Hello fellow AAC Forumites,

I could use your help with the following:
To save money setting up an electronics test bench I've been trying to convert a 220v to 120v, 50 Hz, 3kVA consumer grade step-down transformer into a true tech-type isolation transformer. Project seemed simple enough:

1. Disassemble and unwind the tx.
2. Build a new bobbin out of 2 mm FR4 to replace the cheap plastic one that shattered into 8 pieces after I dropped it.
3. Calculate the turns ratio and other parameters using the best online calculator I could find: http://www.dicks-website.eu/coilcalculator/index.html.
(see attached).
4. Rewind the tx.
5. Test for shorts, then test output voltages, current and overall efficiency. Fused variac used for protection.

Notes:
a. Old tx had only one coil configured as an auto-transformer with multiple taps for five secondary voltages. A rotary switch allowed switching output voltages depending on load. This configuration was desired but with new secondary voltages of 103v, 115v, 120v, 125v, and 130v. The new primary voltage is 103v 50 Hz (Japan - east coast region).
b. New bobbin came out perfectly and is MUCH stronger. It is also divided into two sections for primary and secondary windings.
c. Core is E and I with alternating laminations (easy to assemble-disassemble). Appears to be 1.3T silicon steel but unable to verify. Very shiny - no rust.
d. Discovered too late that the core wire is 14 AWG aluminum. As one of the primary goals was to keep costs down I elected to reuse the old wire for both primary and secondary windings. However, calculations were made using copper wire as that was the only option presented by the program.

The results were mixed:
Good:
1. Voltages were very close to calculations.
2. Tx is quieter than it was originally. No hum at all in no-load configuration.
Bad:
1. Core heats up quickly under no-load. Coils remain cool to touch after 15 minutes but core temp is very warm and continues to climb. Tx cannot be left on indefinitely without temperature sensing protection circuitry added.
2. No load current at 103v is 0.99 amp. Power loss is therefore 100 watts - about what the program predicted under full load, but this was with the secondary leads open. Won't test it under load until this is resolved.

So, where did I go wrong? Would rewinding this tx with copper help any? Would PF correction work, and if so, how much?
Thanks!

 

Could be eddy currents. If so you will need to tear it apart to re-insulate the core.

 

Magnetic field density is highest under no load/light loads. Poor or damaged insulation in the core would show up as high heat due to eddy currents in the iron.
Where does the heat concentrate? In the core or windings?

 

Magnetic field density is highest under no load/light loads. Poor or damaged insulation in the core would show up as high heat due to eddy currents in the iron.
Where does the heat concentrate? In the core or windings?

Problem shows up in the core and not the windings. The core has 126 laminations, all in pristine condition. Lamination resistance test appears inconsistent. If I press hard with the probes from one side of the stack to the other I can get the meter to go from infinite to no resistance intermittently. However, the tx's prior step-down configuration did not overheat. I'd have expected the eddy current problem to have been present then as well, no?

 

The eddy currents are always there, the laminations are how we compensate. By minimizing their effects.

If you can get a low ohms reading then there IS a problem

Check edges closely a stray scratch there could link all the plates electrically

 

I made an interesting discovery. On a hunch I reassembled the laminations so the sharp edges from the stamping process all faced the same direction. The edges feel a lot like the sharp ridge that forms on a woodworker's scraper after burnishing. The idle current then dropped to .68 amps - still too high in my opinion, but the core took twice as long to heat up as it did before. I may take a file to them next to remove the sharp edges entirely. I really don't have a means of applying an insulative coating on the laminations so this will have to do.

 

You will get it solved, I have no doubt.

 

An update to the above for those interested. My new isolation transformer is working well after easing all the rough edges in the laminations, including the holes. This doesn't seem to have been necessary as even welded transformers (MOT conversions, etc) don't have eddy current inefficiencies enough to worry about. Mine, however, still consumes about 65 watts at 103 volts with nothing on the secondary, about 100 watts on 115V, which I think is what I get for running it on 50Hz instead of the 60Hz it was originally designed for. My internet sources tell me you can go up in frequency (to a point) but not down without overheating the core. A 12V fan hooked to a thermal switch will keep it cool and a second thermal switch set to a slightly higher temperature in series with the primary will protect the unit from a runaway condition in case the fan stops. A small Arduino might provide a more programmable response with digital temperature monitoring in the future but for now this is working fine.