Looking for a design: Mod to a Computer PSU that will provide Constant V / A

[irakandjii]

The Need :  I would prefer to not to spend >$750 (Cdn) for a linear bench power supply if possible.

I need a bench power supply for IoT and embedded system work.  12v max is fine (driving motors, relays) and most commonly I use < 5v to drive components, accurate control of 3.3v is a high priority (Most of my stuff is 3.3v).

I have installed a ATX Breakout Board from Dangerous Prototypes that presents the PSU voltages to some binding posts.  I believe this circuit ensures a proper "load" to the PSU to prevent it from overloading.  This is ok but far from ideal. 

Materials on hand:
I have some High quality (800+ watt) Computer Power supplies that I no longer use, but in great shape.  These supplies are "Gold" standard efficiency, high quality etc. etc.

They provide +12V, -12V, +5V and +3.3V
I have 30+ Amps available on each 12V rail.

Goal:  I would like to build a project which will use the output from the PSU and provide:

Some Points of clarification: (from post below)

I envision a 3 part system:

A) The computer PSU is the back end and is not opened or in any way tampered with, I expected to use the 24 PIN Molex as the Input connector (no wire cutting) it provides power to parts B & C.

B) Power Rail (+12v)  This is a fused rail (or current limited supply) that provides power to mechanical components motors, actuators etc.  (would be a separate battery in the real robot)  The bulk of the PSU power would flow this was up to say ~300 Watts

C) Front end (the project) provides the variable  voltages and currents for the electronics test bed

1) I expected to come off the PSU immediately into a circuit breaker / polyfuse as protection into the project.
2) The output of the project would be max 5 Amps on the 12v and 3 amps on the other output voltages.  I could get away with less.
3) I was looking for a way to make sure any noise/ripple in the PSU was reduced within the project, and not passed to the outputs
4) I was hoping to find a way to make sure the front end was very precise since it controls the electronics
5) "Emergency" shut down capability from the project to shut off all power via solid state relays.  Example: Big red Button and/or triggered from conditions on the outputs

So I look at it as if we have a constant voltage source to multiple precision power supplies.

Part C - The Front End cont'd

1) Constant Voltage output set by a variable resistor .. highest possible with the +/- 12v sources.
2) Variable Current limits on all voltages, essentially a variable fuse on each of the stock voltages (+12v, -12v, +5v, +3.3v)
3) Constant current output using the +12v & -12v sources, Ideally with a settable voltage <= the source.
4) Digital display of settings, (maybe digital control of voltages etc. via digital resistors.).
5) Avoid voltage spikes at all costs (I have spent too much already on blown materials)
6) Need through hole parts, I can't mount tiny things on boards

If anyone knows of an efficient robust design for this please provide a link so I can find it.

Background: It is not for lack of trying that I come to this request ...

I made a few "Internet Projects" that have failed dangerously,  one in a spectacular fashion with an explosive mosfet! (parts bounced off my eyeglasses) others just overheated, smelt bad and ejected magic smoke.   

So just "googling it" has presented a lot of "solutions" but, I do not have sufficient knowledge to choose between the alternatives. As-in what is a professional safe design vs some young players daydream.  I have seen many .. many .. internet solutions based on LM317 or 78xx voltage regs but just as many conflicting opinions about using them.   I have also seen folks say just use a transistor with no further detail, quite frankly if I understood what to do I would not need to make this post.

I have watched Dave's Videos for a PSU, but it seems the designs evolved so much that I am not sure what portions are applicable to my need.  It may be that what I need is a sub-component of his design.

Finally, I am interested any pointers to a really solid and current educational resource on precision power supply.  I have become so "confused now" and I can't see the forest for the trees.   

Thanks in advance

[Kleinstein]

The computer power supply is not a good starting point unless you want something like 12 V and 25 A with lots of noise. As a side note, the -12 V is usually only very low power, maybe 100 mA if you are lucky.

A better starting point would be an old laptop supply - these are usually something like 20 V and 3 A. So an output up to 16 V or so might be realistic. Dave's design is no that great - using an voltage regulator and adding current limit as an afterthought is problematic.
 
There are linear lab supplies for a reasonable price if you don't insist on high power and resolution. The main drawback with the cheap Chinese ones it that they tend to overheat at there rated current. So as long as you stay away from the maximum current they can be reasonable good.

