 Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me. |
Forum Index : Electronics : 8 KW Inverter Build
  Page 12 of 22   |
|||||
| Author | Message | ||||
Madness Guru  Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Yeah I did see that one Mike, it is for continuous gate drive and requires a isolated power supply shown in the next diagram. The circuit on page 14 uses a charge pump instead to produce the voltage required to drive the high side gates. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
The situation with switching a mosfet is that when it is "on" the voltage across the mosfet is low but the current is high, and power dissipation within the mosfet is manageable. When the mosfet is off, voltage is high, and current and power are virtually zero. While switching you may go through a point where there is half the voltage and half the current dissipated as heat. A 5Kw inverter may be running at 50v and 100a dc input. There may in theory be a point where each mosfet has 25v across it and 50 amps flowing. That is over 1Kw of flash heat generated EVERY time it switches either on or off. At 20Khz it switches 40,000 times per second with these mini heat flashes, and its a Kw of heat generated every time. This is "switching loss" and the less of it we have the better. Now the only way this is going to work is by making the switching times as fast as possible. Its still a one Kw blast of heat, but if it only lasts a few tens of nanoseconds the average heat released is not that great. The mosfet itself is pretty fast. The problem though is the very large input capacitance that must be charged and discharged every time you turn the mosfet either on or off. These larger 320 amp rated mosfets have an input capacitance of 22nF which must be charged and discharged by about 12 volts as fast as possible. That takes a lot more power to do than most people realise. The more current you can pump in and out of that gate capacitor, the faster the voltage on the gate rises and falls, and the faster the mosfet will switch. Doing that can greatly reduce the switching loss and heat generated within the mosfet. General purpose gate driver chips are usually rated for about an amp or amp and a half maximum output switching current. That is more than enough for most applications driving a typical single mosfet. If you are doing something pretty extreme like switching hundreds or thousands of amps, something much better is well worth the effort to reduce switching times. Gate driver chips are now available rated to twelve or more amps with very fast outputs, which is what we need to get the best out of our monster mosfets with the very high input capacitance, or for driving a large chain of parallel smaller mosfets. One recommended chip is the Microcip TC4452 which is a non inverting driver that you place between the IR isolation chip and the gate. Its best to put this as close to the gate as possible with a one ohm damping resistor in series to both limit the peak gate current and reduce ringing. It will also require at least a 1uF tantalum capacitor placed very close to the gate driver chip. http://ww1.microchip.com/downloads/en/DeviceDoc/20001987C.pdf Success in all this depends as much on the physical layout as it does on the actual schematic and component selection. The faster you can make it switch, the more efficient it will be, but layout becomes more critical with increased speed. There is also a potential problem with voltage spikes being thrown back into the H bridge. The internal diodes within the mosfets will protect the mosfets, but only if you can clamp the spikes directly across the dc input to the H bridge. There must be a very large and very effective capacitance as close to the H bridge as close as you can get it. I mean like within an inch if you can do it. And use a decent low ESR capacitor with a sufficiently high ripple current rating. Cheap Chinese aluminium electrolytics are toys when switching really high power. If your mosfets are rated to 100v, and your dc input is 56 volts, any voltage spikes that go higher than 100v are going to be pretty lethal. And those spikes can have a lot of amps behind them, so a wimpy capacitor is not going to do much. This may be another reason why inverters can suddenly go bang, especially with Chinese mosfets. Good mosfets these days are avalanche rated look for that in the published specifications. That means if its rated for 100v, a voltage spike higher than that will cause the mosfet to go into conduction non destructively and clamp the voltage. That sounds great, but the heat generated doing the avalanche thing is fearsome. If the mosfet is already running nearly flat out, the extra heat will kill it. So try to manage voltage spikes some other way by keeping lead lengths very short, and using a really good quality pulse rated capacitor right across the H bridge as close as you can get it. Getting back to switching speed. If you only have a couple of small mosfets it may work fine. As you add more without changing anything else, you are probably slowing down the switching speed. The explosive flashes of heat get larger, and may be very unevenly distributed. One mosfet finally gives up, and the whole thing then self destructs. That is what I think may be the problem. The cure is switching as fast as possible, and improve the matching between mosfets. And the best way to do that is use a very few big ones rather than a lot of small ones. If you can do it with only four mosfet power blocks, the manufacturer will have already solved most of the die matching and thermal matching problems for you. Cheers, Tony. |
||||
Grogster Admin Group  Joined: 31/12/2012 Location: New ZealandPosts: 9308 |
