Experimental Power Supplys
October 06, 2012 07:48AM
A quick poll and discusion.

I have been looking recently at abundance of cheap switchmode PSU's aimed at 12V halogen lighting and wondering about repurposing them for RepRap use.

In many cases we only actualy need 12V.

Their output is very poorly regulated as normaly they are operating loads that dont care.

A feature of this is pronounced a 100/50Hz ripple as it is switching a rectified line signal. I guess in the US it will be 120/60Hz.

Out of curiosity has anyone done some work on repurposeing these ??

So far I have considered:-

1. Modifying the PSU to add capacitors across the rectified line voltage so that there is reduced ripple. But ultimately discarded this as a DIY option purley on the grounds that I need to sleep at night knowing I have sugested soemthign very risky to a wide ranging skills set.

2. Adding Smoothing/Reservoir capacitors across the 12V line between PSU and Controler. They will need a large capacity for the currents encountered (risky arc welding potential) and this is likely to be expensive enough to reduce the cost advantage of the cheap supply.

Thinking a little deeper about the nature of what we are powering:-

We have spikes of high current loads (A lot at first as things heat up) followed by lower running currents.

So whilst the peak demands are quite high the average should be a bunch lower. It is something i could do to get around to measuring at some point, except that I don;t yet have a heated bed.

Where I am sort left at is perhaps using some form of battery as a Smoothing/Reservoir capcitor. Maybe a small lead acid or alternative.

Anyone else done any work on this ??

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 08, 2012 05:58AM
How cheap are these? I mean, a 12 volt power supply from ebay is ridiculously cheap. I see no need to spend hours/days of time in finding another concept to save in on a part you can get for 5 dollars...
Re: Experimental Power Supplys
October 08, 2012 07:21AM
I guess if ebay has the answers for one it may have the answer for the other...



Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 08, 2012 11:22AM
aka47 Wrote:
> I guess if ebay has the answers for one it may
> have the answer for the other...
> [www.ebay.com]
> sformer-Power-Supply-Driver-80W-12V-for-MR16-MR11-
> /221027629675?pt=LH_DefaultDomain_0&hash=item33764
> 5fa6b
> 1$

$3.58 to the US
Re: Experimental Power Supplys
October 08, 2012 11:46AM
Sure, you have carriage to add on top of what ever you buy mailorder.

The question remains ??

Anyone done anything with this yet ??

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 08, 2012 11:47AM
aka47 Wrote:
> I guess if ebay has the answers for one it may
> have the answer for the other...
> [www.ebay.com]
> sformer-Power-Supply-Driver-80W-12V-for-MR16-MR11-
> /221027629675?pt=LH_DefaultDomain_0&hash=item33764
> 5fa6b
> 1$

That is a 12 AC power supply. Simple filtering and smoothing won't give a usable result.

Help improve the RepRap wiki!
Just click "Edit" in the top-right corner of the page and start typing.
Anyone can edit the wiki!
Re: Experimental Power Supplys
October 08, 2012 12:02PM
Sure, it was illustrative of comparative pricing using ebay.

Pick one you like better.

I am more interested in the question, than either pricing or source.

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 09, 2012 10:45AM
I use a regulated 30-40A source i did myself out of a UPS power unit, using its transformer and case. Normally shouldnt need to rewind the transformer, best chances are it can be used right away. I did rewinded mine, but not because of the voltage but because of the current which i needed larger as originally i made the supply for some other purpose (high current linear rf supply). I really dont recommend rewinding the trafo, but thats an ultimate thing only if it really can be dealt with.

So if you need to repurpose something to get a psu, i highly recommend getting the hands on a broken ups unit with a working transformer. That solves the trafor problem and the case, i had even repurposed the connectors at the back, so my mendel uses normal ups cables. One way to have caps voltage directly to bed and regulated voltage to rest so this way it would waste less in dissipation, which is kinda important overall, especially if its linear like mine, but also for the regulation capacity which then can go down considerably once the bed current is out.


1. Smoothing out the mains voltage rectifier -> almost impossible to do that in an acceptable margin. Big capacitance, and at 400v rating it would be impractical in size and costs. If nobody does it, must be a reason why. Also all chances are you will never get rid of the 120Hz this way, unless you make something which i believe is named a resonant supply, which is also kind of impractical nowadays, e.g. you would probably need an inductance at the size of another 2nd repurposed transformer to use only one widing as an inductor / choke.

