 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 : Various aspects of home brew inverters
  Page 33 of 48   |
|||||
| Author | Message | ||||
| Solar Mike Guru  Joined: 08/02/2015 Location: New ZealandPosts: 1138 |
Hi ItoPower, you could try one of these for your choke Big Powdered Iron Core The link gives its specs.. Cheers Mike |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Two ways to do it. First fill your steel choke core with maximum turns of a wire gauge that is not going to burst into flames at full power. Fit some kind of starting air gap like a 2mm spacer to give a 4mm total air gap. Gradually increase the load on your inverter noting the peak to peak high frequency ripple voltage on the output side of the choke. Try a different air gap and repeat the measurement. If the air gap is too large, there will be insufficient inductance and series impedance that is too low, creating a high ripple voltage. If the air gap is too small, the choke will go into saturation at higher power levels, reducing the inductance, creating a high ripple voltage. At some "optimum" air gap you can find a definite minimum ripple voltage at full load. And that is the maximum dynamic inductance that core and winding are capable of. If that is insufficient, start again with a physically bigger core, that will fit on more turns of the same gauge wire. Method two requires an inductance tester to measure the rate of current rise (amps per uS) at a given applied voltage. From that, its easy to both calculate real inductance, and see how far that inductance falls with rising current. If there is still some sufficient minimal inductance at peak dc current, and that corresponds to max inverter power plus some safety factor, you have fabricated a suitable choke. That is by far the fastest and safest method. If your inductor tester cannot reach say 200 amps, just wind on higher turns than you eventually plan to have. Three times as many turns will saturate at 1/3 the current and have x9 times the inductance. Its still a perfectly valid test of the core and gap. Once you know the ampere turns to saturation, and the inductance for one turn (Al) its easy to calculate the inductance and final saturation for any number of turns. If the required number of turns indicated by this will not fit with suitable gauge wire, start again with a larger core, or multiple stacked cores. At lest you will have a starting point to work from, and further testin should get you something that will do the job. Cheers, Tony. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Wow, those are the biggest Micrometals make. Chinese clones of course, but probably o/k. Anyhow ran a few quick and dirty numbers. Hole diameter 57.2mm, with luck we can fit about 11 turns of 10mm O.D. wire through that. Eight turns around the core, plus an extra three stuffed through the middle. If we assume 100 amps dc that is 1,100 ampere turns. That works out to 55 Orsteads with a magnetic path length of 25cm for that core. 55 Orsteads with 26 grade material gives us 40% of the initial Al value of 262. That is 0.1048 uH for one turn. And 12.68Uh for 11 turns (at 100 amps dc) Four or five of those cores stacked with 11 turns of 25mm squared cable would probably be about right for 50uH to 63uH at 100 amps dc, with a fairly soft saturation characteristic. Cheers, Tony. |
||||
| ltopower Regular Member  Joined: 08/03/2019 Location: United KingdomPosts: 64 |
Thanks for all the feedback and help. Starting to realise that scaling inverters has real limit issues with filtering... the 'art form' really does seem true. Could it be that all larger DIY builds should have more than one LV winding.. Mike, thanks, amazed at the Chinese powdered iron core pricing and shipping cost to the UK. Very cheap for the weight, interesting.... 5 of them stacked is 6.5kg and still only 100A... 5kVA ? Will study those calcs, thanks Tony. My transformer is wired 4x30V (asked for 29V and arrived seemingly 30V) around 32A per winding at 6kVA, so thought is I have to keep the windings separate before the chokes because the current limit issue is then not as large... Also seems to indicate that single high amp windings are not the way to go on DIY builds due to the filter size requirements ? Or is there a separate issue that would occur with inductor imbalances ?? Single winding can make the transformer build easy, but after that it seems like it could create a challenge ? 4 x E70 (2.5mm gap, 11 turns) 52.6uH. At 45 A each one is indicating 106mJ - 4 windings 45A each at 48V = 8.5kW. From the magic newly found charts (1097 page pdf, ferroxcube specs) that seems ok ? This seems like my first option... £45.60 in cores. Can't source the 2.5mm gap so I need to machine each one down and will have to tie the hand held grinder to the CNC mill head to get consistency on all 4 cores... hmmm.. Trying to scale up, seems to be where the 'Big man sized steel core, a decent air gap, lots of turns.' applies and one for further research and testing. Found this one... holly grail or just a rusty old bean tin ? Schaffner RD8147-64-0M2 Mouser No: 631-RD8147-64-0M2 Mfr. No: RD8147-64-0M2 Datasheet : https://www.mouser.co.uk/datasheet/2/355/rd_series-1201815.pdf 4 x 200uH Chokes in one package Paralleled up to give 50uH Each rated 64A Total 256A.... 12.3kW Saturation curves seems ok, soft ? If using 4 parallel inductors with tollerances +/- 20% will one tend to saturate first and then overheat as a result of passing higher current than the rest ? |