Thanks to all who replied. I now have a much better understanding of this. I had totally forgotten about the gate capacitance. Special thanks to Warpspeed for his excellent and lengthy post above, which I have read a few times now to fully understand it. Sounds like four big juicy MOSFET's are the best way to do the H-bridge, then you only need one stage. I have downloaded the Microchip 12A driver PDF - thanks for that link - and I will have a read. Perhaps THAT driver, coupled to the 8010 controller, and four big MOSFET's is a good yet simple combination? Warpspeed - that seems to be what you are suggesting, yes? Smoke makes things work. When the smoke gets out, it stops! |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
That is how I would go about it. The 8010 control board is a completely unknown quantity, and it may have some unresolved problems, I really do not know. There is no real way of knowing how it actually works, and I have never actually seen or used one myself. But anyhow, the IR drive isolator should be fine, and beefed up with some decent high current gate driver chips, would be about the best you can do as far as gate driving goes. Four big mosfets mounted on a very large flat sided heatsink with very short interconnections made from copper bar or 3mm aluminium plate, and with a large electrolytic with screw terminals bolted down to that directly, all with absolutely minimal lead lengths would again be about optimum. The only problem then is not to overstress the mosfets. The mosfet voltage rating is fairly obvious, and its a hard limit that cannot be exceeded without failure. Voltage spikes absolutely must be taken into account, and compact physical layout has more to do with that than anything else. Current is less obvious, but realise that the mosfet maximum current rating is a fictional rating based on a 25C junction operating temperature that is physically impossible to achieve. Its just the industry standard way of doing things which bears no relationship to reality. Maximum safe current rating is based on die temperature and bonding wire temperature. Here is a thought experiment for you. Suppose you take a normal ordinary ten amp fuse and slowly increase the current until it blows. Suppose it goes pop at twelve amps. What has actually happened ? As we slowly raise the current, the fuse wire heats up to melting point where it open circuits. Now suppose our fuse melts at a hypothetical 500 degrees C (This is just a guess). So at room temperature our fuse can safely carry ten amps continuous, and that is how we rate our ten amp fuse. Now what do you think would happen if we put this fuse in an oven and run the oven up to 400 degrees Celsius ? Its going to reach the 500 degree failure temperature at a lot less than ten amps. Maybe only a single amp or less might blow our ten amp fuse. So we have to derate our ten amp fuse with an increase of operating temperature. At room temperature the current rating will still be the basic ten amps. At 500 degrees our fuse rating will have dropped to zero ! It will very likely fail and go open, even with no current flowing through it. We have to derate semiconductor devices in the exact same way. For silicon devices the ultimate failure temperature is always 150 Celsius. As with our hypothetical ten amp rated fuse, we rate semiconductor maximum current at 25 Celsius which is the industry standard. So our 320 amp rated mosfet can only carry 320 amps safely at 25C which is impossible unless you cool the heatsink to well below 25C. Its the junction temperature that matters, and that is always going to be hotter than the heat sink when its carrying current. At the high end, if we run our junction temperature up to 150 C the maximum current rating falls to zero, and the device will fail without any extra help. So we need to think about all this and if we know (or can estimate) the junction temperature under operating conditions, we can establish what might be a safe maximum and stay well below that. Half way between 25C and 150C is 62.5C which is probably about as hot as you would want to go, but at that temperature, our maximum (failure point) current has also halved from 320 amps to only 160 amps. A SAFE maximum to work at would be something below that. You can probably see that a larger heatsink, a fan, or lower switching losses enable a higher safe current to be used. But it must always be below the rated maximum which is the anticipated failure point. How much power ? I really do not know for sure, but measuring rms current waveforms and measuring heatsink temperature might give a pretty good idea. Then you work out if the safety margin is something you can feel comfortable with. If you can keep the heatsink below 60C (very difficult on a 40C day) failure point might be 150 amps maybe. Probably 5Kw from a 48 volt system might be about the limit if everything else is working right. But those 320 amp mosfets also have bigger brothers with higher ratings. Its just a case of money.......... I would start out with the small ones, at least to begin with. Surprises will be less expensive that way. Cheers, Tony. |
||||
| Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308 |
Okey dokey, so therefore the chain is: 8010 controller-->IR2110 initial driver-->TC4452 gate driver-->BIG MOSFET final drive. I am on the verge of buying some of the 4452's and a few of the big beefy MOSFETS. If money was no real object, what IXYS MOSFET block would you go for? Smoke makes things work. When the smoke gets out, it stops! |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
