I am relatively new to MOSFETs and am using 20 in a project I have been working on. I built a CD spot welder based on schematics that I bought from someone online. He also supplied boards and has helped me TONS through out the process, but me being new and anxious to learn, I suspect he is tired of me bugging him! I have done a lot of reading, but these are hard to understand when looking at all of the specs on the datasheets...

The issue: I "suspect" that I have blown some MOSFETs (18 IRFP2907s to be exact) and am looking for advice.

First off, can a MOSFET latch open due to a faulty driver or some other aspect of the circuit? Common sense says that when the circuit is powered off, voltage at the gates would have to be zero (n-channel mosfets) and therefore they would not conduct. However, even when power is off, there is continuity between source and drain. In my experience, these things show physical signs of damage when they are fried (read: they shoot shrapnel across your living room when they "pop"), but none of the 18 seem physically damaged, which is the only reason I ask about latching open.

Secondly, what do I need to pay attention to when seeking alternative replacements to the MOSFETs? The IRFP2907s are what the designer called for and recommends, but they are quite pricey and take several weeks minimum to get as they are only bought from overseas sources. Digikey stocks several different models that seem suitable, but all are made by IXYS. As stated, I am new. Is there a good reason to go with IR stuff over these?

In particular, would the model below be a good replacement? The gate charge and input capacitance are significantly lower than the IR pieces, but power dissipation is relatively the same and cont. current ratings are higher (not that that matters here).

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=IXTH240N055T-ND

For reference, these are being used in conjunction with 2.4 F capacitance at ~ 15 volts being discharged through 1/0 AWG cabling in ~ 15ms bursts.

Date Sheets:
IXYS possible replacement: http://ixdev.ixys.com/DataSheet/51abd158-5540-4eba-b7c7-c2a768f073d0.pdf
IRFP2907: http://www.irf.com/product-info/datasheets/data/irfp2907.pdf
UC27322 Driver: http://focus.ti.com/lit/ds/symlink/ucc27322.pdf

Thanks in advance for the help!

 

A common failure mode is for MOSFETs to fail closed, that is, shorted. Short gate to source and run the continuity test again. Also, test in reverse, to measure the body diode.

 

I'll quickly reply to one part of your question for now.

When you say "latch open" do you mean it's stuck in the "off" or non-conducting state, or do you mean it's latched "on" or conducting? Just asking because your questions seem to imply the thing is stuck "on".

Okay, here is my 2 cents. Just because you remove power from a circuit doesn't guarantee that the MOSFET is off. The gate of a MOSFET is a capacitor. Unless there is a circuit element to drain it, like a pull-down resistor for an N-channel MOSFET, it is quite possible for a MOSFET to retain a gate charge and be "on".

To test if they are *really* blown in that sense, try shorting the gate to source (through a 1K resistor if you want to be gentle). Then test the continuity from drain to source. Also note that MOSFETs contain a body diode, so it may make a difference which direction you test continuity as the multimeter may register continuity due to diode conduction, depending on how the multimeter circuit works.

You could also set up a MOSFET in a test circuit with the source on ground, and the drain to a positive supply through a series resistor to limit current. Say, 1K Ohms. Then vary the gate voltage (also through a series resistor like 100K Ohms) from 0 gradually upwards and see if it starts to conduct and then fully conducts. You can check that:

* off is really off
* on is really on -- estimate the on resistance [Rds(on)]
* that it turns on around the expected threshold gate voltage
* there is no significant gate leakage current

That's my 2 cents. Hope it helps.

And remember, MOSFETs are delicate. A common way they can blow is ESD. The gates are fragile and a static discharge can blow them apart, when they're free-handled. When they're soldered down, they are typically safer from that.

 

Yes, now that gate charge is mentioned, I seem to remember reading about that back before I had my issues. That bit of info in itself answers many of my questions.

In reference to "stuck open", that is my uneducated way of saying they are stuck closed. ie, the gate is staying charged. Bad habit. I'm working on it.

Now that I think about it, I remember that I accidentally had my board grounded to the rectifier's sink (they were accidentally touching) the last time it was powered up. That could very well explain why the driver was acting up and left the gate's charged.

I'll discharge the gates when I get home and try again.

What spawned all of this was my rectifier overheated and blew my main 15a fuse. Originally, I contributed it to the rectifier not getting enough cooling, since it was relatively far from the fan and the sink was relatively small. After seeing continuity between source and drain, I made the assumption that the rectifier overheated because the mosfets were open and my rectifier was essentially being shorted through my electrodes.