Building you own supply is more about learning, not so much about saving money.

[Avacee]

One reason computer PSU's are not recommended as bench power supplies - even though there are a LOT of projects out there when you "google it" - is that if anything goes wrong a PC PSU can supply a lot of current. You mention 800W PSU's and the 12V rail having 30Amps. That's a LOT of current to blow through You and/or your IoT/Embedded system when something goes wrong.

You mention a budget of CDN750 so may I suggest looking at a "proper" bench power supply with all those protections build in.
You can then use that proper bench PSU with its protection as a source whilst you learn to build your own power supply.
Something like the popular http://www.rigolcanada.com/products/dc-power-supplies/dp800/dp832/

Converting a PC PSU into a bench power supply is a fun project, and I've done it myself, but if you're currently blowing stuff up whilst learning then blowing stuff up with a PC PSU behind it it can blow up big and bad and worse, hurt you at the same time.

Learn how to do constant current, voltage, overload protection, voltage spikes,etc using a small power supply first (ie a wall wart,laptop charger/a proper bench psu).
When you have those fully understood then move up to something can hurt you

Not to put you on a downer as a self-built PSU is a fantastic project for learning and its highly satisfying to use your own gear - but please don't learn with something so dangerous.
Another great project when building a PSU is a DC Load to test it with - how will you know your PSU control circuitry is working unless its tested with a known load?

Peter Oakes, imho, does a good job of explaining the basics of building the control circuitry for a Power Supply and DC Load (as see Dave's EEVBlog #102).
Work through learning PSU design with a wallwart/laptop charger/proper bench psu first
https://www.youtube.com/playlist?list=PL_atu5RtEPi4aNzoMtZ5_S6ruhFR98T_p

[BrianHG]

Even a crummy supply like one of these is a safer bet than retrofitting a PC supply (I listed 10amp and 30amp units):

http://www.ebay.ca/itm/Low-Cost-0-30V-10Amp-Bench-Power-Supply-CSI3010SM-/221720976295?hash=item339f999ba7:g:MqwAAMXQeW5TYsax
http://www.ebay.ca/itm/CSI3020SW-0-30V-0-20A-Switch-Mode-Bench-Supply-/381200355498?hash=item58c14fd0aa:g:3NkAAOSwofxUdggQ
http://www.ebay.ca/itm/Hi-Resolution-0-01V-0-001A-Mini-DC-Power-Supply-Output-0-15V-0-10A-AC110-220V-/302236440409?hash=item465eb21759:g:A~kAAOSwjDZYY8tO
http://www.ebay.ca/itm/Industrial-10-Amp-30-Volt-Adjustable-Digital-DC-Power-Supply-Box-Kit-w-Clip-Cord-/201702004870?hash=item2ef6605886:g:bCwAAOSwB09YE5Yn
http://www.ebay.ca/itm/Leader-718-10D-Laboratory-DC-Power-Supply-18-Volt-10-Amp-/142314771443?hash=item21229f13f3:g:wwsAAOSwLF1X2YKw
http://www.ebay.ca/itm/Hi-Resolution-0-01V-0-001A-Mini-DC-Power-Supply-Output-0-30V-0-10A-AC110-220V-/282376251674?hash=item41beefa91a:g:KMEAAOSwjDZYY8w5

Though most of these are Chinese crap, they are safer to develop with compared to the raw output of a PC power supply.

[Monkeh]

The Need :  I would prefer to not to spend >$750 (Cdn) for a linear bench power supply if possible.

Okay. How about <$100?
https://www.circuitspecialists.com/bench-power-supply-csi1802x.html
https://www.circuitspecialists.com/csi3003sm.html
https://www.circuitspecialists.com/csi5003xe.html

etc etc.

[farzadb82]

If you only need a single supply at any given time, I would highly recommend the Korad KA3005D (http://www.ebay.com/itm/120922717980?_trksid=p2055119.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT) as budget power supply. Dave has done a video on these, which is what prompted me to get one. I would not recommend the dual output version, however. The power supply itself is fine, but the UI is somewhat kludgy.

