 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 10 of 22   |
|||||
| Author | Message | ||||
oztules Guru  Joined: 26/07/2007 Location: AustraliaPosts: 1686 |
Mad, Further testing with the inductor gives me a huge range of weird stuff. It looks like the noise in my transformer is induced by the inductor. By changing the inductor form iron powder to ferrite, from ecore to ring and changing the turns.. the noise and waveform imperfections changed radically. On one setting with the ring and 10 turns the tranny was so quite I had to put my ear next to it to see if it was running. With a different ring and 10 turns, the wave from looked like your trace at idle, and sounded like a chaff cutter at idle, but straightened up a lot when loaded. So it looks like that with removing the 4u7 and fiddling with the inductor, we can change the waveform markedly. It may be I am on the edge of decency with regards to drive, but gee, it gets very good by fiddling the inductor, and removing the 4u7 on the VFB line. more to follow as I find out more.... gee it would be nice if I knew what I was doing too  ...........oztules Village idiot...or... just another hack out of his depth |
||||
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Looking at that spec PDF in the link above I notice a diagram with one input and two outputs from that chip. So, with using a EG8010, I assume one requires 4 of these chips to drive the Mosfets? Then, what is one going to connect to the extra outputs from the TC4452. Just curious and trying to learn new electronic tricks. Klaus |
||||
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Thanks OZ and Warp, The Aero-Sharp GTI could be sick, after all they were all recalled units. I have plenty of spare bits to play with. When it was working the watts it was supposedly making did not seem to match up with its daily total, ie it was making around 300 plus watts for 2.4 hours but total daily output was 0.2 KWH. One thing that was interesting is while it was connected it made the OGI 50% quieter, this happen as soon as the AC was connected and prior to it making any power. Another thing I noticed was the power meter I have in the OGI was showing a power factor of 0.6. Interesting about the Inductor, the one I am using ATM came from Tinker which is 3.5 turns of 4 pieces of enamelled copper bar. It is definitely a major source of noise, sounds much better now as I have removed it from being bolted to the inverter case where it sounded like it was possessed. I have the 100 mm Ferrite ring that I have played around with 6 turns on it that seemed to work well and was quiet. I have been trying to wind it with multiple strands of 2mm wire but that gets real messy. I might go to a welding supply shop tomorrow and get some welding wire for it. 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 |
Klaus, It has one input, and the usual internal N channel pull down mosfet, and internal P channel pull up mosfet. In just about all CMOS chips the drains of the two complimentary output mosfets are tied together internally, to produce a rail to rail output swing on one output pin. But in this chip, the two drains come out to two separate pins. Those two pins can simply be joined together and used in the normal way. Having two individual outputs (one pull down, the other pull up) enable you to use two separate gate drive resistors which can be made very different values. You can arrange it to have very different turn on and turn off switching speeds for your large power mosfet. That is no real advantage in most circuits, but for things like current mode flyback and boost converters, it can be handy to have a fairly leasurly noise free turn on, and the fastest possible turn off. You might want to do that if you are monitoring source current with a shunt or current transformer. You may not want to have a 12 amp gate drive pulse put a 12 amp turn on spike into your source mounted current shunt. So very limited gate drive current at turn on fixes that articular problem. Cheers, Tony. |
||||
| Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904 |
Did you locate the exact source of the noise? I should think the enamelled copper bar is stiff enough to resist any vibrations. The E core halves *must* be clamped very securely otherwise they definitely are noisy. I use 1mm neoprene rubber pieces under the top clamp to get an even pressure without stressing the ferrite. Also, the 3 1/2 turns, whilst they are locked together with epoxy also should not move in any way on the core once the choke is assembled. I did that on mine by epoxying dowel halves (cut like a D) to the inside of the coil turns since with such stiff copper bars there is no way to wrap them tight around the bobbin. I will experiment more with my choke once more Mosfets arrive to do some testing. I would say using insulated welding cable can make a tighter wrap as the many strands are much more flexible but its at the expense of a fair bit less copper cross section area. Interestingly, on the chokes I removed from that 'well known Australian manufaturer's inverter they