If I was doing this I would start out with four small ones and first get it working at low power. Gradually increase the load while monitoring all the voltage waveforms, current waveforms and heat sink temperature. If something is obviously not right, I would stop at that point and try to work out why what I am measuring is not what I expected to see. Go at it slowly and get a feel for it under continuous full load. That might be difficult if you are battery capacity limited. I would try to tie all the numbers together, input power, output power (efficiency) and various temperatures and voltage drops. Try to work out what the mosfet junction temperature is, because it cannot be measured directly. If at some point it blows up, I would probably at that stage have at least some evidence as to possibly why, and make whatever changes. Suppose for instance the heat sink temperature is rising dangerously. Fitting higher rated mosfets only means it will go slightly longer before blowing up anyway. A bigger heat sink would be a much better solution, or maybe fitting a thermal switch and a fan. If everything is running great, but you still need a bit more power, larger mosfets may be the answer, but only if everything else is working properly. If the transformer is almost on fire, or you have massive voltage drops around the system, or some other limitation, that is what needs the attention. One other point worth mentioning is test the entire drive and control system with a very slowly rising and falling supply voltage. Some things work great at +5 volts. But what happens to a microcontroller at +4.5v or +4v or +2v ? If its designed properly it will start up and shut down cleanly and safely. If its crap, it might start do some randomly awful things just before it dies. Check the gate drivers, most have an under voltage shut down features, check any protection systems, over current, under voltage input and so on. You want your inverter to be absolutely bullet proof when it starts up and when it shuts down, some things can go seriously wrong at very low supply voltages. And the supply voltages are always low as they come up from zero and fall back to zero at turn off. Supply voltage sequencing can be critical too. So think the whole thing through very carefully, how it is supposed to start up and shut down safely. It is a very different thing to it just running continuously with all the various dc voltage regulators working. Cheers, Tony. |
||||
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Thank you for all that info warpspeed, it is much appreciated. With your input I might get that inverter beast going yet.  Its now a two pronged construction, one with 4 x 4 HY4008 mosfets and the other with 4 IXTN320s. Both on the big aerosharp heat sink and the better one gets the job. Grogster, RS components stock the TC4452, they come in a bag of three??? So I ordered 2 bags (6 pieces) which came to just under $AU20.- I'm going with the IXTN320N10T which should be good for 5KW, perhaps a little more at turn on peaks. That's plenty for my little house & workshop. Klaus |
||||
| Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308 |
Hi Tinker.  I will probably use the IXTN360N10T, which is about US$1 more then the 320's, cos the 320's look about to become obsolete. The TC4452's are about US$2 a pop from Microchip Direct MOQ of 5pcs though, so that us US$10, which works out as AUD$13 There are several different package types to choose from, but I will probably go for the standard DIL or maybe the SOIC - not sure which package just yet. I have visions of a small PCB that bolts directy to the MOSFET block's gate screw terminal, with the resistors and diode along with the 4452 to keep all the gate-drive parts as close to the MOSFET block's gate as possible, as Warpspeed suggests. Smoke makes things work. When the smoke gets out, it stops! |
||||
| Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Grogster, don't forget the shipping costs, they are usually the killer with stuff from the US. I very much doubt you can get 5 pieces for 13 bucks AU$ landed at your door. I will do a similar PCB, power tracks are 1.5mm thick copper strips which screw directly to the terminals. The big caps also have terminal screws so they will be easy to connect to the strips. I think layout is everything here, suggest you read as much on the subject as you can get hold of and don't be afraid to do several versions of that PCB. Also, it might require a beefy 12V supply for these gate drives, I will use a separate one from the controller supply. Klaus |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
I suggest you go for the DIP package and put it into a good quality machined contact socket. It will make any future repair and fault diagnosis much easier. Cheers, Tony. |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Thanks for that link Grogster, if I ordered 8 today they say they would arrive on Thursday so only 4 days with those coming from Thailand. Price is as per below  However RS are a little more expensive but have free delivery, if they had 9 (lots of 3) in stock it would arrive the next day but the still would have them here in 5 days.  Then there is Aliexpress, but you need a lot more patients.  There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
The power required for these drivers is absolutely minimal. Yes its ten amps plus, but only for some tens of nanoseconds. Just hook up the 4452 to the same voltage sources that the output side of the IR isolator chip uses. The same voltage pump that powers the high side of the IR chip can power the high side 4452 driver. The 4452 will need at least a 1uF tantalum capacitor directly across its supply pins. Power from that (ten amps) is used to charge the 22nF gate capacitance. As the 1uF is around forty times the gate capacitance there will be hardly any voltage dip on the 1uF. The average power required to keep the 1uF fully charged is very small. So don't worry about it. Just keep the track lengths between the 4452 and mosfet reasonably short because any extra inductance there will promote ringing. A 1 ohm gate resistor will limit the peak gate current and help damp out any ringing. Cheers, Tony. |
