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.
oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686
Posted: 02:16am 11 Sep 2016
Copy link to clipboard
Print this post
Eh..another day... another explosion ..... same ole same ole...
If your working with mosfets, it is best to buy them by the sack full...
And it goes without saying, that you should do so in a sound proof basement... so the neighbors can't hear em scream.
...........oztules...
Village idiot...or... just another hack out of his depth
Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308
Posted: 02:20am 11 Sep 2016
Copy link to clipboard
Print this post
You masochist!
However, no inverter progress without pain.....
Smoke makes things work. When the smoke gets out, it stops!
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498
Posted: 02:50am 11 Sep 2016
Copy link to clipboard
Print this post
No pain, no mains (off grid doesn't rhyme)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
Posted: 04:08am 25 Sep 2016
Copy link to clipboard
Print this post
So I connected my inverter to the battery bank and switched it on today.
Well, it soft started and settled on a steady output voltage (235V).
I had an oscilloscope probe connected to the transformer secondary and another probe to the output of the inverter.
The upper trace is the output while the lower is the input to my filter circuit, for which I posted a picture of earlier.
So, despite oztules misgivings, my filter idea seems to work.
That was with no load connected.
I then connected a 800W bar heater as load and instantly got a spectacular bang, flash and magic smoke from somewhere on the Mosfet board.
Hmmm, it looks like this thing needs more work.
After extracting the heatsink and attached PCB's, which is quite easy with my design, I looked for the cause of that magic smoke.
Not very obvious at all, at first I thought one of the 10,000uF capacitor had gone as it has a small dent in the can. But maybe not so, checking with a pencil torch showed scorch marks around the legs of one mosfet. Perhaps a stray solder blob stuck there, I will have to unsolder this board (one of 4 mosfet boards) to inspect it closer.
Klaus
oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686
Posted: 10:50am 25 Sep 2016
Copy link to clipboard
Print this post
Something was very wrong in there, that wave is staggered, should be perfectly smooth, and the secondary trace is very messy, should have been near perfect there too.
There is some wicked noisy silicon or feedback into the wrong places ( noise... floating gate in there somewhere?).... will post traces at various places later today, but none of the traces I have posted previously had any filters... and they don't show any fluffy waveforms on the secondary like that.... check you signal lines waves, and at the gates.
Wondering if your filter is causing this ( extremely unlikely... but... ).... so when you fix the fet/s, try without the filter connected, and look at the waveform, should be very good anyway. It may be masking a problem, that will be easier to see without it connected.
The raw wave with the 2uf shaping caps will look very good on that definition, and will only show a fine sawtooth imposed on the sine wave..... that bottom one is just plain wild.
There is some serious instability there to find.
Do you have the shaping caps/s there installed? ( the 2-4 uf across the secondary ).
........oztulesEdited by oztules 2016-09-26Village idiot...or... just another hack out of his depth
oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686
Posted: 12:17pm 25 Sep 2016
Copy link to clipboard
Print this post
Here is a pic from the one I am playing with for current shut off... single fet/bank
The fluffy appearance is not there in reality, but at this frequency, the scan is flickering in my old girl... it is actually a clean scan... but you will get the idea.
This is the direct output of the transformer with a 2uf cap shaper only, no other stuff connected.. compare this to your secondary scan.... something is very wrong there.
When the batteries in the camera recharge, will do a few more from different places in the circuit if I get a chance.
If you got flash, and no fet has opened up or exploded, then there is a chance it is foreign material causing the bang, but the wave concerns me.
.....oztulesEdited by oztules 2016-09-26Village idiot...or... just another hack out of his depth
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904
Posted: 08:53pm 25 Sep 2016
Copy link to clipboard
Print this post
Thanks oztules, makes perfectly sense what you say.
Perhaps the "2uF shaping capacitor" is too far away from the secondary, its connected actually after that filter....
It may be a few days before I get time to unsolder the copper busbars to see the mosfets close up.