used many in hand of relatively thin wire to cover the ring evenly all way around in a single layer. The whole lot must have got rather hot by the burnt appearance. The ring cores - 4 stacked ones - were perhaps powdered iron, they did not look like ferrite. I tried them with several turns of multistrand wire but had noise issues and a higher idling current. Klaus |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Hi Tinker, The inductor is all rock solid, It has wood pieces in each end and epoxied so there is no movement in it. The level of noise dropped by a very large factor by having it suspended by a couple cable ties rather than clamped to the Inverter case. So the noise did not come from something loose in the inductor. I am going to use the welding wire on a Ferrite Ring 100X65X20, it could take wire up to 30MM diametre. Using this with a single core Toroid I could get idle current down to 0.26A with 6 turns on it. I will get 90MM square wire which will wrap around the ring very easily. Once I get this Inverter sorted out I am planning to build another and use 90 mm wire for the primary, the 70mm I have on it ATM has room to spare and the wire can be compressed to fit more turns around the in circumference if needed. The OD between 70 and 90 is only a few mm. 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 |
That's interesting madness, I assume there was magnetic coupling to the metal case that made the noise. My E-core choke is bolted to a 12mm plywood base which in turn screws to the case. But so far the inverter did not run long enough under load to worry about noise from that part .I just ordered a ferrite ring too, a little bigger (120x80x25) than yours and a little more expensive. This one has a fair number of buyers for some reason. More room to wrap wire around too and another excuse for more experimenting. I wonder what happens if there is a choke in each primary leg? A new batch of HY4008's turned up today, all measured 3.0 mOhm and passed my 75VDC drain/source test. I did not push them to 80V as at 75V a tiny current (O.0000001A) started to flow which is as per spec sheet. We'll see if I get that thing to behave as expected with these Mosfets. Klaus |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Hi Tinker, I think the noise is a physical connection, plus the case is made of 2 Aero Sharp cases welded together so the part where they meet could be vibrating against each other making it worse. Yes it's frustrating when you spend a heap of time on something and as soon as you power it up it sh@ts itself. The Hy4008w's I got tested the same 3 mOhm also but they behaved just the same as the 4110's which is wy I am now working on the drive side as they all work fine with 3 FETs on each leg but not 6. I got some 95 mm welding wire today, that fits comfortably with 6 turns in the ring. But I have not had time to run it yet, I have Wednesday off so will get to it then. In the EG8010 data sheet they show using two chokes for a bipolar setup rather than the unipolar setup we are using. If you have not done it have a look at either side of the choke with your CRO, there is some serious white mans magic happening those things. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686 |
Have not progressed too much further, the darn thing works too well. With the 4u7 clipped off the VFB, and a small ring with 5 turns for the inductor, the thing ran 5kw and was virtually silent. 6 fets/leg 4011 No I don't know why, it is just perfect. This is using the inspire boxes... the curves were a problem but not any more.  It is running over 5kw, very very quiet, and the temp seems much lower than I remembered before the small changes. Will try to get more sens out of it/me tomorrow. ......oztules Village idiot...or... just another hack out of his depth |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Hi Oz, I have time tomorrow to do some more fiddling also, so off with the 4u7, I gave my new Ferrite ring choke a quick run but it was like a stair case on leading side of the wave. Suspect the 95 mm wire I used is not wound tight enough, so I will see if I can get it tighter, if not I will put cable ties on it to pull the wire on the inside of the ring out against the ferrite. Also I have a board ready to try with speed up caps across the gate resistors to see if they have any positive effect. Another job is to swap the main PCB in the Aero Sharp and see if it play more nicely. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686 |
