Parallel Boost Converters

[@evildaddysteve]

When using non-isolated dc to dc boost converters, I use separate power supplies for each converter. This allows the outputs to be combined. The ideal solution is to use converters with isolated grounds. Then they can be fed by a common power supply.

[@celioantonio7943]

When using converters in parallel, it is not necessary to use separate sources, only use in the case of series converters.

[@TheEmbeddedHobbyist]

The problem is the one that the converter with the lowest setting will stop providing power to try and get the output voltage to drop. Which means the the one with the highest feed back setting will then be taking the full load. Putting a diode will not make a lot of difference to the power sharing. it will just stop one back feeding the other. Paralleling power sources can be very tricky, even removing the feedback form one and using the same feedback to feed both might work. Done this with very high power alternators and it can be very expensive to provide all the required control signals so the big buggers share the load.

[@JulianIlett]

Yes, exactly :)

[@ernieschatz3783]

Yup!

[@d614gakadoug9]

The output current of a boost converter, unless it utilizes an additional switch, cannot be limited to less than the supply voltage divided by the load resistance, ignoring the resistances in the circuit itself, normally dominated by the inductor. Boost converters without additional switches therefore cannot be made short-circuit proof. Paralleling constant voltage power supplies of any sort generally works poorly without additional circuity. Degrading the voltage regulation by reducing the error amplifier gain can help, but that of course degrades the overall voltage regulation precision. Another alternative is adding output ballast resistors, but that also degrades voltage regulation. Paralleling constant current supplies generally works reasonably well, each supply delivering current according to its setpoint. (remember that an ideal voltage source has zero source impedance while an ideal current source has infinite source impedance) When the current sensing is in switch circuit, the current regulation is not accurately reflected in the output circuit. This is simply a function of the fact that the regulation is normally based on _peak_ switch current, not average switch current.

[@sc0or]

You've just said that a current through a battery connected to a certain resistance cannot be limited without an additional switches. Sure! It cannot. This is an Ohm's law )

[@michaelkaliski7651]

The sensible option would be to use a single control circuit and parallel the input and output components. So the parallel inputs and outputs are reconfigured to accept the control signals from a single board, while the other control circuit is removed or disabled. Otherwise one circuit will always attempt to drive the other at the output due to component tolerances and variations on the voltage/current limit potentiometer settings.

[@snakezdewiggle6084]

Right idea. I thought he knew electronics !

[@pritambissonauth2181]

@@snakezdewiggle6084 I am surprised too, does he know what he is really doing!

[@korishan]

I'm thinking that you could either 1) use diodes on the output Pos lines, it'll keep both units separate 2) separately set both converters to the exact same voltage and then connect them in paralle 3) use a high frequency 1:1 transformer to completely isolate the outputs and the parallel the isolated outputs after the smoothing caps

[@SuperBrainAK]

Simply putting diodes won't work, what you want is resistance so that slight changes in voltage set points will equal smaller differences in current. The diodes will just be forward biased the entire time and contribute nothing but resistance and power loss which can be more efficiently done with low value resistors. 1:1 transformers might work, replace the inductors with something more like a SEPIC converter but that might just turn into a project of "just increase the power rating of the converter by doubling up the components".

[@SuperBrainAK]

That is intriguing behavior from those converters. I remember the last video discussing this and I said you had to have precise control of the voltage which is certainly the case here. I am also a huge advocate for positive current shunts. Now they are still negative shunts but not in the usual way. Which I can't say many bad things about. I just think about input voltage fluctuations changing the current output. Your idea of just setting the current limit to what the units can handle at maximum is a great solution.