Now my theory is that the original problem WAS due to cooling issues, then upon relocating my rectifier, I caused the short which made my driver act funny and leave the gates charged. Hopefully that is the case, because I do not want to spend another $100 in mosfets.

Sage: Thanks for the info on how to set up a quick and simple test. If it comes down to it, that would be a great way to know for sure.


For my own personal fulfillment, did either of you look at my possible replacement? I was looking at the data sheets and saw that my proposed replacements would have a max pulse charge of almost 200A less than the IR pieces, so they are out. I am still interested in what to look at when considering replacements though.

 

Here's a list of potential replacements:

http://uk.farnell.com/jsp/search/br...e&locale=en_UK&catalogId=&prevNValues=1004185

Based on Id (±30A), Vds (±5V), and type (N channel).

However they are all still pretty pricey. Unfortunately those kind of power FETs are. I don't know much about gate charge etc. so I'm not going to narrow it down on those characteristics.

 

Here's a list of potential replacements:

http://uk.farnell.com/jsp/search/br...e&locale=en_UK&catalogId=&prevNValues=1004185

Based on Id (±30A), Vds (±5V), and type (N channel).

However they are all still pretty pricey. Unfortunately those kind of power FETs are. I don't know much about gate charge etc. so I'm not going to narrow it down on those characteristics.

None of those are the right package (TO-247) and the only have about half of the power dissipation of the one's i'm using. Does this matter? If I used the TO-247s again, I wouldn't have to re-drill/re-tap my sinks and things would just be easier.

Also, will it hurt to use mosfets with higher current and dissipation values than called for, as long as the driver can trigger them? Price is relatively the same, so it would just be extra cushion. This is assuming I even need them, which I am hoping I do not.

PS: Your signature is something I am all too familiar with as a result of this project's temper! haha

 

It's completely harmless to overspecify your devices, I'd go as far as to encourage it.

I didn't realise package was a concern, so it looks like Farnell have no others. You could try Digikey, but I assume you've already done that. Unfortunately these particular MOSFETs are quite expensive to produce.

I'm not sure but how practical is paralleling cheaper FETs? I guess this means increased gate charge and all those parameters I don't know much about, so I wonder if it is practical.

 

It's completely harmless to overspecify your devices, I'd go as far as to encourage it.

I didn't realise package was a concern, so it looks like Farnell have no others. You could try Digikey, but I assume you've already done that. Unfortunately these particular MOSFETs are quite expensive to produce.

I'm not sure but how practical is paralleling cheaper FETs? I guess this means increased gate charge and all those parameters I don't know much about, so I wonder if it is practical.

Digikey has 9 suitable for my application, 6 of which are in stock. More powerful IXYS models are still significantly cheaper than the IR's I have due to the IR's availability. I'm now leaning towards one of IXYS' 75v 240A+ models.

 

My search came up with the same as you -- several IXYS parts.

I'm curious about how this circuit works.

We're talking about *extremely* low resistances.

2.4F of capacitance at 15V -- how much ESR?
(Is this electrolytic or supercaps stacked in series? Are there many caps in parallel?)

AWG 1/0 cable. I'm assuming it's copper. That's like 100 mOhms per 1000 feet. I'm guess you have 2 or 3 feet of cable. That's not even a milliOhm.

Maybe there's a welding tip of tungsten or something? Maybe a mOhm?

Then there's the junction to the material being welded, and the material itself -- who knows how many mOhm -- and it heats up and changes resistance.

So are all these MOSFETs in parallel and turning on simultaneously?

Let me just say, you better have a good gate driver, because you don't want the MOSFETs to be in partial conduction for any more nanoseconds than necessary. There's going to be a lot of Miller current. That's where a lot of the heat probably builds up. And looking at the gate driver, I see it handles 9A. I assume there is one gate driver per MOSFET right?

How does this circuit work? Does it turn on all the FETs at once? Does it turn on some fraction of them to adjust how much power is used? Is there a microcontroller involved? What gate resistance is used?

What's the duty cycle on these 15 ms pulses?

Anyway, I think you're looking at the right datasheet figures.

However, this is such a *large* amount of current that there are probably things going on that we can't predict. Things not in the data sheet. Effects we don't think of.

For example, if the Rds(on) of the FETs is a big limiting factor of the current -- meaning the ESR of the caps and other resistances are very low -- then the lower Rds(on) of the replacement may cause *higher* currents to flow and may blast out the FETs or something else.

 

(So wear a welding mask when you try it out!)

 

My search came up with the same as you -- several IXYS parts.

