What's inside an ACE ATX psu

Hi All

After much discussion of power quality on other threads I decided to look inside my PSU. Warantee I hear you cry, well ...... that I reply. Seriously they only cost a tenner.

While I was at it I decided to snip off all those annoying disc drive cables.

Wow what alot of stuff for a tenner and probably a fiver at the factory gate.

The mains is full wave rectified and applied to two 470uF 220V caps, in series I hope, through a largish inductor, well thats a turn up for the books, since when was an inductor cheaper than adding more cap?

As for the rest 5 power devices on heatsinks. 4 transformers. 3 torroids one double wound.

On the 12V output there was guess what a 2200uF 16V cap.

All in all astonishing value for money for a tenner.

Rory

Quote
Rory166
The mains is full wave rectified and applied to two 470uF 220V caps, in series I hope, through a largish inductor, well thats a turn up for the books, since when was an inductor cheaper than adding more cap?

The 2 x caps in series is a standard arrangement for 115V/230V power supplies. When running from 230V, the bridge rectifier works normally, giving about 320V. For 115V, a link is added between the junction of the caps and one side of the mains, shorting out 2 of the diodes in the bridge rectifier. The remaining 2 diodes act as half wave rectifiers, giving +160V and -160V, so again 320V in total.

The inductor is needed to meet emission regulations, to prevent the PSU switching current feeding back into the mains.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Yes, well, that's what you can do with the pricing for commodity items! However, I'm not writing this PSU off entirely yet. I've been powering Ormorod with one of these (relatively) expensive Bench Supplies today:


I also connected it with some gigantic speaker cable to get a minimal volt-drop (<200mV with Bed Heater on - nice to not hear the fan RPM drop!). So, would Ormorod still lock-up part way through printing? Unfortunately, yes.

Out of interest these were the currents I measured:
Duet only (no motors or heaters) 0.25A
Hot End 4.5A
Hot Bed 11.3A

---

RS Components Reprap Ormerod No. 481

Hi both

Yes DC I am familiar with the 110v arrangement but I had not seen the inductor's reappearance, takes me back to the valve days.

Sorry to hear still no reliability Radian. The ATX supplies 4 12v cables to the two sockets we use, it would be nice if the bed supply could be fed seperately to the contol circuit rather than being combined at the connector board. Then we could try a seperate psu on the bed.

Rory

Edited 1 time(s). Last edit at 12/28/2013 07:09PM by Rory166.

If the hot bed takes 11.3A then it deserves some serious decoupling caps right between the +12V feed to the hot bed and the source terminal of the mosfet. As opposed to no decoupling caps at all, as on the present design

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
If the hot bed takes 11.3A then it deserves some serious decoupling caps right between the +12V feed to the hot bed and the source terminal of the mosfet. As opposed to no decoupling caps at all, as on the present design

I also thought you'd mention a snubber network across the load ;-) 1uF/10R seems about right

---

RS Components Reprap Ormerod No. 481

Quote
Radian
I also thought you'd mention a snubber network across the load ;-) 1uF/10R seems about right

10R would be too low, 2R would be more like it, that would limit the peak mosfet drain voltage to 12 + (2 * 11.3) volts. But since we don't need to reduce the rise time of the drain voltage, just reduce the amplitude of the spike, a flyback diode is simpler and avoids increasing the turn-on transient.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Hi Radian,

I've been thinking about your suggestion of a snubber network, and I think a diode-capacitor-resistor snubber could work well. I've connected a flyback diode in parallel with my bed heater (see this post). The flyback diode stops the mosfet avalanching (which wasn't really a problem anyway), however it does nothing to avoid the transient on the 12V supply due to sudden cutoff of the 11.3A load.

The snubber network would be wired like this: Schottky diode anode to mosfet drain, diode cathode to capacitor, other side of capacitor to ground (as close to the mosfet source as possible). Also a resistor in parallel with the diode. When the mosfet switches off, the capacitor charges through the diode, reducing the rate at which the load current falls. When the mosfet turns on, the capacitor discharges through the resistor, ready for next time the mosfet turns on. A 1uF capacitor would reduce the rise time to about 1us. Then the Schottky diode needs a surge rating of 11A for 1us. I might try removing my flyback diode and installing this instead.

Another way to avoid the sudden turn-off transient would be to slow down the mosfet turn-off, as suggested by bobc here. This could be a simple as adding a resistor in series with the power mosfet gate. However, that isn't easy to do as a mod to the existing board. An easier possible mod would be to replace the BS105 driver mosfet (Rds(on) <= 0.25 ohms) by something with higher Rds(on).

Edited 1 time(s). Last edit at 12/29/2013 07:21AM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

DC42

Now in posession of schematic, I don't think changing the driver will have the effect you desire as it is not in series with the gate, however putting a cap on top of R20 300K whch connects the gate to ground should slow up the switching. Still a bit early in my day to work out a value to give the apropriate rise time, bearing in mind it is not the rise time on the gate that matters but the time betseen an OFF level and an ON level. The RC time constant will be with R20 1K and not R20. Not really sure what R20 is for but ESD protection will be enhanced by it. Will have to look up the device which is probably designed for switching. Slowing down the switching will increase the power dissipation in the device, not a problem for the current bang-bang regulation but if PWM were implimented as the diagram suggests it might, then this would need consideration.

I notice that there is an output for an external FET and I like the idea of using an external switch so that the Bed could have its own dedicated supply and not share a common ground return with the electronics. I would possibly design my own ATX power board using the 4 way plug for bed supply, with 12V and ground, X 2 it would be ideal. I have yet to find a source of the pcb power sockets, ideas anyone?

Rory

Edited 2 time(s). Last edit at 12/29/2013 04:58AM by Rory166.

Rory,

If I change the driver transistor for one with a higher Rds(on), then that Rds(on) will limit the rate at which the driver mosfet sucks charge out of the gate of the power mosfet to turn it off, and thereby slow down the turn-off. However, I now think it would be possible to cut the power mosfet gate leg and solder a 0603 resistor between the leg and the pad. By my calculations, a 100 ohm resistor would slow the turn off to about 500ns.

I don't think R20 serves any useful purpose. Resistors are often put in that position when the mosfet gate is driven directly from a microcontroller, so that the gate potential is defined when the mcu is powering up and the pin is Hi-Z; but that is not the case here. Putting a cap in parallel with R20 would further slow down the power mosfet turn-on time (which is already quite slow, about 2us) but will do very little to the turn-off time because the Rds(on) of the driver mosfet is so low.

The bed heater mosfet displays a similar 50V spike and avalanching when the mosfet turns off, and its rise time is likewise about 2us. Power dissipation does not seem to be a problem in either the bed mosfet or the extruder heater mosfet, despite the slow turn-on times. So increasing the turn-off time to 500ns or so is not likely to cause any problems.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].