So, you have found the inductor makes a big change too. I suspect the transformer may too ( only small I think), will test tomorrow or next. This unit was bit noisy compared to others I have built, ran very well, and kicked 20kw, but a bit noisy with a glitch shown before. As the glitch changed shape and size, the noise changed appreciably. Now, very quiet, except for heat gun and freezer. Like I said seems cooler for the same loads by about 10C... not sure if it's the weather or the 4u7 and inductor combo... or something else, but cooler is betterer. All units out in the field have had no failures, and some have been out there for near a year, with no interference from me. Every one seems to now own jugs, pie makers, microwaves and electric hot water services, NBN satellite modems ( constant 50 watts 24/7 ) and toasters etc...... and they don't seem to be afraid to use em. One of the island ones will soon have an extra 1.5kw all day for the desalination unit out there..... all very civilized. Hopefully that picture above will help those using inspires... the curves are not a problem.... just chop them off, and join like the aerosharp boxes. ..........oztules Village idiot...or... just another hack out of his depth |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Hi Oz, This photo shows one definite difference between your boards and mine. Those long legs, the board I have that is working the best is TO220 4110 FETs with legs about 12 mm long. All the rest are TO247 and they are in the board as far as they can go which leaves only 5 mm of exposed leg above the PCB. All these boards work fine with 3 FETs on each leg but with a full house they have sharp spikes in the waveform. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
You are right again Oz,  (don't know what you are doing??? must be hard to be humble when you are perfect in every way, at least I think that is how that song goes) finally was able to pay around with the Inverter again after injuring my back and was out of action for a week.By removing the choke all together I got a really good waveform regardless of the load on it. But any of the chokes version I tried today gave various levels of blips at zero crossing through to quite large spikes that affected the waveform all the way to the peak. I will spend some more time on it tomorrow to see if I can get to the bottom of it. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
The Inverter is running again, still not perfect as per the picture below and this can only be achieved with the maximum dead time. Short dead time results in nasty spikes and MOSFETs blowing to pieces.  It has been running like this powering the house for 4 days now without a hiccup. It even runs my abrasive metal saw perfectly that my old Trace 4.5KW refuses to start and shuts down with an overload error. I even did some welding today but only 2mm plate so not big current draw. But I have had it running 5-6 KW continuously for a couple hours with my 10 KW (3KW consumption) AC and 2400W hot water system running. The 4 X 140 MM fans were running but not fast enough to hear them. The vertical configuration seems to work well with natural convection, running up around 2 KW with 25 degrees ambient the fans are not quite on but the natural airflow out the top is quite obvious if you put your hand above it. Mid afternoon with 35 degrees ambient will get the fans running with 1 KW continuous load, fans are set to start at around 45 degrees. Overall efficiency ranges from around 90% to 95%. Next step is to add the Totem Pole drivers to the power PCB MOSFET gates and see if that improves the waveform. Nothing else I have found has resulted in a nice clean wave other than half the MOSFETs which is why I am suspicious that it is it something do with the gate drive.  The 100mm x 65mm x 20mm Ferrite Ring with 6 turns of 95 square mm welding wire has worked well and gives me about 20% less idle current than the E type which is now >40 watts with the big double core toroid. It is quite a bit bigger than it appers in the photo. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
yahoo2 Guru Joined: 05/04/2011 Location: AustraliaPosts: 1166 |
I agree, I know these transformers are stiff but I see inverters with butt-ugly output waveforms run just fine with all sorts of inductive loads. If you guys were measuring both voltage and current and could show things getting out of phase I might buy it but a few blips on a sine wave close to the crossing point is not doing it for me. I think you should be looking for something that is messing with the triggering timing on the mosfets. Maybe low current or a runt signal or something that gets them out of sync or maybe a secondary triggering. I'm confused, no wait... maybe I'm not... |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
I can get a really good wave with half the FETs, that is why I am suspicious of the gate drive signal.  I have all the parts now and have etched the PCBs to add the Totem Pole drive to boost the Gate drive current as shown on page 14 of this topic. When I get a little time I will put them on one of the other powerboards I have and see if that makes an improvement. Somehow this all just works for Oztules but I am not that lucky or it could be I have done something slightly different but am running out of options to improve it. However, it is running quite well as it is and the wave is certainly a lot smoother than the old Trace inverter. I am not losing any sleep over it as it is now. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Tried the Aero-sharp GTI again with a different PCB and still no luck. It works until it is making more power than what is bein used by the house and it spits the dummy. Tried a 5KW HF GTI I bought from the tip shop for 5 bucks and it trips the AC breaker between it and the inverter every time. I tested it connected to the grid and it works fine. So there must still be something that is not right, hopefully I will have the Totem Pole gate driver boards on test by the weekend. There are only 10 types of people in the world: those who understand binary, and those who don't. |
