ETD49 Overheating

[wally7856]

The only factors that effect primary turns are voltage, frequency, Flux level (Gauss), and the Ae of the core.

With an etd59, Ae 3.68, 70khz, 340vdc, and 28 pri turns, you are around 1200 G.

I think you should try changing your frequency as low as 50khz. This will give you some forgiveness in your wire size. But raise your flux level to 1650 Gauss an acceptable level. Fortunately this is a easy RC change on the controller chip. If i remember correctly the frequency of the controller is divided by two for the transformer, so your RC on the controller should be set for 140khz for 70khz on the transformer. Try changing the RC for 100khz for 50khz on the transformer.

 

[Jagd.Panther]

I think you should try changing your frequency as low as 50khz. This will give you some forgiveness in your wire size.

Are you sure that could reduce copper losses for 0.25mm wires?

 

[badboy_6120]

Hi Badboy,

Go to farnell.com / digikey.com / element14.com / etc and order some EDT cores made of better material. Ferrite material, its permeability, losses, maximum induction, play big role in trafo design. The "better" the material the less turns you need, so you can use thicker wire and cut down copper losses.

Also you can look at toroid cores, do you have anything over 50mm in diameter in local shops?

I will check and inform you
Thank you

 

[badboy_6120]

Does it make any difference to use 0.4mm wire instead of 0.25mm??
Like 14x0.4mm for primary and 12x0.4mm for secondary???

The "ExcellentIT ver 7300" on another forum here suggest 0.5mm for 70khz and 0.56mm for 50khz

 

[wally7856]

Thinner strands are always better than thicker strands, as long as the total copper area adds up.

 

[badboy_6120]

So you think 36x0.25mm for primary and 30x0.25mm for secondary enough for 1400w output power?(160v x 8.7A)

 

[wally7856]

BadBoy give me a few minuets. I see i made a mistake in my Gauss calculations. When using a half bridge you use 1/2 of the input voltage for the transformer. So 340vdc / 2 = 170vdc on transformer. Assuming you have 240vac input.

I am also working on your wire calculations.

 

[wally7856]

Your primary winding's assuming 500 Cmills/A and assuming 70% efficiency are good for 829W.

The secondary winding's are good for 5.88A x sec voltage = Watts.

Let me recalculate the Gauss level now.

 

[wally7856]

170vdc x 10^8 / 4 x 70000 x 28 x 3.68 = 598 gauss

170vdc x 10^8 / 4 x 40000 x 28 x 3.68 = 1,031 gauss
170vdc x 10^8 / 4 x 30000 x 28 x 3.68 = 1,374 gauss
170vdc x 10^8 / 4 x 25000 x 28 x 3.68 = 1,649 gauss
170vdc x 10^8 / 4 x 20000 x 28 x 3.68 = 2,062 gauss

Right now at 70khz you are at a very low gauss level 598 Gauss, this is to low. As a general rule you should be around 1600 Gauss. I would say 25 to 30khz is your best frequency for 28 pri turns.

BTW, as far as i know this smps was never designed to run at 1400W, that was just a test to see if it held up under stress.

 

[wally7856]

After thinking about what you have i think you should try running the transformer at 27.5khz. This means the controller should be double that freq.

I think this will give you the coolest transformer and most power. Very easy to try, just change the RC on the controller.

 

[Jagd.Panther]

As a general rule you should be around 1600 Gauss.

Eh? The general rule you choose flux density according to core loss at desired frequency and desired flux density to have reasonable temperature rise.

The DYI way of doing this is very simple: wind test primary, run trafo on idle for like 15-30minutes, check core's temperature. Usually you are splitting core/copper losses 50/50. So if you design transformer for +40degC temp rise you don't want the core to heat more than +20degC on idle. If the core temp rise is higher : add more turn and/or increase frequency. The If the core is stone cold: decrease frequency/decrease number number of turns. Ideally you should check magnetizing current, if it's too high (like higher than 1/10th of maximum primary current) then you should either peek a core with higher Al (permeability) or use smaller core.

The 'scientific' way: get Rth from manufactures catalog, for given temp rise calculate maximum core loss, then knowing core volume and desired frequency peek max B according to material data.

