Power Inverter Transformer design discussion

20 hours ago, Sid Genetry Solar said:

I'm still learning more about how transformers function...workin' on that 12kw transformer, trying to figure out how to get a design that will do 12kw without overheating, without going to a bigger core, etc., etc.

At least as far as I'm aware at the moment, the transformer core does not actually handle electrical power, but rather a magnetic field.  (As such, core losses are best measured at no load.)  The #1 limitation for how much power a core can handle seems thus far to be how much wire it is possible to get wrapped around it.  Too little wire = too much heat and/or excessive losses.

The bigger the core, the more magnetic flux it can store--and also due to the increased surface area, the more magnetic flux can be transferred to the windings (resulting in a higher volts/turn).  The higher the volts/turn, the less wire required to reach the desired voltage (to a point, obviously!)--and the less wire required, the lower the losses, etc., etc.

I'm not too familiar with what the PJ-spec transformers using the ASL3 / ASL4 cores are capable of--but I do know that the ASL3 and ASL4 cores are basically the same size +/- 10mm.  In other words...a very minimal difference between the 2.

I attached a spreadsheet to a previous post somewhere that I used to calculate what core size would give what max power without saturation, and to calculate the required windings.

I don't fully understand magnetic theory myself but the basic formulae to calculate everything you need can be found somewhere on the net

1 hour ago, Paul said:

I attached a spreadsheet to a previous post somewhere that I used to calculate what core size would give what max power without saturation, and to calculate the required windings.

Already got the winding spec / core voltage figured out...actually was pretty fascinating to graph the core voltage efficiency out and see exactly where the sweet spot was.  I think my (big) error was in calculating the battery side amperage by dividing expected input watts by battery voltage.  Transformer is an AC device with a significantly lower voltage (i.e. 42-64vDC battery voltage, but transformer spec 30-32vAC)...and calculating battery-side winding by the TRANSFORMER voltage specification SIGNIFICANTLY increases the calculated battery-side winding amperage.  Pretty sure that's where I went wrong...

...when I could only account for less than half of the losses in the transformer, I was starting to wonder if the core was carrying electrical power, and as such had an electrical resistance (=losses) that we were up against.  But from what I've read, the core pretty much only handles a magnetic field (and it's actually easiest to measure said core losses at no load.)  Recalculating losses with TRANSFORMER voltage specs gave much more believable power loss estimates (i.e. accounting for ~75% of actual losses).

I'm not convinced yet that a core can saturate with "too much power"--UNLESS the core voltage is increased (in which case it's game over very quickly, per my core voltage graphing tests!)  If the core voltage is constant and/or decreasing under load, I don't see how a core could saturate.  Usually, the losses in the windings become too great to maintain the desired output--not to mention heat losses going through the roof...but the core shouldn't saturate.  Unless I'm missing something?

1 hour ago, Paul said:

but the basic formulae to calculate everything you need can be found somewhere on the net

Yah...IF I knew every single specification about the tranny cores, how they were manufactured, with what materials, processing, etc., etc.  Unfortunately, I know nothing about them except what I can measure on my bench...so a bit of wheel reinventing is kinda necessary.

21 minutes ago, Sid Genetry Solar said:

Already got the winding spec / core voltage figured out...actually was pretty fascinating to graph the core voltage efficiency out and see exactly where the sweet spot was.  I think my (big) error was in calculating the battery side amperage by dividing expected input watts by battery voltage.  Transformer is an AC device with a significantly lower voltage (i.e. 42-64vDC battery voltage, but transformer spec 30-32vAC)...and calculating battery-side winding by the TRANSFORMER voltage specification SIGNIFICANTLY increases the calculated battery-side winding amperage.  Pretty sure that's where I went wrong...

...when I could only account for less than half of the losses in the transformer, I was starting to wonder if the core was carrying electrical power, and as such had an electrical resistance (=losses) that we were up against.  But from what I've read, the core pretty much only handles a magnetic field (and it's actually easiest to measure said core losses at no load.)  Recalculating losses with TRANSFORMER voltage specs gave much more believable power loss estimates (i.e. accounting for ~75% of actual losses).

 

I'm not convinced yet that a core can saturate with "too much power"--UNLESS the core voltage is increased (in which case it's game over very quickly, per my core voltage graphing tests!)  If the core voltage is constant and/or decreasing under load, I don't see how a core could saturate.  Usually, the losses in the windings become too great to maintain the desired output--not to mention heat losses going through the roof...but the core shouldn't saturate.  Unless I'm missing something?

 

Yah...IF I knew every single specification about the tranny cores, how they were manufactured, with what materials, processing, etc., etc.  Unfortunately, I know nothing about them except what I can measure on my bench...so a bit of wheel reinventing is kinda necessary.

