Merging power supplies

[axero]

Let's say that I have a bunch of DC power supplies with the same voltage and a something that I want to run with these power supplies. The problem is that this load takes more amperes than each of these power supplies can run.

So, is it ok to just merge them all into one unit and let them together supply the load? Or do I have to protect the outputs of each supply with power diodes?

What if these supplies have a slightly different voltage, or if they have different power/ampere ratings, would that matter?

[alxnik]

Generally speaking this is not recommended.

Having said that, add a diode on each power supply and ballast resistors. The idea is that the resistor adds some voltage drop on each output so if one power supply provides more current than the rest, its voltage dips more than the others so they kick in too. I have played around with this solution in the past with 3 identical PSUs and 0.5ohm ballast and again I had some difference with the supply current of each of the PSUs. Remember to add a proper rating diode/resistor for the power you are expecting to provide.

[Simon]

some supplies are built to do this but you have to be sure, use the diode and resistor as suggested. but remember that a higher voltage power supply will try to supply more unless you balance the resistors accordingly.

[axero]

Interesting, there are quite a few amps involved here (20-25A total) so I would probably need quite fat diodes I think. Could I use a coil/inductance instead of a resistor as a shunt? Would there be any advantages to using a coil as a ballast?

The input of the load is rated at 12V so I'm not sure if the load could take a voltage drop. However, when looking into the circuitry of the load; the power regulators (FDMF6823A DrMOS modules)  should tolerate 3-16V input if I interpret the datasheet correctly.

How about different rails inside an ATX computer power supply unit? Do they need diode/shunt protection or is that taken care of in the ATX specification?

[alxnik]

Interesting, there are quite a few amps involved here (20-25A total) so I would probably need quite fat diodes I think. Could I use a coil/inductance instead of a resistor as a shunt? Would there be any advantages to using a coil as a ballast?

The input of the load is rated at 12V so I'm not sure if the load could take a voltage drop. However, when looking into the circuitry of the load; the power regulators (FDMF6823A DrMOS modules)  should tolerate 3-16V input if I interpret the datasheet correctly.

How about different rails inside an ATX computer power supply unit? Do they need diode/shunt protection or is that taken care of in the ATX specification?

I'm not sure you understand the concept, and I think what you want to do is move a house with hot air baloons. It's messy, it needs a lot of preparation and it might result in disaster. 25A are a lot and not something you want to experiment with because you could end up with a fire much closer that you would like

The idea is that the resistor adds some voltage drop on each output so if one power supply provides more current than the rest, its voltage dips more than the others so they kick in too.

The voltage drop is not a side effect, it's the reason why the ballast resistor is added. If one psu produces 12.11V and the other 12.05, without a ballast resistor the first one could theoretically power 100% of the load and the other 0%. If the psu doesn't have proper safety design and is overloaded it can very well overheat and catch fire. An inductor does not produce a voltage drop hence it's useless in this situation.
You have to add one diode on each of the power supplies, so if for 25A you are going to use 200 PSUs, you might as well use 1n4148s.

There are PSUs that are designed for load balancing, hence they have proper support for it, even proper feedback loops but most aren't. ATX power supplies have different Voltages on different rails so it is simply impossible to connect them in parallel. Parallel PSUs MUST have the same output voltage

I would strongly suggest that you look for a proper single power supply to power your load. If you really want to use parallel power supplies, find a sharing controller like http://www.mini-box.com/Y-PWR-Hot-Swap-Load-Sharing-Controller or try fully understanding the concept and experiment with much lower current before moving on with a solution

[AG6QR]

Could I use a coil/inductance instead of a resistor as a shunt? Would there be any advantages to using a coil as a ballast?

Absolutely not!  A coil/inductor looks like a dead short at DC.  Pop quiz time:  If you connect an ideal 12.05V DC voltage source to an ideal 12.06V DC voltage source with nothing except an ideal inductor between them, how much current goes through the inductor?

V=IR 
I=V/R
I=0.01V/0
I=infinity

Infinite current is not good for power supplies or anything around them.  Of course, the supplies and/or the inductor and/or the connections between them will show some non-ideal behavior, so the current won't really reach infinity in practice (and besides, the ideal inductor will take infinite time to allow the current to ramp up to infinity, though it will allow the current to grow without bound).  But the point is, it is very likely to do something that's not pretty.