Alternatively, if you want to go even more budget, you could go with something like the DPH3205 (http://www.ebay.com/itm/32V-5A-Adjustable-Digital-Programmable-Buck-Boost-DC-Power-Supply-Module-H9H0-/132030603424?hash=item1ebda320a0:g:60kAAOSwA3dYS5S6). I personally have not used them, however, there was an article on hackaday a little while ago about how "hackable" these were and a small community has started to grow around modifying them to provide nicer features. Pair this up with a nice laptop power supply that can provide 19v and you should easily be able to get the voltage range you need and for < $100.

[BrianHG]

More 20 amp cheap junk:
https://www.aliexpress.com/item/PS3020-DC-switching-power-supply-30V-20A/32757587746.html?spm=2114.01010208.3.45.ZWkm76&ws_ab_test=searchweb0_0,searchweb201602_6_10065_10130_10068_433_434_10139_10136_10137_10138_10060_10062_10141_10056_10140_10055_10054_10059_10099_129_10103_10102_10096_10052_10053_10050_10107_10142_10051_10106_10084_10083_10080_10082_10081_10110_10111_10112_10113_10114_10078_10079_10073_10070_10122_10123_10124-10102,searchweb201603_9,afswitch_1,ppcSwitch_5,single_sort_0_default&btsid=6db8291a-7e75-45cc-9549-d2388ba4201a&algo_expid=761181d4-f710-452f-9d3e-59a95f77e1a7-5&algo_pvid=761181d4-f710-452f-9d3e-59a95f77e1a7
https://www.aliexpress.com/item/Precision-PFC-Compact-Digital-Adjustable-DC-Power-Supply-OVP-OCP-OTP-32V-20A-220V-0-01V/32328835922.html?spm=2114.01010208.3.46.XxWfZ7&ws_ab_test=searchweb0_0,searchweb201602_6_10065_10130_10068_433_434_10139_10136_10137_10138_10060_10062_10141_10056_10140_10055_10054_10059_10099_129_10103_10102_10096_10052_10053_10050_10107_10142_10051_10106_10084_10083_10080_10082_10081_10110_10111_10112_10113_10114_10078_10079_10073_10070_10122_10123_10124,searchweb201603_9,afswitch_1,ppcSwitch_5,single_sort_0_default&btsid=14268e1a-bde6-487c-9ad3-f3c370d96bc6&algo_expid=f1bac658-e2f9-4810-9f86-ba617608a05f-5&algo_pvid=f1bac658-e2f9-4810-9f86-ba617608a05f

Dont forget to verify your AC input voltage support.

[james_s]

My most used bench PSU is one that I built using a 32V power brick from one of those printer/scanner/copier things that I got for free and a $5 Chinese buck regulator module with current limiting. I mounted it all in a box I salvaged from work and put a cheap LED V/A meter on the front panel. Removed the pots from the regulator board and replaced them with panel mounted pots. Output is a bit noisy, expected since it's a switcher, and the current limiting oscillates a bit but does the job keeping the magic smoke from getting out. I have less than $20 invested in the whole thing and it gets used a lot because it's compact and very light weight.

If you want a linear PSU there are some simple designs out there that have adjustable current limiting. Power supplies make good projects IMO, they're fun to build, not too complex and you can never have too many.

[elecman14]

You could hook something like a DP20V2A (multiple current and voltage rating modules are available) to your computer power supply. Not sure how much headroom you would be left with if you used the 12V rail though. Might be better with a laptop power supply. You also could add fuses or buy a product with poly fuses like this: http://dangerousprototypes.com/docs/ATX_Breakout_Board

[Smokey]

I'm always amazed that people think they can desolder a magic resistor or add a magic capacitor or something and turn a PC power supply into a high precision, professional, programmable, Ethernet connected, LCD displaying, bench supply.  It's even more silly to think there is a magic circuit you can solder in somewhere that will only cost you $3.50 that does the same thing.  And it only takes 5 minutes to upgrade!
There are guys that spend their entire career designing circuits to do this stuff.  There is a very fundamental lack of understanding on how power conversion works here.  Might as well be asking how to make a Mars Rover.

Buy a power supply and have fun building whatever you are planning on building.

