Magic smoke! What happened to the JK-BMS? Still reliable or bad design?

After we saw some magic smoke from the JK-BMS in the last video ( • The JK-BMS 4S-8S Editi... ), we want to find out what exactly happened and why the BMS exploded... but still works.
After pulling just over 200A for only 5 min, one of the MOSFETs died when the BMS disconnected the current. What a scary moment! I really thought, that this is the end of the BMS!
But hey, it is still working fine and nothing obvious happened. So let's open it up, sniff a bit around and see if we can find any suspicious stuff!
Can the JK-BMS be saved?
Take US$10 off if you spend more than US$100 in the Hankzor store one AliExpress. Use the link down below and the coupon code OFFGRID at checkout.
JK-BMS 4s-8s Edition: off-grid-garag...
Please subscribe and join me on my journey of the Off-Grid Garage!
For more information, please check out my website with links to all the batteries, materials, devices and products I have used in my videos: off-grid-garag...
The 'Buy me a Beer🍻'-link is here (on the website, scroll all the way down):
off-grid-garag... Thank you for your support!
You can also click on the Thanks button under the video to support the channel. Thank you, if you have already!
T-Shirts, caps, mugs and more in our merch-shop:
off-grid-garag...
**** These are the batteries you are looking for ****
🔋➡EVE LF280/304 LiFePO4 cells 3.2V/280+304Ah: off-grid-garag...
🌞➡Solar charge Controllers, Inverters and accessories: off-grid-garag...
🔧➡Tools, Analysers and Testers: off-grid-garag...
📈➡ Battery Tests and Results: off-grid-garag...
🔌➡Batteries, charger and accessories: off-grid-garag...
⚡➡Cables, Inverters and Electronics: off-grid-garag...
Get 5% off when buying Bluetti all-in-one Power Storage Products!
off-grid-garag...
Great "Next Gen" AOLithium Batteries (4000 cycles, Bluetooth, CALB cells)
aolithium.com/...
The Sunfunkit DIY battery. Great product to build your own 12V battery. Use OGGA at checkout to get 6% off. Video Review: • The Sun Fun Kit Experi...
New to AliExpress? Use my referral code below to register and earn US$19 store credit. What a great start! a.aliexpress.c...
Take a live sneak peak at the Off-Grid-Garage in the Victron VRM World:
vrm.victronene...
My other KZbin channels:
/ offgridgarageaustralia

[@PeterVegh1]

Hi Andy, I’m using 2pcs of the same JK-B2A8S20P from JK. According to the specs, it is designed to withstand up to a 350A continuous over-discharge for up to 2 minutes (aka 120 seconds). But for reasons I do not understand, JK delivers them with the over-discharge protection parameter set to 300 seconds. So if you do not change the setting to 120 seconds or less, you will eventually toast your BMS. So I’d say that the problem was not a faulty MOSFET, or a bad design, or ...., it was simply a stupid factory default setting from the factory. The same goes for the cell under temperature protection parameters - the factory settings are -20 with reset at -10 degree Celsius. If you don’t change them, you will ruin your cells eventually if you try to charge them in sub zero conditions. From my experience, it’s a great product, but you cannot relay on factory settings. Greetings to sunny hot Land Down Under!

[@A2an]

You have 2 solution, exchange the FET or leave it be but NOT entirely. If you decide to exchange it, you start by cut the 2 legs of and desolder the legs so you don't pull of the traces from the PCB. Then place som solder on the top of the heatsink, heat it up until you can remove it with a tweezers then clean the surface for the heatsink. Place fluent solder paste, consisting of a lot of small solder ball with a lot of paste, plate it on the PCB where you want to place the FET. Then solder both legs to the PCB so the FET doesn't move while heating the heatsink. Then heat up the heatsink and be ready to push down the FET when the solder start melting, keep pushing down on til the solder is harden. Clean up with isopropyl alcohol. Second option, cut of the to legs and de solder them. It is important that you do remove them since nobody know if it is making a load on your gate driver, if it is driven in parallel with the other FET's, then you might have it conduct between gate, drain or source, so when it is switch off it is still conducting slightly and nobody know how it will develop into the future. PS: don't worry about the temperature, it can take a lot of heat, just don't pass 350°C remember to have a good heating conduction, use solder. Good Luck, what ever you doo 😉

[@howardschlunder9754]

As a general rule, semiconductors are extremely reliable within their design specification envelope and manufactured with extreme precision and very repeatable processes. The MOSFET therefore shouldn't be assumed at fault. Repeat the same 203A disconnect on a new BMS after 5 minutes of self heating and you will very likely have one MOSFET pop again, and most likely it'll be the same or immediately adjacent one that pops. Yours was in the middle of the bank, which means that MOSFET had the most adjacent heaters around it and combined with it's own heating, that MOSFET had the highest temperature and therefore slowest switching speed. Matched switching speed is critical as all the colder/faster ones in parallel switched off slightly faster and started forcing all 203A of load current into the slower ones, and when those slow ones are starting a few degrees hotter initial die temperature, it becomes quite trivial to overshoot the MOSFET thermal limits by multiple orders of magnitude. At microsecond timescales needed for switching, thermal conduction carries no heat away, so it's like getting hit with a 2.5kW laser pulse in a really, really tiny area. You should lower your over current threshold to 160A or less and reduce the delay to ~30 seconds or less to avoid repeats. The MOSFETs chosen on this BMS have a remarkably tiny die relative to their RDSon and voltage specifications, which means they are rather fragile during the massive power dump that can happen at switch on and especially switch off. The 200A 24S JK BMS's use half state of the art/tiny die MOSFETs and half old generation/big die MOSFETs for the very purpose of making the "multi-kilowatt laser pulse" far less powerful by spreading it over a substantially larger silicon area.

[@teardowndan5364]

I doubt it has much to do with heat. I'm suspecting the unlucky FET's avalanche energy (wiring+load inductance on the DC bus) or SOA gets exceeded during turn-off.

[@OffGridGarageAustralia]

Thanks Howard, great explanation. That makes sense. I will repeat the test again with the existing BMS and see what will happen.

[@SVAdAstra]

Howard's assessment sounds entirely plausible. However, lowering the over-current threshold may not help if the BMS has to interrupt a high fault current. Short circuit protection is one of the requirements of a good BMS, and it appears this BMS would interrupt a short circuit only by frying FETs, and then only if the FETs don't fail in a shorted state. "Not fit for service" until proven otherwise.

[@teardowndan5364]

@@SVAdAstra The first two questions to ask are: 1) how big is the input choke in the inverter and 2) does it have a working suitably-sized input protection diode to provide a freewheeling path for the choke current? Under full load, input current may crest as high as 400A and if the inverter has a 1mH input choke, that is up to 80J for the FET array to absorb on turn-off. 4J per FET is enough to blow up just about anything available in typical board-mounted form factors and would almost guarantee at least one more dead FET per full-load turn-off event.

[@williamluke9677]

@@OffGridGarageAustralia Maybe with a thermal camera?

[@PowerPaulAu]

As our testing guinea pig, I think it would be great to run the test exactly the same way again, to see if a second BMS does the same thing. That would help remove the question about if it's a design fault or if it was just an individual component failure. I already have a couple of those BMS's on the way for my own testing, so it'll be interesting to see if any more issues happen in the meantime.

[@OffGridGarageAustralia]

I'll certainly repeat the test with the existing BMS...

[@lexicase8805]

​@OffGridGarageAustralia did you ever end up doing this test? Im very curious to know the result still

[@smak201]

Nice to see the BMS is still working, and we will get new videos about it :). There are many reasons why FETs can die. Could be a faulty component from the factory, but my guess would be something to do with the board design. Probably with the gate drive circuit or snubbers. It could be too weak to turn of the mosfet fast enough and that will generate too much heat. They do not like to be half turn on under such load. Alternatively, the gate drive can be too fast and cause ringing on the gate with stray inductance. That will cause the transistor repeatedly to turn on and off destroying it under load. And so that it wouldn’t be too easy there is also problem with fast switching of high current. The inductance of the cables and loads can generate large transient voltages on the transistor that can be way more than 40 V and cause the transistor fail. It is possible to solder a new transistor there, but it won’t be easy using only a soldering iron. The board is designed to take the heat away from the transistors and will do the same for your iron. You should heat up the whole board on a hotplate near the melting point of the solder and then use hot air (or iron) to take of the broken mosfet. Personally, I would probably leave it like it is. There is plenty of current carrying capacity left for normal use, like you said. Of course, it is possible that other mosfets have failed but less visible. It will let you know under next load test if that is the case, and you will get a nice thumb nail for your video. Good luck for your next experiments!

[@dennis_m]

Hi Andy, great details and diagnosis on the failure. I also think it is not a major failure, if the mosfet failed in short circuit (very common) then it would have been a huge problem. No need to take the mosfet apart but I would highly recommend you disconnect the source and gate legs of the damaged mosfet, just can cut it off or de-solder. This will make sure that no further problems arise in the future. Otherwise there is a risk of the malfunctioning mosfet conducting or passing high voltage back to the gate signal. Thanks.

[@dc1544]

Yes you don't want any arcing to start a fire.

[@russkubes]

I was just going to comment this suggestion as well. If not replace, at least cut the legs off. We don't know how big the gap is from when it blew, or if there's any pieces inside that may become loose and eventually cause a short. Don't just trust the magic smoke to have disconnected the component.

[@upnorthandpersonal]

Compared to the other JK's, the continuous current for the MOSFETs in this one is a bit low. The 100A 48V version for example has them rated at 180A (G035N10N). You can replace the MOSFET if you want with a SMT rework station (hot air). If you wouldn't be at the other side of the world I'd do it for you...

[@denniskenny72]

glad I'm not the only electrician that uses the scientific sniff test :)

[@OffGridGarageAustralia]

That's what you learn in the first month, right?

[@ZecxyPumpkin]

It failed open, just for safety desolder/cut the two legs. You never know what could happen in another surge. Today i bought the same BMS, it's the product i have been waiting for, i can finally remove the balancer, the heating thermostat, and the mobile phone i used to remotely control it. Also i can connect the new bms directly to the Victron portal via the rs485 I have 100A loads maximum so i think im more than safe.

[@kustomzone]

If you de-solder the two legs you'd be able to see if it's glued to the board by lifting it slightly. If that works, continue flexing (up & down slightly) and the very edge of back tab should tear cleanly. Solder new mosfet over the existing tab? (heatsink might require removing it also)

[@MeineEnergiewende]

It does not say how long the pulse can be but the pulse was to long" ...... you made my day Andy :) nice video as always. Thanks a lot.

[@OffGridGarageAustralia]

Thanks Jens! 😂

[@matija3791]

It would be interesting what happens if you try the same testing method on the same BMS multiple times. Does every turn off @200A decrease the number of MOSFETs by one xD?

[@HellTriX]

You can use a hot air soldering to replace the mosfet, they can tolerate the heat for the short time required to solder at those temps.

[@off-gridretreat9293]

I agree with most here. Speculating is good, knowing is better! Since Hankzor gave you the BMS for testing, just continue doing that to narrow down why the MOSFET blew up 1. cut or desolder the two legs of the faulty MOSFET 2. Lower the 200A max setting by 5% (in order to be fair since one MOSFET less now) 3. Run the same test (now 192A with 300sec delay) a few times again to make sure whether the issue was caused by a faulty MOSFET or if it's a design issue (design doesn't match ampere rating and/or delay duration, so MOSFETs overheat, hottest one switches slower and blows) Anyway it shouldn't have happened, seller/manufacturer should take responsibility and exchange faulty BMS if it wasn't for testing purposes.

[@Sprengi86]

Rule of Thumb: If you buy a BMS for a 200A continuous application, get one that is rated for 250A+. It's rarely a good idea to run anything at it's maximum capability continuously.

[@tiloalo]

For sure, but the high current disconnect is supposed to be there to protect the system if something goes wrong. The suicide disconnect is probably not the best safety...

[@evil17]

Ur absolutely right ICHBINS, a good rule of thumb, but as this was a test situation, so I guess we may tend to push things a bit to check manufacturing claims.

[@kustomzone]

For a timed over current cutoff it should be easy to for the BMS to start dialing the current back down prior to disconnect for 30 seconds to level out any temp/current imbalance between mosfets. Any reason this is not standard practice?

[@junkerzn7312]

@@kustomzone The BMS has no ability to dial the current back slowly. You would need a voltage converter to do that. The MOSFET can only be on or off. Any gate voltage that puts it in an intermediate state for too long (probably just a few milliseconds would do it) will dissipate the difference as heat and blow the MOSFET to smithereens.

[@smak201]

@@junkerzn7312 That is the purpose of snubber circuits. I don't know how (or if) it is implemented in the BMS. Snubbers help to reduce switching losses, slows down the current change and absorbed the voltage spikes to protect the switches.

[@retrozmachine1189]

As mentioned by another person inductive spike may be the cause but Andy's comment about load sharing is also possible. It may simply be that the failed FET had slightly different switching points / times compared to the others. It would be interesting to desolder the other FETs and put them through a tester to characterise them to see if they are well matched or not.

[@immrnoidall]

Maybe it's my sick personallity but, that slow mo clip of the magic smoke and you going OHHHHHH. IS SO FUNNY. You need to get that out as a meme. "wait for it... ohhhhhhhh".

[@spoonwzd]

Just a faulty MOSFET which can happen to any device that uses them, not necessarily a sign that JK BMS are bad. I would be tempted to remove the faulty FET before putting the BMS back into use, ideally replace it.

[@dig1035]

Been waiting excited for this follow-up video! Thanks Andy!

[@vin6455]

not fit for purpose...need to have them replace..thanks for great vids Andy

[@bigshotdadz]

My first JK-B2A8S20P unit did something similar when I manually turned off the discharge MOSFETs while drawing less than 20A, except I had burnt traces, shorted transistors and mine was completely dead.. I contacted JiKong, explained the issue in detail, sent photos of the damaged components and they sent me a replacement unit with zero drama. The new one performs flawlessly.

[@millzee60]

May I ask what kind of load you had on it? Was it inductive like an inverter?

[@bigshotdadz]

@@millzee60 Yes, a Xantrex Prosine 3.0

[@OffGridGarageAustralia]

Inverters are particular different for a BMS due to very fast changes in current. But good for testing ;)

[@millzee60]

@@bigshotdadz I'm no expert but I would think a voltage spike would have caused the MOSFET to go short circuit but with all the others turned off it was forced to carry all the current and then burnt out like a fuse.

[@dc1544]

I think you are correct. It was bad from the start and couldn't handle the load. It does show how strong they build the JK BMS so if 1 blows it still works. I would not try to replace it yourself. having them send you a new one so you can redo the test is what I would do.

[@paramax55]

You can replace that MOSFET - no problem. I've done it a bunch. Those things are oven soldered, so the whole board gets to soldering temperature. Take it off just like you were showing. Put a drop of solder on the exposed pad and spread it out. Put flux on the back of the MOSFET and heat it up the same way until the solder melts and squishes out from underneath. It works fine. You can even solder the legs first to hold it in place.

[@740GLE]

Dig more into why JK says their BMS aren’t rated for parallel configurations. Your 16s JK seems to be fine in your rack set up with the other 2 makes of BMS.

[@OffGridGarageAustralia]

I will, video is coming...

[@MartinFreund]

The board has a huge very conducting thermal mass due to the copper planes and the MOSFETs. Preheating will be needed. Use a reflow (hot air) soldering station with tin foil heat shielding or a high-power temp-controlled soldering iron with a large thermal mass. Heat and lift the pins first. After removal apply flux and solder wick to clean the pads.

[@tonydickerson999]

Use Kapton tape to help protect surrounding devices, use might even get away with just using a hot air desoldering gun

[@ugomanfredi2147]

The delay in sec before disconnecting could also be related with temperature: it is possible that for long delays an additional heat dissipator is needed to keep under control the temperature. The default value in secs should be much shorter than 300 sec to avoid overheating

[@pragmaticcosmic2826]

You'll probably find the MOSFETs are arranged in pairs of series back to back(drain to drain) so as there is true isolation when they switch off as this is needed due to the source drain diode within every MOSFET. So this means that there are only 10 MOSFETs effectively sharing the current. Also the RDSOn will be double due to the series connection. The individual MOSFET has to withstand the current squared RDSOn value and this cannot rise faster than the heat can escape from the die else the magic smoke appears.⚡💨

[@volodumurkalunyak4651]

You are slightly wrong. There are in total 40 power MOSFETS across 2 board sides on that BMS, NOT 20 total. 40 MOSFETS in total are connected in 2 drain to drain having 20 of those to share the load, not 10.

[@jackoneil3933]

I'd say your analysis is pretty much on Andy. Lucky I guess the over-current was not a bit lower and it latched closed. As it tripped on cut-off, and from a recent experience I had with a cheap battery spot welder that was blowing Mosfets. I found several videos online saying that if the the gate voltage was low or the gate circuit had residual voltage, some the Mosfets might not switch off at the same time. With 200a of current, I wonder if you could have have got a voltage spike of well over Max VDD, and the weakest of the lot went first. It would be interesting to put a digital storage scope on the Gate and see what voltage looks like when switching, and see clean or noisy it is. From this side of the screen I'm catching a whiff of more than blue smoke, and with a brand new design as a manufacturer, supplier or critical application user, I'd want to do more reliability testing to get a better sense if it's a random failure or a vulnerability. Also, I'd be inclined to test it as is and see how it holds up with the blow Mosfet. If it holds under test then maybe a good design, if it blows or has other issues, then time to re-evaluate and re-calibrate.

[@TrevorFraserAU]

Awesome work mate! Thank you Andy.

[@OffGridGarageAustralia]

No problem 👍

[@geraldhenrickson7472]

Thanks for the explanation. I will still be installing my JK bms as planned.

[@robvandeschepop8595]

First: Thank you for all information you provide each video. Next a question: Are you sure the other 19 mosfets are still ok? They did not explode, but one could just stopped working? Magic smoke is not always provided... Then some nitpicking: talking about a 'mosfet transistor' is double: 'metal-oxide-semiconductor field-effect transistor transistor' Finally: I am waiting for your next video!

[@markobotz2890]

Hi Andy as always a nice vid 🙂 I personally would reduce the delay time from 300 to 180 an that's it and you are probably right with the 19 remaining Moffetts, but you have to remember it when you work with this bms if you go to the max currency. Best regards from Nörvenich

[@jmaus2k]

You can repair using a hot air soldering station. Not that expensive and common for surface mount component repairs. Get some low melt temp solder and have that flow in to move the heat, and out it will come.

[@Steven_Root]

In addition, it is better to isolate other mosfets from the heat by using the kapton tape on them while doing this.

[@roadeycarl]

It's a tricky one. They do advertise the bms to operate at 200A continuous. They do also have a setting for a 5min (not sure why it's that long) delay for over-current protection... based on your test,to me, the bms failed and isn't fit for the purpose of it's specification. I do think it could be an unfortunate malfunction of the mosfet but the bms manufacture should be under-rating their specs so they are confident their product will meet the specification requirements. I'd be very interested to see you repeat the same test!

[@geraldhenrickson7472]

Its not tricky at all. Stuff fails. Especially when the continuous rating is exceeded. Moving on.

[@roadeycarl]

@@geraldhenrickson7472 you're very forgiving! It's great Andy shares these videos so we can all learn and take something away from them.

[@rcinfla9017]

Couple of other observations. Cannot identify a thermistor temp sensor from video. Often this is on a small flex circuit that folds between MOSFET's drain tabs. Not having a temp sensor near MOSFET's would be a significant design deficiency. The 4/8S version in video seems to be missing the BMS turn-on pre-charge surge current limiter MOSFET and series resistors that the 16/24S version has. There are unpopulated pads for it on the PCB at the inside end of MOSFET string. Compare it to your 16S BMS video. This circuit activates for a fraction of msec before parallel MOSFET's string are turned on to allow some inverter capacitor charging to reduce the very high turn-on surge current that can happen. It can actually blow out the metallization and wire bonds on the MOSFET die. They may have slowed down the MOSFET gate drive ramp up to slow the MOSFET turn-on. This method is not as controlled and effective as having the separate pre-charge circuit, but it saves costs.

[@qfour22]

Andy: I definitely enjoy the weather reports in your videos. :-)

[@sebydocky5080]

Hi Andy, great video. You raised an important question : "why some BMS can died when installed in // with some other ones ?" I am planning to add another 16s LFPO with another JKBMS..... Can be an interesting topic for a futur video.

[@mfgxl]

Another reason I have said since the beginning of the first build that cells should be put in parallel, instead of putting batteries in parallel.

[@niclas.lindstrom]

What do we mean by "BMS'es in parallel"? Are we talking about using multiple BMS'es on the same battery string to increase capacity, or are we talking about paralleling battery strings with one BMS each? That would be a huge difference and I can only imagine an issue with the first case.

[@michaelseitz8938]

The first thing I did with my (JK) BMS was replacing the thermal pads with copper/aluminium spacers. It's a shame that JK not only uses thermal pads, but that the pad on one side is extremely thick. The circuit board is well designed and they use fairly good components ... only to completely mess up the thermal design with crappy "rubber insulators"!

[@MrRossi1805]

Great review and very gentle! Good reputation for thie jk bms! Danke

[@rcinfla9017]

Total BMS series resistance at 75 degsC is about 0.45 milliohms which yields about 18 watts of heating at 200 amps. Heatsinking MOSFET's from epoxy side is very poor at transferring heat out of MOSFET's. The flat aluminum plate is not much of a heatsink either. If the system does not take heat away fast enough the MOSFET die can get damaged from excessive heat buildup. Current rating of MOSFET is not the important factor. It is the Rds_ON resistance. Neither the 16/24S or 4/8S JBD active balancer BMS meet their claimed 0.3 milliohm spec, not even at 25 degs C. At 200 amps it should get pretty hot with 18 watts of heating, but BMS MOSFET temp sensor should shut BMS down when it gets too hot. Over-temp sensing should take over regardless of what user puts in for max amperage and max time for over amperage to save BMS from damage.

[@uksa007]

The only way to be sure if its a component failure or a design problem is to run the test again with same BMS and see if more mosfets die, to confirm run the test again with a new BMS that would confirm it. I have a hot air reflow station and could try to replace the mosfet if you have a spare one. I'm local to you. Cut the two legs off the faulty mosfet so it's isolated.

[@FutureSystem738]

As I had already mentioned to Andy, mine has failed after just a few days of running on my work bench at only VERY low power, with just a constant beeping and will now not connect to the App. The first time it reconnected after a total disconnect, but now it will not at all. They’re totally aware of the problem, (which is apparently common), and now I have to send it back to Shenzhen when they will only then fix or replace it. (Postage from Australia is expensive and very likely slow!) They won’t send me a new one- very disappointing as this means a very big delay, probably literally months, in my upgrade to my Rv. I’m not happy. 👎 I would have expected better service (such as an immediate replacement in the mail) for an issue that they already know about in the design, (apparently being an incorrect surface mount resistor.)

[@OffGridGarageAustralia]

Others have reported the same. I'll talk to Hankzor and see what they will do to rectify this.

[@OffGridBasement]

I say put it back together and continue using. I would reduce the max settings in the BMS by 5% to compensate for the loss of the mosfet. Good luck and thanks for the video!

[@matija3791]

Stray inductances of the setup probably sent the MOSFETs into avalanche and this one was the one with the lowest breakdown voltage. Still, if 200A continuous are stated for real applications, then the product must also be able to turn off 200A multiple times without taking damage.

[@roadeycarl]

I completely agree

[@teardowndan5364]

Thing to keep in mind: although the BMS was under 200A DC load, pushing 2500W out of an inverter is 5000W instantaneous at AC peak which would call for ~400A peak at 13V. If the invertner's DC-in filter takes care of half of it, that still leaves 200A peak-to-peak current ripples on the DC bus and the BMS may have disconnected at ~300A.

[@matija3791]

@@teardowndan5364 when you look at the datasheet of the MOSFETs, there is a SOA graph. The MOSFETs can take multiples of 300A (20 MOSFETs) for a short amount of time.

[@teardowndan5364]

@@matija3791 The point I meant to make is simply that the 200A DC readout may have been misleading regarding the amount of current and energy the BMS attempted to break.

[@lnxpro]

hot air station. they are not that expensive and a must have for SMD repairs and some leaded solder. it just flows better and perfectly safe as long as you don't breathe in the soldering fumes daily :)

[@gumpster6]

Thanks for sharing what you found regarding the failure. I think 5 minutes at 200A might be a bit too long (but if the BMS spec says it should hold up that long then it clearly failed). I agree with some others that you could replace the MOSFET with a soldering heat gun. I don't think you'll have much luck with a standard soldering iron. If you plan to leave it in I would snip off the two exposed legs or you may end up with more smoke from the same spot. I would reduce the over current setting by 10A and test away.

[@gregnorman3183]

Hi Andy, I always look forward to your videos! If you're going to test 4s why not look at Electrodacus SBMS0? No MOSFET!

[@SeaMaster101]

There is many articles and videos about running mosfets in parallel. It is not as simple as it seams... there is a difference in the time it takes for different mosfets to turn on or turn off. With simple words - 19 mosfets could have turned off faster from the failed one and it was still on last, thus taking the entire 200+A load which it could not handle and "shit itself" (I liked that description and understand better why the sniff test...)

[@MrSummitville]

The "Pulsed Current" rating is like *570* amps !!! So, yes the last MOSFET to turn off should handle the entire load for a very short time. That is how it works. You need to check the SOA.

[@olivierrossard5705]

If you do not have an hot air station, you should leave it as it is, I guess, thanks for all your video, it's really interesting

[@jerseyjeeper1575]

If you leave it, clip the leads off. To remove it without hot air just turn your iron up put solder on the tip until a molten drop is hanging off and lower it to the soldered tab. The large drop causes a quick heat transfer. Just hold it there while wiggling the package gently with small needle nose pliers. To replace pre tin board and reverse procedure. I would just buy and smd station off eBay.. worth every penny

[@kswis]

I'm not for sure but I've heard mosfets fail in different ways. If it had failed in the closed position I don't think the bms would function, it blew open? I digress thankyou for showing how you looked into it and cheers from the west coast of the states

[@dirkaust5107]

Usually MOSFETs are dieing with a short, that means the safety switch off doesn't work anymore. Repairing the PCB would be a little nightmare, because of the density and the doublesided population

[@OffGridGarageAustralia]

And I don't have the equipment or experience with this. Not the time...

[@SylwerDragon]

Hello Andy. Shortly to answer your concern /questions.. No probably with your soldering iron it is impossible to do ..you need special one that will allow high amount of heat to build on it's thick end point.. also there are special preheating ovens or such that are used for this..ofc you need low melting solder and also similar equipment..I would suggest to cut legs of that mosfet. Secondly as you can see there are 20 mosfets..but they are probably in H bridge configuration..or maybe not..anyway.. it looks like there is no overvoltage protection at all. BEcause current in that point is not important..what you worry that time is voltage spike accors whole board..also reverse voltage should be handled by external fast diode so it would clamp down this voltage..also some low pass filter should be considered to increase time those mosfets are attacked by.. All those mosfets are working in paralen..on one way or another..but still it is more tha capable to handle it all.. But you should Understand that even it say on Mosfet it can handle 100Amps each.it isn't actually the case..it can handle much less .because of package restrictions and also temperature restrictions..also i did notice they are using pretty thick thermo pads..there are not the best for this purpose..

[@lyfandeth]

When you see the rating for any component, remember that is plus or minus a tolerance. For common resistors and capacitors that is often 10%. If the MOSFETs have the same tolerance, that means a "200"A MOSFET can only be relied on for 180A. To rely on the parts, you need to underrate them. And then underrate them AGAIN because the ratings may be at 20°C.

[@tullgutten]

Each side have 20 mosfets but two are back to back so if one is removed or dies then the one paired with it also goes. So the current is shared with 10 pairs of 2 in series. The mosfets have a diode in them so even when off it will always be able to conduct one way, so they have one N channel and one P channel back to back to break the current both ways. I think it might be a little too weak pull down resistor on the mosfet keeping it partly on for a short while and if so there is a HUGE power in the mosfet (current and voltage drop). Mosfets can handle hundreds of amps at full on but at lets say 50% on when turning on or off there is a large voltage drop on them that gives a lot of heat and most of them only handle 1 to 20 amp for a short while. Either the control circuit was to slow turn off pulse or the pull down resistor is to high resistance so it took too long time to turn off

[@paulanderson5022]

Andy, in the profession it's called a 'dark brown smell'! ... 🙂

[@OffGridGarageAustralia]

Dark brown sound weird!

[@paulanderson5022]

@@OffGridGarageAustralia It sounds a bit but you get the message pretty quickly! 😵‍💫

[@AlexanderBurgers]

The mosfets are in back-to-back pairs for charge/discharge, so every mosfet is in series with one other mosfet, and then those pairs are 20 in parallel. You may wonder why, it's because of the body diode of the mosfet, the one mosfet turns off to block current in one way (let's say discharge), but when a charger is connected, the current can still go the other way through the disabled mosfet. It's partner mosfet is positioned the other way around to block current coming from the other direction.. That's how you can control both directions individually.

[@williamhustonrn6160]

Andy, more than likely that mosfet was not making great thermal connection to the heat sink through the thermal pad, they usually only crack and pop like that when the mosfet overheats. The reason it overheated is unknown really if lack of cooling or faulty manufacturing. I would use a hot air soldering to heat the mosfet up enough to flow the solder for removal and hot air flow the new one on. Typically mosfets are good for up to 400c soldering, but many recommend controlled soldering of up to 260c for up to 10 seconds, I’ve seen a few say 300c up to 10 seconds also. I’ve repaired many with just hot air soldering without any issues.

[@OffGridGarageAustralia]

Could have overheated while turning off. If they don't drive it to the close position hard enough the internal resistance increase and therefore the heat. That's my understanding so far.

[@johno186]

I was working a electronics repair shop and I did the sniff test. My coworkers thought I was crazy, but I found the faulty parts quickly. One of my coworkers tried it and was coughing and retching. Seems he was leaning over the board in a test rig sniffing when a cap blew.

[@marcusbinner9656]

Hello. Like others mentioned before the MOSFET failed due to overheating. So surrounding temperature is one factor. Usually for ratings 23°C are used. Other factor which is used is √2... So I personally I would go with a Maximum of 140 A @ 23°C environment and max. 70 minutes before failure. If you get enough heat away then... If the MOSFETS would start @23°C I would set the 200A for max 141 sec. Just some thoughts from my side.

[@stevensteltenpool904]

I guess they use this many mosfets to get the "on" resistance extremely low, and not necessarily current handing. So no problem losing one.

[@teardowndan5364]

If the failure mode is avalanche, he will be losing a FET each time he's unlucky enough for the BMS to turn off under a load that exceeds the weakest remaining FET's avalanche energy.

[@lisathomas8498]

If you are getting a replacement unit anyway Andy, just give this one away as a prize. I’d gladly take it off your hands. 😀

[@OffGridGarageAustralia]

Hahaha, yeah, that would be something to consider...

[@glennsonne9786]

Andy, personally, I'd test it at 190 amps and if it worked fine, I'd leave it alone. I believe that w/o current, soldering this unit will be fine, cooling it as soon as possible and letting it return to room temp before testing... I have soldered a LOT of computer memory boards in telephone switchrooms and never had a problem as long as I didn't "bake it" w/ excessive heat.

[@edwardvanhazendonk]

Hi Andy, you can use miss piggy to do the desoldering.

[@edwardvanhazendonk]

One extra note, when performing the same procedure, please use a breaker at around 300A, if the mosfet goes bananas you are covered when it short circuits, that would give a bang, now it went into high impedance, a short circuit is also possible which is very bad without a breaker

[@davestech6357]

You could replace the fet with your soldering iron and heat gun together. The problem is you cant do it fast enough without disturbing the fet on the other side of the board without proper equipment. Just cut the gate and source off and you will be fine. The failure might be because it did not turn off as fast as the others.

[@OffGridGarageAustralia]

Yea, I'm worried about this as well. It is not easily done as some suggest here. I probably just leave it in and cut the pins of the broken FET.

[@DG3ACM-Marc]

Yepp, I think it might have been a FET produced at the border of his specs and it probably cut off slower then the others. Hang up the two legs and have fun testing with the other FETs. Solder will probably only be possible if you are using a solder iron which is normally used for rain gutters. I can`t wait to see more about the 4S JK because I plan to setup a 280Ah 4S for our RV.

[@leocostantino3184]

Did you ever set up that 280ah 4s for your rv? I'm setting up the same ah and bms and was wondering how yours is performing.

[@trevortrevortsr2]

It looks like that mosfet latched momentary or lagged slightly closing - it would only have to have been for a few micro seconds and all the current the others were taking would have surged through that poor mosfet, killing it - mosfets used in parallel to take high current have to be well matched so their on resistance is the same - if that mosfet was from a different batch or its gate had slightly higher capacitance it could be enough - I would cut off the 2 mosfet legs so the damaged mosfet is isolated and continue to use the BMS perhaps de=rating it to 180amp

[@EngelOfWar1]

That's my thought too - that one was simply slower than the others when switching gears and that's why the entire voltage flowed through it before it "secured" itself.

[@trevortrevortsr2]

@@EngelOfWar1Howard raises an interesting issue - its possible that as the mosfet was towards the middle of the pack it ran slightly hotter and that slowed the switching speed compared to the outer edge mosfets - heat being energy would have inadvertently excited the semi conductor slightly causing it to conduct and cause an avalanche effect as the current surged through

[@dig1035]

Quality control, Send it back, test new one! Version 1? It did magic smoke but kept working, wow!

[@wayne8113]

Thanks Andy

[@EddywaKenya]

Have been around test equipment and components. The weak link in componnts like mosfet is the legs. Too small. Cant hold that much current. Compare a 300Amp fuse to one leg of a mosfet and you will see the problem.

[@anthonyevans7911]

Great video. In this scenario, where the BMs controls the cut off via its own settings, why then cany it not be programmed to ramp down the power from 200A to 50 or 10 amps over the time of 30 seconds then disconnect at the lower current? Id assuem that software change would solve that issue. However, if you have external cut off relays to cut the load, or you had a power outage / trip, then of course the bms cannot control that scenario and the mosfet will have to endure that high current shut off. Id much rather control the shut off via software of possible, cold be easily achieved.

[@MrSummitville]

Explain how you are going to "magically" ramp down the current?

[@gelecopter]

in my personal opinion this is mosfet driving issue. If MOSFET stays in its linear region for too long it will overheat and die. To put theory aside - you have to forcibly close them by applying negative voltage to the Gate (for NchMOSFET) ideally or at least short the Gate electrode to Source. Gate electrode are isolated from the rest of the transistor and have significant capacitance which charge must be discharged to close transistor. As for desoldering - better use hot air or hot plate and soldering iron. Preheat with hot air (max flow and about 200C) then use soldering iron with chisel tip to desolder the FET. Just crank soldering iron up to about 400C and be careful with burned leg - it can came off along with PCB trace. As for soldering - its the same but whick old solder and apply new 63/37 to the Tab pad (while keeping other two pads flat w/o solder) press new transistor with soldering iron (again 350-400C and wide tip) on its tab then two other terminals and voi là!

[@SeaMaster101]

I second that!

[@junkerzn7312]

Yup

[@jasondevine6014]

Yes faulty fet or a inductive spike. Not temp or even current.

[@normalminimal2647]

We will know for sure If only one MOSFET was the problem, when you run the 200A disconnect test again :-D

[@OffGridGarageAustralia]

I will!

[@MR-ub6sq]

I'm thinking that it's possible that one of the mosfet "exploded" was caused by the second of the two connectors you connected to the B port. When excessive current heats up the connector, it expands, causing its tip to overshoot to give an uncontrolled overshoot via air directly from the connector tip to the mosfet leg.

[@OffGridGarageAustralia]

Thank you.

[@repairman2be250]

Specs from last video rated it at 200A continuous and 350A for 2 minutes. Your setting was at 200A discharge and time was set to 5 minutes (300). So you went only a few Amps over the 200A mark. Makes me think even if your time was set to a few seconds instead of 5 minutes this still could have happen. Mosfet's in parallel is not a simple thing to do achieve. Component tolerance might have been a big factor here. Or it was simply a weak mosfet. If one needed a BMS that needs to handle 200A continuous, a 200A rated device is not good enough. As you said cheap chinese, though JK BMS as you have proved already are a very good and with lots features, you take the rating down 25 - 50%. Disconnect the faulty mosfet from the circuit.

[@SailingCatamaranElement]

Looks like the JK BMS app would benefit from a system/component status check, that runs a quick diagnostic on the board to make sure all components are fully functional - especially if the BMS is able to continue running with components damaged. Detecting a blown mosfet will prove to be quite difficult when my BMSs are secured within production style plastic battery cases on my sailboat.

[@OffGridGarageAustralia]

I don't think it is possible to detect a faulty FET if it is in parallel with 19 others. They don't provide any kind of control circuit or so... If the BMS is packed away, you won't even noticed if something like this happens.

[@timeconomu7395]

Pretty sure after sleeping on it...probably not an overcurrent event. If so the whole unit would have been quite warm if not hot. I'm leaning more overvoltage. Yeah I hear you, the system voltage was too low for an overvoltage event, but maybe not. Only way to prove overvoltage is to do the test again, disconnect at 200 amps and put a scope triggered at 40V, which is the VDS of those MOSFETs, connected to the MOSFET drain. My guess is that the cabling inductance created a very high voltage spike and that you should probably improve your cable routing. If you look at the current each of those MOSFET's carry, at 200 amps, each of them "share" about 20 amps each. Not enough to blow any one of them. t

[@davidpang8773]

I have seen a diode which already crakcked but still works perfectly fine. I found it in a welding machine, and I think this situation can happen among MOSFETs.

[@MrSqueegey]

The old saying if it aint broke dont fix it,well it is broke,but still working.keep running the bms til it finally gives out.

[@williamfairley2502]

Andy my new jk 8s bms ran OK for a week then stopped working , in started saying cell count was wrong then it would not connect to the app . It is being returned to Hankzor store for replacement. The older 8s jk bms that I purchased 6 weeks ago still works ok

[@ugomanfredi2147]

The possible causes of failure are 1) Defective synchronization of parallel MOSFET, leaving a few active before disconnection, and thus bearing a very high current 2) Inductive spike 3) a mix of the two. The heating due to the long disconnection period, could make the MOSFET more sensible to avalanche damage and less resistant to voltage spike. If the circuit below BMS is inductive, it seems very easy, starting with 55V of battery voltage, to have a spike that exceeds 100V and can perforate the MOSFET, the last one interrupting the load. An inductive spike, could extend its damage also to a connected inverter, if it is not adequately protected from spikes in its battery input. To get rid of inductive spikes possible damage, on BMS side and downstream to the inverter side, the best solution seems to have a surge protection device (SPD) attached in between. DC SPD exist even rated at low voltage, like 48V or 60V. In case of significant spike, the SPD will cut it to tolerable levels protecting both sides of the electronic.

[@vaneay]

you got lucky the mosfet fail open

[@OffGridGarageAustralia]

We could have seen bigger smoke otherwise.

[@pragmaticcosmic2826]

You need a hot air rework station to replace the MOSFET.

[@chuxxsss]

Andy, with the pulse load, check out the pulse timing diagram mate. You need a soldering air heat gun. Plus some solder paste. Plus some alcohol Plus toothbrush to clean afterwards.

[@OffGridGarageAustralia]

I need alcohol for sure 🍺

[@chuxxsss]

@@OffGridGarageAustralia It's pure alcohol so pop it in an orange for a week. ;)

[@gregb1714]

Well, other mosfets could have been damaged/overstressed even though they work ok now. Some may have a shorter lifespan now. I would cut the 2 leads on the bad one and be done with it. If you want to replace it, a hot air gun would be the way to go.

[@Trent28888]

Ignore the pulse current,its a charger and ya gotta remember they are in parallel,it was a unmatched mosfet and the main problem is that they dont have any proper cooling cause the metal tab is used for it not the front so it got to TJmax and then some

[@garys-half-baked-offgrid-dream]

Charge point boardom busting again thanks Andy.

[@Trent28888]

Ignore the pulse current,its a charger and ya gotta remember they are in parallel,it was a unmatched mosfet and the main problem is that they dont have any proper cooling cause the metal tab is used for it not the front so it got to TJmax and then some

[@rklauco]

I would not repair the mosfet. This just happens, the board is overengineered a bit anyhow, so no big deal. I just realized - disconnecting 200A by such a small component, that's crazy.

[@OffGridGarageAustralia]

Yeah, I'll disconnect the legs and leave it on.

[@guntherdrehsen]

All electronic components have minimal differences even if they are from the same batch. Heating up the components will spread the differences. Also the max. ratings will lower if heated up. As you said in the video it seems like the burnt chip was the slowest when turning off the current. A normal fault would have been a short circuit. Well, but at 200A there was no chance to keep a short, so the internals of this chip have just been vaporized and been blown through the crack on the side to the other chip where you have found them. The BMS should still be usable at a lower current. To cut off the two pins might be a good idea, but i think it is not necassary since the internals have already been blown out. It is always a bad idea to switch off high currents. This should only be done in a case of emergency. Overcurrent protection is just a safety feature and nothing for controlling current. After this relays have burnt contacts and as you have shown MOSFETs might be broken and/or shorted if they have to switch off a current higher than they are designed for. ⚠ To replace the broken chip with a normal soldering iron might be possible but is not easy, I wouldn't do it. To heat up the big ground plane fast enough to desolder the broken chip without heating up the other components too much is quite a challenge. Soldering the new component is a little bit easier since you can apply a little pressure on the chip while heating up the solder. If it sinks in a bit you're done. As others have mentioned, you should let someone with a SMT rework station do the job. 😉

[@centerrightproudamerican5727]

A few points: 1) It is possible that this is a one-off failure due to a bad FET, not a fundamental design flaw, but it is still a black mark on the BMS. That should not have happened since it was operating within the advertised parameters of the system. 2) Overall, I am disappointed in the failure, but it still seems like a good BMS. 3) I would not bother with trying to change out the failed FET. Trying to change it is just as likely to make things worse as to make things better. 4) I would not design a system that ran any BMS at its full rating. I like to have at least a 25% margin between the rated current and the max it will be run at. The target sustained current for a 200A BMS should be closer to 150A *or less*. However, If the problem was a bad/slow FET, the problem might have happened even at 150A. (This is *not* a ding on what Andy did. He drove the product to its specified limits as part of his evaluation for us, his channel viewers. He even said in the video he would not normally be running it that high)

[@teardowndan5364]

If there is too much inductance in the wiring + load, the FETs' SOA or avalanche energy can be exceeded by the installation and blow perfectly good FETs too. In that case, you can expect to blow FETs on a regular basis from disconnecting under heavy load. Keep in mind that a 2500W inverter operating from 13V will have 50-100A of ripple current on the DC bus, so the likelihood of blowing FETs will be a function of how close to peak current the disconnect occurred.

[@OffGridGarageAustralia]

Fully agree with all 4 points. I'm the same.

[@offpro]

JK BMS has an RS485 port. How about sharing the information between your BMS to raspberry pi? In a perfect world, the Shunt would not be necessary anymore, you could read cell information / warnings / health out of reach of BT, the BMS will get an active part of the PMS and much much more. I can see why you do not have to worry about a relatively "simple" installation with only solar energy and have low C-rates. But in an RV or a boat that has charging from solar, wind, shore connection, generator and alternator, and often several at the same time, this will be a valuable feature. Getting the RPI to read the information is easy, but being able to process and integrate the data is perhaps the big challenge?

[@Chris-hy6jy]

Might be a good idea to cut the source and gate legs of that MOSFET just in case it starts to short out in the future.

[@xx3stksm459]

I guess "magic smoke" was triggered by an overcurrent shut down(both happened simultaneously!). As long as continuous current flow goes through MOSFETs, 200 amps are divided by 20 MOSFETs, where you can have a safe operation(no smoke). In case of an overcurrent shut down, an unbalanced cutoff time between MOSFETs can have the situation where only one MOSFET has to carry everything(200A), even if in a tiny duration. My hypothesis is an unfortunate cutoff time difference between MOSFETs is criminal. The broken MOSFET's source pin pad apparently has a slightly different color. That can be the footprint of 200A current. If I design a BMS with a max of 200amps, I definitely do several tests to see if MOSFETs are safe in such a high current. But cutting off is a little bit of a blind area. I suspect an unbalanced cutoff time can be fatal only if a very high current (more than 200A?) is used. 😅 I don't recommend removing the broken MOSFET with a heat air gun because recent PCB uses non-lead solder, which has a high melting temperature. It's not easy for a beginner to have a successful job. In my opinion, cutting off the gate and source pins are enough for safe operation.

[@D9ID9I]

It might be the issue with the MOSFET driver and developers need to find better one.

[@GregOnSummit]

Try a big 'ol 300 watt soldering iron with a chisel point. I have an old craftsman one that will melt steel .... Not really, but it puts out a lot of heat in a fairly large area.

[@glottis512]

Excellent content of your channel! Even if i live on oposite of the earth :) I have a suggestion for another test. BMS with mosfets have a one problem. Becouse a mosfets have a body diod by design, you need two mosfets in series to switch of current in both directions. And this is interresting point. At example i switched off charging, i can discharge but current is flowing through one mosfet (ON - mili ohms) and one diode (maybe 1 volt drop). With discharge current 100A and 1V drop = 100W of lossis. Another interesting point is oposit configuration. Charge is ON and discharge OFF. 100A, 100W of heat but 1V lower voltage on battery for charger. For right voltage you need smartshunt between battery and BMS and export battery voltage to mppt or BMS comunicating with mppt. This is theory. You can proove it. Or maybe you already done this?

[@glottis512]

One more thing. This could be the reason of so manny parallel mosfets. Spread this waste heat to bigger surface

[@tiloalo]

I think it might be a design issue, the last MOSFET to close takes 200A and dies... You could try again with 190A limit and see what happen when the BMS disconnect

[@OffGridGarageAustralia]

I'll try again with 205A as before and see what happens.

[@radod6147]

I'm not Chinese, but you can't put a blame on the manufacturer. There are many products of very good quality coming out from China as well.

[@OffGridGarageAustralia]

Hence I said, it's not Hankzor's fault.

[@excillisbank2611]

Oui je suis d'accord avec toi ANDY, c'est juste une coïncidence malheureuse d'un MOSFET avec une fabrication sensible. c'est meme extraordinaire car moi j'ai déjà eu la meme situation avec un onduleur et je peux te dire que tout les mofsets de mon onduleur on brulés ! je pense qu'avec un déssoudeur électronique à AIR CHAUD cela devrait peu être possible de changé le mosfet.

[@krg038]

I would leave the mosfet alone. You were using it in a test. Probably will not use normally as in the test condition (I'm guessing). Great information on how the mosfets/bms works