Excellent video! But at 24:18 the Open mode caps are the cheapest, not the most expensive. The 3 stars mean "excellent from a costing perspective", not "a lot of cost".
you had professors at school? wow. here in europe we only have teachers. and at learning facilities where professors teach it doesnt make a big difference because people are going there because they WANT to learn.
>>you had professors at school? we call all of it school in the US >> people are going there because they WANT to learn confirmed, definitely not from the US.
You can want to learn something with a passion and still fall asleep while learning it if the professor teaching it has zero enthusiasm for the subject themselves :/
This was amazingly fascinating!! Probably explains a lot of "I dropped it and it just stopped working" type of failures in modern electronics containing SMDs. Thanks for going through the trouble of doing so much research, and giving us the summary in a format of a well made youtube video. :)
Yeah... Might be. XD Though sometimes it's something else... (well, no duh. But anyway) After dropping it several dozen times my nokia 520, which had otherwise been one of the toughest phones I've ever had (barely even cosmetic damage even from drops on concrete and asphalt), finally gave up the ghost. So... I took it apart. (as you do). Initially couldn't find anything, but the way it worked intermittently led me to investigate the power... Sure enough, one of the 3 battery terminal prongs was loose. Broke the solder joint. Just shows what mechanical force can do. Surprisingly everything else seems intact. I guess the fact that the battery is a heavy, loose component might put extra strain on the connector... I'm sometimes surprised not just by what fails when dropped, but also what survives...
@Kuralthys, very true.. I should have said "explains a lot of drop failures which are NOT due to broken solder joints or traces", to be more clear. :) These three failures probably explain a very large percentage of drop failures. I just didn't know how fragile the ceramic SMD caps are, until now.... A good video.
I fixed my mother's wireless mouse which was dropped and had stopped working. The lovely lead-free solder on the crystal gave way, causing it to flap around in the breeze partially disconnected. Re-soldered and it has worked ever since.
I've been working as a component engineer for a couple years now, so while I have seen this kind of stuff out there, it is nice for someone to put it all together. Interesting stuff.
Splendid having all these hints, tips and tricks in one place! When putting boards through shake&bake testing and HALT, at one employer we often did a final test on our "sound stage", where we attached linear coil audio-band actuators to each mounting post, and activated them in combinations to induce all mounted vibration modes while monitoring for performance deviations due to MLC failures. Found so many issues for one product that we built a sound stage into the final production board test station, so boards were sound tested before any other tests. We used MLCs because of their otherwise great environmental ratings, especially radiation hardness and vacuum tolerance. They're amazing little wonders!
No, it was for friggin' military vehicles. Worse than helicopters when it comes to shock and vibe. Most vendors surrender much of their volume and mass allotment on isolation mounts. Our MEs and EEs, together with our fab house, came up with a new strategy that's still a trade secret over a decade later. The MLCs were the "weakest link" in the system, but we couldn't get rid of them. Fortunately, boards that passed our testing had long lives in the field. But the in-house infant mortality rate was fierce until, one by one, many of the MLC tips and tricks Dave covered were finally incorporated. One trick we tried, that Dave didn't mention, was to put clusters of MLCs on a daughterboard sitting atop a high-density Hirose connector, with the assembly retained with nylon fasteners. It works mechanically and electrically, though the COGS hit is terrible, but possibly acceptable for military and extreme environment systems. It also works for one-off or low-rate boards, particularly prototype and debug boards with BGAs, where it can help get lots of decoupling caps in very close using larger MLC packages.
I remember several fellow EE & Comp E students complaining about having to take a basic mechanical engineering class. This video is a good example of one (out of many) reasons such classes are actually quite useful.
SAI Peregrinus A lot of cross discipline classes tend to be far more useful and transformative than they may at first appear to be. I was a CS major, years ago, and I would have to say my most transformative class was without a doubt a single class in anthropology -- which I took my very first semester, who would have guessed, but it fundamentally changed my perceptions of others and how I relate to other cultures and ideas. But it is always helpful to be exposed to other ideas and disciplines as they can transform how you perceive concepts and ideas as well as how to communicate them with others.
Wow I never really realized or thought that these small SMD caps were so fragile in cracking like they are !!! and the many ways that you can design board layout to reduce the chance of a failure, as well as specify chips designed differently to avoid cracking failure !!! Real INTERESTING, THANKS DAVE !!!!!!
This truly is one of those things you don't learn from standard electronics design textbooks. Serious gotcha, so it's great to see some of this more obscure knowledge for us inexperienced hobbyist designers...
I'm a PCB Design Engineer in the automotive industry. Whenever MLCC's are on any circuit that's connected directly to battery, the capacitors must be in series. Also, the two caps have to be 90 degrees to each other. This is done with the thought that the board will typically only flex along one plane. Most auto companies don't seem to use those Megacaps very often because they're too expensive.
Same here, I can agree to that. I would also add that fexible termination caps are widely used now, almost exclusively. In my experience the "megacaps" are a rarity and are only used in very dense designs where the smaller footprint size for the same capacity is more important than the higher price of these parts.
It is easier to talk about good pcb layout practices than wrong. Ask not electronic fellow of yours to route your PCB and you can easily make a video how bad PCB design can be made.