Yes I'm certain you've done a lot more research than I have on this matter. Making a 'real' 12kVA transformer is no simple task and I admire your persistence in researching this. Cooling the core is also a big challenge especially when it has very thick layers of windings on it.

My 4kVA transformer build was completed with 'finger in the air' estimate of required core dimensions, then find what I could obtain cheaply close to that size, then calculate how to wind it using basic formulae.

The core came from an audio testing power supply transformer so although I don't know the material used and the exact properties thereof, I can safely assume it is of high quality material and construction.

18 hours ago, Paul said:

Cooling the core is also a big challenge especially when it has very thick layers of windings on it.

So that's what I was wondering with the first 12kw test...whether the core was generating a lot of heat at full load.  But the best I can find online seems to indicate that the transformer core strictly carries magnetic flux, NOT electrical energy--and as such is not a significant part of transformer losses.  My tests after correcting the winding loss estimates seems to corroborate that statement.  Still not 100% sure, but I did have significantly undersized primary (low voltage) coils...which were generating a lot of heat.

1 hour ago, Sid Genetry Solar said:

So that's what I was wondering with the first 12kw test...whether the core was generating a lot of heat at full load.  But the best I can find online seems to indicate that the transformer core strictly carries magnetic flux, NOT electrical energy--and as such is not a significant part of transformer losses.  My tests after correcting the winding loss estimates seems to corroborate that statement.  Still not 100% sure, but I did have significantly undersized primary (low voltage) coils...which were generating a lot of heat.

Yes in theory as long as you don't go close to the maximum magnetic flux density of the core (i.e. saturation) then there should be little or no heat generated in the core itself assuming it is of good quality. Most formulae recommend toriodal transformer should be wound for a maximum magnetic flux density of around 10,000 gauss, which is what I did with mine. The only heat generated in mine when I tested it was the slightly undersized primary wire gauge.

However if you get close to the maximum flux density for any given core CSA (typically around 14,000 for most good quality oriented-grain silica steel toroids) then some heat will be produced by the core due to saturation.

The turns/volt ratio is fixed for any given core. You can lower the maximum gauss by using more turns than required on each winding, but then additional resistance and tighness of winding of the wire brings its own problems and you may run out of space in the centre of the core, as well as reducing air flow through the centre thus hindering cooling

Obviously a larger core would overcome this issue but my understanding is the magnetising current is proportional to the core cross sectional area of the core and so idle current will be increased and once the SPWM carrier has been removed by external chokes the idle current cannot be reduced any further.

So here we have a large number of contradicting factors which make designing a high power transformer with a low idle current for an inverter extremely difficult.

If it's not a 'trade secret' I'd be curious to know what configurations of core dimensions, and windings you have tried and what the outcome was. I certainly don't need to build one myself (4kVA is enough for my needs!) but I like to gain more understanding of the concepts as I think I just 'got lucky' with my design.

24 minutes ago, Paul said:

The turns/volt ratio is fixed for any given core.

Ummm......the turns/volt is determined by the winding specification.  The practical LIMITS of the turns/volt ratio is determined by the core. 

This is why I currently believe that knowing the "ideal core voltage" of a transformer core is of far more practical value than a scientific laboratory "gauss" number that requires expensive equipment to measure--only to provide a value that's not very friendly to a DIYer.  Once you know the ideal core voltage, you can mathematically determine exactly how many turns of wire are necessary to reach the desired winding voltages at the most efficient core voltage.

28 minutes ago, Paul said:

Most formulae recommend toriodal transformer should be wound for a maximum magnetic flux density of around 10,000 gauss, which is what I did with mine. The only heat generated in mine when I tested it was the slightly undersized primary wire gauge.

However if you get close to the maximum flux density for any given core CSA (typically around 14,000 for most good quality oriented-grain silica steel toroids) then some heat will be produced by the core due to saturation.

I have absolutely no idea how to measure core flux in gauss or otherwise...I'm a simple country bumpkin with a DMM, clamp ammeter, and an oscilloscope.  (Oh, and an AC variac.)

Determining the optimal turns/volt for the core was actually pretty simple once I got the math figured out...basically, I took multiple measurements at increasing voltages on the winding under test.  Each measurement basically comprised of the resulting core voltage, and the AC current being used by the transformer at that voltage.  Dividing one reading by the other gave an unscaled "delta" that showed a very interesting "peak" at the most efficient turns/volt rating.  (And of course, at the high end when the core began to saturate, the no-load current went up exponentially--though it is worth noting that the turns/volt continued to increase even during saturation.) 

I repeated the test on a much smaller transformer core--and found that it saturated considerably quicker than the big core.  (In other words, there was very little "wiggle room" with the winding specification on the smaller core.)