I already have plans to make that access a little easier.
And I have a 2uF cap spare so I'll locate that one right at the secondary terminals.
I did not expect that experimental version to work perfectly right first try and it did not let me down there .Klaus
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498
Posted: 09:31pm 25 Sep 2016
Copy link to clipboard
Print this post
Oz have you tried the cap that is across the Toroid in the Aero Sharps? I think that is a 5uf?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
Posted: 10:24pm 25 Sep 2016
Copy link to clipboard
Print this post
Mad, no but it should be fine EDIT!!( the 5uf ones, not the 10 uf ones.... (white) from the 1.5kw units... or 2 tens in series.
Tinker.... I suspect thats the problem, the VFB was getting 20khz noise as a first harmonic... and this would have confused the cycle spwm control for voltage.... your measured 240v or whatever, was controlled by noise, and when you connected the load, it probably damped it, so the control circuit took off trying to make up for the lost spikes.... full spwm into the stiff transformer may be the answer here.
The spwm would have driven those multiple noise lines because of the wild VFB signal perhaps... only conjecture, but seems to fit the wave form in theory.... maybe...possibly....
Can't see how else you would have such strong hash, it was as strong as the primary wave in amplitude.
.......oztules
Edit IMPORTANT!! Yes the feedback ( VFB ) MUST come directly from the secondary wire terminals themselves... or it WILL blow up when load is applied.
By putting the feedback take off after the filter, you have introduced damaging hysteresis, causing extreme over shooting in both directions... seen as the fluffy wave forms before the filter. Edited by oztules 2016-10-01Village idiot...or... just another hack out of his depth
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904
Posted: 12:11am 16 Oct 2016
Copy link to clipboard
Print this post
It would be nice if I could report success at last but no, another setback occurred . More about that in another post.
Suffice to say that my inverter is again in bits now and I'm testing each individual part - with surprising results.
In a tread a while ago it was mentioned that the 10.000uF/100V capacitors we use are not the best choice. So I tested each of mine (I have 2 batches, one Rubicon and the other Nippons)with my hi Volt DC supply.
Connecting DC to a capacitor should result only in a charge current flowing, once fully charged the current should stop.
Not so with my capacitors, I ran each one up to 90 Volts and more than half started passing current above 70 Volt. Some over one Amp above 80V, my hi Volt supply was maxed out there.
Not good when the operating voltage is around 60V, not good in my book anyway. So I ordered the Elna type (with screw connections). These, being about twice as expensive, I hope will pass my test, time will tell.
Just in case anybody is wondering, yes I did monitor the DC voltage on my CRO. There was a slight ringing noise, a few mV @ ~20us/div which is from the hi volt DC/DC converter. That noise was a lot worse in the bad capacitors.
Another test that might interest someone here is how I set the C/T trip point. I decided on a 30Amp setting which should cover two or more inductive devices switching on exactly at the same time. I do not have a 240V device that would draw anywhere near that much current, beside my 10KWh lithium battery bank would not last long then.
Anyway, I needed 30A AC to set the C/T. It occurred to me that my old stick welder might do the trick as these can handle a short circuit at the secondary. So, at the lowest setting (it is a tapped coil device)I got 60Amps AC short circuit current. Too much for my test.
But when connecting a 2.4KW fan heater in *series* with the welder's primary the available settings got in range. On one setting I got 26.5A and the next gave 31A current - perfect.
So I set the trim pot to reliably trip the SCR when 31A flowed through the C/T and never trip when 26.5A flowed.
This test is quite easy, nothing got hot (I had used decent size wires) and I could take my time doing the adjustments. It tests the complete SCR circuit, including flashing LED and reset button. I did *not* plug in the EGS002 board for this test.