I'm curious about how this circuit works.

We're talking about *extremely* low resistances.

2.4F of capacitance at 15V -- how much ESR?
(Is this electrolytic or supercaps stacked in series? Are there many caps in parallel?)

Two 1.2F caps in parallel with ~.007 combined ESR.

AWG 1/0 cable. I'm assuming it's copper. That's like 100 mOhms per 1000 feet. I'm guess you have 2 or 3 feet of cable. That's not even a milliOhm.

Maybe there's a welding tip of tungsten or something? Maybe a mOhm?

~ 3' of cable from cap positive to 3/16" copper elctrode tip and same amount from mofset board to tip.

Then there's the junction to the material being welded, and the material itself -- who knows how many mOhm -- and it heats up and changes resistance.

Welding .01 Nickel tabs to A123 cells.

So are all these MOSFETs in parallel and turning on simultaneously?
Yes, all in parallel.


Let me just say, you better have a good gate driver, because you don't want the MOSFETs to be in partial conduction for any more nanoseconds than necessary. There's going to be a lot of Miller current. That's where a lot of the heat probably builds up. And looking at the gate driver, I see it handles 9A. I assume there is one gate driver per MOSFET right?

How does this circuit work? Does it turn on all the FETs at once? Does it turn on some fraction of them to adjust how much power is used? Is there a microcontroller involved? What gate resistance is used?

One gate driver firing all. That is (1) total driver on the board. I didn't design it, so I don't know exactly how it works, but the board runs on 24vac through a rectifier and is triggered by a 3-5V input. There is an LM317K on the board, two 6800mF caps, various small resistors/capacitors, a random zener diode and a 20ohm resistor between the driver and gate(s). There is a separate control board that uses a processor, three pots to control voltage and pulse widths, and an LCD to see all the settings.

What's the duty cycle on these 15 ms pulses?

Triggered by a foot switch.

Anyway, I think you're looking at the right datasheet figures.

However, this is such a *large* amount of current that there are probably things going on that we can't predict. Things not in the data sheet. Effects we don't think of.

For example, if the Rds(on) of the FETs is a big limiting factor of the current -- meaning the ESR of the caps and other resistances are very low -- then the lower Rds(on) of the replacement may cause *higher* currents to flow and may blast out the FETs or something else.

Supposedly the board is designed for up to 3F capacitance, and I have some large sinks on the FETs (3 3"x4"x1.25" sinks, 1 per 6 FETs). They were never even warm, even after I thought they blew following about 100 welds within 10 minutes. The fan blows directly onto these sinks. I'll be sure to check with the designer regarding a lower Rds(on) before I jump the gun.

Would you mind going a little more in depth regarding driver selection and why you think 9A max is low for the driver? How did you come to that conclusion? The board works fine and I don't think there is an issue, I just want to know for my own knowledge!

 

Okay anyone who knows, correct me if I am wrong. I've one some research on miller capacitance. To clarify, the Miller Effect is essentially an additional capacitance between the gate and drain that begins to charge when the Vgs is half way charged, meaning it essentially pauses Vgs charging when the FET is halfway open, causing excessive heat if not "broken through" quick enough by the driver, making significant driver current important?

So, in picking MOSFETs, Lower capacitance values are better, specially when driving multiple FETs from a single driver. This would also allow faster turn off times as well, right? Would one need to change the gate resistor upon installing MOSFETs with different capacitances?

Based on what I've learned, these are what I have chosen. Having trouble narrowing it down:
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=IXTH220N075T-ND
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=IXTH240N055T-ND

They have lower Cgs values (in comparison to the IRFP2907), lower total capacitance, higher cont. current ratings, higher power dissipation rating and lower max pulse current ratings. One has slightly lower and one has the same Rds(on) values (sage, you scared me about the power thing).

I don't know which to choose. I think the Rds is going to have less of an effect than I think, so I am probably going to choose the one with higher max pulse current ratings.

The IR mosfets have 870A pulse ratings though and the highest that IXYS has to offer is 650A. 650x18=11700A max pulse rating, 3960A less than the total max of the irfp2907s (15660A).

Would these in effect limit the current, or would they just blow should current exceed that rating?

 

Would you mind going a little more in depth regarding driver selection and why you think 9A max is low for the driver? How did you come to that conclusion? The board works fine and I don't think there is an issue, I just want to know for my own knowledge!

Matt, thanks for the answers to all my questions. Now the picture is getting clearer of the actual application.