2. If we talk about a linear psu, caps on 12v line is somewhat better. But there is a limit to it. If too many caps, the transient curent will blow up the bridge rectifiers at start up, which i think its called "dead switch" or "blow switch" or something like that. So the instant bridge current must be higher than the transient made by the caps at startup. Surprisingly once you do the math, a linear 12v or 24v psu at say 20 amps, needs quite alot of caps and stuff, for an acceptable ripple. I have one with 80ooouf and one with 120ooouf. I had to put current inrush limiting coz they were keep blowing the bridge rectifiers. Even like that when the current of the bed switches on and off then we can expect a voltage jump of 1-2v or smth like that, for small time, even if its regulated the loop wont be like instant as its post-active rather than pre-active.

3. If we talk about switching psu output caps, thats a little different. We dont just need a capacitance there. Also need a certain value for the caps series resistance. And usually a voltage rating 1,2-1,4-1.6 than the input at least. So all critical parts have multiple ratings rather than just one. I think there are different generation switching DC-DC medium current converters, so some chips datasheets will say you need low ESR caps, others you dont need them, some will say use or dont use ceramic caps, etc, sort of depending on when was chip made or what generation it is, weird stuff like that. Most of manufacturers applications for these designs will either give you the exact type of cap and manufacturer and series of the part, or will say you need 100uF wich has 0.143mohms esr rating.

Edited 1 time(s). Last edit at 10/09/2012 10:53AM by NoobMan.
Re: Experimental Power Supplys
October 09, 2012 12:13PM

That is an interesting point, repurposing a UPS. Had'nt though of that one. I have lost prints a number of times through power intruption. Often man made. Anoying as my laptop or netbook keeps on going but the print has died.

Quite a few of the smaller UPS's are 12V so should works as a direct feed in to run a machine.

I think the battery in a UPS should cope with the peak demands reasonably well if it was in circuit.

On your comments re input capacitance to fix ripple. On switching psu's many are in effect DC DC converters, switching rectified mains at high frequency. High frequency makes it more responsive and means it's transformers are a couple of magnitude more efficient (Read smaller and cheaper). Given that the switching is translating a high voltage to low works reasonably well with some degree of input capacitance.

Having had a nosey at the output of the elcectronic halogen PSU's and read an article or two about repurposing them. I have been pondering on how best to put them to the task, and what sort of power rating is realy needed if averaged.

Looking at the waveform it is like one big 50/100Hz ripple switched to 12V.

Here's an interesting doc from motorola


This one appears to have current limiting function. But shows how the mains is recitifed before being switched.

There could be some mileage in repurposing them to act as a fast charger to a 12V battery which is simultaneously running the machine. Sort of like a UPS but with the battery in circuit.

I guess it may be possible to add in an Aux input to take solar or something if available.

It would probably need to be a simplish add on circuit, and cheap, unless the UPS function would be considered to be a useful somethign to have that gives advantages over a re-purposed PC-PSU.

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 09, 2012 12:23PM
Here is a link to a guy that has been experimenting with building his own (not reccomended due to the voltages etc) but it has in it useful info. He has repurposed SMPSU transformers for his experiments.

The waveforms are shown at the bottom of the article.


As I say, very risk and not to be messed about with.

My thoughts were to use one as a black box and deal with only the output. Leaving the internals well alone. Treating it very much like an unusual but cheap transformer.

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 09, 2012 12:35PM
Here's some more notes and some piccys particularly on the last article in the forum.


Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 09, 2012 01:22PM
If the cheap AC units will handle a rectifier it may work financially to use it for hot end and HBP while using a smaller regulated power supply for the motors, and to run the 7805 voltage regulator off of. Or even use the AC directly for the heaters, but then you'd have to custom design the control circuits as you couldn't just use FETs like everyone is now using. It makes sense to split the heater and motor supplies as the motors benefit from higher voltage enabling faster movement, while the heaters really don't care much what you feed them with.
Re: Experimental Power Supplys
October 09, 2012 02:02PM
The ups batteries are 12V, but charging a 12V means 14,4 V or more. Most often there are 2 batteries of 12V in series and the charging current is usually somewhere at 26-27V at least, i think 26.7 typical. I have used a bunch of trafo from these and ofc they have multiple outputs, but only a power one usually with a middle-tap. Usually their output after rectification and smoothing is about 19-21V, with a serious bank of caps, otherwise a bit lower than that. This 20V i personally find it fairly common for the high current transformers i can get my hands on, except ofc those of microwave owens.