The idea behind diodes and resistive shunts is that they limit and balance the currents even with mildly mismatched power supplies.

Look at the data sheets for your power supplies, figure the worst possible mismatch between them, over the range of loads you'd expect; then calculate the ballast resistors so that all of the currents will be reasonable in the worst case.

Usually, unless the supplies were specifically designed with parallel operation in mind, it's probably better to avoid hooking them together in parallel.  It's generally cheaper and simpler to get a single power supply that's rated for the task, or else a bunch of supplies that are specifically designed to be operated in parallel.

[axero]

I'm sorry, meant ballast I somehow confused it with shunt, I apologize for that. As a ballast it sounds like a good idea to use a coil to protect against sudden inrushes of current as the caps are charging initially. But maybe it wouldn't give a proper protection against a sustained voltage difference between power supplies even if this difference would be quite small.

A load of ~300W@12V means that the total impedance of the load is about .5 Ohms. Using .5 Ohms ballasts would give rise to a considerable voltage drop, so I don't think it is a good idea to use such a high resistance. If we have n ideal power supplies (with perfectly equal voltages u) and assume that all ballasts have equal resistances where each resistance is a fraction q of the load impedance then I find that the voltage drop at the merged output is u*n/(n+q). This means that we want to make q as small as possible if we want minimal voltage drop. On the other hand a smaller resistance of the ballast will give less "protection" against voltage deviations of each power supply. So I guess that there is a trade-off to make here. If we have many power supplies we can use a higher resistance but if we only have a couple of them, we don't have as much "playroom". On the other hand we also know that each power supply is behind a diode so we might not need that much protection.

I need to think further about how much resistance is needed. Perhaps the non-ideality of a coil would suffice to serve as a resistive ballast...

[alxnik]

 
Seriously...did you read what everybody wrote or are you just repeating what you are convinced is true? Please read them carefully again.

1. Ideal power supplies with equal voltages don't need ballast resistors nor diodes

2. A coil on DC is effectively a piece of wire. You will achieve absolutely nothing. Soft start (current inrush limiting) is a completely different matter, and it is probably handled by the PSUs already with NTCs.

3. Nobody said anything about 0.5ohm ballast. You have to calculate the thing according to the expected maximum voltage difference between PSUs, average current and internal resistance of the PSUs. You don't want a minimal voltage drop, you want the smallest possible in order to achieve balance. If the PSUs are so mismatched that you need 1k resistor to achieve balance, then that is what you need to use.

So in conclusion. Keep a fire extinguisher close.

[Jon86]

I think you should give up with your ATX supplies and find a good quality single one to do the work.
ATX PSUs are shit. The output is unstable and noisy, the cheap ones break down so easily, they don't like being loaded on one rail, they don't like switching on and off under load, and if they're not balanced perfectly one will get overloaded and shut down.
If you need any sort of reliability, this is going to give you hell. 

[axero]

I understand that the resistance should be in such a way that the difference between the voltages shouldn't yield a higher amperage than what the higher voltage psu is rated for. However, if the voltage drop due to the ballasts gets to high, the device to be powered may not like it. If the device has a voltage monitor such as an INA3221 it may say; "Power supply is bad, will not power up the device!" if it is programmed to do so.

I was tempted to use Zener diodes to regulate the power (together with some good caps), and that could work I guess. However, the true danger I think is not the slight voltage differences between the PSUs but a possible scenario where one PSU starts to fail. If one PSU is starting to drift down say below 5V whereas the other are maintaining 12V then I agree that we are in danger here unless the PSUs have built-in overload protection.

A bronze/silver/gold/platinum certified ATX PSU rated at 500W+ ought to have sophisticated circuitry for protection. Some PSUs do have several +12V rails so I was thinking that I might want to merge them together to get higher amperage.

Using ideal power supplies was a ceteris paribus assumption made to analyse another part of the circuitry, in this case total voltage drop from the ballasts. Ceteris paribus assumptions are pretty common in many fields of research...

Edit:
Btw, I think this is a good thread and I feel like I'm learning a lot here, I've been wondering about this for a long time now. Load share controllers sounds really interesting, but I guess that they are not cheap at higher amperages.

