Last saturday I put 2 amplifiers to work hard with +/-45V rails. One of them had FQP46N15's and the other one had IRF640 (not IRF640N).
Both of them were playing music at full power the whole day on 4 ohm speakers.
The second one had a smaller heatsink and the temp rise was about 20 deg higher than the other, I have to test both in the same conditions to see what the difference really is due to the bigger Rds(on) but smaller Qg.
Best regards,
Pierre
Both of them were playing music at full power the whole day on 4 ohm speakers.
The second one had a smaller heatsink and the temp rise was about 20 deg higher than the other, I have to test both in the same conditions to see what the difference really is due to the bigger Rds(on) but smaller Qg.
Best regards,
Pierre
FQP vs IRF
Hello all.
This saturday I made some tests with two equal prototypes, one with IRF640 (not N) and another with FQP46N15. It was a surprise for me that, when playing LOUD for some hours, the fairchild one got a higher temperature (around 25ºC above ambient), than the IRF (4 degs less).
The IRF one has far more resistive losses (180mohm vs 42mohm), but less gate charge. This proves that switching losses are dominating my total losses.
The gate resistors were 10 ohms in both cases. I think I will try to reduce them to about 6 ohm, perhaps that should speed up turn-on and reduce losses, at least in the IRF variant.
BTW: I have played both modules for about 15 hours total at near full power (but +/-45V rails and music, not sine wave), and none of them have failed. I think that these mosfets ARE reliable, much better than my old NTPs from On-Semi.
Best regards,
Pierre
Hello all.
This saturday I made some tests with two equal prototypes, one with IRF640 (not N) and another with FQP46N15. It was a surprise for me that, when playing LOUD for some hours, the fairchild one got a higher temperature (around 25ºC above ambient), than the IRF (4 degs less).
The IRF one has far more resistive losses (180mohm vs 42mohm), but less gate charge. This proves that switching losses are dominating my total losses.
The gate resistors were 10 ohms in both cases. I think I will try to reduce them to about 6 ohm, perhaps that should speed up turn-on and reduce losses, at least in the IRF variant.
BTW: I have played both modules for about 15 hours total at near full power (but +/-45V rails and music, not sine wave), and none of them have failed. I think that these mosfets ARE reliable, much better than my old NTPs from On-Semi.
Best regards,
Pierre
Some curious questions:
Did you use the amps in the meantime ? Are they still performing properly ? Did you modifiy anyrthing else than just the output FETs, like snubbers, antiparallel diodes etc ?
Are you still using the post-filter NFB ? Is the "ringing" problem still there ?
Regards
Charles
Did you use the amps in the meantime ? Are they still performing properly ? Did you modifiy anyrthing else than just the output FETs, like snubbers, antiparallel diodes etc ?
Are you still using the post-filter NFB ? Is the "ringing" problem still there ?
Regards
Charles
Thanks for the interest, Charles.
Yes, the amps are performing well, I have one with FQP46N15's and two with IRF640 (not -N by the moment).
I haven't added snubbers and yes, I continue taking post-filter negative feedback. The ringing is still present.
I have three amps, two with the same layout and another one with other layout (designed only to improve thermal transfer to the heatsink). The ringing is most noticeable in one of the equal modules. Perhaps the phase margin is just at the edge.
Now that the reliability is less a concern I will concentrate in removing that ringing completely. I still have to try the trick to add phase margin you recommended me.
I have decreased the gate resistors in the newer amp and I expect to have faster edges and hence lower losses. By the moment, rise/fall time is 30ns, with 6.8ohm gate R's and IRF640.
Best regards,
Pierre
Yes, the amps are performing well, I have one with FQP46N15's and two with IRF640 (not -N by the moment).
I haven't added snubbers and yes, I continue taking post-filter negative feedback. The ringing is still present.
I have three amps, two with the same layout and another one with other layout (designed only to improve thermal transfer to the heatsink). The ringing is most noticeable in one of the equal modules. Perhaps the phase margin is just at the edge.
Now that the reliability is less a concern I will concentrate in removing that ringing completely. I still have to try the trick to add phase margin you recommended me.
I have decreased the gate resistors in the newer amp and I expect to have faster edges and hence lower losses. By the moment, rise/fall time is 30ns, with 6.8ohm gate R's and IRF640.
Best regards,
Pierre
I think that the problem was my previous mosfets, NTP35N15. A pair of FQP46N15 failed as well, but I am almost sure that it was due to an accidental shortcircuit in the lab. Now I am using FQP46N15 and IRF640's with no problem.
The fact is that the circuit has been running with music at full power during hours with big speakers and none of them have failed so far.
However, I have prepared a new board with very improved routing, specially the gate loops and switching node, just to be sure that there are no important parasitics effects that can affect reliability.
Best regards,
Pierre
The fact is that the circuit has been running with music at full power during hours with big speakers and none of them have failed so far.
However, I have prepared a new board with very improved routing, specially the gate loops and switching node, just to be sure that there are no important parasitics effects that can affect reliability.