[irakandjii]

You could hook something like a DP20V2A (multiple current and voltage rating modules are available) to your computer power supply. Not sure how much headroom you would be left with if you used the 12V rail though. Might be better with a laptop power supply. You also could add fuses or buy a product with poly fuses like this: http://dangerousprototypes.com/docs/ATX_Breakout_Board

The DO20V2A is the sort of circuit I image, but with 12V in so it is the correct concept

Thank you, the ATX Breakout Board is in fact the board I have now.

Some Points of clarification:

I envision a 3 part system:
A) The computer PSU is the back end and is not opened or in any way tampered with, I expected to use the 24 PIN Molex as the Input connector (no wire cutting) it provides power to parts B & C.

B) Power Rail (+12v)  This is a fused rail (or current limited supply) that provides power to mechanical components motors, actuators etc.  (would be a separate battery in the real robot)  The bulk of the PSU power would flow this was up to say ~300 Watts

C) Front end (the project) provides the variable  voltages and currents for the electronics test bed

1) I expected to come off the PSU immediately into a circuit breaker / polyfuse as protection into the project.
2) The output of the project would be max 5 Amps on the 12v and 3 amps on the other output voltages.  I could get away with less.
3) I was looking for a way to make sure any noise/ripple in the PSU was reduced within the project, and not passed to the outputs
4) I was hoping to find a way to make sure the front end was very precise since it controls the electronics
5) "Emergency" shut down capability from the project to shut off all power via solid state relays.  Example: Big red Button and/or triggered from conditions on the outputs

So I look at it as if we have a constant voltage source to multiple precision power supplies.

 


I keep reading in the eevblog forum that lots of power supplies have big start-up spikes.  I was hoping to avoid this.

[irakandjii]

I'm always amazed that people think they can desolder a magic resistor or add a magic capacitor or something and turn a PC power supply into a high precision, professional, programmable, Ethernet connected, LCD displaying, bench supply.  It's even more silly to think there is a magic circuit you can solder in somewhere that will only cost you $3.50 that does the same thing.  And it only takes 5 minutes to upgrade!
There are guys that spend their entire career designing circuits to do this stuff.  There is a very fundamental lack of understanding on how power conversion works here.  Might as well be asking how to make a Mars Rover.

Buy a power supply and have fun building whatever you are planning on building.

Oh you assume too much my friend.  I do not think this is easy as you should see in my post, nor do I think it is a $3.50 exercise.  I expect that a proper and safe design would take some effort to do.  As it is the PSU's are worth over $300 given what I want to do $100 in parts would be fine.

I am hoping that one of the folks that has spent 30 years in the field would have some designs to offer up to the public domain as I have on the software front.  Good folks are out there.

But thank you for taking the time to express your opinion. 
I am sure your approach will help someone else justify throwing money at a problem rather than be open and learning of a way to do better.
I prefer to learn and make something right. 

[djnz]

Have you taken a look at this:

DIY SCPI programmable dual channel bench PSU 0-50V/3A (now EEZ H24005) by prasimix

https://www.eevblog.com/forum/projects/diy-programmable-dual-channel-bench-psu-0-50v3a/

[Avacee]

Your requirement sounds like a smooth linear 0-12V,0-5V 0-xA with typical bench psu features and the reason you haven't found a design from someone with 30 years experience whilst googling is because those designs do not exist as people with 30 years experience recognise that a PC PSU is, quite simply, the wrong tool for this job. You are trying to fit the proverbial square peg in a round hole.
If such an instructable / circuit existed you'd have found it.

Safety aspect. If something goes wrong your 12V 30A rail is capable of 30A sustained. It's peak when something goes tits up will be much higher before the PSU's safety trips. That's beyond a slight tingly feeling in the end of your finger. Along with a melted/ignited breadboard. You mention an "oh-shit switch". It's going to have to be a bloody good switch. Wrong tool for the job.

You can't do 0-12V from a 12V rail simply because you'll drop some voltage over the transistor/mosfet/whatever.
So you would need to boost the voltage to higher than 12V and then drop it down again. Efficiency issues. Wrong tool for the job.

The ATX Spec details a tolerance of 11.4V-12.6V with a max ripple of 120mV on the 12V rail land 4.75V-5.25V with max ripple of 50mV on the 5V rail.
So your boost converter is getting a noisy rippletastic input and boosting it to something that's going to also be noisy which you then have to smooth.
People with 30 years experience will just get a smoother power input at more than 12V and avoid having to deal with both the PSU's noise and boosting it. Wrong tool for the job.