[axero]

Ok, I have another scenario where one might want to merge several power supplies. But in this case we need not worry about a load that is higher than the power supplies are rated for.

It's about a UPSes. The ones that you buy in a consumer store are not designed to sustain a voltage for very long durations whenever there is an outage. So it is tempting to swap the small lead-acid batteries with car batteries or marine/deep-cycle/AGM batteries that are considerably larger. For simplicity I assume that this "cheap" UPS can handle longer durations than it is rated for and sustain a given load for extended periods of time during outage, when using the larger batteries.

A PSU also have a battery charger, depending on how it is designed, this charger may not like to handle bigger batteries. A common concern is that a smart battery charger may have circuitry that monitors the health of the batteries. A larger battery takes longer time to recharge so a battery health monitor may assume that this longer duration is due to a bad battery.

A solution to this I figure is to add a smart charger circuit onto the batteries inside the UPS. I myself have an APC SmartUPS 5000 and a Mikrobak 2k1 (nice toys, I know ). The APC has four banks of lead acid batteries where each bank consists of four 1280 batteries connected in series, yielding 48V per bank. That gives about 1.5k Watt hours minus inefficiencies in the inverter circuit. The Mikrobak has one bank of four NP38-12, NP40-12 or 1240D batteries. That gives about 1.9 kWhrs minus inverter losses. Also, the APC is rated to sustain a load of 5000VA and the Mikrobak to sustain a load of 2100VA.

I figure that I could replace these banks with one bank consisting of four 75Ah AGM batteries. That would about more than double that capacity of those UPSes, but their charging circuitry might not be so fond of this replacement. So a solution could be to add a 48V smart-charger on top of this bank while putting the smart-charger behind a rectifier circuit to prevent current from flowing the wrong way.


How does that sound as a solution? Perhaps if a self-diagnostics test is made on a load of bad batteries, the smart-charger may not be so happy about it. I think at least the SmartUPS does that routinely once per week.

[Jon86]

You're really starting to confuse me, what are you asking? You were asking about load balancing supplies now you're asking if you can add batteries to a UPS?

[axero]

I'm asking whether one can merge a battery charger with the charger/inverter circuitry of the UPS the way described in the prior post without complications.

[Jon86]

So charge the battery with a separate charger AND the UPS circuit at the same time? No. Not unless you switch between them.

[axero]

It's not very practical to routinely have to remove the batteries from a UPS to recharge them. A UPS is supposed to be a "maintenance-free" solution.

[Jon86]

Well you're just answering your own question. No you cannot charge with two chargers at the same time. Therefore you would have to change between them.

[axero]

Why not? An overload protector circuit may temporarily disconnect the charger while the UPS goes into self-diagnostics mode.

[Jon86]

So you're going to put the UPS into overload protection on purpose? 
If you screw around with the charging like this you could easily damage the batteries AND your chargers. If you want to increase the capacity then you can try replacing the batteries and see how it goes, but I don't think you even want to attempt charging with two supplies. They're just going to be fighting each other, feeding in too much current... Bad idea.

[dr_p]

this is funny, mate. I'm sticking around for the show.

[axero]

The overload protection is not for the UPS, it is for the charger that is also behind some protecting diodes. I can't see how the batteries would get damaged. The float charging voltage is about 12.5-12.9V, the alternator in an auto-mobile can be upwards 14.5V. So I can't really see how two merged chargers would damage a battery. A higher amperage may wear out the battery faster but that's it. The chargers on the other hand may interfere with each other. If there is a floating voltage supplied by one charger, the other one may think that the battery is full.

I think there is a good solution to this however. I may keep the batteries inside the UPS as they are and add the extra bank of 200Ah worth of AGM batteries to it with a smart battery charger permanently connected to the bank behind an overload protection circuit and diodes.

Between this bank and the internal batteries or battery terminals of the UPS is a big (rated for 200 Amps or something) inverted relay. So whenever there is a power surge the relay closes and the batteries get the extra capacity. When the power goes back on the relay reopens and so prevents interfering with the internal circuitry of the UPS.

With that solution we can forget about what I wrote in the first paragraph, pretty much.

[BravoV]

1. Are you looking for helps or assistances on your problem here ?


... OR ...