On the gate driver -- I think it's a fine driver. I was just saying "I hope you have a good gate driver..." and then I checked you do have 9A driving. However, this will be distributed through all the MOSFETs.

Now you mentioned that there is 20 Ohms of gate resistance. If anything would be adjusted, I think that resistor value would be the thing. In the IXYS data sheet, they mention a 5 Ohm external gate resistor. What do they mention in the data sheet of the original part?

I didn't mean to scare you!! Let me state that I have never designed anything like this. I've never designed anything to pass that much current through a FET! I've done solar power stuff of 20 Amps and low-power stuff in the microAmps. This project of yours is a great opportunity for me to learn about this kind of application, as well. I'm bringing up possible things to look out for. Someone who has designed welders would know a lot more than me about what is relevant. Together we can reason out answers.

Like for example, I don't know if the Rds(on) of the MOSFETs are limiting the current or not. But I suspect it's not, since the ESR of the caps is around 7 mOhms, and the FETs are all in parallel, and therefore have a lot lower combined resistance, on the order of a fraction of a mOhm. In that case, a lower Rds(on) FET would be a fine choice, and lead to less heating in the FETs. (Although you reassuringly mention that they don't even get warm.)

And after all is said, you can try it and see. I hope you report back to us on the results. That is, if you're still alive. Just kidding

 

I tested my fets and found that of the 18, only one was bad. The one closest to driver and electrode connection.

As for gate resistance, the IXYS models call for 3.3ohm resistance while the IR parts I am currently using spec 1.2ohms.

By that, I would need to change my 20ohm gate resistor to ~ a 55-60ohm resistor? Correct me if I am wrong.

 

I tested my fets and found that of the 18, only one was bad. The one closest to driver and electrode connection.

As for gate resistance, the IXYS models call for 3.3ohm resistance while the IR parts I am currently using spec 1.2ohms.

By that, I would need to change my 20ohm gate resistor to ~ a 55-60ohm resistor? Correct me if I am wrong.

Yes, I think you're wrong about that. I don't see in either data sheet where they spec the gate resistors you mentioned. I did see mentions of 25 Ohms in the IRFP2907 datasheet, and 5 Ohms in the IXTH240N055T data sheet. However, these are not firm specs saying "you must use these values". Rather, they are stating what they used for testing to give the other values they claim.

In fact, in the IXYS part, you can see figures 15 and 18 giving various switching speeds based on different external gate resistor values.

Here's my 2 cents on the gate resistors. I'd love it if other people with more experience will chime in.

Firstly, it's not essential to even have a gate resistor. The driver data sheet, for example, mentions this. A gate resistor is partly to limit gate drive current draw to protect the gate driver circuit, and may also be to slow the actual switching speed for some applications. However, in your case, you want switching as fast as possible.

On the other hand, you're using a single 9A peak driver for 18 MOSFETs. That is 0.5 A per MOSFET. Then, it is also limited by the gate resistors. Also, some of them will hit the Miller plateau at slightly different times / voltages... and then will pass it. The single driver will be pushing current as fast as possible into all the MOSFET gates at the same time. If you have a gate resistor for each one, then it will limit the current into each one, and the pool of drive power will be more distributed, and a somewhat better chance for the MOSFETs to switch on at the same time.

At these high currents, some things may come down to slight differences in current path and resistances, and very slight differences in timing. I think it's telling that the MOSFET closest to the driver and electrode connections blew first. Possibly it was seeing the most current and possibly was also switching on first.

I feel like the gate drive is somewhat weak, at 9A for all 18 MOSFETs, plus 20 Ohms in the path of each MOSFET gate.

My gut feeling is that you should keep the gate resistors but lower their values to 5 Ohms, and you should double up the gate driver chip so 2 or more run in parallel. you could even piggyback another one on top. And also, be sure there is plenty of bypass capacitance at the supply of the gate drive chip. Very close to that chip, there should be substantial high-current power capability, in the form a of a capacitor of many uF.

Hope this helps and as always hope those more knowledgeable than I will chime in.

 

One more question -- you said "two 6800 mF caps" -- are you sure that's not nF, and what's it near?

 

Firstly, it's not essential to even have a gate resistor.

One other thing the gate resistor does, is damp ringing on the gate. Since the gate is a capacitor to ground, and the trace leading to it has some inductance, you have an LC circuit. If you turn the gates on & off quickly (as you're doing) you can get ringing, and it can be large enough to turn the FET on (when you intend for it to be off) or off (when you intend for it to be on). This can increase dissipation in the FET. With a resistor, the ringing is damped faster. (The trade off here is, larger resistor damps faster, but slows the on & off time.)