Also, the battery cant be used as a capacitor for smoothening the ripple. Think of a heated bed taking out 10+ amps on and off all the times, this is fairly hard to cope with. A battery capacity is measured in Amps*hours, the ups ones should have like 7-12Ah. A battery is charged at max 10% of their Ah number, so to "smooth" 10 amps, with a minimal de-rating, you will need 100Ah, lets say with some derating that is 150Ah. That is 3 full sized car batteries in order to be able to put back the current you take out, and do that in same amount of time. Not to mention battery usage voltage and charging voltages are different. Cant use batteries as capacitors. Maybe can use a battery to ensure a bottom feed for an ocasional small drop in voltage, yes, but i dont see battery smoothening a sine wave.

If you want a straight forward diy route for a psu, check out this one DIY:

Stage 0: Mains power emi filter plus an NTC current limiter -> to trafo.
At least NCT part can be typically be repurposed from a bad atx psu and the Cx-Cy caps for the emi filter, these caps being somwehat special. But the inductor for the emi filter is usually missing on the atx psus, even sometimes there is place on the pcb, but emi inductor its just not there.

Stage 1: trafo-> rectifier-> cap_bank: draw that say ~19V directly for heated bed, perhaps the motors too if they use serious current. Carefully size the caps, calculate inrush current, and compare to the bridge max instant current or "surge" rating. E.g. KBPC3510 diode bridge (check datasheet) 400A surge rating should be (blind guess here) about 60-80.000uF capacitance max. Also for DIY comments, about routing, make sure capacitors are equally used.

Stage 2: about 4-5A at least DC-DC buck converter, preferable switchmode, for extruder. This way the extruder is adjustable, mine is usually at 14V with the typical 6r8 resistor. The inductors, diode, and perhaps even the radiators can be repurposed from an atx psu, at least the inductors from atx should be overkill, just make sure they have the inductance otherwise need to rewind new wire on that core. The diodes in atx are actually very good to-220 diodes, just check the parts search their datasheets and you are bound to find a to-220 schottky diode. Choosing an regulator ic switchmode frequency should be bigger than 200khz. I would personally suggest ti's LM2679-adj, or at least something with current limiting, but there is a tone of these types around.

Stage 3: a number of smaller, preferable linear voltage regulators lets say for a 7-12V margin to feed a microcontroller card like the mega, and a 5V generic supply for accessories and etc. In a hurry can use a 7805 with a voltage divider/pot to make it adjustable (multiturn pots pwn).

At the current levels for a heated bed and rest of electronics, i would say that a switch mode solution efficiency is 60-70 low side to maybe 80% or something. Also complex and all parts need to match each other in 2-3 criterias, which probably takes it out of DIY range. In comparison a transformer efficiency should be, say 98%+.

This is what i would think of a fair way to do this while its still "recommendable and diy-able".

Better solution would be to make a flyback converter with the parts that can be found in atx psu, that is basically doing it again with just different operating points. Also perhaps have the bed driven directly from mains power. Either one is not really "recommendable", as in better dont have other ppls on my conscience.

All those stuff you guys linked is 50W, and we need like a 5-600W or more for a reprap with a serious bed.

I think the 12V voltage in reprap is exclusivelly because of the abundance of 12V atx psus. But its not really all that good, to have the overlook, most of the stepper driver ics will go up to 36-40v. That should be the upper limit of what we actually need. So, 36-40V max voltage, i want some kind of over voltage protection to that limit, that means at least a 4v or more drop, so the actual good voltage for a reprap i think should be 30-32V, that would be the target imho - not the 12v.

Edited 3 time(s). Last edit at 10/09/2012 02:20PM by NoobMan.
Re: Experimental Power Supplys
October 09, 2012 02:28PM
Hmm spliting the power supply by useage perhaps has some mileage too, nice sugestion. I like that one.

Divide and conquer, good strategy.

Consider. The greatest power draw is for the Heaters. (Happy to debate that one).

I have been having a side discusion vie email with Nophead, he sugests that the duty cycle for the heaters settles to aprox 40%. But has'nt studied it too closely as yet.

The heaters dont actualy need smoothing at all. They arguably need a rectified supply. If onlsy so that they can be switched by the Fets. Smoothing out 50Khz+ if we needed to should'nt be too much hassle. The 50/100Hz component can stay as is. It switched through the Halogen transformer OK, It will switched through the heaters OK.

