Can the regulation of the ATX PSU be improved?

The low-cost ATX supplies that have been shipped with Ormorod so far don't hold up their 12V rails very well under load, hence we hear the fan speed fall and rise as the heated bed switches on and off and some people struggle to get enough power into the heaters.

As I understand it, the fixed resistor dummy loads on the 5V/3.3V rails are needed to keep the PSU from going into fault-shutdown but do they also form part of an active feedback loop in the PSU? In which case the PSU would think the load on it was constant even though the 12V load isn't.

If so then perhaps an active dummy load could be developed using a power MOSFET on a heatsink as a variable resistor, feeding back the effects of the loading on the 12V rail into the lower voltage rails?

I support RRP's thinking on using a COTS PSU; it's so very close to providing the ideal price/performance ratio. If a couple of MOSFETs and an OP-AMP instead of the "watty" resistors could improve the regulation, I think it could make a huge improvement all round.

I could do some tests to confirm my assumptions but it would shave some time off the exercise if anyone already knows how the ATX internals operate.

---

RS Components Reprap Ormerod No. 481

Adrian

Can't be of much help although I did have a cursory look inside my ACE. It would be good to have confirmation that the Alpine is indeed identical. It'd probable be cheaper and easier to buy a regulated 12v supply if one is that concerned. The main reason the kit vendors like the ATX so much is the ready done approval and safety for the mains connection. As well as being dirt cheap.

It does seem strange to buy a 12, 5, 3.3 and then generate the last two from the unregulated one. I believe they wanted a single rail capability.

I doubt the mosfet idea is that practical as the current drawn from 5v would probably need to be the same as from 12v to achieve the regulation required. We would require approx 50W load at 5V, it should probably go under the Y-axis to help keep the build area warm.

Rory

You could use the output meant for an external bed heater mosfet to turn on an additional mosfet that loads the 5V or 3.3V rail. That should help keep the 12V rail up. But there will be a lot of heat to get rid of from the extra load. Also, part of the problem is the resistance of the wires between the ATX PSU and the Duet.

The only real problems I have seen described on these forums are from people who have the Ace PSU (not the Alpine) and are trying to print ABS. I suspect that most Ormerod users will only ever print PLA and won't have a problem with the PSU. Those who are serious about printing ABS may want to get an industrial PSU instead. These are not expensive (I bought a 12V 300W one for £21-80), but do require the addition of a switch, cover and mains wiring.

---

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].

Maybe I have it wrong but the entire machine runs off the 12v line, none of the others are used. Basically the power box is just used to bring the power on line to gnd and a small load to ground on 5v most likely to make the psu think there is a load. Don't have it in front of me now to look but pretty sure that will be the setup.

I think in the psu, one would need to drill the feed back track from 5v or 3.3v to opto circuit in order to break it, then connect the feed back line into the opto circuit to 12V rail via split resistor network to gnd in order to bring the voltage down so that unloaded it still think it's reference is from the 5v or 3v3 line. The opto feed back circuit will be designed for input only from 5v or 3v3, basically we trick it.

The only downside is the 5v and 3v3 lines will be unusable I suspect as by doing this it will ramp the voltage up to the transformer for the load on 12v line causing 5v and 3v3 to over shoot but we don't use those anyway.

All pure simple theory! In practice, may be very very different.

Feel free to rip the idea a part!

Dieter

Ormerod #257

Thanks for all your thoughts on this. My take on it is that RRP already have to supply an ancilliary PCB and case with "watty" resistors which aren't zero-cost items in themselves. This item will always have to ship with ATX's so a re-design of the board to use semis instead of dumb resistors could keep pretty well within budget, but make the PSU behave more like an industrial one without the cost and engineering downsides.

dc42 - I like your KISS idea of wiring back from the bed heater external switch to apply extra loading when the bed is on. Although for three or four resistors and an OPAMP we could keep it all on the one board (no extra wiring) and compensate for all loads.

---

RS Components Reprap Ormerod No. 481

@dieterzar - Yes, at the moment everything runs from 12V. The Duet was designed to encompass a wide number of environments, and it was felt that having a single 12V supply was a sensible choice. However, there are also 5V input, and attaching pin headers is possible. There are also pin headers for supplying 5V on the ATX power PCB; I think markbee has been running his Duet like this. There is no 3.3V input on the Duet, though.