 

Yes, I think you're wrong about that. I don't see in either data sheet ... those more knowledgeable than I will chime in.

I just ordered 20 of these:

http://www.mouser.com/Search/Produc...Tvirtualkey65120000virtualkey747-IXTH240N055T

In response to the replies above:

First off, the circuit only has one single resistor between the driver and the shared gate. All of the driver's current passes through a single 20-ohm resistor and then goes to the gates simultaneously. The mosfets do not have individual gate resistors. That is why I was calculating the resistance the way I did.

Secondly my main concern, in the IRFP2907 Data sheet it specs a 1.2-ohm resistor in the conditions column on the second page, next to it's time specs. The IXYS model I quoted resistance for was the wrong one, so ignore that value. The correct value from the IXYS data sheet is 5-ohm. Comparing those, what would be a good resistor value?

As for the caps, I meant Micro-Farads. I guess I am used to the labels on the packaging at fry's. They label Micro-Farad as MFD. Yet another habit I need to kick. Caps are very near rectifier, but traces lead to other side of board. One of the caps traces through a diode and straight to the driver, so I assume that it what you were asking about. The other seems to have something to do with the LM317. There is also a small surface mount cap right next to the driver.

In regards to which FET blew, now that I look at it, they are all perfectly symmetrical to the gate resistor. So, (I believe) if it did fail due to firing time discrepancies, it was purely by chance. I believe it would more likely be due to being close to the electrode and therefore over-driven, but they usually catastrophically fail under over-driving conditions, right? One other thing to note, during deconstruction of my board I noticed a dry joint on the gate of one of my MOSFETs. Could that mosfet not firing have caused my problem? It seems as though if the circuit was designed to fire 18 FETs, one can't expect reliability when only firing 17.

Spacewrench, is the ringing even a valid consideration here, given that the mosfets are only fired in groups of two pulses? The gate traces are very wide, very straight and very symmetrical about all the FETs.

After i finish this project and get it working, I'm going to have to come up with a new project to put some of these spare FETs to good use! Possibly a MOSFET controlled 12v regulator/rectifier unit for a motorcycle.

Thanks for sifting through my lengthy replies and helping me out guys!

 

First off, the circuit only has one single resistor between the driver and the shared gate. All of the driver's current passes through a single 20-ohm resistor and then goes to the gates simultaneously. The mosfets do not have individual gate resistors. That is why I was calculating the resistance the way I did.

You hope, but that won't ensure they turn on at the same time. Oscillations between the different mosfets could be problematic. What will happen when just one turns on before the rest?

It is normal practice to use individual gate resistors. Check out these references:

http://www.irf.com/technical-info/appnotes/para.pdf
http://www.irf.com/technical-info/appnotes/an-941.pdf (page 3/7)

AN-941 is a bit more readable; the other is more theoretical. There is also a Microsemi app. note (APT0402) that I can't get to load today.

Gate turn-off is also an issue, particularly, if you are turning the mosfets off during the time they are conducting a high current. You may want to consider an anti-parallel diode on the gate, rather that a large resistor to ground, if you want fast turn off. That diode is discussed in several of the IR application notes as well.

Can you post a schematic of what you are trying to do? Without it, we are all shooting in the dark. We may detect lots of potential problems as you mention various details, but without the whole picture, it is impossible to do much more than that.

Finally, do you really need 20 high-current mosfets? Their single-pulse current rating is several times their continuous rating. I suspect that at some point, there not only is diminished returns, but perhaps disadvantages from imbalances

John

 

John,

As mentioned, I didn't design this, so I do not know the theory behind the decisions made to have only one resistor, 20 FETs or anything else. As for a schematic, all I can do is to post a picture of the board. Let me know if you would like me to.

However, all I am really interested in is what goes into choosing a gate resistor. Do I simply choose the value the manufacturer lists under their test conditions, or is it even that important? Most of the research I have done (including reading several documents from IR) suggests that the resistor is more or less guesswork. Honestly, I'm leaning towards leaving the 20k gate resistor in for the new mosfets. My pcb is already designed and built, so I am now aiming to make the best of it.

As for discharge time, there is a diode and resistor parallel to the gate resistor that is there to control turn off times. After reading the second article you posted, I see how important turn off can be when dealing with parallel mosfets. It seems to me that you would almost not want any resistance at all on the diode, to allow them to turn off even faster since miller charge can tend to keep the quicker to turn off FETs to stay in linear mode for longer, creating heat.