||||
| oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686 |
Dammit, this just makes no sense. I spent a day connverting a 3kw aerosharp to run off a 3kva generator using capacitors to soften the output so the mppt would work ok. The power factor must be horrible, but with 6x50uf caps, we get avbout 1.5kw back into the batteries. The inverter has been running the house for a month or two, and is a standard oz inverter with the 8010 and no fancy drivers.... It works fine no matter the house loads from 6kw down to 170 watts, and the continuous 1.5kw from the aero sharp runs the house, and charges the batteries. depending on the house needs. Max back into the battery was about 1.3kw, as the house had all the fridges and freezers and other loads quite at the same time, this seems to leave 170 watts that appears to be the minimum. It just works flawlessly, and just cant see how yours does not. The reverse feedback should be a walk in the park My wave form is good, but for that little glitch shown previously.... I'm not ferral about it, and agree with Yahoo about the wave forms we have got. Your inverter must be every bit as good, as it runs higher loads for longer. Mine does it's 15kw or more to start the big motors, but spends most time less than 3kw..... 6 and 7 for 5 mins perhaps when jugs go on with hot water etc. I'm just not getting it. The aerosharp made no noise or growls or anything else, neither did the inverter... something is wrong up there, and I can't put my finger on it...... the rotten things just work so well down here. .......oztules Edit.... the reason the grid tie has the caps is so I can run with any 240v generator without loading them down too far... particularly the AVR types which would probably head to stall, and blow the avr controller from over current trying to pull the voltage back up too hard. With the caps, I can use any ac input and get a result.... and by using the 2 mppt inputs with 150uf on each inpiut we can limit to 750 watts for even the little 2 stroke CMG 850 watt units.... but the power factor brings the usful input down to 670 watts max on testing. This will be used with the salt water desalination plant out on an outer island, and the wool shed where the desal is, is over 100 yards from the battery bank up at the house..... so we want to add panels at the woolshed, to help with the 1k4kw needed, and have a 240v back up a long way from the house, so it won't be heard in the dead of winter as it runs the heat pump if needs be... thats the plan, and it should be wired in next quite weather so I can get out there. Village idiot...or... just another hack out of his depth |
||||
| Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498 |
Yeah, I am not getting it either, right now it has around 5 KW load on it with AC and oven on as well, AC has been on for the last 2 hours and oven for the last 30 minutes. I went out to check it just a few minutes ago, the temperature in the shipping container where it setup is near 40, the inverter is in the mid 50's. As I have said you must be Irish and just plain lucky. As I have mentioned it is running with maximum dead time otherwise the waveform has several nasty spikes and blows the MOSFETs to pieces if left running like that. It is running everything fine and the panels I was trying on the GTI are a 1500W array. I normally just leave them connected directly to the battery bank in 4 strings of 2. For the GTI I change that to 1 string of 8, the end result is about the same amount of power even though their maximum power point is quite a bit higher than the battery voltage. So there is no urgency to sort it out but I will be working on the gate drivers very soon. 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 |
Its really impossible to speculate about all this, which I why I have kept quiet. Only solution is to hook it up to some really good test equipment and gradually work it up from no load to flat out. Make sure any protective circuitry actually works. Sometimes things are found to be working in a completely unsuspected mode of operation. It runs, but not in the way you think it is running. There can be some very high stresses somewhere that you may be completely unaware of. Maximum ratings of components cannot usually be used in combination. Everything needs to be run well within safe limits. Unfortunately most of us don't have the resources to really do all this this R&D in a proper methodical manner. Without the right gear you are really flying blind through the fog. I wish there was a simple answer to all this, but there isn't. Cheers, Tony. |
||||
| Page 10 of 22 |
|||||
 Print this page |