[@JulianIlett]

Thanks :)

[@d614gakadoug9]

The failure to share the load when the supplies are operating in constant voltage mode is absolutely predictable. The worse the individual supplies are at accurately regulating voltage, the better they share. This is true of switchers and linear regulators. It is simply a result of the fact that constant voltage supplies have very high gain error amplifiers and they are trying to behave like they have zero source impedance. In constant current mode supplies try to behave like they have infinite source impedance, thus they inherently share according to their setpoints. edit, forgot to mention: High-side shunts are a lot harder than low-side shunts. Depending on just what you are doing you have to select a shunt amplifier that can handle voltages at least to and possibly greater than its own supply rail. That limits choices significantly. You also have to have good common mode rejection to keep the error manageable. An integrated differential amplifier can be a good though expensive solution. If you try to make a diff amp with discrete resistors you may find you need to use resistors of 0.1% tolerance or better, depending on the output voltage of the supply and the voltage across the shunt.

[@dand8282]

You can add some small balancing resistors to the output to share I in CV mode, just like driving parallel transistors with slightly different hfe figures; the voltage drop gives you a bit of voltage margin to fall within.

[@d614gakadoug9]

That can work but at the expense of efficiency and voltage regulation. Trying to accomplish balance within a few percent is likely to result in unsatisfactory performance on both parameters.

[@toreediassen144]

It can be good to use cables of the same length (and the same quality) from the converters. Then these wires will act as small resistors, and the small voltage loss will cause the load to be distributed more equally (if the converters are otherwise the same). A bit hard to explain without a picture.

[@JulianIlett]

I get what you mean - but the important thing is to get the converters into current limit. When they're voltage limiting, it's virtually impossible to get them to current share.

[@toreediassen144]

It should be possible to get 2 identical converters to deliver approximately the same amount of current. At low loads, there will likely be a difference, but at high loads it "should" be fine. Of course, it depends on the quality of the converters, and how much voltage loss you accept in the resistors (wires)...@@JulianIlett

[@ernieschatz3783]

​@@JulianIlettThis! The outcome of this test was known before the start of it. It's a good illustration for the "show me" people.

[@SotirisPiloridis]

Friend Julian, thank you very much for the video, you saved me from about 150 euros that this experiment would have cost me. It's lucky that you did it. Thank you very much. I know if it was already successful thank you very much from Greece Be well Julian thank you very much

[@murraymadness4674]

A long time ago, I created a dc-dc converting using a lm2576 (I think) when they just came out to power wireless equipment. Since I had a lot of them I tried to use multiple in parallel and it was a complete fail. I didn't have current limits on them, so there was really no chance, one does all until it blows up. lol This was when I was an electronics newbie, so I learned something from this experiment. What I looked into was using the chip to drive an external mosfet and bigger coil to get more power, and I determined I should just buy a different chip that was designed to do that instead of the lm2576, as the control chips particularly now are a small part of the cost. As you mentioned, the '1200' watt units sold are really not even close to that rating. I want to use a 12v battery and generate 36v to run a e-bike or other things, like large displays or spotlights with leds in series and they won't really do it.

[@DeckCain]

You can add diodes to the output so that the second DC-DC converter does not receive the voltage from the first one. This way, you can effectively balance them even when in voltage limiting mode.

[@DiyNuke]

Thought the same but no spend many hours on it and every time one would do the work until you set the second one high enough but then it would runaway and burn the fuse

[@d614gakadoug9]

That will not work. Assuming perfectly matched diodes, the supply producing the highest voltage ahead of its diode, even by a few millivolts (depending on the gain of the error amplifier will take 100% of the load until it transitions to current limiting.

[@neutronstorm]

The problem with this is that the current measurement is on the negative side for those devices. Basically the current from one returns back through the other causing that one to keep turning down its current until its effectively off.

