Protect A transformer from a DC voltage

You haven't mentioned the load current, in addition to Kevin's list of missing information.

The problem could be :

• primary inductance too low (not enough turns on the primary),
• core saturation (too many turns on the primary) or
• both at the same time.

Since you can't simultaneously increase and decrease the primary turns, the last of these requires either a higher switching frequency, a bigger core, a different core material, or a combination of all these.

I'm suspecting core saturation because of your comment that heating worsens the problem (you'll probably find core permeability reduces with increasing temperature, reducing the primary inductance thus increasing the current, causing thermal runaway). Does freezer spray reduce the current?

Once the basics are right you can further improve efficiency by (for example) increasing wire diameter if I^2*R losses are important, or interleaving primary and secondary windings if leakage inductance is too high, but you're not at that stage yet.

Better information as Kevin asked will help improve the answers. At least, tell us

• the switching frequency,
• the primary inductance (measure across both legs of primary with secondary OPEN circuit)
• the leakage inductance (measure across both legs of primary with secondary SHORT circuit)

and anything you can about the core.

^{simulate this circuit – Schematic created using CircuitLab}

Figure 1. Transformer tap selector.

This isn't a complete answer but will give you some ideas to work on. Whatever electronic solution you come up with is going to need a steady voltage to work with. This makes it awkward to derive from the transformer as you have a full-wave bridge rectifier on it. A commercial product would add an auxiliary winding to the transformer to give, for example, a 12 V supply that doesn't change with the tap selector. Since this is a one-off project we can add a second transformer (ex wall-wart) to generate this.

How it works

• The main supply is derived from XFMR1 via two tap-selecting relays. If both are off the 5 V tap is selected. If MED is energised the 12 V tap is selected. If HIGH is energised the 24 V tap is selected.
• XFMR2 and the bridge form a secondary supply which remains steady.
• R3 and 4 divide the output voltage down to < V+AUX. A 4:1 ratio would be fine. CMP1, a comparitor, compares the divided down output voltage with the reference set by R7. When Vout > Vref the comparitor output will switch low energising the MED relay.
• A second comparitor circuit (not shown) is required for the HIGH relay.

Comments

The circuit is messy. Some hysteresis may be required to prevent the relays switching back and forward when close to the transition points. Vaux will droop every time a relay picks and this will upset the reference, etc.

All these problems can be designed out but a switching regulator solves them all with far less complexity.