Mains Isolation. It's recommended that a bench psu is isolated from the mains which a  PC PSU isn't. Wrong tool for the job.

Right tool for the job for developing your IoT/Embedded systems = buy a proper bench PSU with all the protection built it that is isolated from the mains and make one of your projects building a power supply that isn't powered by a ripple-tastic massively over-powered mains connected PC PSU.
12V 3A can be provided by a Laptop charger. I use an IBM one that's 16.8V 3.33A. The 16V can be dropped down to 12V in one go without needing boosting first and it handles 3A ok.

When is a PC PSU the right tool?  Powering motors/lots of LEDS/other high current devices/etc at 5 or 12V.
There's lots of examples on how to put a PC PSU into an enclosure and wire up some banana jacks. You've probably seen them with your previous googling and that's what the breakout board provides.
Those tutorials will also explain about the dummy load requirement that's needed for an SMPS to achieve regulation.

Here's a project using a buck/boost converter if you are dead set on doing it:

[irakandjii]

Thank you for taking the time to answer in detail.  I really appreciate the effort you took,  but it seems my requirement has been badly written or misunderstood.

Your requirement sounds like a smooth linear 0-12V,0-5V 0-xA with typical bench psu features and the reason you haven't found a design from someone with 30 years experience whilst googling is because those designs do not exist as people with 30 years experience recognise that a PC PSU is, quite simply, the wrong tool for this job. You are trying to fit the proverbial square peg in a round hole.
If such an instructable / circuit existed you'd have found it.

Yes but to be clear: The component I need is a precision V & A controller with multiple 0-5A outputs in the 0-9V range AND a (protected) 12V power bus that takes its source from a 12V dc supply that happens to be a Computer PSU which is unadulterated.   I see the PSU as a big cost optimized over millions of units rectifier or DC source.    I was unaware that electrons had achieved sentience and knew what source they came from.


Safety aspect. If something goes wrong your 12V 30A rail is capable of 30A sustained. It's peak when something goes tits up will be much higher before the PSU's safety trips. That's beyond a slight tingly feeling in the end of your finger. Along with a melted/ignited breadboard. You mention an "oh-shit switch". It's going to have to be a bloody good switch. Wrong tool for the job.   BTW I have a lot of 12V marine switches, relays etc designed for 200+ amps.
Agreed but how is this anymore dangerous than using a 12V (14v nominal) 150 Amp Hour deep discharge battery with the potential for > 600 amp current?  That is what I use now 

To clarify: 
1) The PSU is unchanged internally or externally. 
2) For the 12v power rail,  I am taking 12V from the PSU immediately into a fast blow fuse of "breaker", most likely 12A.  From there this power bus will connect to a "controller" designed for use with 12V mechanical elements.  Motors linear actuators etc.  These controllers would be managed from the electronics under test via a serial bus. (Likely CAN bus, but TBD)

The idea is this power circuit (item B in my spec) is taken off the 12v rail and is parallel to the "front end control" circuit I need the design for.

The "oh crap" switch would likely be a normally open relay, sized to properly disconnect the bus in case of errors in the control unit.  It is a primarily mechanical safety device not an electrical safety component.  The mechanisms to trigger the "relay" would be encoded in the "DUT" as part of the error routines for exceptions or faults or out of spec conditions.  The "front end" simply needs a set of connection points exposed to the user,  the mechanical "big red button" is simply another way to trigger the relay.  (Used for ... HEY! you are about to fall off the bench you silly robot! "press button now")

You can't do 0-12V from a 12V rail simply because you'll drop some voltage over the transistor/mosfet/whatever. 
I know about the voltage drop but, 0-12v is not part of the spec for the "front end" electronics portion.  A rheostat or pwm motor controller is sufficient for the motors.  I should have been more precise.  The design I need is for multiple 0-9v,  0-5v, 0-3.3v outputs (probably 3, maybe 6) for the precision front end part.

So you would need to boost the voltage to higher than 12V and then drop it down again. Efficiency issues. Wrong tool for the job. Assumption!
No I don't as explained above.