2. Are you looking for a confirmation on what you believe and sort of expecting other people to nod it for you ?


If the answer is no 2, then go try it, we really-really like to hear the outcome. 


And if we dont hear anything from you for quite some time, then it means ...  Err... naah, just forget bout what I said. 

[Jon86]

Look, just go try it out then get back to us with some pictures of your flaming UPS. 

[axero]

Ever heard of brainstorming? I'm not expecting people to agree with my opinions but a "you can't do that" without a proper argumentation is a pretty bad answer. I present ideas and hopefully someone more experienced who have been in a similar situation before may evaluate it or even come up with alternative solutions. That is what I expect. And sure, I like to take the subject to extremes just to get acquainted with where the limits go.

And, as for these "facepalm" posts and posts implying that I should reread prior posts are not very constructive. They only bring a bad atmosphere to the thread which tends to bring the attention away from the subject at hand.

[Jon86]

You're got to be shitting me, you've literally ignored everything people have said so far, and pretty much just talked to yourself about things.
If you want to come here and ask a question, then please do. But if you're just going to totally ignore feedback then why bother?

[axero]

Ok, let's go back to ordinary power supplies. I've read a prior comment that ATX PSUs are bad. Sure a cheap one is bad but a certified one must be good ríght? There are bronze, silver, gold and platinum certified PSUs. I don't know much about them more than that a higher certification implies better components inside the PSU and higher energy efficiency.

If a lot of switching goes on in the supply voltage from the PSU then I suggested in a prior post to use a capacitor to even the load. But this should be present in the ATX supply, at least if it is certified? I heard somewhere that higher quality supplies are one-rail (at least when it comes to 12V), so perhaps certified PSU are one-rail then?

[Jon86]

To find if you've only got one rail, you'll just have to read the specs, normally they're single 12v rail I think.

[Jon86]

I think the problem you're going to have with these ATX supplies is that you're not using them for how they're designed, therefore you're not going to have a great amount of success or reliability.
If you can afford it, you'd be much better off just getting a dedicated 12V supply that can handle what you need in one unit, then you won't have to screw around with balancing.

http://www.ebay.co.uk/itm/12V-10A-120W-16-7A-200W-20-9A-250W-29-2A-350W-PSU-Power-Supply-Transformer-/190899007235?pt=UK_BOI_Electrical_Test_Measurement_Equipment_ET&var=&hash=item2c7277af03

[axero]

I know I'm seemingly jumping between subjects here but that is because I've been wondering a lot about this.

A common scenario where it is tempting to merge power supplies into one is for low-voltage halogen lights. I think it was this particular scenario that started these thoughts. There are mechanical AC transformers and there are electronic transformers. Many of the electronic transformers for halogen lights are really bad and they even make the filament burn out prematurely. I think one should be careful with the electronic ones even though many of them do have overload protection. The mechanical ones which are just simple transformers should be similar to the DC transformers when it comes to merging as long as they are using the same phase. But with AC voltage we can now use coils and capacitors as ballasts if we want to. I'm open to input about this. But then again, power supplies for halogen lights are not that expensive...

The main reason for starting this thread is for a project that I'm discussing in this thread

https://www.eevblog.com/forum/projects/acpi-power-saving-circuitry-for-150-w-pci-devices-%28ie-gpus%29/

I wanted to get through the basics here before moving on with that project. There are pretty good posts in there already. I myself always use good PSUs for computer hardware but I cannot assume that *everyone* is. So therefore it is good to know what is required to meet the plain vanilla ATX specifications and what is not so that I cover as many user scenarios out there as is reasonably possible. There are good posts in that thread already discussing this but maybe I should give the ATX whitepaper a good read too. The intention with merging the input terminals as described in the project is to save on components.

There are PSUs that are merged into one when I come to think about it. They are found in enterprise-grade hardware where you can hot-swap each should one fail. Most likely a load sharing controller is being used to put them together.

[AG6QR]

Ok, let's go back to ordinary power supplies. I've read a prior comment that ATX PSUs are bad. Sure a cheap one is bad but a certified one must be good ríght? There are bronze, silver, gold and platinum certified PSUs. I don't know much about them more than that a higher certification implies better components inside the PSU and higher energy efficiency.