We would love to provide a better power solution, at reasonable cost, if there is one. Replacing the ATX power PCB with something better would be great, and a sensible and easy place to start. But... we're not electrical engineers! If there is a way for the power to be better managed, is there any chance you could come up with a schematic? We can prototype boards to test.

Ian
RepRapPro tech support

Hi all

Just been looking on the web for any info on ATX psus. Found an American article [www.directron.com] which is slightly interesting. It sets out to show that you can judge the quality of a PSU by its weight. Obviously not a generally applicable concept but they have found that amongst mass market ATX offerings those with properly rated components and adequate heatsinks weigh more than those lacking these benefits.

Rory

Lol, maybe lead-shielding would be a good marketing ploy in that case

Quote
reprappro
We would love to provide a better power solution, at reasonable cost, if there is one. Replacing the ATX power PCB with something better would be great, and a sensible and easy place to start. But... we're not electrical engineers! If there is a way for the power to be better managed, is there any chance you could come up with a schematic? We can prototype boards to test.

I think the best place to start is to choose carefully which brands and models of ATX PSU you ship. From the limited information I have seen on the forum, it sounds to me that the Ace ones are mostly OK for PLA but no good for ABS, whereas the Alpine ones are just about OK for ABS too, although it takes a long time for the bed to get up to temperature. This makes sense, because a side-effect of the higher current rating (32A) of the +12V supply on the Alpine PSU is that it is likely to have a lower output resistance, so the +12V output will drop less with increasing load than it will on a lower-rated supply. This is something you can easily test if you have access to both types of PSU.

---

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].

Although the ATX PSU has a lot of benefits, one of the things they are not so good at is dumping loads of amps onto a 12V rail. For me, that is the #1 requirement for a PSU... so I stopped using ATX PSU a while ago.

I can see the advantages of a cheap CE/consumer friendly box, but unless you get a tailored PSU with user friendly connections there will always be some consumer dissatisfaction.

---

What is Open Source?
What is Open Source Hardware?
Open Source in a nutshell: the Four Freedoms
CC BY-NC is not an Open Source license

btw I think I should point out that the output regulation of the ATX PSU is a smaller problem than the voltage drop in the cables. I took some measurements:

With no heaters turned on:

PSU output voltage: 12.19V (measured at one of spare output connectors)
Voltage at input to Duet: 12.18V

With both heaters on:

PSU output voltage: 11.53V
On ATX distribution board: 11.23V
At Duet input: 11.14V
At bed heater terminals: 11.10V
At bed heater: 10.78V

So with both heaters on, the PSU output voltage is 0.47V below the ideal 12V. However a further 0.3V is lost in the ATX PSU cables, 0.09V in the cable linking the ATX board and the Duet, 0.04V in the mosfet, and 0.32V in the ribbon cable + connectors. Total cable loss 0.71V. So even if the ATX PSU had perfect regulation, you would still only get 11.25V at the bed heater. This is with an Alpine PSU.

I have a Corsair CX430M ATX PSU in the cupboard, which is also rated at 32A on the 12V output, so I may take some measurements with that tomorrow.

---

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
bobc
Although the ATX PSU has a lot of benefits, one of the things they are not so good at is dumping loads of amps onto a 12V rail.

Hi bobc, from what I've see by googling around for ATX PSU schematics I think that few (if any) have closed-loop regulation on the 12V rails because I haven't found one yet. But even a cheap one like the Alpine can deliver tens of Amps, albeit with a lost Volt or so. This is still a couple of hundred Watts for a fraction of the cost of an industrial PSU. Unfortunately, given a fixed load resistance, all this power isn't quite readily available.

dc, I also found that the main volt-drop is in the (unavoidable) wiring. Because of this, just putting a better 12V PSU up front won't make a big difference unless it's set to a higher voltage to compensate. This is why I'm itching to see if applying a dynamic load on the 3V3 rail can be used to steer the 12V in the right direction. If it works, it's like having a separate sense wire at the load which is only found on the fanciest of supplies. I won't be able to get my hands on this until the weekend but I'd start with something simple like bridging the 3V3 dummy resistor load with a 50W halogen lamp and note any increase on the 12V rail. If it doesn't rise, the scheme is probably doomed.

