 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 : Nano Power Inverter - Roll Your Own Style
  Page 4 of 16   |
|||||
| Author | Message | ||||
Murphy's friend Guru  Joined: 04/10/2019 Location: AustraliaPosts: 648 |
Hi Wiseguy, your design looks like I could copy some aspects of it, if I may. Can you tell me the values of D1,D2,D3,D5,D6 on the power board please? On your Nano board what EXOR? gate did you use for your clever opto interlock? I was wondering how to do that warpspeed trick with the Nano, thanks for figuring it out  . |
||||
wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
Hi Murph - Feel free to copy whatever you can make use of, I'm happy to share. The EXOR I used is an MC74AC86 which can source and sink 24mA for each output. I wired it so that the nano drive outputs travel through a gate and therefore all have the same delays. When the nano is "stopped" the outputs of the EXORs are all held at zero. From memory I believe all the output FET gates are held off like this and there are no spurious events to date during power up. I hope to finally put a few kW of power through my inverter this weekend. Looking at the circuit now I'm having trouble remembering it all, but confident it will all come back. If you want to experiment with anything you see send me an address and Ill send you whatever you want at original cost. (the main power board is the most expensive item ~$18) The nano control board $2 all other pcbs ~ $1-2 Edited 2019-10-30 19:59 by wiseguy If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648 |
Thank you for this offer but I'm a bit like you and "roll my own PCB's". If I may ask another question, if you were to use snubbers on the mosfets what value resistor & capacitor in series would you use? I note way back Oztules used snubbers on his ozinverter but I am not sure if the nano runs at the same frequency nor have the knowledge to calculate snubber components. |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
Sorry Murph only answered half the questions, D1,2,3,5,6 are all 1N5817 Schottky diodes. Snubbers are a subject all of their own. They are typically used to modify fast transients and slow down their rate of rise, or to kill ringing. This may also be to help reduce the D-G coupling, or may be just to reduce emi radiation. I currently dont have any snubbers fitted and am not sure if I will need them at all yet. Instead I have the fast diodes across the FETs internal diodes in an attempt to direct fast transients back to the supply rails. Depending on the reason for using snubbers there are some mathematical approximations to help chose appropriate parts and there are empirical methods, by observing the wave shape whilst trying various values to tame the transient & trying to achieve a critically damped result. Having an assortment of a few values between 470p and 10n and a few resistors of 4.7R to 100R is a fast way to get you in the ballpark of a correctly snubbed waveform. if you burn your fingers on a resistor whilst doing this the "c" is probably too high (along with the resistor value). Professor Warp is a great resource for other approaches to consider on various subject matters and probably will no doubt add a few more gems on snubbers from his vast experience. If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| BenandAmber Guru Joined: 16/02/2019 Location: United StatesPosts: 961 |
glad to see new people on here Some of the best the most talented people in the world are right here be warned i am good parrot but Dumber than a box of rocks |
||||
| Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648 |
Thanks wiseguy, interesting to learn that your fast diodes (D4, D7, D12- D21) do the trick instead of the snubbers. What type # diodes did you use and am I right in assuming 3 in parallel is to increase their current capacity? |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Snubbers come in various types and configurations to perform some very different functions. First thing is to work out why you need to have a snubber there in the first place, and what it is supposed to actually be doing. Three basic types. A series inductor to limit the rate of current rise, usually into a highly capacitive load when the voltage is very suddenly switched on. A shunt capacitor across something, usually to limit the rate of voltage rise across an inductive load when the current is suddenly switched off. A series resistor placed into a circuit to damp out ringing. Both the inductive and capacitive types of snubber always have a resistor associated with them to discharge the stored energy before the next switching cycle, and sometimes a diode as well. The values of all these components need to be very carefully chosen to suit the voltage, current, the absorbed energy, and the switching frequency. Like choosing any component, you first need to decide what it is supposed to do, how it is supposed to work, and then calculate (or experiment) to determine some appropriate values and ratings. So its really a question that has no very simple answer. A snubber is the solution to solving a particular problem. Before a solution can be suggested, the problem needs to be understood. If you can tell us what this snubber is for, why you need it, and something about the circuit it will be operating in. Edited 2019-11-03 06:46 by Warpspeed Cheers, Tony. |