Best regards,
Pierre
Hi,
Good to know, thanks. I guess we can put it down to avalanche then for the NTP35N15? If we had to guess that is..
It will be neat to see how the new board does Vs the old version too.
It's a little off topic but.... you haven't mentioned sound yet, how is it? It must have been worth it?
Cheers,
Chris
Good to know, thanks. I guess we can put it down to avalanche then for the NTP35N15? If we had to guess that is..
It will be neat to see how the new board does Vs the old version too.
It's a little off topic but.... you haven't mentioned sound yet, how is it? It must have been worth it?
Cheers,
Chris
Thanks for your interest.
Yes, I think that the NTP35N15, that weren't rated for avalanche and even for max V/ns in the body diode were failing.
Just today I have been doing more listening tests, using IRF640. I have to say I am quite impressed with these little mosfets, they are cheap, easy to get and give excellent results. Resistive losses are not so important, definitively, but efficiency will increase when I put IRF640Ns (lower Rds(on))
BTW, the sound is excellent, not only IMO but all the people that have heard it is impressed. Very low noise floor also (a bit contaminated by the 100Hz hum from the supply, but as soon as you short the input just at the board, it disappears.
There is still some ringing at very high amplitude levels, no noticeable at all by ear, of course, but still present and worrying me.
Best regards,
Pierre
Yes, I think that the NTP35N15, that weren't rated for avalanche and even for max V/ns in the body diode were failing.
Just today I have been doing more listening tests, using IRF640. I have to say I am quite impressed with these little mosfets, they are cheap, easy to get and give excellent results. Resistive losses are not so important, definitively, but efficiency will increase when I put IRF640Ns (lower Rds(on))
BTW, the sound is excellent, not only IMO but all the people that have heard it is impressed. Very low noise floor also (a bit contaminated by the 100Hz hum from the supply, but as soon as you short the input just at the board, it disappears.
There is still some ringing at very high amplitude levels, no noticeable at all by ear, of course, but still present and worrying me.
Best regards,
Pierre
reverse recovery problem on 520ns dead time gate drivers
I'm beginning a SMPS/motor driver project with half-bridge gate driver IR2111. It has an internal fixed 520ns dead time and 40ns fall time / 80ns rise time. The MOSFET is IRF540 or IRF3205; the power supply voltage is 24V (up to 30V).
520ns dead time seems long enough to saturate the body diode and cause reverse recovery dv/dt.
The question is, can the body diode reverse recovery generate a dv/dt high enough to kill itself? Are antiparallel diodes necessary?
Another problem: it's said that dv/dt from motor brush noise can cause MOSFET dv/dt failure --
http://www.4qd.co.uk/serv/MOSFETfail.html
It seems incrediable. Is some EMI protection against that needed?
I'm beginning a SMPS/motor driver project with half-bridge gate driver IR2111. It has an internal fixed 520ns dead time and 40ns fall time / 80ns rise time. The MOSFET is IRF540 or IRF3205; the power supply voltage is 24V (up to 30V).
520ns dead time seems long enough to saturate the body diode and cause reverse recovery dv/dt.
The question is, can the body diode reverse recovery generate a dv/dt high enough to kill itself? Are antiparallel diodes necessary?
Another problem: it's said that dv/dt from motor brush noise can cause MOSFET dv/dt failure --
http://www.4qd.co.uk/serv/MOSFETfail.html
It seems incrediable. Is some EMI protection against that needed?
Hello Kenshin, a good thing about it being motor drive is that you do not have to be concerned much at all about distortion as in audio reproduction. So though the 520ns is long enough to allow the body diodes top enter full conduction, it could also be long enough to allow them to recover after the current stops flowing. It depends on your circuit. You may even want to add more dead time if necessary.
One good thing about fighting problems from noise spikes is that the higher the frequency of the transient, the easier it is to snub. Charles aptly commented similarly.
One good thing about fighting problems from noise spikes is that the higher the frequency of the transient, the easier it is to snub. Charles aptly commented similarly.
Returning to my Class-D design, I must say that I am quite happy with IRF640 mosfets. I am currently using a IR2110 to drive them and 6.8ohm gate resistors, with antiparallel schottky. The rise/fall time is about 30ns by now, but I think I can improve this further by reducing the gate resistor to, say, 4.7 ohm or even less.
I can buy IRFB38N20's but I really doubt it worths the pain the price and difficulty to obtain, as the main difference is Rds(on). But as my previous FQP46N15's have similar figure and they heated up more, I think that resistive losses are negligible compared to switching losses at powers around 200-400W.
Has anyone of you experience with IRF640(N)'s and do you think it worths the pain to use more expensive ones?
Just thinking loud again...
Best regards,
Pierre
I can buy IRFB38N20's but I really doubt it worths the pain the price and difficulty to obtain, as the main difference is Rds(on). But as my previous FQP46N15's have similar figure and they heated up more, I think that resistive losses are negligible compared to switching losses at powers around 200-400W.
