So I got a 19″ rack with 3U from work in which I installed a transformer. Since I did not want any AC lines on the backplane I had to build a power supply. Linear power supplies would be way too inefficient, and i wanted more power, so I chose switching mode regulators.
Then I thought about which voltages I wanted on the bus. A 5V rail makes sense, since I wanted to smash in lots of Arduinos. 9V were also added, just in case I need it (spoiler alert: never happened). Both voltages were regulated by a LT1074CT each. I made the inductors myself (Amidon T106-26 as core) with an inductivity of 150µH. According to the calculations of the step-down converter (buck converter) on the website of Schmidt Walter the current ripple is down to about 0.14A at the specified 5A. The current is higher than the transformer can provide, since the clocked inductor is providing all the current needed. Clock frequency is here at about 150kHz and both voltages are converted from a 12VAC line each.
For opamps I wanted to add a symmetric voltage. I settled for +-12V, made from a 20-0-20 secondary coil from the transformer. As switching controller I used the L4960 providing up to 2.5A which is more than enough for my uses. The inductors I made for these voltages had 68µH and the clock frequency was about 100kHz.
All voltages got a fuse on their input side just in case. Grounds were connected to each other, too.
After some tests I recognized heavy interferences and noise on the supply voltages. I measured up to 20V peaks on the 5V rail which is more than enough to kill every microcontroller connected to it. For filtering out these peaks I designed a small low pass filter board with one filter for each voltage. The filter was designed as a passive second order low pass filter (LC-combination) with 8µH inductors and 220µF capacitors. With these filters the peaks almost vanished completely and Vrms got down from ~35mVpp to ~12mVpp which is pretty nice.
In order to see how much current is drained I also added a small shunt board with 0.1Ohm 5W resistors. The voltage drop is negligible but I still can sense it and calculate the current.
As usual here are the schematic and layout for the SMPS-board
Here the filter board:
And the shunt board:
And for you to see the finished thing, here is a picture of it.
You can see the transformer, the power supplies and all the small boards. I also added NTCs to the rectifiers since they are the ones getting hot. A small fan was also added for ventilation and cooling. But not really the prettiest project…