||||
| Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648 |
Its for the nanoinverter, across the Mosfets. I do not know if I need them. Oztules used snubbers for his EG8010 based inverters (47R in series with 0.01uF) but I'm not sure the values he used apply for the nanoinverter too. Wiseguy uses parallel fast diodes instead, happy to try that if I knew which type to get. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
I have no idea why Oz or Wiseguy fitted these extra components, they obviously had some reason. Better to ask them. Cheers, Tony. |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
On page 1 I wrote why I did it, so Ill just copy and paste it here: "Why the 3 paralleled diodes between drain & source? Are the diodes not already inside each mosfet?" "I have made no inclusion for snubbers in my design - it will probably require some and they will be essentially in parallel with the diodes in question so there are holes available to put them in (or legs to solder them to). There is another benefit in not relying on the internal diodes of the FETs to conduct current internally in my opinion. The FETs are already shouldering the bulk of current switching, for every watt dissipated there is a penalty of higher RDS on generating more heat. I wanted to experiment with external diodes such as MUR420 with lower foward voltage drop to minimise heat generation inside the FET by passing current externally to the FET when it is not on & minimising the internal heat. I am also aware that in a final design the MUR420s may be under rated for full load and the TO247 U30D20C or Motorola equivalent MUR3020 will probably be required) Note: The MUR420 has a peak current of 125A so 3 can handle ~ 375A peak, with the body diodes of the FET also in parallel they should be able to tackle any nasty inductive current spike and then some. Another good fast (low Vf) diode is the BYV32-200. This is directly associated also with dead time. The diodes conduct during dead time. Dead-time is essential to ensure no shoot through. Excess dead time I consider an enemy. Efficiency is essentially maintained with excess deadtime but the cost is extra heat in the switching FETs. The sooner the FET is turned on after deadtime the better as then the diode (inside or external to the FET is not conducting & generating heat which is wasting energy. It doesnt actually matter as much that the external diodes may get hot, they become more efficient as they get hotter as the Vf value reduces further." Whether my final design keeps them in or not will be determined during rigorous testing and evaluation - remember this is a Rev0 board. Edited 2019-11-04 16:36 by wiseguy If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648 |
Thank you wiseguy, that explains it very well. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
O/k thanks for that Wiseguy. For interest I just looked up some data on a typical HY4008. https://datasheet.lcsc.com/szlcsc/HY4008W_C111008.pdf On page three it claims a typical turn on rise time (40v/100A) Tr of 18nS. And on page two the diode reverse recovery time (100A) Trr of 30nS. Both are very fast indeed and show how the technology has improved in recent times. Cheers, Tony. |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
Finally found the time to wire up my version of an inverter and see what happens. For a power source I used an Eltec Flatpack2 HE 48V 2KW power supply - actually they are an industrial charger for 48V systems, their voltage as delivered is ~ 53.2VDC. I have 2 of the charger units and I am building up a can bus controller for them as a work in progress. Eventually I will have 2 in parallel and can then bench test up to 4kW. I bought 2 Chinese DC circuit breakers one 32A the other 63A, as each Eltec supply is rated at 37.5A and I am just using the one, I used the 32A breaker, later when I add the second supply in parallel I will use the 63A breaker. The toroid I used was one of my ebay purchases and the choke is 3 of the hiflux toroids with 2 x 8 turn windings. For first power up I opened the breaker and using clip leads I wired across the breaker a 5A fuse which was promptly dispatched on power up - I thought the Eltec would have a soft start but obviously not soft enough. So then I used a 10A and powered up again, I was greeted with a nice sinewave that rapidly turned to crap just after the inverter soft start signed off and blew the 10A in a few seconds. Here is a picture of the distortion which was rapidly getting worse just prior to the fuse blowing.  I tried extra series choke but it had no positive effect current rapidly increasing with a growl from the toroid. Here is a picture of the setup.  