Has anyone of you experience with IRF640(N)'s and do you think it worths the pain to use more expensive ones?
Just thinking loud again...
Best regards,
Pierre
After 5 hours at full power (music) today, one of the amps has failed in a curious way:
When I have returned to see how it was (I couldn't bear the loud sound ;-), I have found that the positive rail fuse had blown and a small capacitor had exploded.
After examining it, I have seen that the IR2110 driver had evidend signs of failure and the upper mosfet was shorted (three pins together). The heatsink also had an evident spark mark just by the side of the capacitor.
The capacitor was (and I say _WAS_) a 47uF/16V in parallel with the 12V supply of the driver. My theory is that, after overheating (a wide track from one of its pin was direct to the low side mosfet), it exploded, shorting the positive pin to the heatsink, that is connected to GND. That caused the full 40V going to the driver, frying it and also blowing the upper side mosfet (¿?)
Do you think this is feasible? I think the solution is to put the capacitor far away from the heat source and keep the mosfets cooler, but perhaps using a 16V capacitor for 12V is also risky.
I would like to hear your opinions.
Best regards!
Pierre
When I have returned to see how it was (I couldn't bear the loud sound ;-), I have found that the positive rail fuse had blown and a small capacitor had exploded.
After examining it, I have seen that the IR2110 driver had evidend signs of failure and the upper mosfet was shorted (three pins together). The heatsink also had an evident spark mark just by the side of the capacitor.
The capacitor was (and I say _WAS_) a 47uF/16V in parallel with the 12V supply of the driver. My theory is that, after overheating (a wide track from one of its pin was direct to the low side mosfet), it exploded, shorting the positive pin to the heatsink, that is connected to GND. That caused the full 40V going to the driver, frying it and also blowing the upper side mosfet (¿?)
Do you think this is feasible? I think the solution is to put the capacitor far away from the heat source and keep the mosfets cooler, but perhaps using a 16V capacitor for 12V is also risky.
I would like to hear your opinions.
Best regards!
Pierre
Hi,
You should be using good computer grade caps that are rated for high ripple and high temp. For such a small cap there's also little reason for not using a very wide safety margin on their voltage rating, consider 50 to 60% as a bare minimum and if it doesn't make it too large even 2 or 3 times the required voltage is fine.
I'd imagine you can also take into account how well the voltage feeding it is regulated, If it's just a zener with a series pass a much higher rating might be a good idea, if it's a linear reg. IC with all kinds of built in protection and tight tolerance it wont' be so critical an issue.
You should be using good computer grade caps that are rated for high ripple and high temp. For such a small cap there's also little reason for not using a very wide safety margin on their voltage rating, consider 50 to 60% as a bare minimum and if it doesn't make it too large even 2 or 3 times the required voltage is fine.
I'd imagine you can also take into account how well the voltage feeding it is regulated, If it's just a zener with a series pass a much higher rating might be a good idea, if it's a linear reg. IC with all kinds of built in protection and tight tolerance it wont' be so critical an issue.
Thanks for the opinions.
In fact, it comes after a 7812 regulator, so it should be quite clean.
What I discovered comparing the failed board with others that haven't failed is that there is a thick plane at the - terminal of the capacitor connected to VSS very near the lower mosfet. Presumably that has conducted a lot of heat from the mosfet to the capacitor, heating it and making it explode.
Other modules have 25V capacitors. That should reduce the internal temperature of the component, right? If so, I will for sure use a 50V cap if it helps !
Best regards.
In fact, it comes after a 7812 regulator, so it should be quite clean.
What I discovered comparing the failed board with others that haven't failed is that there is a thick plane at the - terminal of the capacitor connected to VSS very near the lower mosfet. Presumably that has conducted a lot of heat from the mosfet to the capacitor, heating it and making it explode.
Other modules have 25V capacitors. That should reduce the internal temperature of the component, right? If so, I will for sure use a 50V cap if it helps !
Best regards.
Defective electrolytics you say? NEVER..... how foolish.
Seriously I recently found out about a very common problem for computers, the caps used for (I guess) filtering on the local CPU regulators are often pron to leaking, swelling, or even exploding from hydrogen build up.
Someone in around 2001 stole a dielectric formula which was incomplete and all kinds of cap manufacturers bought this stuff because it was cheaper which led to a whole slew of defective caps.
The missing ingredients in the dielectric formula had to do with preventing internal electrolysis, hence the hydrogen produced.
Apperently they can fail either way, open or shorted.
Seriously I recently found out about a very common problem for computers, the caps used for (I guess) filtering on the local CPU regulators are often pron to leaking, swelling, or even exploding from hydrogen build up.
Someone in around 2001 stole a dielectric formula which was incomplete and all kinds of cap manufacturers bought this stuff because it was cheaper which led to a whole slew of defective caps.
The missing ingredients in the dielectric formula had to do with preventing internal electrolysis, hence the hydrogen produced.
Apperently they can fail either way, open or shorted.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.