||||
renewableMark Guru Joined: 09/12/2017 Location: AustraliaPosts: 1678 |
So are you planning to make one inverter to handle all that? Or make two? Nice size battery bank BTW, bit pricey for me, you must have deeper pockets. Cheers Caveman Mark Off grid eastern Melb |
||||
| ltopower Regular Member Joined: 08/03/2019 Location: United KingdomPosts: 64 |
Bought the cells when they were only $24 each... price now is silly and they are no way economical at the higher pricing. Was just after they reduced the subsidies in China end of 2017 so a brief excess of cells floated around for a few months. Economically they had to make sense to me for a 15yr timeframe before I would buy them, they should pay back plus a bit... more I use them less the per kWh storeed cost and they still have well over 10k cycles left. One inverter at the moment, looking like a second is going to be built for the shed and then a third for 'portable' use. Then probably one or two for my brothers house... looking to build this one with minimal tuning so it's easier to build the next 4 and they can all share parts for repairs. Idea originally was 5kVA (turned into ordering 6kVA) toroid and then oil cool it, few litres of mineral oil circulated, so it could then be run for extended time for maybe 9-11kW with higher losses. Mainly a forward plan for coping with EV charging at 7kW (of which I don't have an electric car at the moment)... When I was about 8 I played around with a small 12V transformer in a tub of old engine oil in my bedroom, cutting up tinfoil. Maybe the oil cooling is more about being 8 again.. lol. That was all the plan until I started working out the rest of the details and fell into a rabiit hole. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
[quote]Can't source the 2.5mm gap so I need to machine each one down and will have to tie the hand held grinder to the CNC mill head to get consistency on all 4 cores... hmmm..[/quote] Forget about grinding your own cores, just use a 1.25mm spacer between two unmolested core halves. Exactly the same as one 2.5mm gap, and a lot more easily adjusted and consistent, especially where there are multiple cores stacked. Beware of all commercial chokes !!! You may see something advertised like 50uH 100 amp choke, and think that sounds ideal. It probably is not. Its very likely designed as an EMC noise reduction filter choke for some low voltage dc power supply. It would work fine for that application where there may only be a volt or a few hundred mV of noise crap that needs filtering. Feed 48 volts worth of constant high frequency PWM into it, and the core is very likely going to smoke and burst into flames. The ac voltage across an inductor is important. Here is a thought experiment. Suppose you have a transformer that is 1:1 ratio and it says on it 12V 50Hz 1:1 ratio to 12v on the rating plate. And you connect it up to 230v 50Hz and expect to see 230v on the output. Its going to saturate and die. There absolutely must be sufficient core area and turns to support high voltages. An EMC choke is designed to only ever see very low noise voltages across the choke. We need something that can support 48 volts of PWM continuously without the core saturating or overheating. It comes back to having a man sized core cross sectional area, and lots of turns. And its going to be big, and possibly expensive. No short cuts unfortunately. Cheers, Tony. |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Mike's monster toroids should work, but you may need to juggle the figures. Eleven turns on five cores might be a first best guess, and should have acceptable core heating and temperature rise. More turns would increase the inductance, and reduce core heating, but it would saturate sooner. If you need a higher dc current rating, reduce the number of turns, and stack more toriods, with possibly thicker wire. Cheers, Tony. |
||||
| Solar Mike Guru Joined: 08/02/2015 Location: New ZealandPosts: 1138 |
Tony, I tried some of those aliexpress sendust Cores type 225-125A initially as they were going cheap at the time, for memory I needed about 15 stacked for 100 odd amps, using the inductance saturation tester with about 7 turns 20mm wire. In the end the silicon steel "C" cores picked up on EBay and using copper strap as the cable worked out the most efficient. What would those T400 cores be like if the pwm frequency was increased to 50 Khz... Cheers Mike |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Those sendust cores are a fair bit smaller than the T400 but the material would appear to be a better grade and less lossy. The T400 is enormous but the 26 material is a low cost general purpose material, not particularly special in any way. I think the temperature rise at 50Khz might be at least double what it would be at 23Khz. I just took a quick stab in the dark at those T400 cores. Lowest loss frequency specified for T400-26 is 50Khz and 17 volts per turn giving a rise of 15C.  I guessed about 30v/turn might be about right for 15C rise at 23 Khz. The previous design suggested 11 turns and five toroids. 11 x 5 x 30v/uS per turn suggests we might get about 1650v/uS for 15C rise. A 48v inverter at 23Khz at 50% duty cycle would be about 48v and 22uS say 1056v/uS. Big guess here but we might be looking at 10C core rise in temperature + copper losses. That is not huge, and pretty confident that would be quite o/k. What I cannot be sure of is that the Chinese clones are anywhere near as good as the US made Micrometals T400-26 cores for loss. Cheers, Tony. |