I will now take a leaf out of oztule's book and build me a 4 Mosfet inverter for testing purposes. I do have a small toroid, out of a 1.2Kw/48V Latronics inverter that should do nicely. Blowing 4 Mosfets is certainly less stressful than blowing 24 if something lets the magic smoke out.Klaus
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498
Posted: 01:55am 16 Oct 2016
Copy link to clipboard
Print this post
Hi Klause,
Your not the only one to hit a brick wall, I have blown up quite a few MOSFETs.
First off my CRO was not working and I had picked up somewhere the wrong values for the snubbers(103 3KV cap and 2W 47r is correct), had a few spectacular failures from that. Flying blind made it even worse, I even made up an easy way to swap MOSFETs without soldering.
Since doing this I have only had to change the MOSFETs once and that was due to over heating with no heat sinks. I know it looks like crap, that board was actually in the bin at one point, my PCB making skills have improved a 1000% since)
I think I am on to what the issue is, if you look on the design made by Clockman based on the Chinese boards there are 4 tracks that run around the left-hand side of the PCB and 2 drive the MOSFETs, the other 2 are connected to the left and right bridge. When I first saw that and was redesigning for my own reasons I thought why would they run those tracks to the left & right bridge when they are so close to the connector on the PCB. Now believe there is a reason, here is what I am getting from the board shown above where the tracks just take the shortest path.
One is from L output and the other is from N output, another board with a different layout similar to the Chinese boards but not with all those 4 tracks up the side is working. However, it has a similar wave-form to above but also has a bit of squiggle in the sinewave on the slope up from zero crossing. This resulted in fairly loud buzzing in the Toroid and choke, with the little test board and the wave-form shown above it has a very slight hum.
It has come to a standstill at the moment as I am waiting for more MOSFETs and blank PCB to arrive so I can test my theory, here is the design I will trying next.
I was waiting to try it again and post the results then.
Gary Edited by Madness 2016-10-17There are only 10 types of people in the world: those who understand binary, and those who don't.
Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308
Posted: 02:09am 16 Oct 2016
Copy link to clipboard
Print this post
Pain the posterior.
My batch of ten of the Rubycon ones arrived last week. I have not installed them yet though, as I am waiting on the power-board to come back from the PCB house.
While I wait, I will run tests on my ones.
That means they are crap if they pass that kind of current while still being inside the voltage rating printed on them. As you no doubt know, you should be able to run them right up to 100v DC without them sucking any current once charged up. If they are drawing amps past 80v even when charged, then the cap is breaking down under the voltage - certainly means they are NOT 100v rated if they do that. Perhaps they are really 63v caps, marked and sold off as 100v ones.....Smoke makes things work. When the smoke gets out, it stops!
oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686
Posted: 10:15am 16 Oct 2016
Copy link to clipboard
Print this post
You sure don't learn anything when thongs just go right from the get go.
I used the power jack power boards with the resistor and diode mods for my first power experiments, and all went swimmingly..... didn't know how lucky that decision was to be.
Had I had the problems you boys have experienced I don't know how far I would have gone, thinking it just might not work... and gave up on it.
As it is there are plenty of running version around the place now...... one that has been running for months now on a remote island.. very glad that one works flawlessly,as it runs the whole shebang down there.
It is looking like the power board may be the key... just hard to believe with the run of success I have had.... Even the two or three experiments with the flat heat sink designs worked fine up to at least 7kw surges, and 4kw continuous.... tests finished as I decided to go with the PJ style.
Sometimes it pays to be lucky
The new boards I got made ( basically the same as the pj 8000w units) work fine too..... so we know that design works perfectly well also.
Thats this one:
.....oztulesVillage idiot...or... just another hack out of his depth
Pete Locke Senior Member Joined: 26/06/2013 Location: New ZealandPosts: 181
Posted: 07:08pm 16 Oct 2016
Copy link to clipboard
Print this post
Interesting thing with Electro' caps. If they have sat for an extended period of time with no charge in them, the electrolyte barrier shrinks. Your, for example, 100uF 63v cap can test up to ten times it's stated value. But with a HUGE reduction in working voltage. We strike this with older VSD's that have appeared from a dusty corner somewhere. Providing the cap' isn't shorted, it can often be rejuvenated by sitting it quietly somewhere with low voltage applied for a couple of weeks. Just because you bought them new, doesn't mean they haven't been sitting on the store shelf for some time. Also interesting is to Google 'Capacitor Plague' something we have also been victim of with old school 24v instrument power supplies. Cheers Pete'.
Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308
Posted: 08:20pm 16 Oct 2016
Copy link to clipboard
Print this post
Oh yeah, the "Leaky caps" problem.
That was rampant in the days of the 1GHz or so single-core computers, but I figured that they would all have vanished by now, but you make a good point - perhaps some of those are still being on-sold in China as old stock. Would not surprise me one bit.
I will run some tests on my caps ASAP, but yeah, that could be why Madness is seeing such a large conduction above 80v. If the caps use crappy electrolyte and/or incorrect construction for that voltage rating, that would explain the rather large leakage current. Still should not happen though, if the cap was truely rated as per the label.
I bought my caps from AliExpress, but I might just have to bite the bullet and get my caps from Element-14. Much more expensive, but I guess they are fully factory-approved genuine parts - you get what you pay for, as has been said many times before......
I will let the thread know the results of my testing - as soon as I can find my variac.Smoke makes things work. When the smoke gets out, it stops!
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904
Posted: 11:40pm 16 Oct 2016
Copy link to clipboard
Print this post
Thanks for your input oztules, always appreciated here.
Well, I know the circuit works for you but find it difficult to believe that running the tracks differently should upset it.
I do have all my gate drive signals running the same physical length (mostly shielded cable)and the power tracks as short and beefy as possible.
I also know that the latest failure was most likely caused by myself by not noticing the tiny craters in the top of the driver & controller IC's and plugging in the same EGS002 board that was used on failure #1.
As you see, its a big learning curve and, since this is a non urgent retirement project, part of passing the day here.
So I have a question for you. As mentioned above, I'm building a 4 Mosfet inverter for testing purposes. It will closely resemble the layout of the big one and use the same auxiliary boards to drive it. I removed the wrapping from my little 1200VA - 48V toroid and did some testing to see if it is suitable for that.
This thing has 65 turns (2 in hand of 2.3mm dia wire) for the primary. Powering it from the 240V winding I see 35V AC on that 65 turn winding. This is a lot higher than the 26V AC my double stack big toroid generates.
So, do I just unwind 16 turns to reduce the primary voltage to 26V?
I'm almost certain the answer is yes but its always good to have that confirmed by the forum's expert on this matter.Klaus
oztules Guru Joined: 26/07/2007 Location: AustraliaPosts: 1686
Posted: 12:21am 17 Oct 2016
Copy link to clipboard
Print this post
Expert?????... thats a bit generous I think.
Not sure about shielded cable for a gate drive, any EE's out there with an opinion?
I think remove 13 turns and go to 28v for testing, slightly less stress on the fets, but more drive on the driver chips. The chinese with the 002 units like 26v, I still think 28v is acceptable too for most battery voltage in the 48v band.... my heavy power is during the day, and b+ will be 55v or more as a rule.
I cant say how much layout is actually effecting the outcomes here, but some of you out there have had a bad run of outs, that I thankfully didn't see... and I don't know why at all now.... but there are quite a few running here now in actual off grid applications, running whole houses..... I struggle to see why the differences... and the 002 board has less reason to fail, as there are less parts attributabe to the builder than the 8010 boards... ie if it cant be the 002 board, then it must be the fet board or the drive lines in between.
The driver chips are the iffy part so far.. I have had a few duds out of the box
The 4 fet card should help get it sorted I suspect, as it will do 1kw at least.
Let us know what you find works... that did not before.