[@d614gakadoug9]

@@neutronstorm Unless the circuit is _very_ unusual, the voltage across the current sense resistor is irrelevant to the failure to share current when the supplies are operating in constant voltage mode. That failure is due to the fundamental nature of a voltage regulator - it is "trying' to behave like a source with zero impedance. It does this by means of an error amplifier that has very high gain, typically in the ranges of hundreds of thousands at DC. Error amps are typically run with no local feedback at DC, so you get the full open-loop gain of the amp.When you connect two power supplies operating in constant voltage mode in parallel (call them A and B) and A is st (say) 10.000 volts while B is set to regulate at 9.995 V, the error amp in supply B will tell the power delivery part of the circuit (linear or switcher, it makes no difference) to turn off completely. You might get both A & B to share with very careful adjustment, but changes in temperature could easily destroy the balance. If the total load current cannot be delivered by one supply, then one supply will go into current limiting and the other will deliver the balance of the required current. There definitely can be issues with non-isolated supplies that are paralleled with current going through the "wrong" return path. And of course if the current sense resistor is between the output [-] terminal and the circuit's negative and you return the load current to the input supply negative, the current sense resistor knows nothing about the output current at all.

[@IFix-Things]

@@DiyNukeI’ve experienced the same issue

[@boblewis5558]

I think people are confused ... you can limit current but voltage is just set to a level, not "limited" per se. When the current limit is tripped then the voltage drops correspondingly but as a result of the current limit not any kind of voltage limit. To get the supplies to balance requires careful monitoring of all the outputs and multilateral lateral synchronius control otherwise it's the case of "most wins" when it comes to voltage. It's just like when there are two 12v batteries connected in parallel, one will stop the other providing any current until the load resistance drops low enough. Worse, if they are more than a few hundred mV different in open circuit volts, current will flow from the higher voltage battery into the lower voltage one as soon as they are connected in parallel. To prevent back feeding, yes, diodes COULD be used but would have to be rated for the max output current-plus. BUT ... that also means wasted power ... 0.7v x diode current per diode. So if just 5A per diode then 7 watts of pure wasted heat ... with just two supplies. A more robust and vastly more efficient reverse blocking mechansim can be achieved using series MOSFETs. However the only way to achieve full balancing is via more complex power monitoring and feedback circuitry coupling the simultaneous control of each supply. That is the ONLY way to balance the individual supply loads properly. Otherwise, as already stated ... "most wins!" 🤔🙄

[@d614gakadoug9]

Sorry, an earlier reply I'd posted was intended to respond in a different thread. Linear Tech has some "ideal diode" controllers that are very cool. They use external MOSFETs "backwards." I used them in a lithium battery charger I designed years ago to allow connecting batteries of different charge states (actually each battery was two completely separate batteries internally - these were made for the military market and sold for about $400 each; they were about the same physical size as a 6 A•h 12 V sealed lead-acid battery) Making constant voltage supplies share reasonably equally when they are connected in parallel, without compromising efficiency and regulation accuracy, is indeed a significant challenge and invariably requires some carefully designed circuitry, even when the necessary "hooks" are designed into the individual supply modules. Trying to retrtofit supplies without the necessary designed-in features is really daunting.

[@wilvandendoel4289]

Hi Julian, nice test, I wanted to parallel two of the 1800 watt models, but another guy on you tube blew them up, doing the same thing. I didnot understand it, but probably this was caused by the shunts in the negative, as you pointed out. I am only going to parallel the outputs to my battery bank, the inputs are two seperate toriods, with bridge rectifier and elco, so the inputs are isolated, that should work. As for the CV state, connected to the battery, regulation will be slower, load sharing is probably not perfect, but that does not matter.

[@ernieschatz3783]

As general rule, power supplies, dc to dc converters etc don't work well in parrellel unless designed to do so.

