Active pre-charge/inrush control

Only the first fet is shorted, the 0.5 ohms didn't save it. Maybe the second fet didn't turn on, but even that shouldn't kill it. M1 is rated 160 amps continuous, should switch in a few microseconds, well before the SOA says there's trouble. Peak possible current was 216 amps with the 108V pack I used. Looking back, I've never hard switched into the inverter without failure, wonder what is happening, could there be a high voltage kickback spike? I rushed the testing, didn't even turn on the scope, traveling soon, wanted to try it first. I'll go much more slowly for round two, maybe put several ohms in series until things look good. Might look at waveforms with a contactor switching the inverter too but with some series R to prevent damage. Still don't see my gate drive as a weakness, but I've been wrong before...
 
I was thinking a larger resistance might be good, at least for my application where it's mainly the capacitors that are the load.

Your gate drive is probably OK but always nice to measure with a scope.

Is there any way to delay the inverter startup so the caps get charged before it tries to turn on?
 
I haven't considered all possibilities but I think a low resistance is needed so that if the output is shorted the time spent at high current is quick, same as the time to charge the load through low resistance. This is to stay within the SOA with high current for a short time. This circuit uses load voltage rise versus time to detect a fault. A more sophisticated fault detection would measure or simulate fet dissipation, but that gets complicated and needs to be very fast.

I can just make sure that the inverter is switched off, but my goal is a "foolproof" switchless connection, so there is no possible wrong operation sequence. Remains to be seen if that's actually possible.
 
There might be a way to insert a delay timer into the inverter itself so it doesn't start running for a second or so after getting power.

And while you circuit may have failed for your application, I think the general topology will be good for typical ebike controllers.
 
Finally got back to troubleshooting my crispy critter guinea pig Hot Box. Found a major error caused by the PC layout software. The footprint that I used for the PNP transistors is wrong, CBE instead of EBC, Q3, Q6 and Q8. That's why the protection was triggered, probably why it blew up too. We'll see if that's all it takes to make it work soon.
https://endless-sphere.com/forums/download/file.php?id=230174
 
KeithT said:
The footprint that I used for the PNP transistors is wrong, CBE instead of EBC, Q3, Q6 and Q8. That's why the protection was triggered, probably why it blew up too.

I hate it when that happens... I'll be interested in the results.

I took a stab at designing at a similar circuit using a FAN3224 dual gate driver chip. Each section has a RC delay on the input to give enough time for the 12v regulator to get up to voltage. These drivers also have an inhibit input that can be used to prevent the main switch from coming on if the precharge fails to bring the load up. I put a 1 ohm in as a precharge resistor, which should be able to survive a small constant load in addition to precharging the caps. These driver chips also have a UVLO, so during power down, the gates get turned off when the 12v supply drops to the trigger point. The 12v regulator is not drawn out, but it can be a simple linear regulator as the driver chip quiescent current is quite low.

Automatic Precharge 5 prelim.jpg
 
So my crispy guinea pig is now black and charred. I got my circuit working properly and increased the load in steps. No problem driving a 500W halogen with inrush. The switch turned it on and off reliably and there was no arcing on connect or disconnect. I then connected the inverter through a 4 ohm series resistor. Again, no problem, the inverter charged and worked with the series resistor to limit peak current to around 30 amps. I dropped the series resistor to 2 ohms and things seemed ok but then the inverter started acting strangely. It pulsed on and off about once per second, beeping it's low voltage alarm. Apparently the series resistance made it unstable. So I connected it directly and switched it on and off a few times until I heard "that sound", this time from the inverter, and smelled burnt parts. At the same time I saw that I could not turn off my circuit. So now I have another blown first fet, and my inverter is also fried. It has 8 10 amp fuses, four are blown, and one arced and burned the circuit board. Looks like two of the 8 main fets are gone, it uses FDA50N50's, a 500V 48A part.

So I'll try to repair the inverter and think again about how to make this idea work. Maybe a three stage precharge drive is needed. I have no doubt that the inverter is more than just a capacitive load, there's something else happening that's killing the first fet. But how to handle that? The linear ramp circuit works fine with the inverter but so far no hard switched fet can take it. I could give up, but I probably won't. I don't think the inverter is much different to drive than a motor controller but who knows. I do think things get much more difficult over 100 volts. And a long slow precharge followed by hard switching may be the only way. But that isn't easy with a simple hot connection.

That gate drive chip does handle a lot of details nicely. The handoff from the precharge fet to the main ones is fixed by time constants rather than voltage dependent like my circuit. That's probably fine if the load is known and unchanging. I hope your luck is better than mine! By the way I found two four leaf clovers this morning, didn't help.
 
Fried_guinea_pig2.jpg

Fried guinea pig award hasn't been seen for many years, but you earned it!

Bummer on the smoke release. If the UPS failed and shorted, it may be the reason the switch circuit blew up. Maybe something I'm not understanding, but if it worked with a 4 ohm resistor, why not stick with that? Your application is pretty extreme compared to a typical motor controller.

On my circuit, it would be possible to trigger the main switch by sensing the precharge has finished. The inhibit feature, which is the transistor part, would prevent the main switch from coming on if the precharge failed to bring the load up to voltage. I could just use that instead of the RC delay. I think they have versions of the gate driver with inverting inputs that might eliminate the need for the transistor. The initial time delay is intended to allow hot plugging the circuit with the switch on so it has enough time for the connector to get fully seated.
 
Looks yummy, not! I considered posting a pic like that after the previous round. Hard to say what sequence the failures occurred. Post mortem on the inverter found a few things. Only one fet failed, I think. How many to replace is never clear, I'll do at least two. Weird design. I was surprised when I bought it that it came with spare fuses, 8 of the typical 10amp red automotive type. Well they have those fuses connected in parallel (!) in four pairs. And of course they are not rated for 120v DC which has a remarkable tendency to arc and burn metal that I now know really, really well. So it will take some doing to replace the fuse connectors that vaporized. Bottom of the board was not pretty either, sloppy soldering and flux that was not removed. Ah, you get what you pay for. On the other hand I haven't found any other inverter that has this input voltage 90-120v range. So if the apocalypse happens soon, I'll only have DC for light and heat, no 60Hz for a while...

No way to run the inverter with any series resistance, too much voltage drop to be usable, it needs over 20 amps at full power. I did a really careful analysis of the hand off to balance the heating between the two fets, letting the first handle the voltage with resistor limited current and then shorting it with the second fet getting about the same current surge but with less voltage. All very nice in the simulated world.
 
So there is some hope for my poor fried critter. I blew another first fet today, had no more spares so I replaced it with a larger type that I did have. So far, the new fet has passed every test. Is this the solution? Well, I still don't have an inverter that works. But, I've connected the HotBox to a variety of loads that I do have and it now handles them as expected, no arcs, no sparks and no blown fets. I don't have enough capacitors to match the inverter load of 2500uF, but I have connected 1500uF and switched it many times, in parallel with a light bulb so I can tell it is working, discharge the cap quickly and get a good inrush too. A friend loaned me a really big capacitor, 8,000uF so I tried that. As Spice predicted, that trips the protection circuit because it can't charge fast enough through 0.5 ohms at more than 200 amps. So that part of the circuit has been tested and passed. And since the large capacitor is almost the same as connecting with a short circuit load, I tried that too. It worked! Red light, no smoke, all good.

Today's fet failure happened while I was using the switch. Possibly the current spike was happening too often for the junction to cool off, losing some SOA. It was working fine during connect and disconnect. But the new fet seems fine with the switch or the sequential connector. I replaced IXFH160N15T2 with IXFK220N17T2. That gave me this possibly important improvement in SOA.
 

Attachments

  • 220-160-SOA.PNG
    220-160-SOA.PNG
    16.4 KB · Views: 2,540
I'm used to devices in the under-100v range with very low Rds.
The IXFK220N17T2 is pretty good for a 170V rated part.


IXFK220N17T2 SOA.JPG
 
I just ordered a similar mosfet, but with lower max voltage and higher Rdson. More expensive but maybe worth it. How can a part with lower performance cost more? SOA. This is a Polar rather than Trench design. The market pressure for lower on resistance has led to failures due to the Spirito effect at high voltage and current. Since the pre-charge fet doesn't need low resistance at all, this should improve reliability significantly. Here's a comparison of my first, second and now third fet choice. Green is the 220N15P(ordered), red is the 220N17T(tested and works so far), and blue is the 160N15T that I have now blown four times. The Polar Fet 100uS rated current at 100V is more than 60 amps higher. I will need to be sure that the voltage rating is not exceeded by something weird from the inverter, but that seems very unlikely. Since the second fet doesn't turn on until the voltage is low, a trench design with very low on resistance is ideal. These are expensive fets, but that seems to be required for this application. 220PT-SOA.png
 
Manny said:
I used a linear tech part for my inrush limiter. cant recall the part number right now.
it a chip controlling a low side mosfet with a current sense resistor. Used a constant current source ic to supply the LT part because it uses a internal zener. could post a schematic if there is interest.

Manny

Hey Manny, I think your post got overlooked, but I'm guessing many of us would love to see your schematic for this. Can you also give a little more detail on the application and how it's working for you? Thanks!
 
in my build log there is some more info. https://endless-sphere.com/forums/viewtopic.php?f=12&t=83316#p1291037

so the chip is a LTC4252 Hotswap controller
used a LR8 as a constant current soure to power the LTC4252
the current source is better than a fix resistor for use with a big voltage range.
 
Thanks so much for posting the add'l info Manny! Been trying to figure out if one of the LTC ICs would make it a whole lot simpler to design/build a reliable pre-charge/inrush control. Glad to see you've had success with it.

Any chance you have a schematic you could share? I'm not sure my circuit design skills are sufficient to sort it out on my own...
 
Hello

I have built the "inrush limiter 3". It's quite simple and works just well. Thanks to Richard Fechter.
Now, I would like to use the same circuit to connect it to a buck/boost converter. So imagine
just replacing the Ebike controller by the converter. By the way, it's for an indoor project, not an ebike.

My problem: as it is now, the converter kills the MOSFET, probably because of inductors generating high voltage. MOSFET is rated up to 100V.

MOSFET: IRFB4110
Battery: ~48V, LIPO
Converter: dph5005 (can't find any schematics)

Just wondering if there is an easy solution to this problem.
Thanks for your help.

Walter
 

Attachments

  • schema_inrush_limiter3.PNG
    schema_inrush_limiter3.PNG
    29.7 KB · Views: 2,153
rachdatu said:
My problem: as it is now, the converter kills the MOSFET, probably because of inductors generating high voltage. MOSFET is rated up to 100V.

Do you know if it blows when you turn on or turn off? The problem might be the converter is trying to draw a bunch of current during turn-on and not just charging the caps. There may be a way to put the converter into a 'standby mode' during precharge. Just a short delay before it actually turns on.

Another approach is to just slam the FETs on hard so there is no precharge. This may blow up the FETs too if the caps are too big.
A two stage setup where the caps are precharged through a resistor, then the main switch turns on would probably be the best. The precharge resistor can be a pretty low value, but enough to keep the precharge FET from blowing. Something like 1 ohm.
 
fechter said:
Another approach is to just slam the FETs on hard so there is no precharge. This may blow up the FETs too if the caps are too big.
There is likely an intermediate step (very rapid turn-on) that will avoid both overcurrent due to the precharge spike and overheat from long term ohmic dissipation. Reducing C1 and/or R1 would allow such an experiment. (Existing values will give you about 100ms, which is a long time from a thermal perspective.)
 
billvon said:
There is likely an intermediate step (very rapid turn-on) that will avoid both overcurrent due to the precharge spike and overheat from long term ohmic dissipation. Reducing C1 and/or R1 would allow such an experiment. (Existing values will give you about 100ms, which is a long time from a thermal perspective.)

That might work too. Might be worth trying about 10x less capacitance to speed things up.
 
Ok, Thanks for your suggestions.
I will have a look at that in a couple of days. I still have 10 MOSFETs (blew only 2!) and they are not expensive...
Tomorrow, it's the Swiss National Day. I will take some fresh air up in the mountains.
Walter
 
I replaced C1 with 0.1 micro; it still doesn't work.
Switched it ON --> OK, then I could not switch it OFF; measured the resistance between D and S: 250 Ohm.

I still need to test with the small resistor when I will find one.

In the meantime, I am going to get another MOSFET like the IRLB3036, 60V only, Vgs=± 16V, Id=195/270 A.
If still not OK, then I will live with the sparks.
Any other suggestions?
Thanks

Walter
 
rachdatu said:
Any other suggestions?
Thanks

Walter

Search eBay or Ali express for "solar circuit breaker". These have a good track record when used with bike controllers even at 72v. Not only can you turn off the load, but it also functions as a circuit breaker in the event of a fault.
 
KeithT said:
So there is some hope for my poor fried critter.

So Keith, how did it go? After reading through the entire thread it would be nice to see a conclusion to your story!

And fechter, i must say i really like the simplicity and elegance of the capacitor limiter. It's not without drawbacks but this seemingly simple task really is fraught with peril. I'll be trying my hand at it when the time comes..

From a scientific point of view it would be very interesting to know why the results from hard switching all point to problems, with your welder i think you have shown that it's very possible to manage a similar switching situation but here it seems lika a dead end! Have anyone thought about the energy in the pulse delivered to the caps in comparison to the welder? I'm guessing that the big difference is related to the much higher voltage. Would be interesting to see if a proper gate drive chip would make a difference, i have a colleague who was working on designing UPS units and they had a 15 amp gate driver. Also the very high inrush current probably gives some magnetic coupling potentially messing with switching. An overbuilt(expensive) switch to use for tests, measuring the currents, seeing any spikes and possibly potting it next to a new revision and seeing if it has any EMC effects.
 
I guess my fried critter story is not over yet, but it's become a cold case. I was not able to repair the inverter that blew up. Replacing the output mosfets did not fix it, it blew up again. And I don't have enough information to troubleshoot the drive circuitry and try again. When I looked into buying another inverter, I did not find the same input voltage offered, perhaps they have had too many failures and are less optimistic about the reliability at high voltages. I believe that my two stage design can, does and will work, but until I have a real load to precharge, it remains a valiant effort that did not quite reach completion. I'm retired and when the weather is nice my soldering iron is cold, I have other interests. But I'll get back to this project, someday.
 
Back
Top