25 minutes ago, Paul said:

as well as reducing air flow through the centre thus hindering cooling

Admittedly, I haven't found this to be of any significant benefit.  Bought a PJ 9kw inverter several years back (that's what started the entire Genetry inverter design project actually!)  Found that it overheated at 3kw continuous.  I figured that airflow through the center of the core would help...so I mounted the tranny sideways (no easy feat when the inverter was also mounted sideways!) with one 200CFM fan blowing directly into the center of the core, and a second one "pulling" from the center outwards (to outside the chassis).

Still overheated at 3kw.  In other words, it wasn't a "holy grail solution."

39 minutes ago, Paul said:

If it's not a 'trade secret' I'd be curious to know what configurations of core dimensions, and windings you have tried and what the outcome was. I certainly don't need to build one myself (4kVA is enough for my needs!) but I like to gain more understanding of the concepts as I think I just 'got lucky' with my design.

Ha, well, none of this testing/experimentation has been free (quite the opposite actually), so I'll just provide some principles that I have learned by trial and error.  In essence, if you can calculate the DC resistance of the coils (need a DC clamp meter + a high-output DC power supply that has a constant current limit on it, as well as a meter that can accurately register millivolts), you can use Ohm's Law to estimate the wattage losses at the desired full rated load.  (I still have no idea how to determine the maximum heat dissipation for a transformer--that part's still trial and error.)  It's important to calculate coil current from the transformer coil voltage spec--that's where I went wrong, using battery voltage instead.

I also realized that ideally, if both primary/secondary coils of the transformer are intended to carry an equal amount of wattage, the wire thickness should be proportionate to the voltage ratio of said coils.  In other words: Half the voltage = twice the wire.  (Yeah, *duh*, but...!)

Down at 3-6kw, the scaled losses are considerably less...meaning that there's a lot more "wiggle room" with the specifications.

43 minutes ago, Sid Genetry Solar said:

Ummm......the turns/volt is determined by the winding specification.  The practical LIMITS of the turns/volt ratio is determined by the core. 

This is why I currently believe that knowing the "ideal core voltage" of a transformer core is of far more practical value than a scientific laboratory "gauss" number that requires expensive equipment to measure--only to provide a value that's not very friendly to a DIYer.  Once you know the ideal core voltage, you can mathematically determine exactly how many turns of wire are necessary to reach the desired winding voltages at the most efficient core voltage.

 

I have absolutely no idea how to measure core flux in gauss or otherwise...I'm a simple country bumpkin with a DMM, clamp ammeter, and an oscilloscope.  (Oh, and an AC variac.)

Determining the optimal turns/volt for the core was actually pretty simple once I got the math figured out...basically, I took multiple measurements at increasing voltages on the winding under test.  Each measurement basically comprised of the resulting core voltage, and the AC current being used by the transformer at that voltage.  Dividing one reading by the other gave an unscaled "delta" that showed a very interesting "peak" at the most efficient turns/volt rating.  (And of course, at the high end when the core began to saturate, the no-load current went up exponentially--though it is worth noting that the turns/volt continued to increase even during saturation.) 

I repeated the test on a much smaller transformer core--and found that it saturated considerably quicker than the big core.  (In other words, there was very little "wiggle room" with the winding specification on the smaller core.)

 

Admittedly, I haven't found this to be of any significant benefit.  Bought a PJ 9kw inverter several years back (that's what started the entire Genetry inverter design project actually!)  Found that it overheated at 3kw continuous.  I figured that airflow through the center of the core would help...so I mounted the tranny sideways (no easy feat when the inverter was also mounted sideways!) with one 200CFM fan blowing directly into the center of the core, and a second one "pulling" from the center outwards (to outside the chassis).

Still overheated at 3kw.  In other words, it wasn't a "holy grail solution."

 

Ha, well, none of this testing/experimentation has been free (quite the opposite actually), so I'll just provide some principles that I have learned by trial and error.  In essence, if you can calculate the DC resistance of the coils (need a DC clamp meter + a high-output DC power supply that has a constant current limit on it, as well as a meter that can accurately register millivolts), you can use Ohm's Law to estimate the wattage losses at the desired full rated load.  (I still have no idea how to determine the maximum heat dissipation for a transformer--that part's still trial and error.)  It's important to calculate coil current from the transformer coil voltage spec--that's where I went wrong, using battery voltage instead.

I also realized that ideally, if both primary/secondary coils of the transformer are intended to carry an equal amount of wattage, the wire thickness should be proportionate to the voltage ratio of said coils.  In other words: Half the voltage = twice the wire.  (Yeah, *duh*, but...!)

Down at 3-6kw, the scaled losses are considerably less...meaning that there's a lot more "wiggle room" with the specifications.

Ah ok, I didn't realise the turns per volt ratio wasn't constant for any given core. Just shows how little I know 🙂

I'm also just a country bumpkin who learns most stuff by guesstimate, trial and error - I gave up on A-level physics back in the day! Yep quite understand you not sharing the details, as I'm sure you have spent a lot of money on parts to experiment with - fully understand that.