A certification is only as good as the certifying agency and the certification criteria.  Be careful that you're not putting a lot of faith in a certificate that comes from a thinly disguised marketing agency.  Be sure to read the certification criteria to determine whether it's relevant to your use.

At best, certifications would concentrate on the factors that are relevant to the intended use.  That is, using one ATX supply to power one computer motherboard with accessories.  Good ATX supplies are good at what they're supposed to do, but they may not be so good at operating in parallel, or charging batteries, or providing low ripple, or operating with an external source of voltage on their output.

[AG6QR]

There are mechanical AC transformers and there are electronic transformers. The mechanical ones which are just simple transformers should be similar to the DC transformers when it comes to merging as long as they are using the same phase.

I think you may want to study up on transformers:

http://en.wikipedia.org/wiki/Transformer

None of them are mechanical.  They're all electromagnetic.  Since they work via induction, there are no "DC transformers".

While we're at it, we don't "merge" power supplies.  Occasionally, we may want to operate them in parallel, but that's generally a very bad idea unless they were explicitly designed with parallel operation in mind.

[axero]

Back where I live, transformers for low-voltage lighting are classified either as 'mechanical' or 'electronic'. I can assure you that it wasn't my idea to give them that name. The distinction is made especially for dimmers where some supposedly don't work for the former and other don't work for the latter transformer type.

[tehmeme]

Back where I live, transformers for low-voltage lighting are classified either as 'mechanical' or 'electronic'. I can assure you that it wasn't my idea to give them that name. The distinction is made especially for dimmers where some supposedly don't work for the former and other don't work for the latter transformer type.

where do you live?

[axero]

Ok, look here then;

http://www.el-grossisten.com/pl/Halogenbelysning_Transformer_1407.aspx

[AG6QR]

Ok, look here then;

http://www.el-grossisten.com/pl/Halogenbelysning_Transformer_1407.aspx

In English, those are not "transformers".  OK, perhaps the Danish term "Transformer mekanisk" could be translated to the English "transformer".  I'm not sure what is in it, but it looks like it might be a transformer.  It's not mechanical, though, at least not in the sense of the English word "mechanical".  It has no moving parts; no gears, levers, inclined planes, pulleys, ball bearings, four-bar mechanisms, or similar devices.  It works based on electromagnetic principles.

Similar languages often have "cognates", words that have similar sound and meaning in the different languages.  But beware of the "false cognates", words that sound similar but mean different things.   I could tell stories about misadventures involving the English "embarassed" and Spanish "embarasada" (pregnant); or "excited" and "excitado" (aroused).

I'm sympathetic to the problems of a second language, but when we post on an English language board in English, we should use English words with their English meanings.

[Eddie]

Trying to run power supplies in parallel is asking for trouble unless the power supplies are designed to be used that way (as in a telco installation).

If you need higher amperage than what you can get from what you have buy / build a power supply that can supply what you need for your project. Remember linear power supplies are easier to make but thay weigh a lot more and tend to be more reliable (in my experience) switch mode are nice and light but can have voltages I shy away from now a days, also switch mode power supplies can have harmonics on the output and can have other undesirable (for RF people like myself) frequency radiated from the cct.

What are you trying to run and what amperage are you looking at needing for your project?

Cheers Eddie

[lapm]

Ok, let's go back to ordinary power supplies. I've read a prior comment that ATX PSUs are bad. Sure a cheap one is bad but a certified one must be good ríght? There are bronze, silver, gold and platinum certified PSUs. I don't know much about them more than that a higher certification implies better components inside the PSU and higher energy efficiency.

A certification is only as good as the certifying agency and the certification criteria.  Be careful that you're not putting a lot of faith in a certificate that comes from a thinly disguised marketing agency.  Be sure to read the certification criteria to determine whether it's relevant to your use.

At best, certifications would concentrate on the factors that are relevant to the intended use.  That is, using one ATX supply to power one computer motherboard with accessories.  Good ATX supplies are good at what they're supposed to do, but they may not be so good at operating in parallel, or charging batteries, or providing low ripple, or operating with an external source of voltage on their output.

Lol, i was about to ask this too. Certified by who and to do what exactly... Without knowing that certificate is not worth paper its printed on.