There's a lot of details you have to know. Yesterday I had finished PCB that takes me six days - and it is not from the scratch. Can you imagine I have it on video and after each of trace tell "Now it,s better to put this signal to layer 3 as in layer 2 is the local ground plate to shield the oscillator"? I think show fails and it's results is more interesting...
I can't believe it, this is exactly the explanation we are looking for, now I know why the capacitors are failing, I just took out a capacitor that shortened my motherboard from my laptop
Back when I used to do component level repair on cell phones, our quick and dirty check for cracked ceramics was to touch the end-cap of the cap with our soldering iron. If the cap was cracked then the end of the cap will stay stuck to the soldering iron (due to surface tension of molten solder).
A standard MLCC that has 'failed open', has a high possibility of becoming 'failed short' over time due to shock/vibration. So, is 'failed open' really a valid failure mode in this case. There is probably a case for 'failed working', where the layers on each side of the crack still line up and make contact, just waiting for the next bump before deciding whether or not to burn the house down.
Excellent video/article. In my repair of the instruments I work on, shorted ceramic caps have been an issue, especially if a drop is suspect. In the old days of through hole, shorted ceramics was as rare as Jen’s teeth (although I’ve seen two in ~36 years of service). In surface mount, shorted ceramics seems rather common. Watch some of those iPhone repair videos (Jessup at iPad Rehab), and shorted ceramics in iPhone 6+ phones seem the most common failure. I fear dropping my phone now even surrounded with a genuine Outter Box case.
Hi Dave, IC Test Development Engineer here. Probably the best video in the latest months, you clearly have lots of experience on PCB design and sharing it is very interesting and formative!
Very interesting and helpful. As you might know, there is a global shortage of ceramic capacitors. We are constantly having to specify new P/Ns as the ones we are using become unavailable or obsolete. Using special technologies such as Megacap or soft termination may be a good solution to the cracking problem, but relying on a single vendor for a special technology in today's market environment just isn't practical at least for us. I think the best mitigation for us is controlling placement of caps in the layout, using smaller size caps or using non ceramic caps where it makes sense.
I suppose you could also increase the pcb thickness to reduce flexing. Of course like all other options this too has its downsides, but it may be worth considering in some circumstances.
great informative video ! one another possible crack cause is wrong setup in pick and place machine. normaly a pnp machine needs to know component height. many operators doesn't care or bother to check different caps for heights. they just use simple template for 0603, 0805 etc. but in reality 100nf/0603 and 1 uf/0603 caps are in different height. so if pnp machine pushes a little bit much that cap to the pcb. that can cause stress cracks or become more sensitive for future stresses on the pcb.
I think your best videos are the instructional ones. This certainly was one of those. Just a note, this started out as a failure (the magic smoke) but it led to a great learning exercise.
Good video! As mentioned, this is learned in the school of “hard knocks” and (rarely) in any Uni. Had a memory card that was randomly causing the memory chips to “loose it”. Root caused it to microcracks in the mlcc decoupling caps.
Damn, do they seriously not teach this stuff in school? I have no formal education in electronics, everything I know is self-taught on the internet, but it seems like this stuff should be taught in schools. Kinda basic stuff that can greatly improve reliability in electronics.
Nope. most people get to their final year before they realise you need a ground reference as well as Rx and Tx connections between one device and another on a floating power supply :/ EDIT: "most" may be a biiiit of an exaggeration, I exaggerate too much :p
No kidding, that's pretty nutty! I guess it's the same with pretty much any kind of formal education. They just teach you what they "need" to teach you.
What may be worth mentioning is that shorting tends to be caused by silver migration through the cracks under voltage. Clearly, SMT MLCC + lead-free solder is a somewhat... _explosive_ ... combination.
Thank You so much for this video. I recently made a electronics project, but it ended up having a dead short. I took off the SMD ceramic capacitor, and it works great!
My rule of thumb: Choose the automotive-qualified parameter if available in parametric search when all the other technical parameters are given. Even one-off toys then have a better chance of working.
Since I didn't see it in the video while skimming through it: A trick we often use with flex circuits is to make cuts around the component in strategic locations to prevent the substrate from bending there. For example you make a U shaped cut around an island where you place a few passive components. It'd be interesting to see a series on flex circuit design actually, especially given the low price for flex in China these days.
I would like to add additional info about this topic, SMD capacitors should not be soldered with hot iron because it can cause thermal cracks inside ceramic structure because of rapid temperature change. Best way is to use reflow oven if you have it, next option is hot air soldering, worst option is hot iron. I think that some SMD MLCC producers highlight this, I know that Epcos did.
6 лет назад
As a design engineer, one very common failure I see on SMT capacitors is mechanical contact (hits). People handle large board assemblies carelessly and hit them to objects and to each other. I see a lot of cracked capacitors on the board perimiter, as they are more vulnerable to being hit.
Thank you for this Dave. We need more awareness on things like this. It is my hope that soft terminations or something even more impressive becomes the gold standard. Sure demand gets driven up immediately and along with that comes price. However, just like side airbags and ABS we may see this in such high demand that output will try to match up to it. In other words, it should become cheaper in the long run. After all, who wants a product that can catch on fire.
