Adapting HP server power supplies to 3D printing

[salfter]

I just bought an Anet A8 3D printer recently...assembled it over the course of a day and a half earlier this week, and got it up and running since.  While the printer itself seems to run well enough, the 240W "won hung lo" (?) power supply bundled in the box doesn't exactly inspire confidence with its semi-exposed AC input, and it's one of the weak points that comes up as something that needs to be addressed in many discussions of this printer.

I have HP common-slot server power supplies driving my mining rig.  Why not adapt one of those to this purpose?  The smallest ones deliver 460W, which should be far more than sufficient, and you can pick them up off of eBay for maybe $20.  Adapter hardware is already available (have a look here for examples), but I wanted to integrate a 5V USB power supply (to run a Raspberry Pi running OctoPrint), and the 750W (and larger) kits sold to miners would be overkill.

I tracked down the pinout for the power supply's edge connector: 64 pins on 100-mil centers, with 26 pins paralleled for each of +12V and GND.  3A of +12VSB is on one pin, and what needs doing to power them up is reasonably well-documented.  I'll have the primary output going through a set of copper pours into a bank of 8 screw terminals.  The standby output will go into a switching regulator to drop it down to 5V for a USB-A connector...and it's here that a double-check would be nice, as I've not previously implemented a switching power supply.

I originally found someone else's circuit built around an MC34063 with an external FET to boost the output current beyond that chip's 1.5A limit.  Someone else here suggested that was a rather crusty device and suggested building a supply around the LM2678.  A bit of googling led me to TI and its Webench, which offered a configuration for the LM2678, but also suggested some newer chips.  A bit of twiddling led me to this: the TPS54327.

I took the schematic and BOM Webench provided and plugged it into KiCad, and got this:

https://alfter.us/wp/wp-content/uploads/2018/03/hpcs2anet.pdf

I think I've transcribed it accurately, but wouldn't mind a second opinion.  I had a look through the datasheet and it looks like the component values are in the ballpark.  I did substitute two pairs of capacitors (at input and output) with single values: two 10-uF capacitors with a 22, and two 22s with a 47.  Is there any reason why I should restore the design to two 10s and two 22s instead of a 22 and a 47?

Also, while I could just dump the resulting 5V into a USB connector and call it a day, this suggests that signaling 2.4A output with some resistive voltage dividers on the data lines might not be a bad idea.

If I've done my "homework" right, I'll have 5V USB power available whenever the power supply is plugged in and gobs of 12V that can be switched off if the printer starts going haywire. 

[benst]

Carefull with some of those big PC power supplies. Some of them need a minimum load on some of their rails to regulate correctly. Otherwise they are way out of spec (or don't start up at all).

Some info here: http://reprap.org/wiki/PC_Power_Supply

Ben

[NiHaoMike]

All the HP server PSUs I have worked with are perfectly happy with no load. For the 5V, just get a quality car USB adapter.

[james_s]

I've messed with lots of server power supplies, every one that I've tried will work fine at no load. I use some of them to power my RC aircraft chargers, one set I modded to get 24V at something like 48 Amps.