OK, this is starting to look a little more doable.

1. Rectify the Halogen Transformer output. Smooth the 50Khz+ swithching.

2. Take off from 1, the heater supplys.

3. 'T' off from 1 via a blocking diode, supply for the logic and motors.

4. Apply smoothing to the supply derived at 3. we are smoothign a lot less current. so caps are more efective or smaller.

5. Use the supply at 4 for the motors, regulation not needed so much here so not bothered with.

6. 'T' off from 4 the supply for the analogue and logic sections via a 7805 (Or one of the newer switching regulators that are cheap, eficient and can supply aprox 2A).

Hmm I will throw this into kicad and post it back up. Give me a mo.....

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 09, 2012 02:31PM
Just thought, 50Khz+ being kind of high, it may be as easy to filter it with a choke/inductor. or Pi filter type circuit.

Necessity hopefully becomes the absentee parent of successfully invented children.
Re: Experimental Power Supplys
October 09, 2012 02:52PM
Yep biggest consumers are heaters, and by far, the bed. I would say that beside the safety arguments, best way to deal with bed is directly mains power. AC is better, can have a triac with a zero crossing optocoupler so it will switch mains at zero volts, simply no transients to deal with, and triac can be snubber-less type, e.g. wont need snubber. Also probably a sort of full quadrants light dimmer circuit, just to have it as adjustable power, so i would care less about the practicalities about the bed resistance and temperatures.

So on short its a light dimmer from store to adjust power, followed by a solid state relay to switch it on and off with the 5v signal as a fet would do. The only downside would be that zero cross means no pwm (max 50-60 switches per sec), hence it has to be bang-bang. I dont even use pwm with the bed and extruder now, so its ok with me. Another downside is i wouldnt recommend using mains 220v on a movable machine bed, its just simply too many things that can go wrong in the area where we put the fingers all the time. If it would be a x+y+z movement then bed can stay still and then yes i would say the other setup is much better in "reasonable safety" terms. At least could have the 230v area better "enclosed". On the other hand, especially for the safety reason, the 220v on the extruder is opposite, quite a big no-no.
Re: Experimental Power Supplys
October 09, 2012 02:59PM
FWIW: I run my machine pretty much exclusively off 24V, with a 24V stepdown module to run things like fans and the rest of the electronics (arduino) off. The motors (via A4988's, which top out at about 30V), extruder (22R resistor) and heated bed (the MendelMax stick-on one) all run at 24V, which means the current I draw is substantially less than I would have at 12V for the same power output. This pays off in less power/voltage loss over the wiring, particularly to the heated bed.

Note that my machine isn't a MendelMax, but predates it (mostly a Sells Mendel with a retrofitted Prusa X stage). I use the MendelMax stick-on bed as it was better than the initial resistor-heated bed I knocked up.
Re: Experimental Power Supplys
October 09, 2012 03:25PM
I also think 24v is better.

So the classical topology: transformer (ac wave) -> bridge rectifier -> bulk capacitance smoothing (enough to overcome the bed current so when its on the voltage drop wont be too much or wont disturb the set points for the rest of system). Use this (unregulated) output for:

1) bed, perhaps motors (if motors are significant)

2) take it to some voltage regulator (switchmode, max setting at the normal drop + bed on voltage drop), use this regulated output to a) and b)

a)\-> feed extruder b)\->feed 2 other voltage regulators (linear), one adjustable and one 5V

Optional could draw the bed from the trafo before the rectifier, as ac, this way dont need to rectify or smooth the current going there, but it would need a more complicated or expensive solid state relay or a triac+optocoupler which is the same as a solid state relay.

About that transformer type, if it goes to 12V AC the peak dc would be the 1.414 - 2xdiodes voltage drop, thats probably too small. Sort of when the bed will be on then the bed will cause a voltage drop that might put the lines down to a point where the extruder and perhaps the rest of the system wont really work - i think thats the danger here, besides the 100W or the 250W transformer power figure being too low. Again, thats IMHO.
Re: Experimental Power Supplys
October 09, 2012 05:37PM

Starter for now to help with the conversation. This is where we are at.

I think there is a bunch more thinking to go into this, but I can add this as ideas develop.