||||
| Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308 |
Okey dokey, thanks guys. @ Tinker - Shipping(DHL) from Microchip Direct is actually very affordable, as you can see from the images that Madness posted above. You are 100% correct though - that is NOT normally the case, and shipping from USA can indeed be a killer, but as pretty much all the Microchip Direct stuff comes from their warehouse in Thailand, this makes the shipping very affordable - especially if you are ordering several things as I often do with Microchip Direct. @ Warpspeed - I will get the DIL package version then and use machined sockets. Excellent additional advise about the power supply to the 4452's too, I will take it under advisement. Smoke makes things work. When the smoke gets out, it stops! |
||||
| Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Thanks for clarifying this Warpspeed, I shall use my little 60V to 12V DC-DC converter then, to power it. Just out of curiosity, is there a particular reason why a tantalum cap and not an electrolytic one? If I find a bigger tantalum cap (2.2uF) is that even better? Yes, the tracks from the 4452 to the 1 Ohm gate resistor and gate will be in the millimeter region, I learn t the hard way that long gate tracks are not a good idea with power Mosfets. Klaus |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
The application notes (page 10 section 3.1) suggest a 1uF ceramic capacitor as being the minimum. Its plenty for driving a 22nF load at 20 Khz. http://ww1.microchip.com/downloads/en/DeviceDoc/20001987C.pdf These ceramic capacitors are fairly new technology, and are only available in surface mount package, which may be less convenient for us. The older style through hole tantalums should work just as well and are much more common and easier to use for home construction. The main feature of both these capacitor types are small size and very low self inductance. They work exceptionally well at high frequencies. Aluminium electrolytics are of a wound foil construction and become inductive at higher frequencies. They also do not like supplying very high short interval pulse current. They may run hot or eventually go open circuit. A low esr electrolytic would be much better, but still not as good as a ceramic or tantalum for this particular application. Cheers, Tony. |
||||
| Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Thanks for that Warpspeed, its just because on page 11 in the diagram they show a 4.7uF capacitor what made me curious why. I take the point about non inductive capacitors. Another related question, I have some neat 5W (I think) 1 Ohm resistors for the gate, they are quite small too. Would these be suitable or has it to be some non inductive carbon film type? Klaus |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
My Inverter is still running quite happily for over a week and a half now. It has spent quite a few hours above 6 KW for several hours each time while running in a 40-degree environment and does not get over 55. I still have not completed the gate driver mods due to other priorities including Air Conditioning the shipping container where the Inverter, batteries etc live. Although the Inverter is happy in the heat I am concerned about how it will affect the batteries. While doing this a 5 KW AC compressor I use as a vacuum pump (yes I know it massive overkill) stalled and tripped a 20A breaker with no bother at all to the Inverter. This was with the house running off the Inverter as well and no dramas it just kept on Inverting. At this stage, I will not be contemplating the 4 big MOSFET approach as the 5 KW limit is too wimpy for me, besides I have heaps of boards already with 24 MOSFETs on them that I am trying to get running right. I have ordered some 4452 driver chips also as they do the same as the driver boards I am building with 4 MOSFETs but in a much more compact and simple layout that could easily be added to the 24 MOSFET power boards. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Yes, those TC4452 chips certainly look a lot simpler to implement than the totem pole drivers you mentioned in an earlier post. I was looking at those totem pole drivers too but fortunately the subject of compact high powered Mosfets came up here just in time to drop that idea. Klaus |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
The actual capacitor value is not at all critical. 1uF would be about minimum, 4.7uF or 10uF will work fine too. The 1 ohm gate resistor can be an ordinary 1/4 watt metal film. It absolutely must be a carbon or metal film type, not wire wound ! The power during normal operation is absolutely negligible, so the smaller the better. Cheers, Tony. |
||||
oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686 |
Mad it is good to see your units showing the same toughness mine have.... but your are running much harder for longer than my units... I'm impressed. I too will stay with the multi smaller fets. I like the hy4008 so far no problems at all. I tried the longer dead time, but the wave looked worse for wear, so went back to fastest. You really work them hard. ..........oztules Village idiot...or... just another hack out of his depth |
||||
| Page 12 of 22 |
|||||
 Print this page |