||||
| Solar Mike Guru Joined: 08/02/2015 Location: New ZealandPosts: 1138 |
Thanks Tony, I went to the Micro-Metals site and down loaded their data sheet. It looks like the 52 material is better at 50KHz and above, I might get a couple of these and try them out. Who knows if they are as good.... Mike |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Its slightly better in every respect but especially in core loss. And the price is right ! The 52 material allows 27volt/uS per turn for 15C rise. The 26 material allows 17volt/uS per turn for 15C rise. If its running hot, definitely a step forward. One problem with all these toroids, the hole is relatively small. Steel cores are almost always designed for transformer applications, where there is both a primary and a secondary. The window area is usually large enough to allow a much higher number of turns. That mitigates the higher eddy current loss of steel at around 20Khz. But at much higher frequencies, the powdered iron toroids start to look more attractive, especially at Chinese prices. But I still think you are probably going to need a bunch of them. Cheers, Tony. |
||||
wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
Hi Tony, I have been playing with a Toroid from a commercial 4KW inverter (possibly originated from Sola) 72VDC input Sinewave 240V output. The Toroid is a stack of 3, A-866142-2 Blue Toroids which I think were Arnold cores, it appears Micrometals number is MP-300125-2. The cores are each ~ 80mm OD, ~50mm ID and ~13mm High  This core has a total inductance of ~70uH with 15 Turns & has very low losses in the saturation tester to 60A, equating to ~ 900 Ampere Turns. Saturation is relatively soft to beyond 100A albeit with increasing losses. It is much better than the large 70-54-32 Ferrite cores I was investigating. They are also dead silent unlike the gapped Ferrites which were more like a loudspeaker...... What information did you use to determine the temperature rise of the 26 & 52 material toroids - can you apply the same magic to these cores ? Data P64 Here Also I have been considering a scheme not running 4 FETs in each bridge leg in parallel. Instead, run 4 separate bridges (1 FET each leg) from the same drive signals and drive 4 smaller (toroidal?) chokes in parallel and then re-sum them together at the main power toroid. Itopower has also mentioned a similar scheme for his 3 winding primary. It would appear to have a few benefits, solve the issue of 1 FET carrying too much current due to on or off speed mismatching when all in parallel. The chokes are also easier to wind and 4 smaller ones are simpler than 1 big choke. The oscillations from paralleled Mosfets should disappear also. I look forward to your comments. If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406 |
Micrometals publish loss curves for all their powdered iron grades, but it gives figures of milliwatts per CM cubed which is not very helpful as far as estimating temperature rise goes. In the previous picture I posted (above ^), down in the bottom right hand corner are some figures for volt microseconds per turn ratings for 15C rise. That is for 26 material. We can assume 48 volt square wave 23Khz which is 43.4uS per cycle. Maximum loss will be at 50% duty cycle or say 21uS. So the core sees 48v for 21uS in one direction, then 48v for 21uS in the other direction. That would be 1,008 volt microseconds. Its not actually not going to be quite that bad, because the PWM is a modulated sine wave. Anyhow we could say that the picture above tells us that a T400 core will rise by 15C at 50Khz at 17 volt microseconds per turn. Our requirement has 1,008 volt microseconds. So the minimum number of turns required for 15C rise would be 1008/17 or 59 turns (on one core at 50Khz). Clearly we are not able to do that. But we could stack five cores and have 12 turns and be spot on for 15C rise. That may not suit our dc saturation requirements, but we can juggle turns and number of toroids to get a workable soloution. Powdered iron is good material for very high dc current, but its lossy, and does not like high ac voltages or high frequencies. Temperature rise is always a concern. Molypermalloy is a much better material in any grade, but its much more expensive than powdered iron. The ac losses will be very low in any choke application, and not even worth bothering with to work out in most applications, except at very high voltages and very high frequencies. Ferrite has even lower losses still, but its hopeless for dc applications, it just saturates far too easily. Its much better for high frequency inductors and transformers where its all ac, with zero steady dc bias. If you have molpermalloy or can get it free, you are in luck. If you have to buy it new in the larger sizes be prepared for a shock. Cheers, Tony. |
