Interestingly, I designed a similar circuit myself early last year. Few variations with mine. Used two TO220 regs. (LM7805 & LM7812 1 amp packages) Plus I also have a bit more available current due to full wave rectification. This arrangement also makes the circuit a little safer too. The low voltage rails are tied close the Neutral line of the 240VAC mains, so provided you connect to a correctly wired GPO (240VAC socket), most of the circuit is nominally at low potential. (*Claimed by Silicon Chip publications)

Anyhow, I discovered a whole bag full of problems with this circuit on the bench. In theory, it all should have worked. In practice, I found that the 5 & 12V regs were unable to deliver the goods. The circuit was pre-calculated and should have been good for about 1.2W of power- 100mA on the 12V rail. The 12V rail failed to provide even a third of this. I have no explanation to the problem and my thoughts are entirely inconclusive. I discovered only about 20mVPP of ripple in the raw DC rail supplying the regs. This is more than an acceptable level of AC component.

In the end I settled for two zener-based shunts to produce 5 & 12V rails respectively. Circuit works well.

Trent Jackson