One way of reducing mechanical stress is by mounting the capacitor perpendicular to the likely bending forces as Dave showed, that way the bending force is more on the pads than the capacitor itself. Also good design practices suggest adding more mounting points if you have heavy components on a PCB. Also you can get MLCCs in a reversed connection, where the long sides are where the pads are. These are also good for extremely low ESR and inductance. One big trap is also the voltage rating of MLCCs. Some dielectric materials can cause dramatic drops in capacitance dependent on voltage. Also, I don't like those MLCCs with metal wings on them, they are so expensive! I suppose if I was making something for high reliability, it would be worth it. But in a regular consumer product, not so much.
I wonder if board flex from dip soldering processes was ever a problem for ceramics. I was really surprised when I visited a local manufacturer by how much a populated PCB will warp as it's dipped into the solder bath.
we had big problems with Murata's SMD multicap capacitors. Very often they are broken when the device has fallen off. The manufacturer has then installed again individual capacitors in the device
Actually Digi-Key sent me a catalogue item while ago about Samsung soft termination MLCC caps and when I've seen your video about that power supply catching fire I was thinking of letting you know about them. You have firgured it out yourself obviously. Cheers
Hey Dave, you mention changing output voltage from references due to board stress, but it's also worth noting that things like op amp offset voltage can be affected as well. In most general-purpose cases, this isn't too much of an issue, but in very high precision applications, it can be a big deal, especially if you get board flex due to thermals. Good video, and very informative!
I had a similar problem a few years back. The burn glass fiber was conductive and red when power apply. i used a alumina sand blast (pen style sandblaster) to remove all the burn area and fill the hole with regular epoxy i even place back the pad on top of the epoxy and solder the capacitor on the pad. The equipment is still working.
Hi Dave, Great video. At 24:24 you mention that the 3 stars mean the most expensive. The stars are from an engineer's perspective and are meant to represent lowest cost.
For the voltage REF, the cut-out is for thermal stalility, the bandgap might have heating element near it for improved stability, and the cutout is needed to beter thermal isolation,...not mechanical stress
Thermal expansion was my first idea. If a brittle Cerco is mounted direct between two solid Output terminals and the solder cools down it shrinks again and it could pull the capacitor apart. I have the same power supply and I will remove the cap before it burns.
I had an MLCC that failed once. It was used as an output cap. for an LDO. At first I didn't know what was causing the problem to my D.U.T., the power supply limit indicator kept blinking red and I thought the low-side ESD diode was shorting the output to GND. But when I replaced the output cap. of the LDO, everything was hunky-dory again. Turns out the output cap. which failed as a short circuit was causing the unwanted pull-down of the output of the LDO to GND.
Have you seen these same failure modes from tantalum caps? What is your take on the tan caps historical cracking issues (usually through mass reflow) and their impact in a high rel environment? Many of the major suppliers have run life tests that seem to point towards these cracks being insignificant, however some of the destructive analyses we've done shows the cracks propagating to the inner slugs, even prior to mass reflow.
You can also put a fuse (or fusable resistor) in series with the cap. That can be a cheap way to provide protection when you need a large MLCC (for low ESR/ESL with high capacitance or such). Film and electrolytic caps tend to have higher ESR, which can lead to thermal runaway in some cases.
Prehistoricman They certainly tend to be. They do make internally fused tantalum capacitors, though they may be fused more by the buildup of heat in a failing tantalum such that they disconnect internally before combusting. Though in some filter applications some added resistance or inductance could be used as part of the filter design.
Another way to deal with this is in track layout. Use a short, thin copper track to the component. Being short, the resistance will be low. Being thin, it will blow like a fuse in case of short, whilst causing only localised limited damage. Equipment will likely operate afterwards if just de-coupling power supply. Are there good formula for track fuses?
Move them to a point not under such stresses or use a ceramic disc cap vs smd cap for that spot... put extra support for the terminals or put the terminals on a lead not attached to the board...
Very interesting video. But I'd like to note that tht devices almost don't suffer from this issue, even the cheapest ones in which the leads are too thin. Tht devices are also visually identified easily. The board that burned on you was probably hybrid smd/tht already if it was a power supply.
16:02 Apparently the somewhat ropey and by now quite aged capacitor you find connected between live and neutral in old valve amplifiers is affectionately known as the "death capacitor".
I did not understand much. But enough, that i checked the caps in my cars electronic box, because i thought it may be a problem with the vibration and/or flexing of the board and the caps broke ... and it was exactly that. thanks! Now it was not a 600 € repair, it was a ~1€ repair :3
The initial thought is that its a capacitor manufacturer issue ..by the time you add flex , pcb flex , thermal differences , manufacturing process, it has such a disastrous consequence of fire it really has to be dealt with better in production of capacitor such as a flexible factor in ceramic construction or fail safe design. perhaps the caps should have legs like a silicon chip to absorb stress of pcb and also thermal
It is fascinating that 107's exist in MLC SMD form, but THT caps may not be so bad after all. Do machines exist that mount and solder THT automatically?
Great video overall, but I think the choice of 0402 size as example at the end was not the best idea, as to get significant bend over 1 mm of length (lateral size of 0402 is 1 mm) you need to apply a serious force which would probably cause some damage within the board itself too, not to mention that it will surely crack any leadless parts (the likes of QFN or BGAs) so at that point the board will probably be destroyed beyond repair. But this also proves that any board needs to have a fuse next to it's connection to the power source (be it power jack, USB port or whatever else) as a simple fail-safe in case some kind of short does occur. Luckily low-current SMD fuses are fairly cheap, so it shouldn't be a big cost factor. Also it would be a fun experiment to solder MLCC caps of various sizes on test boards and try bending the board to see what happens :)
Structural hint: If you need a connector on the board, consider a straddle-mount on the edge. The spacing and length of pins of some through-hole connectors are amenable to straddle mounting onto surface-mount PCBs. [ Don't expect automagic assembly machinery to cope with soldering the connector.] Twisting of the connector can still impose stressses. Structural trickery by milling slots is to have only narrow bridges for traces between any connectors and the other components on the PCB. The strain (deflection) of the PCB is isolated as the narrow bridges are more flexible and will stretch accordingly with the more rigid part of the PCB on the other side of the bridge. Keep in mind that that extra flexibility means more flexing; which can be an issue in applications subjected to vibration. Some connectors such as Molex's MX123 series convey the intent to (rigidly) support the board *only* by the connector'pins, with the connector housing rigidly connected to the housing. Any strain at the connector due to the harness being (dis)connected or moving otherwise is thus largely contained. The tiny residual strain produces only a small movement of the attached PCB. Environments subject to vibration or shock still require some stress and strain analysis due to the inertia of the PCB and the components soldered onto the board. A thicker PCB substrate can save your bacon.
Thank you Dave!! Something i have seen often in my career. Watching your video made me thinking, would it help if they stack the plates vertically or basically take a normal capacitor mount it sideways? When the board bends, both ends of the capacitor are fixed to the board where the x and y direction motion is close to zero, (unless you are twisting the board) but the z-axis will move with the bend. The movement on the z-axis cracks the capacitor, when the board return to its normal flatness the crack can’t re-align itself therefore it will either be short or open. If mounted sideway or the plates are stack vertically, the movement on the z-axis could still crack the capacitor, but no movement on the x and y to cause the plates mis-align. (again, unless you are twisting the board) Will this help, has anyone tried?
I seen stress failures with SMD components in the old c. 2001 hybrid component large receivers when these were new. One had a SMD resistor in the bias circuit in the middle of a large sparse board between the output and driver transistors. This board had flexure caused by heat differential between the driver transistors and by simply vibrations caused by shipping. High failure out of the box. This resistor would crack open and put max bias on the output transistors and blows the fuse. Replaced with a 1/4 W leaded resistor. Dodgy solder flow opens on that chassis with the regulator and driver components attached to each own heat sinks made the repair often interesting.
Excellent video! But at 24:18 the Open mode caps are the cheapest, not the most expensive. The 3 stars mean "excellent from a costing perspective", not "a lot of cost".
Oh, doh, that makes more sense!
One of the most informative Electronics Engineering videos so far, this is gold info mate!!! =]
Thanks
I thought you like the display of POWER! ?
I've learned more in this 28 minute video then a year in school..thanks Dave..
If only professors would teach with half of the enthusiasm and knowledge as Dave here I'd go back to school tomorrow.
you had professors at school? wow. here in europe we only have teachers. and at learning facilities where professors teach it doesnt make a big difference because people are going there because they WANT to learn.
>>you had professors at school?
we call all of it school in the US
>> people are going there because they WANT to learn
confirmed, definitely not from the US.
You can want to learn something with a passion and still fall asleep while learning it if the professor teaching it has zero enthusiasm for the subject themselves :/
This was amazingly fascinating!! Probably explains a lot of "I dropped it and it just stopped working" type of failures in modern electronics containing SMDs. Thanks for going through the trouble of doing so much research, and giving us the summary in a format of a well made youtube video. :)
Yep, could certainly be one of the reasons why stuff fails.
Yeah... Might be. XD Though sometimes it's something else... (well, no duh. But anyway)
After dropping it several dozen times my nokia 520, which had otherwise been one of the toughest phones I've ever had (barely even cosmetic damage even from drops on concrete and asphalt), finally gave up the ghost.
So... I took it apart. (as you do).
Initially couldn't find anything, but the way it worked intermittently led me to investigate the power...
Sure enough, one of the 3 battery terminal prongs was loose. Broke the solder joint.
Just shows what mechanical force can do. Surprisingly everything else seems intact.
I guess the fact that the battery is a heavy, loose component might put extra strain on the connector...
I'm sometimes surprised not just by what fails when dropped, but also what survives...
@Kuralthys, very true.. I should have said "explains a lot of drop failures which are NOT due to broken solder joints or traces", to be more clear. :)
These three failures probably explain a very large percentage of drop failures. I just didn't know how fragile the ceramic SMD caps are, until now.... A good video.
I fixed my mother's wireless mouse which was dropped and had stopped working. The lovely lead-free solder on the crystal gave way, causing it to flap around in the breeze partially disconnected. Re-soldered and it has worked ever since.
I've been working as a component engineer for a couple years now, so while I have seen this kind of stuff out there, it is nice for someone to put it all together. Interesting stuff.
Splendid having all these hints, tips and tricks in one place!
When putting boards through shake&bake testing and HALT, at one employer we often did a final test on our "sound stage", where we attached linear coil audio-band actuators to each mounting post, and activated them in combinations to induce all mounted vibration modes while monitoring for performance deviations due to MLC failures. Found so many issues for one product that we built a sound stage into the final production board test station, so boards were sound tested before any other tests.
We used MLCs because of their otherwise great environmental ratings, especially radiation hardness and vacuum tolerance. They're amazing little wonders!
Brings back memories!
BobC Space?
No, it was for friggin' military vehicles. Worse than helicopters when it comes to shock and vibe. Most vendors surrender much of their volume and mass allotment on isolation mounts. Our MEs and EEs, together with our fab house, came up with a new strategy that's still a trade secret over a decade later. The MLCs were the "weakest link" in the system, but we couldn't get rid of them. Fortunately, boards that passed our testing had long lives in the field. But the in-house infant mortality rate was fierce until, one by one, many of the MLC tips and tricks Dave covered were finally incorporated.
One trick we tried, that Dave didn't mention, was to put clusters of MLCs on a daughterboard sitting atop a high-density Hirose connector, with the assembly retained with nylon fasteners. It works mechanically and electrically, though the COGS hit is terrible, but possibly acceptable for military and extreme environment systems. It also works for one-off or low-rate boards, particularly prototype and debug boards with BGAs, where it can help get lots of decoupling caps in very close using larger MLC packages.
I remember several fellow EE & Comp E students complaining about having to take a basic mechanical engineering class. This video is a good example of one (out of many) reasons such classes are actually quite useful.
SAI Peregrinus A lot of cross discipline classes tend to be far more useful and transformative than they may at first appear to be. I was a CS major, years ago, and I would have to say my most transformative class was without a doubt a single class in anthropology -- which I took my very first semester, who would have guessed, but it fundamentally changed my perceptions of others and how I relate to other cultures and ideas. But it is always helpful to be exposed to other ideas and disciplines as they can transform how you perceive concepts and ideas as well as how to communicate them with others.
Wow I never really realized or thought that these small SMD caps were so fragile in cracking like they are !!! and the many ways that you can design board layout to reduce the chance of a failure, as well as specify chips designed differently to avoid cracking failure !!! Real INTERESTING, THANKS DAVE !!!!!!
Glad you found it useful
This truly is one of those things you don't learn from standard electronics design textbooks.
Serious gotcha, so it's great to see some of this more obscure knowledge for us inexperienced hobbyist designers...
I'm a PCB Design Engineer in the automotive industry. Whenever MLCC's are on any circuit that's connected directly to battery, the capacitors must be in series. Also, the two caps have to be 90 degrees to each other. This is done with the thought that the board will typically only flex along one plane. Most auto companies don't seem to use those Megacaps very often because they're too expensive.
Same here, I can agree to that. I would also add that fexible termination caps are widely used now, almost exclusively. In my experience the "megacaps" are a rarity and are only used in very dense designs where the smaller footprint size for the same capacity is more important than the higher price of these parts.
Hi Dave. Have you considered making a video showing examples of really bad PCB layout and design?
That could actually be very educational.
he already has
He does, its called Mailbag. haha, I joke. I hope something of mine will show up there soon.
It is easier to talk about good pcb layout practices than wrong. Ask not electronic fellow of yours to route your PCB and you can easily make a video how bad PCB design can be made.
There's a lot of details you have to know. Yesterday I had finished PCB that takes me six days - and it is not from the scratch. Can you imagine I have it on video and after each of trace tell "Now it,s better to put this signal to layer 3 as in layer 2 is the local ground plate to shield the oscillator"?
I think show fails and it's results is more interesting...
I can't believe it, this is exactly the explanation we are looking for, now I know why the capacitors are failing, I just took out a capacitor that shortened my motherboard from my laptop
Back when I used to do component level repair on cell phones, our quick and dirty check for cracked ceramics was to touch the end-cap of the cap with our soldering iron. If the cap was cracked then the end of the cap will stay stuck to the soldering iron (due to surface tension of molten solder).
Yep, that can work a treat.
A standard MLCC that has 'failed open', has a high possibility of becoming 'failed short' over time due to shock/vibration. So, is 'failed open' really a valid failure mode in this case. There is probably a case for 'failed working', where the layers on each side of the crack still line up and make contact, just waiting for the next bump before deciding whether or not to burn the house down.
Stress = load / area , i.e. pressure. The correct term to use is probably Strain, which is change in length / original length.
Nice video, thanks.
next up tantalum capacitors? voltage spikes, current spikes etc. and you get thermal runaway and an explosion
Excellent video/article. In my repair of the instruments I work on, shorted ceramic caps have been an issue, especially if a drop is suspect. In the old days of through hole, shorted ceramics was as rare as Jen’s teeth (although I’ve seen two in ~36 years of service). In surface mount, shorted ceramics seems rather common.
Watch some of those iPhone repair videos (Jessup at iPad Rehab), and shorted ceramics in iPhone 6+ phones seem the most common failure. I fear dropping my phone now even surrounded with a genuine Outter Box case.
Hi Dave, IC Test Development Engineer here.
Probably the best video in the latest months, you clearly have lots of experience on PCB design and sharing it is very interesting and formative!
Very interesting and helpful. As you might know, there is a global shortage of ceramic capacitors. We are constantly having to specify new P/Ns as the ones we are using become unavailable or obsolete. Using special technologies such as Megacap or soft termination may be a good solution to the cracking problem, but relying on a single vendor for a special technology in today's market environment just isn't practical at least for us. I think the best mitigation for us is controlling placement of caps in the layout, using smaller size caps or using non ceramic caps where it makes sense.
I suppose you could also increase the pcb thickness to reduce flexing. Of course like all other options this too has its downsides, but it may be worth considering in some circumstances.
great informative video ! one another possible crack cause is wrong setup in pick and place machine. normaly a pnp machine needs to know component height. many operators doesn't care or bother to check different caps for heights. they just use simple template for 0603, 0805 etc. but in reality 100nf/0603 and 1 uf/0603 caps are in different height. so if pnp machine pushes a little bit much that cap to the pcb. that can cause stress cracks or become more sensitive for future stresses on the pcb.
Now I will be worried about every capacitor...
We lost something when we moved away from all through-hole mounted components.
@@richardsandwell2285 Yes. We lost a lot of wasted space.
And our houses because they burned down
I think your best videos are the instructional ones. This certainly was one of those. Just a note, this started out as a failure (the magic smoke) but it led to a great learning exercise.
Good video! As mentioned, this is learned in the school of “hard knocks” and (rarely) in any Uni.
Had a memory card that was randomly causing the memory chips to “loose it”. Root caused it to microcracks in the mlcc decoupling caps.
“Please excuse the crudity the model, didn’t have time to build it to scale or paint it”
Reminds me of Dr Emmett Brown 🤔
Lol
Be aware, he might have a flux capacitor as well. I hope so.
Damn, do they seriously not teach this stuff in school? I have no formal education in electronics, everything I know is self-taught on the internet, but it seems like this stuff should be taught in schools. Kinda basic stuff that can greatly improve reliability in electronics.
Nope. most people get to their final year before they realise you need a ground reference as well as Rx and Tx connections between one device and another on a floating power supply :/
EDIT: "most" may be a biiiit of an exaggeration, I exaggerate too much :p
No kidding, that's pretty nutty! I guess it's the same with pretty much any kind of formal education. They just teach you what they "need" to teach you.
What may be worth mentioning is that shorting tends to be caused by silver migration through the cracks under voltage. Clearly, SMT MLCC + lead-free solder is a somewhat... _explosive_ ... combination.
This finally explains the isolation slots around big mlcc caps that I see on some lcd panel controllers, tnx Dave!
I've got an idea! Put wire pigtails on both poles of the caps then solder the pigtails to the pads. Maybe I can patent the idea! :-)
Thank you for spending so much time on this in depth discussion of your crack!😈
Awesome information. Not very often someone would run across this kind of issue and know why. Thanks for going into detail, very informative!
Really like this little mini-series :)
Lol, same here
Thank You so much for this video. I recently made a electronics project, but it ended up having a dead short. I took off the SMD ceramic capacitor, and it works great!
Excellent info, thanks!
Dude, this is a ridiculously useful info when designing the pcb layout... thank you so much!
No worries.
My rule of thumb: Choose the automotive-qualified parameter if available in parametric search when all the other technical parameters are given.
Even one-off toys then have a better chance of working.
Great timing as I have been struggling with a product I designed which has had a few failures like this on high voltage / high current supplies.
Since I didn't see it in the video while skimming through it: A trick we often use with flex circuits is to make cuts around the component in strategic locations to prevent the substrate from bending there. For example you make a U shaped cut around an island where you place a few passive components. It'd be interesting to see a series on flex circuit design actually, especially given the low price for flex in China these days.
So much I didn't know, so little time. I do now comprehend Fausto.
I would like to add additional info about this topic, SMD capacitors should not be soldered with hot iron because it can cause thermal cracks inside ceramic structure because of rapid temperature change. Best way is to use reflow oven if you have it, next option is hot air soldering, worst option is hot iron. I think that some SMD MLCC producers highlight this, I know that Epcos did.
As a design engineer, one very common failure I see on SMT capacitors is mechanical contact (hits). People handle large board assemblies carelessly and hit them to objects and to each other. I see a lot of cracked capacitors on the board perimiter, as they are more vulnerable to being hit.
"moldy layer ceramic capacitor" ~ Dave Jones ~ 2017
I suppose that once they crack, mould can grow and create a problem.
Actually, I think the mould solves the problem. That's another technique, a secret one that Dave told us in an unsuspicious way.
“Moldymeter” -Great Scott
Thank you for this Dave. We need more awareness on things like this. It is my hope that soft terminations or something even more impressive becomes the gold standard. Sure demand gets driven up immediately and along with that comes price. However, just like side airbags and ABS we may see this in such high demand that output will try to match up to it. In other words, it should become cheaper in the long run. After all, who wants a product that can catch on fire.
Great Video. First time I came across mechanical stress into a package was with a Switched Capacitor Filter. MF10 around 1984.
Great video, even for pros who have been in the field for many years!
One way of reducing mechanical stress is by mounting the capacitor perpendicular to the likely bending forces as Dave showed, that way the bending force is more on the pads than the capacitor itself. Also good design practices suggest adding more mounting points if you have heavy components on a PCB.
Also you can get MLCCs in a reversed connection, where the long sides are where the pads are. These are also good for extremely low ESR and inductance.
One big trap is also the voltage rating of MLCCs. Some dielectric materials can cause dramatic drops in capacitance dependent on voltage.
Also, I don't like those MLCCs with metal wings on them, they are so expensive! I suppose if I was making something for high reliability, it would be worth it. But in a regular consumer product, not so much.
I wonder if board flex from dip soldering processes was ever a problem for ceramics. I was really surprised when I visited a local manufacturer by how much a populated PCB will warp as it's dipped into the solder bath.
Awesome video. Watching all the way from Kenya.
This is the kind of videos I reeaaaally like. Thanks, David!
we had big problems with Murata's SMD multicap capacitors. Very often they are broken when the device has fallen off. The manufacturer has then installed again individual capacitors in the device
Actually Digi-Key sent me a catalogue item while ago about Samsung soft termination MLCC caps and when I've seen your video about that power supply catching fire I was thinking of letting you know about them. You have firgured it out yourself obviously. Cheers
High value video. Thanks Dave!
The way we find the cracked ones is hot air flow them and push them around with an exacto knife. If the ends move separately, it's cracked.
Remember kids, if it's in two pieces it's cracked
If you can see light through it, the component is probably cracked.
Very informative, thanks Dave! I also really like the new graphics you're using at the start/end of your videos, very modern looking.
Thanks for the invaluable information you present, Dave!
Hey Dave, you mention changing output voltage from references due to board stress, but it's also worth noting that things like op amp offset voltage can be affected as well. In most general-purpose cases, this isn't too much of an issue, but in very high precision applications, it can be a big deal, especially if you get board flex due to thermals. Good video, and very informative!
I had a similar problem a few years back. The burn glass fiber was conductive and red when power apply. i used a alumina sand blast (pen style sandblaster) to remove all the burn area and fill the hole with regular epoxy i even place back the pad on top of the epoxy and solder the capacitor on the pad. The equipment is still working.
I never thought this would get so much critical
This is an excellent reference video! I reckon you should also make similar video about all sort of capacitors and other SMD components.
Learned this 1 Year ago at university
Very interesting, that i learned something that coveres real world Problems
Hi Dave, Great video. At 24:24 you mention that the 3 stars mean the most expensive. The stars are from an engineer's perspective and are meant to represent lowest cost.
For the voltage REF, the cut-out is for thermal stalility, the bandgap might have heating element near it for improved stability, and the cutout is needed to beter thermal isolation,...not mechanical stress
I find this really useful! I would love more stuff like this in the future
Great follow-up of you previous video. You answered all remaining questions I had. Thanks!
Beauty.
Your really getting your value for money out of these modules. :) I'm not complaining mind you.
Thermal expansion was my first idea. If a brittle Cerco is mounted direct between two solid Output terminals and the solder cools down it shrinks again and it could pull the capacitor apart. I have the same power supply and I will remove the cap before it burns.
I had an MLCC that failed once. It was used as an output cap. for an LDO. At first I didn't know what was causing the problem to my D.U.T., the power supply limit indicator kept blinking red and I thought the low-side ESD diode was shorting the output to GND. But when I replaced the output cap. of the LDO, everything was hunky-dory again. Turns out the output cap. which failed as a short circuit was causing the unwanted pull-down of the output of the LDO to GND.
makes me wonder if the longevity of SMD products is going to lead to a giant crisis in a few years!
More to the point; lead-free solder apocalypse just around the corner.
Have you seen these same failure modes from tantalum caps? What is your take on the tan caps historical cracking issues (usually through mass reflow) and their impact in a high rel environment? Many of the major suppliers have run life tests that seem to point towards these cracks being insignificant, however some of the destructive analyses we've done shows the cracks propagating to the inner slugs, even prior to mass reflow.
Thank you for this deep cover of the subject
You can also put a fuse (or fusable resistor) in series with the cap. That can be a cheap way to provide protection when you need a large MLCC (for low ESR/ESL with high capacitance or such). Film and electrolytic caps tend to have higher ESR, which can lead to thermal runaway in some cases.
Wouldn't the fuse be quite detrimental for ESR and ESL?
Prehistoricman They certainly tend to be. They do make internally fused tantalum capacitors, though they may be fused more by the buildup of heat in a failing tantalum such that they disconnect internally before combusting. Though in some filter applications some added resistance or inductance could be used as part of the filter design.
Another way to deal with this is in track layout. Use a short, thin copper track to the component. Being short, the resistance will be low. Being thin, it will blow like a fuse in case of short, whilst causing only localised limited damage. Equipment will likely operate afterwards if just de-coupling power supply. Are there good formula for track fuses?
Thanks Dave. One does not get this from schooling indeed.