I was confident that the inverter power hardware was not to blame but how to prove it. My hunch was that the voltage regulation feedback employed and the software did not like each other much. Then I remembered that Poida created a variac code and invited me to have a play with it. The Variac code does not try to regulate the output it simply dials in a variable PWM% value from a variable resistor that controls the amplitude. That is why there are 2 control boards in the picture above The one with all the bits was told to stand in the corner and the simpler variac controller was built up and duly connected instead. This is the resulting sinewave at 240V - much better.  Before I lose all this writing Ill post it and then add a bit more - the results are outstanding and exceeded my expectations. Edited 2019-11-07 20:36 by wiseguy If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
The nearest dummy load I had to hand was a 2kW fan heater with a half heat 1kW setting. Using 2 Power-Mates to check the 240VAC input energy to the Eltec versus output energy from the inverter gave the following results. Input Power 1088.8W output power 1020.6W giving an overall efficiency from mains in to mains out of 93.7%. The idling power of the Eltec is 24.8W so removing that loss from the figures yields an inverter efficiency of 95.9%.  After running for an hour the heatsink was still stone cold, unlike the Eltec, Toroid and Fan heater.. This is the proof that the hardware is behaving well so far and that the regulation loop & software function to regulate dont play well together. This is early days and I will be trying to sort out the voltage regulation/control before putting more power through the inverter. I was finally also able to test the impedance of the inverter (droop) from no load to 1kW, the voltage went from 240V unloaded to 231V at 1kW load - again better than I expected. Sorry for the big pictures - how do you resize them - I never used to have this issue? If you are skeptical of the efficiency figures, I don't blame you, I will swap the two meters (input/output) around tomorrow and see if I get a different result - however the cold heatsink tells me it might just be telling the truth. Edited 2019-11-07 21:50 by wiseguy If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
Whilst I am confident to make a simple transformer and add a few parts that enable me to have 3 isolated supplies to supply power to the 2 low side FET drive circuits and the 2 upper isolated FET drive circuits, Klaus and others weren't very excited by the idea. The truth is I have a few thousand RM6 core sets that are surplus to my requirements. However winding these little transformers is a pain in the arse. So I decided to employ some planar techniques and see if I could find a simpler way. Here is a picture of the cores against a $2 coin and the PCB coils I came up with.   The picture above has a core outline superimposed over the pcb winding, to illustrate where they fit together. There are 2 slightly different pcbs one has 2 x 8 turn windings the other has a 7 turn and an 8 turn winding. The drive circuit is for the 7 turn winding whilst the 3 x 8 turn windings have a slightly higher voltage to compensate for the diode drops. The intention is to use the 3 x 8 turn windings for the isolated Power board FET driver power. I hope the scheme works ok, I fitted 5 columns x 8 rows in the array of a 100mm sq prototype board. - I get back 200 off 2 x winding pcbs for a $2 outlay for 5 PCBs lol. (+ freight) I think that adds up to a 1c per transformer of 4 windings. Mind you the labour to design took about 2 days of mucking around......... Admittedly I have to cut them out - a cinch with my little diamond circular saw and a couple of small notches for each PCB to allow the ferrites to fit nicely, what a bargain ! The intention is to glue the cores together as the clips cost ~ 25c a pair! There was not much point in pioneering a good workhorse Power PCB if it was difficult for others to reproduce, it also makes it a lot easier for me ! Using either twin row 0.1" connector strips or their R/A version should allow simple connection pins for the PCB coils into the host PCB. Edited 2019-11-07 22:37 by wiseguy If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648 |
PCB coil windings, now there is a *really* clever, time saving idea |
||||
mackoffgrid Guru Joined: 13/03/2017 Location: AustraliaPosts: 460 |
Mike, Your inverter is looking real nice. I have told someone before but will say it publicly that I think your design will become the new standard for DIY SPWM inverters. IMHO, I am in Tony's camp about feedback control of the output voltage. When has AC mains ever been that regulated, my mains voltage varies by 20v at the equipment end  . Can I suggest you continue to test a Forward voltage control (open loop control) as well. Have a version of software available with forward control only. It's one less issue for people to sort out if they start getting distorted outputs.Great work I loooooovvvvve your design on the pcb and RM6 core. It's an example of innovation brought about by the democratisation of technology (easy access to cheap PCB )Cheers Andrew |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
