Realistic power from oil cooled transformer

[noneyabussiness]

I have two old stripped pj transformers (they burnt out so nothing salvaged except core).. they have 70mm "hole" and together 120mm high... i will be oil cooling them, part of a bigger system, now i roughly know there is about 2 odd kw as there current guise... so wondering what the awesomely smart people here think i could push them to ?? If any more info is needed just let me know...

[oztules]

Hmmm.... the core does not dictate the power really.

What it does give you is a winding window. The core will give you the turns per volt at a particular frequency and magnetic strength... thats it in a nutshell.

After that it is up to you and the physics to get the wire in there that will dictate the copper losses... this governs the power you can get in a time period... ie massive power for very short periods... and lower power for extended periods..... and every where in between.

We choose a bigger core for higher power simply to get thicker wire in there. ..... ie the more core...... then the less turns per volt we get, and so the thicker wire and less turns we can get away with in the winding window.

So I will assume that all the losses will be in the copper ( not entirely true... but...), and we are looking at the I^2 R for the watts lost in the winding as heat. This is what will burn it up or not.

If we can get the heat to dissipate faster ( fan cooling, oil and fins etc etc), then we can loose more voltage due to current and the copper resistance, and still keep the temp in the safe range..... ie our watt rating will be higher with better cooling, or thicker shorter windings or both.

The iron loss will be the same regardless of power.... ie once we have wound the thing, and it does not saturate, then higher currents or loading ( even overload ) will not saturate the core... only more voltage or lower hertz will do that... not current... so it is mainly copper losses we have to worry about. Torroid steel is usually very good.

So answer to your question as I see it is how low can you get the copper resistance, verses how fast you can get rid of the heat due to copper losses... otherwise known as how long is a piece of string really.... gut feeling is bigger is better

I  guess if you calculate the resistance of the primary you envisage and the secondary you can get in there, then you will have a good idea of how much power will be converted to heat...... that you have to get rid of at different currents.

It all comes down to efficiency, and getting rid of the heat.

Re reading that, I'm not sure I have helped all that much but the long and the short of it is.......more copper is more goodderer



..........oztules

[noneyabussiness]

Oz, u have a awesome way of explaining things so us simple people understand. .. thankyou.... a site i got reading was discussing this and they put up a spreadsheet for calculating the values for line frequency transformers (mainly E I core) ... i plugged my super core dimensions and played with the "flux ddensity".. i have attached said file with my current info... what u think??

Said website : https://ludens.cl/Electron/trafos/trafos.html

Only thing that worries me is the turns are 51:6 ... now ive read many of your posts and im sure it was 100:16?? I think...

Kinda outta my league here

[oztules]

Good article you point to... but not too relevant to torroids...in almost any way, except they are both transformers.

If you look at your spread sheet, you will notice that with proper care and attention, you have massive iron losses... where there should be almost none... maybe 30-40 watts... not 3-400w..... notice you have 2.5T... which is wicked high. This will mostly be saturated steel......

Do not wind that.

If you don't know the turns per volt of the core, you will have to trial end error. It is fine to use manufactures data... but you have none..... so if you did not count the turns for the volts of the original to calculate the core area and voltage they obtained, then you will have to measure it.

I would have a variable AC 50hz supply.... energise a set number of turns .... and see where the knee in the graph is for a number of measured points of volts versus current. Choose the lower current side of the knee, and use that as a result for turns per volt with small losses.

Then for 48v system, use the 16:1 rule... stack in the most copper you can, and you will have the near optimum combination... whatever that may be.

That article can form the basis of your deliberations, but all the prac is useless to you in this case. The torroid is twice as light in it's steel, as it does not need the return path... ie with E and I... only the central leg is used, all the rest of the steel is only a return path for the flux. In a torroid, there is only the central leg, and very low leakage because of the tight magnetic circuit.  Second, the iron losses are very much less because they can wind it with very thin steel, much thinner than EI can be, as there is no requirement for stiffness to push the laminates into the core, or support the structure in any way, it can be paper thin, as it has no mechanical function.

Once you have the turns per volt or volts/turn, then you wind the maximum copper in there.


...........oztules

[noneyabussiness]

Again. .. thankyou for your sage advice. .. you have saved me many hours...will update you with my progress...and kudos again for explaining it in real terms i can understand. .

[oztules]

Your welcome.... actually if you measure the cross section of the core you will get close with 1turn/volt for every 2800mmsq. It is not scientific at all, just a rule of thumb I have found for the chinese steels.... won't necessarily be perfect, but a good starting point.

.........oztules

[noneyabussiness]

Okay dokay... 145mm od , 72mm id and 120mm high combined. .. i used the formula h x (od-id) ÷ 2  to find out square mm... 4380mm2 ... so for your start rule of 1v turn every 2800mm works out 1.56... so would that mean a good start would be 156 primary 9.75 (say 10) turns secondary  (16:1)??

Is that close, right?? Sorry to annoy, just trying to learn...

[oztules]

ooops.... no due to the fact I'm an idiot, it is 8:1 for 48v.  ( 16:1 for 24v systems) so 156:19... .... so the primary will be 19 or 20 turns.
Sorry about that.

We want the primary to be about 30 to 32v for the PJ style processors, and 28-30 for the 8010 chips for a nominal 48v system.

Don't know what I was thinking...... dementia creeping in now

.............oztules

[noneyabussiness]

Please don't apologize,  thank you for your help, i have the thick primary,  just gotta source the secondary. .. again John, thankyou for your help..

[noneyabussiness]

Oz, thanku again for all your help, so finally wound 159 turns primary  (three layers of 1.8mm, tested in between the way u mentioned in another post, so all three are the same).. and 18 turns primary. .. 27v to 240.... seems very happy... only thing was i couldn't fit the original large primary in, with appropriate turns and layers so i ended up using 12 layers of 1mm secondary(all tested as well, so all the same) .. ive yet to really load test it but, just got another project fall in my lap thats fairly urgent so gotta finish that first now ..