[@stickmenwithrayguns]

The "safe" way to put multiple supply's in parallel is to use one in CV mode and steer the others thru a current limiting network. ( 2 transistors, 1 resistor, 1 diode pr. "slave" supply ). The diode is placed on the output before connecting these together. You will loose a few volts in the current limiting network on the "slaves" but the voltage can be controlled and the current will be evenly shared. IMPORTANT: ( You will also need short-circuit protection like adding a foldback circuit to the "steering" supply. If you don't, during a "short" all the power will be frying the current limiting networks!!! The built-in closed-circuit protection on the SMPS's will not save you, since it "sees" the external current-limiter as a normal load. )

[@TheCommonDanger]

You should be able to use a few diodes on each boost converter output to combine the outputs and yet keeping the boost converters from seeing each other because one see the other it’s reacting to your adjustments on the other one so if you put two steering dies on each boost converter one for negative and one for positive on each boost converter, you bring a negative together and you bring the positives together this might resolve your problem. I used to do this when looking at two different outputs across the single trace scope to see two traces simple way to cheat and neither power supplies saw each other and that’s when I figure out I could bring two power supplies together without affecting one of the other, and have a total combined output

[@bascomnextion5639]

I have two of the units with the heat sinks but they have 3 x 20A fuses and a fan on the heat sink.

[@SergiyGryb]

H-mm, I believe to make them balance the load in Voltage limiting mode you need to re-arrange feedback lines for both: they should get feedback from a single resistor divider located at the output after current metering multimeters right at the light bulb connection... Please, try.

[@JulianIlett]

I can't see that working. In voltage limit, the converters are on a knife edge. Tiny differences in op-amp gain would result in one unit doing all the work, and the other doing nothing.

[@SuperBrainAK]

@@JulianIlett I think that idea has some potential. Having a single high value pot for adjusting the output voltage and 2 smaller ones for very precise tuning to balance between them. Probably have the pots in a variable resistor mode with a fixed divider resistor. There has to be some "linear region" in the feedback somewhere.

[@d614gakadoug9]

@@SuperBrainAK It's all linear, but error amplifier gain is usually very high - typically limited at DC only by the open-loop gain of the actual amplifier, which can easily be in the range of hundreds of thousands. The difference required to drive the error amp from one rail to the other might be a millivolt or less at the sense point. Trying to trim the feedback paths for balance is futile.

[@CornishMiner]

Interesting. A nice bit of playing Julian.

[@sc0or]

Thank you for the interesting info. However this is an expected behavior. Until a voltage drops below a regulator limits, the second part won't start working. But if one unit provides enough power until this happens, what's bad that it works without any help from its "brother"? It will stay in working margins. And when it will try to exceed them, the second unit will help. Sure, it will be needed to set them up as close as possible initially. This is always better to have a "real" architecture with paralleled switches and chocks. However in case you load is increased unexpectedly, to purchase a same unit to solve the problem won't be a bad idea imho if this will be the cheapest or quickest solution.

[@prashanthb6521]

You should watch this guy's video "Виктор Сочи". He has paralleled two XL4016 buck converters by modifying the feedback circuit with only one trimpot controlling both and it works. Video is titled : "Параллельное включение преобразователей, тест и как поднять выходное напряжение DC-DC Step Down"

[@universeisundernoobligatio3283]

With these cheep boost converter looks like they are not meant for continuous power at their full rating, they will run very hot with their undersized traces and coil windings. Maybe run them at 75% capacity and add a fan.

[@MasterIvo]

I wonder if series input will also work. I have two 230V ac to 45V dc which have output sensing for series balanced output of 90V. Meanwell said it can do the series output but not the series input. I still wonder why not. they didn't explain. Could it have to do with the chassis ground?

[@d614gakadoug9]

The voltage across the input of each supply depends on the current delivered by that supply. If one supply momentarily delivers no load current the full supply voltage will appear across the input to the other supply.

[@MasterIvo]

@@d614gakadoug9 thank you. yes. but since the outputs are balanced by the sensor, that wont happen. extra output capacitors should also help prevent that.

