country: gb
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #20 on: November 04, 2017, 12:54:08 PM »
At around 15:00 Dave says that it's set to 10V, but it looks like 18V to me. The 8s and 0s are almost indistinguishable in that font.
 

Online Cerebus

  • Super Contributor
  • ***
  • Posts: 1916
  • Country: gb
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #21 on: November 04, 2017, 01:48:20 PM »
I'm betting on that PCB being charred enough between the output terminals that it's now looking more like a smallish resistor than the highly insulating FR-4 that it once was.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 
The following users thanked this post: elgonzo

Offline MisterDiodes

  • Frequent Contributor
  • **
  • Posts: 317
  • Country: us
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #22 on: November 05, 2017, 04:47:47 AM »
That failed power supply board is still pulling juice with no load - so that leaves the possibility that the output cap might have been the victim, not the perp here.  No matter however many design mistakes were made.  Obviously this is a cheap power supply board, with not a lot of design regard for reliability - but it is still interesting to dig deeper to find the -real- cause and use that as another good example of what can go wrong on a board design.

I suggest it's time Dave pull out his FLIR camera, power up that failed supply board briefly (maybe outside to avoid the stench?) and really see what's getting hot... and perhaps that will lead to more clues as to the real failure cause.  There might be something that failed FIRST upstream of that cap?  Or maybe the failed cap caused a second failure??  Let's take a look!

 

Offline Lockon Stratos

  • Contributor
  • Posts: 18
  • Country: hu
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #23 on: November 05, 2017, 05:00:23 AM »
"Nice", and i was planning on getting one from them. After this i think i wait with the purchase until Dave finds out what the hack is happened...

@Dave

Have ben any response from the manufacturer?
 

Offline drussell

  • Frequent Contributor
  • **
  • Posts: 383
  • Country: ca
  • Hardcore Geek
Re: EEVblog #1035 - Flaming DIY Power Supply
« Reply #24 on: November 05, 2017, 06:47:14 AM »
I'm betting on that PCB being charred enough between the output terminals that it's now looking more like a smallish resistor than the highly insulating FR-4 that it once was.

That was my first thought, as well... 

Dave, why didn't you check the resistance between those terminals in the charred area before you powered it up?  I would expect that area to have issues after the cap blew up.  Alternatively, if you did check and it seemed to be OK, why didn't you mention that in the video?
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf