Yeah, a fet might pass through an embarassing zone of the SOAR curve if you switch it slow.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
So it seems that the PV driver might not to be the reliability bottleneck, but the super-durable 400+A FETs themselves, if driven improperly. Thank you all very much for pointing it out. What reasonable switching speed should I aim at?
Best regards, Piotr
The latter is what I mean.
Probably it causes damage. You'd have to look in the SOA graph of your FET. It'll tell you how long it can spend in the linear region at around
50 amps and cruising through your voltage range. I'd be surprised if 1msec were allowed.
It has to be fast in both directions if there is a tight SOA limit. But only the datasheet will tell.
-- Regards, Joerg http://www.analogconsultants.com/
Oh yeah, and it can go with an impressive kablouie and lots of smoke. I just had a similar event, an UVLO didn't come and rat-tat-tat .. *POOF*
We know it's 17V and 50A but we'd also have to know which FET you are using. That info is (hopefully) in the datasheet :-)
-- Regards, Joerg http://www.analogconsultants.com/
I haven't bought the FETs yet, but was thinking about IRFB7430PBF in TO220 or something bulkier in TO247.
I naively thought that the 1524A max pulsed drain current is enough to survive even a short-circuit if disconnected in the ms range...
Best regards, Piotr
The fet data sheet should have a SOAR (safe operating area) graph, which tells you how much time you can spend at a given voltage*current point before the silicon melts.
To be safe, switch fast, as in very roughly 10 microseconds. That might imply a modest amount of gate current, 10s of mA maybe. You've got to figure out the total gate charge required, and the time you have to switch, to get the current.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
afaict I will barely do 50A@17V for 1ms
won't even do 2A@20V for 10ms
-Lasse
This one is quite impressive:
Figure 10 shows the SOA. For 1msec 17V is at 10-20A but if the load is resistive you won't have that much while the load is zero. If capacitive and discharged you might exceed the limit. Then it may be a close call.
I'd try to switch it faster, like in 100usec.
Those are usually just marketing numbers :-)
Anyhow, a millisecond is a long time in FET-land.
-- Regards, Joerg http://www.analogconsultants.com/
Then youhave two options that I can see:
a. Use a gate drive transformer but build a reliable UVLO circuit on the high side. It must come on with a snap once you exceed 10V or so and when it drops below must instantly short the gate to ground. That will take a little engineering but should be fun.
b. Next is the lazy method but quicker to develop. Buy a little iso module that converts 5V or whichever crircuit supply you have (or your
17V) and make a stable isolated 12V on the high side. Look over your shoulder to make sure nobody sees you doing this, hardcore analog guys can take someone's man card away for that. Place a sturdy gate driver with Schmitt input up there, Micrel or something. Then add an optocoupler to ferry your command signal across. That's pretty much it, should still be under $5, just not as elegant as method a.-- Regards, Joerg http://www.analogconsultants.com/
IR is infamous for claiming absurd currents and powers. Imagine a TO220 conducting 195 amps DC through its source lead!
They also do impossible heat sinking, like testing the fets in flowing, boiling, unspecified liquids.
Claiming 375 watts dissipation in a TO220 is criminal.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Den torsdag den 19. februar 2015 kl. 02.05.30 UTC+1 skrev John Larkin:
yeh, a table of fusing current says a 1.8mm copper wire will melt in
10 seconds at that currentI seem to remember an app note on how the testing is done
the liquid is some DuPont fluorocarbon with a boiling point of 25'C
depends on the time frame :)
-Lasse
On a sunny day (Wed, 18 Feb 2015 15:41:43 -0800) it happened John Larkin wrote in :
mm, he could bridge the MOSFETS by 2 schottkeys
Schott.. U1 -----|>|-------- | | | | | | d s | g |------------- out | | | U2 -----|>|-------- | | | | d s g | This will limit the voltage across the FETS at all times to a few hunderd mV, at 50 A to a few watts, and gives time to switch.
This may have been true years ago, but IXYS has made a nice selection of parts available, and Mouser has a good stock of them. Table 3.6 in AoE-III, page 210, is quite useful, with parts from Infineon, Clare, Supertex, NXP, Vishay and IXYS.
--
Thanks,
- Win
Well, that's pretty unconventional, SFAICT.
If you really do need large-die n-channel parts (consider IGBTs intead?), there are multiple ways to get reliable gate-drive power, as have been spelled out, but also including an attractive optical power source, see AoE-III, Figure 3.107, with 0.5A gate-drive for high-speed switching.
Table 3.5 has these isolated switches working to 1800A pulsed, or 100A continuous. But beasts like this are likely ridiculously over-spec'd for your application.
--
Thanks,
- Win
b'. An isolated, low mA-range, power supply of any kind and Si826x. The part is cheap, easily available, comes in a much better package than similar ADI parts, has an UVLO and a strong gate driver.
I didn't like this option because of the need for a separate supply, but this thread has really taught me something new. Thank you very much!
Best regards, Piotr
Whoa, it looks like there's a misunderstanding here. If you short a strong low-resistance voltage source, you actually want less current flowing, not more.
Let's do the math. The 1524A rating of the '7430 is for the case of a very low voltage across the MOSFET, fully turned on, with only a volt or two of drop. In fact, as the die rapidly heats, it'll current limit at around 1kA, Fig 4, assuming 10V gate drive. Trying to short out 17V, it would dissipate 17 kW.
Fig 14, Effective Transient Thermal Impedance, says maximum allowed square single-pulse duration = 20us.
That's the thermal mass calculation for the '7430.
Let's try a higher voltage rated MOSFET (large-die HV parts will limit at lower currents), say the original IRF1407 version, an IRF1407PbF. It has an impressive 333W dissipation rating, nearly as good
current limit will be about 400A, or 6.8 kW. We
Fig 11 tells us the single-pulse duration = 35us. So the "weaker" part can handle a full short for almost twice as long.
--
Thanks,
- Win
To review, in a robust application, you want fast switching time, well under say 5us, and under 1us even better. This is so you can turn the switch on and check to see if it's at its proper low Vds(on) voltage. If it's not, then you have a load fault, a short, and you want to turn it off. Preferably you want to do all this within say 10us, see the 20 and 35us junction-temp limits calculated above.
There are driver ICs conveniently designed for this purpose. We discuss this in AoE III, e.g., in the Logic Interfacing chapter, page 847, Figure 12.87, which shows an opto-coupled driver IC with a DESAT function. This is an industry-standard method for dealing with short-circuit faults in a robust way.
In Figure 12.87 we suggest Avago's HCPL-316J and ACPL-332J. Both of these driver ICs include an optically-isolated FAULT feedback path, which your driving logic can use to deal with the fault. One way to deal with this would be to retry the power switch after a short delay to let it cool, say a few seconds. If there's still a failure after a few tries, a general FAULT condition is signaled and the user can be notified to fix the problem.
--
Thanks,
- Win
Those are nice. Just keep in mind that they have a lowish UVLO. Also, they seem not to be widely stocked but for a one-off case that's ok.
That extra supply isn't nice but as John Wayne said, "Man has got to do what man has got to do". It doesn't have to be regulated so even a simple gate drive transformer could be pressed into service to make an isolated supply.
As a SW guy once put it, we are all here to serve :-)
-- Regards, Joerg http://www.analogconsultants.com/
With these little TO-packages it is more likely that there would be a bright orange flash, followed by a loud bang, plastic pieces shooting off in all directions, and an "amperage stench" :-)
-- Regards, Joerg http://www.analogconsultants.com/
Where are the sound effects? I want my sound effects!
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.