[@d614gakadoug9]

@@MasterIvo Increasing the output capacitance is unlikely to be of any benefit. You are still up against things like soft-start circuits that could cause a differential in startup time of tens of milliseconds. Even if the startups were synchronous, variations in the input filter capacitors could lead to substantial imbalance. There are considerations like capacitance between the output circuit and the input circuit. It is very common to use a discrete capacitance between the output negative (usually) and either the input DC positive or DC negative. This is done to provide a local return path for RF as part of EMI/RFI management. Things get pretty vague if you have two supplies in series. I take it you want to try to operate these supplies on 480 VAC. If I were doing this I'd use a step-down autotransformer. This is typically something you'd buy as an ordinary transformer with split primary and split secondary, each individual winding rated for 240 V. They're pretty common as "control transformers." Be aware of the fact that if the switchers don't have active power factor correction, the ratio of RMS to average current at their input will be quite high, so don't skimp on the current rating of such a transformer. There was a time I might have tried calling Meanwell on your behalf. It owed me for identifying a engineering/production flaw in one of their supplies (in a bit of circuitry that should have improved reliability of the supply but because of the execution actually significantly reduced reliability). But that was quite a few years back.

[@brauchmernet]

Many commercial power supplies allow parallel units with a Schottky in the positive per supply.

[@JulianIlett]

The handy thing about boost converters is that they already have that Schottky diode in their positive output.

[@SergiyGryb]

Schottky diodes at the outputs would be required for buck converters. Topology of boost converter implies that diodes are already there...

[@brauchmernet]

Yes, and no. There are designs without a (external) diode in boost converters. Some of the LTC Series are essential Inductor only. Then it becomes a question where your voltage feedback is done @@JulianIlett

[@akshathans5459]

The feedback loop senses the voltage after the internal diode. Isolating the feedback loop by adding an external diode on both supply's positive rail will help isolate the feedback loop. Commercial power supplies also uses current sharing controllers which further matches the output voltage for accurate current sharing, even during transients

[@JulianIlett]

Ah yes, my new favourite chip - the LT8705 :)

[@andyguest6543]

Please get some cheerios to show us all as you say your ending to the videos, it would makevmy day 😂😂😂😂

[@borayurt66]

Just put a diode to their outputs so that one can't "see" the other one's voltage, it will work in constant voltage too. You need beefy diodes though.

[@mxmv.g]

Wow, very useful information! Thank you! Have you figured out if the other type with the current sensing on the output can also be paralleled?

[@nicod974]

Do you have to add a diode bridge to parallel 2 of these red PCB boost converters?

[@RND-USA]

Hi from Los Angeles, great video! Have you tested parallel buck converters yet?

[@Patriarchtech]

Julian I love your videos, parallel well of course you can. eg a solar panel is a boost converter....There is no hokus pokus, diodes are key, regardless of what people may tell you. How do you think a Solar charge controller or several of them can be connected to the same battery, and charge the battery at different voltages and amps? Diodes.

[@RixtronixLAB]

Nice video shot, thanks for sharing with us, well done :)

[@johnward7619]

Interesting. Would placing a 0.1 R or similar in each output not help with current sharing?

[@asusdreamland5670]

I have a question: Can we link the 2 output - (negativ) or we must take for output - the input - like in your video ?

[@jimw7916]

Have you tried 2 large diodes (one on each output)

[@JulianIlett]

Yes, they're already part of the boost converters.

[@wouter481]

Current limit solves the problem. But how about a electronic perfect diode, i think that will give you voltagebalance as well. Anyway thans jullian

[@d614gakadoug9]

No, that will not work. The fundamental nature of a constant voltage power supply trying to behave as a zero impedance source, prevents it. Constant current supplies try to behave as infinite impedance sources so they do inherently share.

[@hubert3048]

Does the same applies for buck-converters? Or it is required to add diodes at the outputs?

[@JulianIlett]

I've seen people putting various power supplies in parallel. Probably only works if the power supplies have current limiting, but I've yet to try it.

[@d614gakadoug9]

The problem is the same for _any_ constant voltage regulator, regardless of type. It is a function of the fundamental nature of the supplies. Diodes don't fix anything with regard to sharing.

[@rezapournamdar7275]

What about in series form??