On using AC to run the heaters. I agree with a standard wound transformer this is probably doable. You are going to need to either use a triac or rectify it and then use a thyristor or transistor.

With the electronic transformers though that 50Khz+ signal is going to mess with thyristors and triacs as they will turn off when the power stops flowing. So you may as well get rid of it that 50Khz which means rectifying and then you may as well use a transistor.

The circuit attached may work with current methods. With a little bit of modification.

Necessity hopefully becomes the absentee parent of successfully invented children.
open | download - HSupplyRR.pdf (20.8 KB)
Re: Experimental Power Supplys
October 09, 2012 05:42PM
Spot the deliberate mistake, not. Just tired saw it when i stopped to look.

Necessity hopefully becomes the absentee parent of successfully invented children.
open | download - HSupplyRR.pdf (20.8 KB)
Re: Experimental Power Supplys
October 10, 2012 09:54AM
hmmm the rectifiers in series that sends power to the motor then the logic rails .... is that neccesary?

as in they are in series when you consider that at the last rail, it drops 1.2v ? and current gonna squeeze thru the middle diode. i think u should fan the rectifiers from the reservoir cap (parrallel distro out) instead of series them out from the reservoir cap. not sure if i make sense. and moreover unless the transformer is of high current supply, the voltage output will drop as higher current is drawn. so you will prolly need to end up with a DC voltage regulator somewhere and the trans will prolly have to select something like a 15VAC, so that you can end up with a fixated 12VDC (deduct diode fwd voltage?) ... etc etc

and also i think, cos these machines will draw considerable power, i feel that it is better to use PWM front end, an analogue transformer will always produce waste heat, at 200w draw, we are prolly going to waste 100-150w as heat approx somewhere in there. i wouldnt mind if this is a class A audio power amp tho grinning smiley

so in my setup, i have thought it out such that i got a 12V PWM @ 10A, a 24V PWM @ 10A, and a step up that will take either 1 and ramp it up PWM style into 30-70VDC for misc use. so all in all, i try to keep efficiencies at >80% approx.

to further enhance it, each rail could have their own fuse, an individual flyback protection diode from ground to +Vcc, a possible 1k shunted LED to help indicate what rail is working, and a further 0.1uF to help eliminate small spikey EMIs.

if any of you have access to ARRL handbooks, they have very nice articles on analogue PSUs. i happen to have DLed 1 2011 PDF copy from torrent the other day n it was really fantastic (in chapter 7). i used to read some of these 10 years ago.

Edited 4 time(s). Last edit at 10/10/2012 10:33AM by redreprap.
Re: Experimental Power Supplys
October 21, 2012 05:40AM
The blocking diodes are there to block the reverse flow of current to a previous stage that is less smooth.

This means that each stage can have progresively better smoothing without needing capacitors that are sized to smooth the previous stage. Ergo cost and space savings. That is the theory anyway. The smoothing for each stage is just enough for the needs of that stage.


Stage 1 the Heaters, has next to no smoothing just enough to clean up the high frequency component. Probably does'nt even need this. Maximum current is drawn from stage 1 ie 120W ish for heating beds and extruders, when all on depending on the machine design.

Stage 2 the Motors, Stepper motors less draw than the heaters but need to have less ripple. So we put a sensible amount of smoothing here. A lot of the motor feed is peaky. ie the average current draw if the drivers and firmware are working well together is quite low. By workign well i mean the drivers whatever they are regualte the current acording to what they are doing and the firmware disables the motors when they are not being used. Remember they draw most average current when still if you let them.

Stage 3 the the logic/analog circuitry supply needs to be smoother than all the rest and maybe would benefit from a degree of filtering/decoupling to remove switching noise from the previouse stages etc. However the current draw for this stage is pretty much the lowest of the 3. unlikely to be as much as 1A. If only because the regulators cannot manage more than this. 5W tops.

As the final stage is going to be feeding maybe an arduino or equivalent with an onboard 5v linear regulator then if we have taken a couple of volts off before it gets there it will not make any difference at all. In fact the regulator will perform better/cooler for having less voltage to drop.

Having said all of the above, I don't know of any of the off the shelf controlers that have split the power inputs into the 3 stages to be fed seperately as above. RAMPS 1.3 onwards has 2. So to use this configuration modifcations or custom controlers are required. Happy to be wrong about this.

Necessity hopefully becomes the absentee parent of successfully invented children.
Sorry, only registered users may post in this forum.

Click here to login