Like you say, there seems to be a lot more leeway when designing lower VA transformers and over 6kVA gets exponentially challenging.

It's directly dependent on the "exciting coil voltage" divided by the number of turns of said coil.

Here's a sample table I created with an ASL5 tranny core:

    in volt    in amp     core v       loss        delta
    25vAC    0.015A    0.208v    0.375W    1667
    50vAC    0.027A    0.418v     1.35W      1852
    75vAC    0.036A    0.626v     2.70W    2083
    100vAC    0.046A    0.828v   4.60W    2174
    125vAC    0.062A    1.038v     7.75W    2016
    150vAC    0.092A    1.240v  13.80W    1630    [starting to get noisy]
    175vAC    0.286A    1.450v  50.05W     612    [quite noisy]
    200vAC    2.800A    1.700v 560.00W      71    [loud hum]

where

• "in volt" is the input voltage I provided to a coil (from the variac)
• "In Amp" is the measured transformer current draw (no load) at said voltage
• "core v" is the core turns/volt measured with a single wire loop around the core (put one meter probe through the core, and then short both probes together)
  • Note how "Core V" is always the exact same ratio (+/- measurement errors) to "In Volt"--you can literally divide "In Volt" by "Core V" at any point to determine the # of turns in the input coil (= 120 turns.) 
  • Also note that even when the transformer core is running terrible efficiency due to saturation...it STILL transfers the turns/volt driven by the excitation coil (worth noting that I didn't hold the tranny at 1.7v/turn for very long, and only hasty readings...but it still reports 200 / 1.7 = 117.6 turns)
• "loss" = "in volt" * "in amp" = an interesting graph of core loss
• "delta" = "in volt" / "in amp" = an arbitrarily scaled result that shows the very interesting "peak efficiency" here at ~0.828v/turn

On a bigger core that I tried, the core voltage could be run significantly higher before falling off the saturation cliff.  It also had a higher "peak efficiency" voltage.

These tests can be done on any transformer core...with or without windings.  If you have an unknown core, it's still easier to wind a dozen turns of wire around it for the variac purposes (as it's awfully hard to run a 120vAC variac from 0-2vAC out with any sort of precision!

Here's a sample table I created with an ASL5 tranny core:

That  explain why my ASL 9  transformer ( in volt ) is  36 volt ac .    30vac is better for 48 v dc battery  but the core  do not  have room  for more winding so  I  use  16s lithium ion battery  for dc input .   The ASL 9  run hot  at load over 6kw .    It will be very difficult  to run   12kw continuous .   At this time all my solar panel and   battery   can not run  12kw  for more than 8 hours .   Sean latest  video say he  do not have enough power to run 12kw  continuous  but need  ATS  grid backup .   Thank you for the chart of the ASL 5 .   ASL 5 is a good transformer  running  32vac  in volt .   

Core has plenty of room for more winding on the ASL9 core vs PJ spec....it actually takes LESS wire to run a 30vAC coil than a 36v (fewer turns).  The problem causing the early overheating is insufficient wire on the core.

The problem causing the early overheating is insufficient wire on the core

Will more  pure copper  wires help  with overheating  or is  the ASL 9  core a  poor  design in core  material or using  alumium wires ?  

The transformers are wound with aluminum for cost and weight reasons.  If you rewind the ASL9 core with appropriately sized copper wire, you should easily be able to get more than 14kw continuous out of the ASL9 with minimal heat.  I don't have the patience for that, but...it can be done.

I haven't found a reason to doubt the quality of the PJ tranny cores--at least so far.

Sid and Paul,

If I squint my eyes and hold my mouth just right I can mostly follow your transformer winding discussion.   I very much enjoyed reading your discussion.  Wish I was good (and patient!) enough to try winding my own transformer.

Sid,

I do have some questions, hope you don't mind;

To get a cool running, continuous 12kw inverter trans, have you tried using two of your 6kw trans in parallel?  (I'm having Visions of my older, multiple trans pj's)  I realize it wouldn't be cost effective for production, but it might be a learning experiment.

Are your GS transformers using copper or aluminum wire?  Your above comments to Dickson caused me to wonder.

What transformer voltages did you end up using for the 48v/6kw GS inverter?  32/240?

Do you still think 12kw is a realistic reachable goal (for a production inverter)?  Or would 10kw be more realistic? (I'm still hoping you perfect 12kw!)  In my case, I can run my house split between two inverters if I want, so could get by with smaller units (but would have to run two!  Less efficient).

Do you still think 12kw is a realistic reachable goal (for a production inverter)?

Sid  design  a 14kw  transformer that Sean will test soon according to the latest youtube video  .  It will do 14kw in China but have to see if it continuous .