The effects of changing ATX PSU

I just upgraded the PSU for my Ormerod with another ATX power supply, a Corsair CX430M. It normally sells for around SEK450-500 here (€45-€50), I got it for cheap from a friend who upgraded his PC and apparently needed something bigger.
It's a pretty decent PSU but nothing extravagant, seems to have gotten generally good reviews as being good value for money. 430W, single +12V rail. This is the "modular" version (there is a CX430 "non-M" that isn't, that is slightly cheaper), where you plug in separate cables for hard drives and other extra bits into the PSU, the result being no extra dangling cables since the 3D printer only uses a couple of them. The Ormerod just has the 24-pin big connector + the 4-pin 12V connector, and those two are only ones that are not removable from the PSU, so it all looks neat. The PSU has an 8-pin connector that you take apart, so there is an extra little connector hanging next to the 4-pin part you use.

I did some voltage measurements with the stock "Alpine 600W" PSU, as well as the new one. I measured the voltage at the Duet board (on the screw terminals for the 12V input), at the print bed (on the edge of the board where the ribbon cable connector is soldered), and at the PSU itself (on one of the hard drive power connectors), and I did it at idle (the printer sitting there doing nothing), with just bed heat on, and with bed heat + hotend heat on. No stepper motors doing anything. I did measure voltage (not at the moving bed, just at the Duet screw terminals) while it was printing, but it's hard to tell when the hotend heater is on. In any case, I never saw any voltages below what I got when just heating things. Even though the stepper motors can draw an amp or so each (according to the M906 setting), that is under maximum load (basically just before they start skipping), and that should't happen too often. Under normal use, the vast majority of the power consumed is for the heaters, mainly the bed heater. I never bothered measuring current draw, but it should be in the neighbourhood of 15A (well within spec of the PSU).

I also tested what the maximum temperature of the print bed was (set it to some high temperature and let it sit for half an hour). The print bed is the default setup (cardboard+pcb+heat spreader+glass). Since I did the "final temperature test" with the old PSU a while ago, it is a slightly different setup. I had the standard MDF bed support and kapton tape on the glass, with the new one I had no kapton tape and the aluminium bed support. Since both the kapton and the mdf might provide better insulation, I would guess the old psu might do 0-2 degrees worse with the current setup (no kapton and aluminium bed support). In both cases the printer + psu is sitting on a desk with a lot of "free air" around it, room temperature around 23°C. Temperature is according to the in-bed thermistor, I suspect temperature on the print surface is a bit lower at thermal equilibrium.

I have the default Ormerod power parts (pcb + cable), and the bed is connected via the stock ribbon cable. Also got the dc42 IR sensor, so the three LED's were on all the time and the hotend fan was running when the hotend was. The Duet itself is powered through the +5VSB from the PSU, so it is not using the onboard voltage regulator. Mains voltage is 230V AC, 50Hz.


So, voltage numbers:

Stock PSU, Alpine 600W:

                 @Duet   @Bed    @PSU

Idle             12.37   -       12.37
Bed heat on      11.38   10.87   11.70
Bed + hotend     11.10   10.58   11.59

Final temperature: 109°C

Corsair CX430M:

                 @Duet   @Bed    @PSU

Idle             12.26   -       12.26
Bed heat on      11.62   11.13   11.84
Bed + hotend     11.51   11.09   11.80

Final temperature: 123°C

Some conclusions that can be drawn from this, and some other reflections:

• ATX power supplies do have significant voltage drop. The Alpine PSU is obviously pretty crappy. The Corsair PSU is better, but not awesome.
• Just a little difference in voltage at the print bed actually matters for power output (unsurprisingly, since power is proportional to the square of the voltage)
• Getting to the "final temperature" is very very slow, especially for the last few degrees. In my case, getting to decent ABS printing temperatures took about half an hour with the old PSU. Even if you don't need to get to 120°C, the new PSU means it is a lot quicker to get to 110.
• There is quite some voltage drop and power loss going on. About 0.3V in the power feed to the Duet (through the power PCB and the screw terminals), and anouther 0.5V out to the bed (ribbon cable). Not sure if it's mostly in the connectors, or in the cable itself. I'll try replacing some parts of it with thicker-gauge wires and see if it makes any difference.
• The Corsair 430W PSU is about twice as heavy as the Alpine 600W PSU. Might not mean anything, but it usually correlates well with build quality.
• The Corsair PSU is a lot quieter. The fan turns, but very slowly and I could never hear it. It spins up a little bit from idle when turning on the bed heat, but still practically silent (I never heard it whatsoever). In any case it is completely drowned out by the noise of the printer (steppers + hotend fan).
• According to some tests, ATX PSU:s might behave weirdly when there is asymmetric load, like in this case where there is load on +12V, very little on +5VSB, and none on +3.3V, +5V and -12V. It might be that some "dummy load" on +5V improves things, I don't know.
• While there are good and cheap "non-ATX" 12V PSU:s available, I kind of like this setup where the Duet can turn the ATX PSU mostly off while still running things on +5VSB (which includes a Wifi bridge), so I'll probably stick with this.

Thanks for the measurements. As you say there are two areas; the drops in the cables and the PSU itself. The drops in the cables can certainly be improved by thicker cables and more direct wiring.

The main issue with ATX supplies is that they are intended to deliver multiple voltage outputs. To keep it simple most designs regulate their output by using a combination of the +5V and +12V outputs for the feedback path. This works OK in their intended PC usage where there is significant draw of current from both these rails. In Ormerod usage the power is pretty much exclusively drawn from the +12V rail.This means that when the load changes on the +12V then the feedback to try to compensate is less than 100% effective as the +5V isn't changing. The easiest way of dealing with that is to swap to a dedicated 12V supply where the regulation is not compromised and if it is adjustable that can be used to help compensate the cable drops and even give a bit more than 12V to speed up bed heating. It is possible to modify the Alpine so that it regulates the +12V much better and has adjustable output but that is only for people who feel comfortable working inside a power supply. See Hacking the Apline

Yeah, now that I got something else in place, I was thinking of modding the alpine psu a bit and see what it can do. A volt or two will be enough to make the losses not matter anymore and make the bed heat up a bit faster.

I dont even have the Alpine 600 psu, I got ACE A-550BR, so I suppose its a bit weaker than the Alpine 600W.

I took some measurements on the CX430M too. You can find my measurements here [forums.reprap.org] and the corresponding measurements for the original Alpine PSU here [forums.reprap.org].

---

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].

I know bumping this 2-month-old thread is some serious forums necromancy, but I modded the stock Alpine PSU as per bobtidey's recipe here, and the numbers are at least somewhat interesting. Same test procedures as above. Ambient temperature was a little bit higher, maybe 26°C.

The now variable voltage regulation on the PSU is set to 13.10V at idle. I had it a bit higher, but that could cause the PSU to shutdown under load, presumably due to some over-voltage protection. At 13.4V it would shut down when running both heaters. At 13.2V it would print without issues (with all heaters going), but I backed it down a little bit more just to be safe.

                 @Duet   @Bed    @PSU

Idle             13.10   -       13.10
Bed heat on      12.67   12.10   13.08
Bed + hotend     12.56   11.93   13.07

Final temperature: 132°C

Some interesting items:

• Replacing the voltage regulation makes a lot of difference to how stable the output voltage is under load. The old circuit was basically a "weighted average" of +5V and +12V, and now that it only cares about the +12V for the feedback, it gets a lot better. In all, I get almost a volt more out at the bed with this.
• This in turn should mean that unmodified, the PSU should perform a lot better (less voltage drop under load) with a load of a few Amperes on the +5V rail.
• I do not have any good means of measuring ripple, but with a cheap multimeter-style oscilloscope in AC mode, I could not see anything.
• +5V was at 5.90V, with that rail completely unloaded. +5VSB (which is what powers my Duet) is at +5.12V, and that does not vary at all with load or voltage on the +12V rail, or even if the PSU is on or off. It indeed seems to be a completely separate regulator, which is good.

In all, if you are not scared of modifying electronics like this, the Alpine PSU does in fact become quite useful. I'm using it for the Ormerod now (since I have use for the Corsair one elsewhere), and it is doing well so far.