............oztulesVillage idiot...or... just another hack out of his depth
Grogster Admin Group Joined: 31/12/2012 Location: New ZealandPosts: 9308
Posted: 12:38am 17 Oct 2016
Copy link to clipboard
Print this post
[Quote=Tinker]I know the circuit works for you but find it difficult to believe that running the tracks differently should upset it.[/Quote]
So do I. The actual resistance difference between a short and long track would be negligible I would think, when you are talking about the tiny currents being used for the gate control lines - does not make sense to me either.
[Quote=Oztules]Not sure about shielded cable for a gate drive, any EE's out there with an opinion?[/Quote]
I'm not sure about that either. Not sure in the sense that I don't think you should do it. A MOSFET gate is a very high impedance pin, and shielding the gate line would add a small capacitance to the gate signal. While that may not prove to be much, it might prove to be enough to upset the gate signal. The shield and the core of shielded wire, will form the capacitor. Digital signals are happiest if paired with a matching ground wire(to reduce crosstalk) when run in that IDC cable stuff - I don't know exactly how Tinker was shielding his gate lines though - perhaps he could show us all a photo of his "Shielded" gate lines?Edited by Grogster 2016-10-18Smoke makes things work. When the smoke gets out, it stops!
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904
Posted: 01:17am 17 Oct 2016
Copy link to clipboard
Print this post
Sorry, no pictures at the moment. I used figure 8 shielded audio wire. The shield is connected to ground The other two are the gate drive and the VS feedback.
Keep in mind here that I have separate 4 Mosfet boards, one for each arm of the bridge. So the signals above need to go to each board this way:
1HO VS1 Ground
1LO Ground
2HO VS2 Ground
2LO Ground
I can easily replace the shielded wire with 3 & 2 wire ribbon cable as the wires terminate at each end in .1" header connectors, if I was convinced that it would be a better way. Klaus
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498
Posted: 02:12am 17 Oct 2016
Copy link to clipboard
Print this post
As yet I have not tested my theory, it will be a week or 2 before I get the parts needed. I have tried as much as I can to eliminate all other possibilities by doing the following.
I have tried 3 different transformers
My big double stacked toroid A single toroid wound with 22 turns on the primary A EI transformer from a dead 2KW square wave inverter (also much more forgiving than a Toroid for testing)
All of these give much the same wave form, so can all 3 have the same problem?
Same thing with 2 different chokes, one which the coil was made by Tinker, the other using insulated cable, the one with the cable did give a fuzzy wave-form if the cable was not kept secure.
Also, I have tried 3 different control boards, all give the exactly the same result. So can all 3 have the same problem? One of which came from Oztules.
(BTW I have learnt the hard way, any sort of issue or first use, connect the control board to 48V ONLY and turn on, you have 3 seconds to check wave-forms on the IR2110's before it shuts down due to under volts.)
That leaves the power board, the results of the of the rough test board are shown above. It is fairly good apart from something wrong in the timing or level at the time it crosses 0 volts. That board is a design I have long since abandoned. Another board very similar to the design in red and blue above. It gave a slightly worse result than this one from http://www.anotherpower.com/board/index.php/topic,1116.105.html shown here
In my case, the distorted part at 0 is followed by a by a squiggly distortion of the wave that disappears just before the peak. when running it causes all 3 transformers and chokes to buzz quite loudly. This PCB is exactly the same circuit as the boards Clockman is making, biggest difference is the track for right bridge is connect by the shortest route, the one connecting to the left side goes up to the top left and connects.
I have tried different cables, transformers, control boards, chokes and so on the only common factor is my battery bank.
If it is not something to do with the power board I must be completely insane. It does not make sense to run those tracks up the left side of the PCB to connect to the bridge, it is the long way around, it is much easier to connect to them on the same side as the 10 pin connector. There must be a reason why!Edited by Madness 2016-10-18There are only 10 types of people in the world: those who understand binary, and those who don't.
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.
Print this post
..... same ole same ole...
. More about that in another post.
Perhaps they are really 63v caps, marked and sold off as 100v ones.....