||||
| ltopower Regular Member Joined: 08/03/2019 Location: United KingdomPosts: 64 |
Like the idea of separate FET boards.. really like it... Could have a modular inverter, plug in as many boards as you need for the capacity.. 1kW per board seems about right. 48V / 20A Nominal Single standard E65 choke per board (52.6uH, 10 turns, 1.5mm gap, 30A) part B66387G1500X1N87 Good for around 1.5kW peak per board without issue. Fuse each board separately in hope it could save a FET or two... |
||||
| Solar Mike Guru Joined: 08/02/2015 Location: New ZealandPosts: 1138 |
Nice idea, but what mechanism would you use to ensure each module limited its (Max) current peak to a set value, without that in place, the module with the coolest and lowest resistance fets blows up. For dynamic current peaks, the series choke would perhaps be the limiting factor common to each module, the inductance would have to be high enough such that the rate of change of current was limited by the inductance at any stage in the pwm cycle. Certainly small sized chokes would be easier to manage than the monsters I'm currently attempting to use. Mike |
||||
| wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156 |
In my experience for given on and off times and reasonably matched inductors the currents will ramp/fall & generally share very well. Given the worst case of max output load and highest input voltage we just ensure the inductors dont saturate - that is the mechanism. Im sure you agree that for a given input voltage and on time and inductance value the current ramps should all be equal (ignoring tolerances). In a recent (current fed topology) converter I designed, I had two EE55 transformers fed from the one bridge stage, FETs as output current doubling rectifiers (no TX centre tap) and 4 inductors. The output was 13V6 @ 60A from a 72VDC input, with ~ 95% efficiency at full load. Current sensing was done just on one EE55 transformer leg and was essentially half the total. If we reduce the inverter choke to 4 smaller chokes, as the current is now a quarter of the original the choke cores can be shrunk & the inductance can also be increased accordingly. The power PCB and heat-sink would not be a lot different to the current layout just re-distributing the connections on the PCB to be 2 3 or 4 separate bridges. In the case of separate power modules & heatsinks, the on resistance should only have a very minor contribution to the applied voltage / inductor / current ramp result. I would also expect that due to the positive temp coefficient of the FET resistance should help to ensure they share properly. If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike |
||||
| ltopower Regular Member Joined: 08/03/2019 Location: United KingdomPosts: 64 |
Guess that with a cold start and a large step load change there could be an issue with current distribution in the FET's, but I'm guessing that would only occur if your switching say 4kW on and off for a 6kW rated board cold ? After running for a while would the temperature rises not settle out and even up the current distribution ? Feels like going back to the old days of waiting for valves to warm up.... Over size the boards/split a little to take this into account. My conclusion so far is single winding for the LV on a large inverter is not the way to go even if the FET bridge is not split out.... Even if it is just from the basics of avoiding sourcing a large enough suitable gapped core to deal with the high amps, which seem to be a higher price and more difficult to find.... Separate question... I have 10 E-core E70 units due and needing a 3mm gap. Like the idea of not grinding them (thanks to Tony for the simple solution) I thought about the shims/spacers and wondered if I mixed some cement and iron powder (instead of sand !) if I could not use this to try and fill the outer gaps and leave the central gap open. Guessing this would be a tollerance nightmare unless I can get the mix and gap consistent ? Was just trying to think of a way to avoid leakage that may not be the conventional approach..... more suitable to DIY builds with that extra time put in to make a difference. |
||||
| Solar Mike Guru Joined: 08/02/2015 Location: New ZealandPosts: 1138 |
Using the formula for Inductance L Henries = V * Tseconds\Amps Our battery voltage = 60v max, pwm = 25Khz (40us), current = 40amps Very roughly required L = 60 x 40us\40 = 60uH at 100% duty cycle So to limit each modules current to 40 amps, we need a 60uH inductor that doesn't saturate at 40 amps, 60 to be safe. Quite a small inductor using 10mm wire. In my case I would need 2 low cost inductors 30uH per module (Bipolar modulation) The added complexity comes with having multiple isolated power supplies and opto-drive couplers for the 4-5 modules. This seems a great idea to be able to scale up the power by plugging in more modules. Cheers Mike |
||||
| renewableMark Guru Joined: 09/12/2017 Location: AustraliaPosts: 1678 |
OH F! Are you only meant to gap the middle part of the E with E cores? I have just been placing plastic shims across the entire 3 to gap them. Cheers Caveman Mark Off grid eastern Melb |
||||
| Page 33 of 48 |
|||||
 Print this page |
