Guru
Joined: 02/02/2017
Location: AustraliaPosts: 1432
| Posted: 11:38am 30 Oct 2018 |
And I have done this test against an IRFP4110.
It is just like the others (HY4008, IRFP2907).
These 3 MOSFETs do not like intermediate gate drive voltages.
They want to be either fully ON or fully OFF.
For completeness sake, here is the DSO capture of the 4110 failure.
I am not trying to use these MOSFETs for any novel application involving heating.
I am trying to understand, in a hands-on, experimental way, how they live and die.
Once one small part of the MOSFET starts conducting more than the rest of the chip, this small part gets hotter and hotter, conducting more and more and then it shorts the entire chip. If you look at the specs of some of these MOSFETs you will see the transfer curves have a temperature dependence, if the gate voltage is below something like 4V then heat induced feedback destruction will occur, if gate voltages are well above these sorts of values, there is a small negative temperature driven feedback.
My thoughts on failure modes for my inverters so far are:
1 - gate drive voltages may exceed specifications, which are +/- 20V with respect to Source. I have seen voltages in excess of +20V on the gates of HY4008 under moderate load. How long can the very thin insulation layer withstand over voltage?
(how long is a piece of string..)
2 - gate drive IC have specifications dictating low voltage limits with respect to ground, which is the same as the low side source pin. I have seen gate voltages well below -2 or -3 V. IR2110 specs say no gate output pin voltages less than -0.3V with respect to ground. The high limit is Vcc + 0.3 or about 12.3V How long can these drive chips put up with voltages outside these limits?
3 - putting 50A through one HY4008 does not even come close to producing enough heat to worry the MOSFET. Not by a factor of 10. Heat from Rds(on) resistance is not an issue.
4 - heat from switching ON->OFF or OFF->ON is by far the most important source of potentially damaging heat. Fast switching is a benefit. Slow switching is likely to be a source of problems of the explosive nature.
This source of heat in my view accounts for the vast majority of the heat generation of the inverter.
5 - heat also is generated from the cross conduction that occurs when the high side switch goes ON, pulling it's source to V(supply) at a very fast rate of change. This causes the low side MOSFET gate to rise to voltages within the range of conduction due to the capacitive coupling of the low side MOSFET drain to the low side MOSFET's other terminals. This positive pulse, usually of the order of 4-5V will make the low side MOSFET switch ON at the same time as the high side is ON, causing cross conduction for a short interval. But the low side MOSFET is not fully ON, just in the transition region, where, as I have shown above, these MOSFETs are very weak and can not withstand much abuse. The IRFP4110 could only handle a little less than 2 Amps before destruction. Cross conduction may be the killer and I need to perform more experiments to illuminate to myself how much of a problem it is.
My near term plans include fully populating one of Madness's boards and looking at how much - if any - cross conduction is visible on it. Madness has designed the gate drives to use a totem pole drive, taking it's signal from say, the IR2110 gate drive IC. This design will provide a very low impedance path for the induced voltage pulse on the low side MOSFETs and I hope to see some beneficial effect from this design.
Also, to populate one of his boards requires you to drop about $70 at Jaycar for the bits and pieces, not counting the HY4008 or bulk caps...
Right now I have two home built inverters, based on the Aliexpress 12 HY4008 boards.
I modify them as follows:
- EGS002 has the over current circuit disabled completely. I have other over current protection in place (a 100A fuse) and the faith that no matter what you ask from them, they will survive. We all need to laugh now and then...
Well, I can get away with this if I only ever use a large non-saturating ferrite E core inductor. When I used the Areosharp iron core inductor, failure was only ever a week or so away.
- I pull the IR2110 shutdown pin LOW and keep it there to ensure the IC always stays enabled.
- The HY4008 gate drive outputs are protected with 18V Zeners to clamp high DC volts and Schottky diodes to protect against -ve DC volts. This is for the longest lived inverter. The second one I have just rebuilt after failure utilises 18V TVS devices which are more effective in >18V DC excursion clamping and provide -ve voltage clamping all in one package. I feel the need to protect the gate drive IC outputs as a matter of high priority. This also protects the MOSFET gates from exceeding +20V which is a Good Thing.
These inverters seem to travel quite well under the loads I drive them here at home.
So it's onward towards greater understanding of the various aspects of inverters.
Oh, and here is a pic of the Arduino inverter controller I am using now.
It is much more compact and probably more reliable without so many flying leads going everywhere.
I run the usual 1/2 squarewave 1/2 SPWM signalling, a la EG8010.
I also run my full SPWM on both 1/2 bridges type signalling, which makes for a smoother output waveform and my code has both a soft start as well as a soft stop function.
This daughter board for the Arduino also allows me to run modified code which permits you to have variable output frequency (from 40Hz to 100Hz AC at full power).
Additionally, I can experiment with arbitrary PWM frequencies, continuously variable from 2kHz to 70Khz. This illustrated resonance peaks of the primary choke/capacitor LC filter very nicely. Again I run this at reasonable loads such as 500W.
The device only needs +12V which is referenced to DC bus ground.
The trimpot allows adjustment of deadtime from about 400ns to about 2us.
wronger than a phone book full of wrong phone numbers