After thinking about what you have i think you should try running the transformer at 27.5khz. This means the controller should be double that freq.

I think this will give you the coolest transformer and most power. Very easy to try, just change the RC on the controller.

Badboy, if you follow this advice please lower the frequency gradually. Don't lower it from 70kHz to 27kHz at once. IMHO unless the copper loss is dominated by proximity effect and skin effect (I doubt that because you use 0.25mm wires) you are unlikely to get "cooler" running trafo by only reducing the frequency.

 

[wally7856]

There is ZERO danger lowering the frequency to 27.5khz.

 

[Jagd.Panther]

There is ZERO danger lowering the frequency to 27.5khz.

Absolutely ZERO danger if the core is far from saturation. However taking into account Badboy's experimental data he posted in this thread I have a doubt.

 

[wally7856]

Do you have a post number.

 

[badboy_6120]

170vdc x 10^8 / 4 x 70000 x 28 x 3.68 = 598 gauss

170vdc x 10^8 / 4 x 40000 x 28 x 3.68 = 1,031 gauss
170vdc x 10^8 / 4 x 30000 x 28 x 3.68 = 1,374 gauss
170vdc x 10^8 / 4 x 25000 x 28 x 3.68 = 1,649 gauss
170vdc x 10^8 / 4 x 20000 x 28 x 3.68 = 2,062 gauss

Right now at 70khz you are at a very low gauss level 598 Gauss, this is to low. As a general rule you should be around 1600 Gauss. I would say 25 to 30khz is your best frequency for 28 pri turns.

Thank you "Wally7856" for your explanation
but lowering down the frequency isn't the same as lowering down the turn numbers?
I used 10 turns for primary at 70Khz and that means 1650 gauss, but in this way transformer will heat up even in idle condition (65°c in 1 hour)

BTW, as far as i know this smps was never designed to run at 1400W, that was just a test to see if it held up under stress.

Yes that's true, but I changed the Mosfets to IGBTs and used MUR1520 instead of MUR810 and now the transformer
I don't think that the control unit have any problem with this much power as I tested 1400W for an hour without any problem except for the transformer's heating

 

[wally7856]

Badboy, as far as i know, right now you have 28 turns in the primary. You said 14 + 14 and i assumed in series.

But if your only problem is the transformer getting to hot, put a fan on it like the computer power supply's.

 

[badboy_6120]

The DYI way of doing this is very simple: wind test primary, run trafo on idle for like 15-30minutes, check core's temperature. Usually you are splitting core/copper losses 50/50. So if you design transformer for +40degC temp rise you don't want the core to heat more than +20degC on idle. If the core temp rise is higher : add more turn and/or increase frequency. The If the core is stone cold: decrease frequency/decrease number number of turns

How much temperature rise in cores are good for idle condition??

 

[Jagd.Panther]

How much temperature rise in cores are good for idle condition??

For a design with +40degC target temp rise at full power: < +20degC core temp at idle, measured at hot point (middle leg close to centerline). If you are measuring on the outer surface: I'd aim to +10...+15degC.

If you don't have thermal shutdown/trafo overheat protection then I'd measure at highest possible mains voltage (265 VAC for 230+-15%)

 

[badboy_6120]

For a design with +40degC target temp rise at full power: < +20degC core temp at idle, measured at hot point (middle leg close to centerline). If you are measuring on the outer surface: I'd aim to +10...+15degC.

If you don't have thermal shutdown/trafo overheat protection then I'd measure at highest possible mains voltage (265 VAC for 230+-15%)

Thank you so much for your answer
I'm measuring the core temperature with different primary turn number in idle and will share the results

 

[badboy_6120]

Hi everyone

I wound the primary test winding with 10-12-14-16-18-20 turns one after another and measured the cores temperature in idle condition as below:

10 turns: after 1 hour=87°c
12 turns: after 1 hour=79°c
14 turns: after 1 hour=70°c
16 turns: after 1 hour=60°c
18 turns: after 1 hour=53°c
20 turns: after 1 hour=46°c

room temperature was in the range of 27°c to 30°c

and another thing, one side of cores was less hotter than the other side (about 10°c)
**above temperatures are measured from the most hottest point of transformer cores**
**frequency in above measurements was set to 70khz**
should I go with 18 turns or 20???