The ATX Spec details a tolerance of 11.4V-12.6V with a max ripple of 120mV on the 12V rail land 4.75V-5.25V with max ripple of 50mV on the 5V rail.
Ok and that is a design constraint.  Is there some evidence that this is insurmountable?  We know what it is.  Can we not design for it and minimize it?

So your boost converter is getting a noisy rippletastic input and boosting it to something that's going to also be noisy which you then have to smooth.
Don't need a boost converter.

People with 30 years experience will just get a smoother power input at more than 12V and avoid having to deal with both the PSU's noise and boosting it. Wrong tool for the job.  Just need smooth 12V to feed the 0-9v, 0-5v, 0-5A parts.
I see the PSU as a whacking great rectifier with a lot of the design & construction heavy lifting already done.  And, in many cases with a lot more attention to detail than what I have seen in budget "bench" power supplies from china.  If we know its characteristics why can we not design for it.  The Bench power supplies I have seen for > 300W are really expensive and I do not need "clean" signals for the 12V part. 

Mains Isolation. It's recommended that a bench psu is isolated from the mains which a  PC PSU isn't. Wrong tool for the job. Assumption!
TRUE! but the PSU is the back end and physically isolated from the work bench.  I was under the impression that the outputs for the front end could be isolated with proper and safe design which is the primary ask in this post.  I am a wrong here?  If so how do traditional bench supplies achieve this?

Right tool for the job for developing your IoT/Embedded systems = buy a proper bench PSU with all the protection built it that is isolated from the mains and make one of your projects building a power supply that isn't powered by a ripple-tastic massively over-powered mains connected PC PSU.

The IoT part is the "brains" & "nervous system".  After a lot of research regarding a proper bench supply that meets my needs I have to budget >$1000 (cdn)  it is on my list, but sadly in the future.
 
AND I still need 300W plus at 12 V mechanical rail for this testing bench.  (the final product will potentially have 3kW peak motors)

12V 3A can be provided by a Laptop charger. I use an IBM one that's 16.8V 3.33A. The 16V can be dropped down to 12V in one go without needing boosting first and it handles 3A ok.
Yes, but how is this different electrically? I would be still be using a laptop PSU at 16.8v source instead of 12V.   I would need to drop from 16.8V to 9V at 5A for the 0-9v section (5*7.8~= 39 watts of loss for each 9V output worst case?  
BUT, this is the concept I want!  The circuit after the PSU is the important bit!

When is a PC PSU the right tool?  Powering motors/lots of LEDS/other high current devices/etc at 5 or 12V.
YES exactly!  I just need the precision front end to power the stuff that manages the stuff in the back!

There's lots of examples on how to put a PC PSU into an enclosure and wire up some banana jacks. You've probably seen them with your previous googling and that's what the breakout board provides.
Yes this is what the ATX Breakout Board does! as mentioned in my request

Those tutorials will also explain about the dummy load requirement that's needed for an SMPS to achieve regulation.
Understood, my PSU's do not have this requirement.

Here's a project using a buck/boost converter if you are dead set on doing it:

Thank you for this reference but I wanted something a little more refined.  Therein lies the challenge.

Thank you for this, I edited the quote so I could match my feedback to you comments.

[alm]

Given the task of taking 12V from an ATX PSU to create multiple clean, floating 0-XV rails, I would probably just use the ATX power supply to power an inverter, and from that 110/230 VAC power a traditional bench supply . Going through a bunch of isolated DC-DC converters and filters seems like an awful lot of effort just to save a few transformers and caps.

[Nusa]

Something like this is probably your easiest route to do what you want. The DPS line has gotten reasonable reviews. The hardware has been hacked, and there is an open source firmware you can install if you don't like the original.

https://www.aliexpress.com/wholesale?catId=0&SearchText=programmable+power+supply+module

[irakandjii]

Given the task of taking 12V from an ATX PSU to create multiple clean, floating 0-XV rails, I would probably just use the ATX power supply to power an inverter, and from that 110/230 VAC power a traditional bench supply . Going through a bunch of isolated DC-DC converters and filters seems like an awful lot of effort just to save a few transformers and caps.

  ..  I just don't know what to think, clearly you are trying to be helpful in some way.  I just can't imagine what it is?