---

RS Components Reprap Ormerod No. 481

OK, so I just pulled the case of a PC with some random cheap looking ATX PSU and dumped a 12V/50W halogen on a spare 5V drive plug. The 12V went up by 0.25V! This is interesting.

The lamp is pulling around 2.5A, hence 12.5 Watts. If it's a linearly proportional effect then we need to dump four times as much to restore our missing Volt. An additional 50 Watts is nothing to the PSU, but something to shed as heat somewhere. I can't readily load the 3V3 with this test so I don't know what that would do but the 5V is clearly part of the feedback loop.

To lose 50W a 1'C/W heatsink would be in the right order of things to maintain sensible temps. Actually, there are quite a few cheap extrusions for LED lighting these days but it would be getting on for the volume of the Duet enclosure. Shame the power can't do something useful. As a light, it would go on and off with the Heated bed and drive you batty!

Another possibility is to run a boost convertor from the 5V rail and actually channel the power into where it's needed. Knowing the fixed parameters of the load would allow for a very simple square wave switcher, or for a little more, a voltage sensing boost IC. Maybe an extra 5 Euros but it would be nice and compact - and energy efficient.

---

RS Components Reprap Ormerod No. 481

Radian,

When I posted the pics of the alpine open, I carried on looking and noted the switching transformer is common tapped to gnd. I also know from looking at it that they are doing the feed back sensing on the 3v3 line.

Thinking about it, it all makes sense. The psu is designed for pc use where most load will be balanced and mostly on the 3v3 which needs to have the tightest reference. Since the 12v tap is first, a heavy load on it will cause a change further down the transformer taps but I bet it will be small by the time you get to the 3v3 tap. So it will via the small voltage sag on the 3v3 line try pull up the voltage via the opto feed back of the whole transformer. So I don't think you will ever get the 12v line to bomb out because of this, 11v may be the lowest you will see. I really am guessing here

Problem is loading the 3v3 line will cause the previous taps to also see a drop in voltage and so I don't think it will work. The lower voltages I also took at the bed and so on are all just a function of wire guages as already pointed out. Lowest I got the PSU down to was 11.33V. Could add a few 50watt halogens and see how low it will go and what the voltage looks like on the 5v line and 3v3 line as a test.

Give it a try, I am going to try it on my psu.

Some place on the forum, a suggestion was made to use a switch mode psu used on down lights. It may work and they are low cost except the output is normally very dirty and will need smoothing and so on. Worth a try but will need to be a high powered one too.

Dieter

Ormerod #257

Quote
Radian
OK, so I just pulled the case of a PC with some random cheap looking ATX PSU and dumped a 12V/50W halogen on a spare 5V drive plug. The 12V went up by 0.25V! This is interesting.

As a light, it would go on and off with the Heated bed and drive you batty!

That would be a first on a 3d printer!

Dieter

Ormerod #257

I was thinking of putting an LED light at the top of my z-axis to illuminate the print bed, because the lighting where I have my Ormerod is not good when the sun isn't shining. It sounds as though a few of these [www.ebay.co.uk] powered from the 5V supply may have the beneficial side-effect of increasing the 12V supply.

---

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].

dieterzar, I think you're probably close to the truth about the "mixing" of rails in the feedback path of these supplies. Every manufacturer I've seen appears to use the same trick, otherwise separate transformer primaries and choppers would be required.

dc, don't forget the heatsinks for those LEDs! A few seconds at 1000mA and it will start to fade before your very eyes (not that you should'nt be staring at it!). They kid you about 5W. No sane person would feed it Imax 1500mA.

For any Heath Robinson types that would like some more power in the heated bed (~25W) and some intermittent reading light, this lash-up ought to do it:



The wiring could be done using the spare disc drive connectors to get at 5V and 0V but the losses in those wires detract from the objective. Parallel them up as a work around.

Another edit: This would work, but it's not meant as a serious solution - I can't see RRP adopting it for one second! I think the boost conversion idea has far more merit.

Edited 2 time(s). Last edit at 01/16/2014 05:52AM by Radian.

---

RS Components Reprap Ormerod No. 481

So I tried my Corsair CX430M power supply. It is only rated at 430W (the Alpine is 550W), but like the Alpine, it has a single 12V rail rated at 32A. Here are the results. Figures in brackets are for the Alpine PSU.

With no heaters turned on:

PSU output voltage: 12.05V (12.19V)

With both heaters on:

PSU output voltage: 11.82V (11.53V)
On ATX distribution board: 11.67V (11.23V)
At Duet input: 11.57V (11.14V)
At bed heater: 11.23V (10.78V)

So the Corsair PSU has significantly better regulation than the Alpine, also the cables on it have lower resistance. The drop in fan pitch when the bed heater turns on is barely noticeable. This PSU has the 8-pin version of the ATX12V connector, so if the power distribution PCB had an 8-pin socket instead of a 4-pin one, the voltage drop in the cable would be even lower.

I'll run my Ormerod from the Corsair PSU in future.

Edited 1 time(s). Last edit at 01/16/2014 08:53AM 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].

Hi All

I have just seen photos of the Alpine PSU from Dieter and from memory this is the same PCB as the ACE. This does not mean it has the same components of course. But tends to confirm my suspicion that they come from the same factory. At least we poor ACE owners may be able to upgrade to Alpine spec?

The idea of a boost converter from 5 to 12v seems a bit excessive to me. I wonder if feeding the external switching FET line to a FET in the PSU to frig the feedback on 5 or 3.3 as appropriate would be a cheaper and perhaps better alternative.

Has anyone found any circuits for ATXs ? Please link here.

Rory

Just found overclock selling Alpine psu cheaper than ACE. 500w alpine less than 10 pound including vat.

How about this for an idea? Open up psu disconnect feed back from 3.3 and connect to 12v using a potential divider and one of the 12v supply lines as a feedback line.

Edited 2 time(s). Last edit at 01/16/2014 05:50PM by Rory166.

Quote
Rory166
Just found overclock selling Alpine psu cheaper than ACE. 500w alpine less than 10 pound including vat.

The one RRP has been shipping is the 550W one, £10.99 from overclock. They also have the Corsair CX430 (the non-modular version of the CX430M that I am now using) at £34.50.

Edited 1 time(s). Last edit at 01/16/2014 06:34PM 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].

Hi Rory,

here's a "reference" design from ON semiconductor [www.onsemi.com] (lower power and marketing their components so probably not too directly relevant to any specific implementation, but maybe giving a flavour of the thinking behind - it also refers to intel's atx psu spec and other docs available at [www.formfactors.org]).

There are many schematics - though many are "reverse engineered" I think doing searches like this [www.google.co.uk] then clicking the images link

Well, in other threads there have been discussions of various add-on or back-up PSU options - at prices like these (around £10 per atx psu) it looks to me like starting with the supplied PSU then buying a second one dedicated to providing power to the bed (with switching form the external fet output, or similar) IF you find it necessary, will work out pretty much the cheapest and easiest route for many... Not as satisfying as beating the intellectual challenge that gave rise to this thread though

Hi All

Just had a breakthrough. There is a 3.3V sense wire (brown) disconnect this from 3.3 V in Ormerod power board and feed from 12 v via potential divider. No need to open PSU case. R1 220R R2 82R will hopefully be low enough impedance to feed 3.3v feedback. Values quickly calculated to give 12.15V at Power board.

Rory

Edit


It has been pointed out this will not work because 5v rail is regulated and 3.3v fed from that.

Edited 1 time(s). Last edit at 01/17/2014 04:59AM by Rory166.

I think the discussion about regulation of the PSU is a bit of a red herring. The Ormerod doesn't need a well-regulated PSU because it has 2 regulators on board for the electronics, and the actual voltage on the 12V rail is not critical. The issue is that the output voltage of the supplied Ace/Alpine PSU drops somewhat below 12V under load, and when printing ABS (but not PLA), this and the resistance of the cables means there is either barely enough or not quite enough power to raise the bed to the necessary temperature.

The simplest fix is I think to substitute a better ATX PSU, such as the Corsair CX430M that I am now using, or one that maintains the +12V supply even better under load. Those who are happy doing mains wiring can substitute a 12V 300W industrial-type PSU instead.

---

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].

All the ATX PSU I have seen use a combination of the 5V and 12V in the feed back loop.

The 3.3V is derived from the 5V AC winding and post regulated with its own local feedback by controlling the current in a coil (using it as a magnetic amplifier I think). So I don't think you can control the other rails using the 3.3V sense.

---

[www.hydraraptor.blogspot.com]

Another way to increase the power to the heated bed would be to replace the ribbon cable by something with lower resistance. Three pieces of this cable [www.rapidonline.com] wired in parallel would, by my calculation, have less than half the resistance of the ribbon cable, and increase the bed header voltage by around 0.2V. What I haven't worked out is how to prevent the cables getting tangled. It needs something like hairpin-shaped clips to hold 8 pieces of cable side by side.

---

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].

Naturally there are differening opinions about what constitutes an acceptable "fix" - if indeed there's anything that needs fixing. In my opinion RRP have a clear and sensible objective to provide a safe, simple and certified power source at the lowest cost possible. The ATX is so close. In order to maintain the object and to avoid any health & safety issues I think we should only contemplate solutions that can be engineered to augment the dummy load board.

dc42, obviously there's a finite budget involved and one use for it might be to source a better PSU. I agree with you that very few qualities are critical for powering a 3D printer but there will always be "lowest common denominator" PSU's like RRP ship. Pitching into a slightly higher price bracket won't necessarily return a better PSU. Droop on the 12V rail is all we need to eliminate to give the freedom to print in all common plastics and if a generic fix can be engineered that works with the poorest of examples then we have full control of the most critical factor..

rayhicks, good links to ATX design data. These things are evidently a highly evolved product! In the onsemi quide there's a discussion relating to nopheads's 3V3 regulation scheme. So it may be one of two schemes.

Rory166, interesting find - that brown wire. I don't recall seeing this on all ATX looms but it's certainly there on my Alpine connector. Unfortunately the 3V3 system appears not to influence the 12V on this design. The 5V and 12V systems seem to be linked instead. From what I'm seeing 3V3 typically has a separate regulation circuit.

---

RS Components Reprap Ormerod No. 481

Quote
Radian
dc42, obviously there's a finite budget involved and one use for it might be to source a better PSU. I agree with you that very few qualities are critical for powering a 3D printer but there will always be "lowest common denominator" PSU's like RRP ship. Pitching into a slightly higher price bracket won't necessarily return a better PSU. Droop on the 12V rail is all we need to eliminate to give the freedom to print in all common plastics and if a generic fix can be engineered that works with the poorest of examples then we have full control of the most critical factor.

Agreed, adding £25 to the kit cost for a better PSU to make ABS printing easier would not be worth it for most people. All I am suggesting is that for anyone who wants to print ABS and finds that the supplied PSU is not quite up to it, substituting a better ATX PSU is the simplest fix, especially for those with limited electrical/electronics experience.

I also think that RRP should review which PSU they ship. From what I read on this forum, and my own tests with the Alpine PSU, I think the 550W Alpine PSU is just about capable of ABS printing but the Ace is probably not.

Edited 1 time(s). Last edit at 01/17/2014 04:59AM 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].

Quote
Rory166
Hi All

Just had a breakthrough. There is a 3.3V sense wire (brown) disconnect this from 3.3 V in Ormerod power board and feed from 12 v via potential divider. No need to open PSU case. R1 220R R2 82R will hopefully be low enough impedance to feed 3.3v feedback. Values quickly calculated to give 12.15V at Power board.

Rory

This could be what we are looking for!

Just for information there are some interesting tests and discussions on the DUET board here [blog.think3dprint3d.com] which indicate that will be OK to 24Vs BUT CHECK! So switching the fan to 5V version if a suitable one is available could give greater flexibility on the "12 volts" set value.

I haven't had a good look around the [blog.think3dprint3d.com] blog, but it's well worth the time spent.

---

Ormerod #007 (shaken but not stirred!)

@dc42 - heavier duty (and perhaps shorter) power cabling would make a lot of sense - a cable chain would most likely be the easiest way of taming a bundle of wires eg this one [docs-europe.electrocomponents.com], the links of something similar should be moderately easy to design and print (actually RS hosts models of this belt provided by Igus...)

Ray