[@blackrabbit9601]

How about output using diode plz

[@cccmmm1234]

Well, yes of course you need current limiting to get sharing!

[@williamwright4591]

Put a diode after each volt meter so they don't see the voltage of the other converter.

[@d614gakadoug9]

That does not work.

[@observingrogue7652]

Thank you .

[@G1ZQCArtwork]

When you parallel those boosters with the heatsinks, it blows the input fuses on one of them. No apparent damage though. I have looked at using power diodes to provide isolation, but this is at the cost of losses in the Diodes as they will get hot. Voltage flip is interesting, this occurs in batteries wired in parallel, because of the slightly weaker partner will take voltage from the stronger battery as a load till it gets drained. So, if you fully charge a LapTop battery, then put it in storage, it will be flat or mostly flat when you come to use it next. The cells are paralleled in all to get the higher run times.

[@JulianIlett]

I think there would be a problem paralleling boost converters that measure current in the output negative. Putting those converters in parallel means all the current sense resistors end up in parallel. I think that would upset their current feedback loops.

[@nicod974]

​@@JulianIlettso how can you achieve 48V40A to 72v25A?

[@jacquesb5248]

interesting results

[@panospapadimitriou3498]

piece of cake...!!! but but pffffff every single one i bought was .. eeeeuuuuu single one. they all different .. i guess with some search i need just to make a pair work about 50% to maintain reliability!!! if reliability is something the maker gives on part!!!! noce job and close one .. that last week i m having fun with some dc dc adjustables and wanting to go bigger for a future off grid fridge

[@codebeat4192]

Because you don't use a scope to see what is really going on at the output, you cannot say for sure it is working properly. I mean, these are not the smoothest DC sources, it is PWM. I wonder what it will do with the noise/ripple at the output. Is there some pulse skipping, peaks or other artifacts going on? Do they really balance eachother or do they also work against eachother (for example by not being perfectly in sync)? Do they get hotter than normal? There isn't a feedback so both converters do not know what the other is doing and on what time frame. For example, when out of sync you cannot get the maximum current both converters can provide together because they don't help eachother. In fact this is the same result as using only one converter. When they are perfectly in sync, only then you get the benefits by doing so. It is important to use a scope (or two) at such experiments.

[@alexanderjesmer]

Good one)

[@lezbriddon]

more fun with bucks.......

[@ericwazhung]

:( I'm all for the scientific method and experimenting, but when talking about 10's or 100's of watts, I think folk need a better understanding of how these things work before just trying things willy nilly. Some of the questions in here could be answered with simple worst-case thought-experiments... e.g. imagine your two converters' outputs are connected in parallel... where does the return current flow? Well, surely, back to the converter that supplied it, right? But what if one's wire is 0.01ohms higher than the other? V=IR I=V/R->infinity, so all the current Might go through one converter's return wire, then half would out the back and into the second... this seems like a bad idea. Things change with temperature, humidity, etc... How many comments here rely on *perfect* matching of voltage, etc. via potentiometers whose values will vary day to day, or *perfect* matching of silicon in two chips with different thermal experiences...? Even if an experiment works perfectly one day, at this amount of power, it might be disastrous another...

[@hi-tech-guy-1823]

DC OHMS Being biased Due you cant get 2 or more Exactly Identical components Perfectly Identical in every Way Impurity's and imperfections Change the ohms Value of Every component is annoyingly Unique And more Prone to being DC biased Even the Wire Ohms and Length it bloody fussy and Start's Getting A single Favourite (Due its Shorter or less Resistance ) it why DC Power networks (DC National Power grids ) Dont have 2 or more DC Generators Power stations due DC is so Much of a bitch To Deal with it going biased with just Resistance Distance and ohms slight change and it has a Picky biased personality You see it a lot in Battery cell arrays some Cells Want Chrage super fast and Some take Days ~ weeks ~months to come up even discharged some cells jus want to give up power to fast and others holding on to it