Move them to a point not under such stresses or use a ceramic disc cap vs smd cap for that spot... put extra support for the terminals or put the terminals on a lead not attached to the board...
Very interesting video. But I'd like to note that tht devices almost don't suffer from this issue, even the cheapest ones in which the leads are too thin. Tht devices are also visually identified easily. The board that burned on you was probably hybrid smd/tht already if it was a power supply.
16:02 Apparently the somewhat ropey and by now quite aged capacitor you find connected between live and neutral in old valve amplifiers is affectionately known as the "death capacitor".
Bring back through-hole technology ;)
This is really good to know. Thanks! Dave.
"I woffled on long enough about... cracks" LOL
That would be a transverse stress/load you have to worry about as opposed to a longitudinal. Transverse is usually the one you have to worry about.
Great video, very common failure in laptops, !!!!!
I did not understand much. But enough, that i checked the caps in my cars electronic box, because i thought it may be a problem with the vibration and/or flexing of the board and the caps broke ... and it was exactly that. thanks! Now it was not a 600 € repair, it was a ~1€ repair :3
Nice!
The pricing of the 4 models looks reversed! (24:20) The lead-frame model is the cheapest, but handles the most stress (?).
The initial thought is that its a capacitor manufacturer issue ..by the time you add flex , pcb flex , thermal differences , manufacturing process, it has such a disastrous consequence of fire it really has to be dealt with better in production of capacitor such as a flexible factor in ceramic construction or fail safe design. perhaps the caps should have legs like a silicon chip to absorb stress of pcb and also thermal
This is brilliant !
Awesome video ! Love these informative ones
It is fascinating that 107's exist in MLC SMD form, but THT caps may not be so bad after all. Do machines exist that mount and solder THT automatically?
I really like the new intro overlay.
It they can simply ship incomplete boards with through hole capacitors for user to solder
Great video overall, but I think the choice of 0402 size as example at the end was not the best idea, as to get significant bend over 1 mm of length (lateral size of 0402 is 1 mm) you need to apply a serious force which would probably cause some damage within the board itself too, not to mention that it will surely crack any leadless parts (the likes of QFN or BGAs) so at that point the board will probably be destroyed beyond repair. But this also proves that any board needs to have a fuse next to it's connection to the power source (be it power jack, USB port or whatever else) as a simple fail-safe in case some kind of short does occur. Luckily low-current SMD fuses are fairly cheap, so it shouldn't be a big cost factor.
Also it would be a fun experiment to solder MLCC caps of various sizes on test boards and try bending the board to see what happens :)
I might never find this useful, I did find it interesting though! Thanks Dave!
Structural hint: If you need a connector on the board, consider a straddle-mount on the edge. The spacing and length of pins of some through-hole connectors are amenable to straddle mounting onto surface-mount PCBs. [ Don't expect automagic assembly machinery to cope with soldering the connector.] Twisting of the connector can still impose stressses.
Structural trickery by milling slots is to have only narrow bridges for traces between any connectors and the other components on the PCB. The strain (deflection) of the PCB is isolated as the narrow bridges are more flexible and will stretch accordingly with the more rigid part of the PCB on the other side of the bridge. Keep in mind that that extra flexibility means more flexing; which can be an issue in applications subjected to vibration.
Some connectors such as Molex's MX123 series convey the intent to (rigidly) support the board *only* by the connector'pins, with the connector housing rigidly connected to the housing. Any strain at the connector due to the harness being (dis)connected or moving otherwise is thus largely contained. The tiny residual strain produces only a small movement of the attached PCB. Environments subject to vibration or shock still require some stress and strain analysis due to the inertia of the PCB and the components soldered onto the board. A thicker PCB substrate can save your bacon.
very informative video, thanks
In depth. Thank you.
When life gives you lemons, you make lemonade. This is how you take failures and make them a useful occasion.
Thank you Dave!! Something i have seen often in my career. Watching your video made me thinking, would it help if they stack the plates vertically or basically take a normal capacitor mount it sideways? When the board bends, both ends of the capacitor are fixed to the board where the x and y direction motion is close to zero, (unless you are twisting the board) but the z-axis will move with the bend. The movement on the z-axis cracks the capacitor, when the board return to its normal flatness the crack can’t re-align itself therefore it will either be short or open. If mounted sideway or the plates are stack vertically, the movement on the z-axis could still crack the capacitor, but no movement on the x and y to cause the plates mis-align. (again, unless you are twisting the board) Will this help, has anyone tried?
"And well... it ain't there anymore" lmao
I seen stress failures with SMD components in the old c. 2001 hybrid component large receivers when these were new. One had a SMD resistor in the bias circuit in the middle of a large sparse board between the output and driver transistors. This board had flexure caused by heat differential between the driver transistors and by simply vibrations caused by shipping. High failure out of the box. This resistor would crack open and put max bias on the output transistors and blows the fuse. Replaced with a 1/4 W leaded resistor. Dodgy solder flow opens on that chassis with the regulator and driver components attached to each own heat sinks made the repair often interesting.
It's also bad to drop capacitors on the floor when soldering PCB's. Capacitors accidentally dropped should be thrown away.
Very good video! Please, keep this content comming! :)