This type of asymmetrical waveform is not that uncommon, and has puzzled a lot of people for a very long time, including myself. The fact that Poida's varic software works perfectly, without that odd bump (on one slope only) pretty much confirms that its most likely caused by staircase magnetic saturation of the toroid. One half cycle of your pwm must be very slightly different than the other, maybe a small rounding error somewhere, or perhaps a one bit error where quadrants join up. The toroid is being driven very slightly further in one direction than the other, and the flux in the core very slowly creeps towards saturation in one direction over many many operating cycles. That can produce some rather high peak currents to gradually build up that may be destructive. Or perhaps the inverter will appear to run run perfectly, but with just the curious lopsided voltage waveform. Using pulse transformers to generate multiple isolated dc power supplies is certainly a practical idea, if you can source the transformers at reasonable cost. For a long time I used some commercial SCR gate drive pulse transformers that I already had here in my early Warpverters with success. The difficulty will be getting reasonably equal voltages, and reasonable voltage regulation between the various outputs. The final dc output voltages will be all over the place unless you take very great care. Overshoot and ringing can make a 12v supply go to well over 20v at light load, then drop to 9v with only a few mA of loading. Klaus has had this same problem. My final solution was to use full wave rectification on all the secondaries, along with a zener diode. I used a 16v zener with supplies that were supposed to be nominally 15v. After that it all tamed right down and was reasonably well behaved. These days I much prefer to use the small Chinese postage stamp sized supplies if possible. For less than two dollars I get perfect voltage regulation, much higher available dc power, and the output is short circuit protected. Edited 2019-11-08 05:48 by Warpspeed Cheers, Tony. |
||||
renewableMark Guru Joined: 09/12/2017 Location: AustraliaPosts: 1678 |
Nice work.  If you want to resize pics you can use Paint if you are on a PC (very quick). Right click and open with paint, top left is a resize option, you'll work it out. Cheers Caveman Mark Off grid eastern Melb |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
Tony after a sleepless night I essentially came to the same conclusion. What I expected to see from saturation though was a "clipping" effect of the waveform peak, not near the zero crossing. I think I also get the mechanism that rapidly makes it get worse, the RMS of the distorted waveform is obviously less than the nice sinewave so the feedback voltage must have a corresponding droop. The Nano would then turn the wick up a bit causing a rapid saturation runaway to ensue. I guess that if we consider the slow onset of saturation, that means that the total energy put into the core for a complete 180 degree sinewave must just get to the saturation threshold at the 180 degree end, sure lots of energy is put in at the peak, but the total magnetic energy is the sum of all the energy during the 180 degrees so the final topping up causing the beginning of saturation (the straws that broke the camels back) will start to occur as we get nearer to the 180 degree - zero crossing point ? I dont understand how the feedback is able to skew the energy unequally between the positive peaks and the negative peaks causing a drift towards saturation. I wonder if the maths to correct the amplitude or similar causes a pwm event to be missed or the timing to get to the end of the array means we dont quite make it to the end causing an accumulation of offset - something is going on that I dont get Ok, so now I am considering using the variac code and putting an external analogue loop and circuitry that looks at the pot setting and the feedback voltage and makes an analogue adjustment to the Pot setting voltage to maintain constant output. ( in effect a servo motor on the pot lol). A feed forward from the input voltage should be a cinch to control the output but I would still like a relatively pedestrian return to the target voltage during operation for prolonged changes in load. Andrew I agree about the mains fluctuations and I consider that 240 - 245 volts output at no load and that under full load reduces to say 230V is still a great result. But surely we can get a reliable scheme to control the output whether it is digital/analogue or a combination of both - I'm not ready to chuck the towel in yet. Mark - thanks for the tip I'll try it next time I post a pic. Edited 2019-11-08 15:29 by wiseguy If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| Page 4 of 16 |
|||||
 Print this page |
