Diodes in parallel

[archil]

What happens when two diodes are in parallel? if this is a bad idea, why exists shotkky diodes with common cathode? (i. e. to220 package).

[CaptDon]

Diodes in parallel is bad, one will do all the work the other will do nothing. Why two in one package with common cathode or common anode, to be used as a full wave rectifier of course, and never put in parallel.

[madires]

The one with a lower Vf wins, becomes hotter and Vf will drop even more (thermal runaway). But you can add a small resistance to each diode to mitigate that. Dual diodes with common cathode don't necessarily have to be in parallel (-> power oring) but are quite similar in most cases. So you might get away with paralleling them (and the wire/trace resistance in series with each diode). If you parallel diodes underrate the maximum current.

[JohnnyMalaria]

Diodes in parallel is bad, one will do all the work the other will do nothing.

Huh?

https://www.st.com/content/ccc/resource/technical/document/application_note/a4/ef/bc/7d/78/89/49/f1/DM00098381.pdf/files/DM00098381.pdf/jcr:content/translations/en.DM00098381.pdf

[Dave]

Dual common-cathode diodes in power packages like TO-220 are commonly used in forward and isolated push-pull converters.

Forward converter


Isolated push-pull

[T3sl4co1l]

It's perfectly fine to do, as long as you know what you're doing.

If the diodes match perfectly, at all times, then it is safe to use the combination at twice the rating of a single part, or however many.

If they don't match at all, like one diode is an open circuit and not even there at all, well, obviously the limiting rating is just one diode.

So for two in parallel, you have somewhere between 1 and 2 times the current rating of a single diode.

Where in that range, is the real question.

For diodes that are closely thermally coupled, it can be pretty high, even without ballasting resistors.

Note that pairs are typically two dies, so although their temperatures track closer than separately packaged diodes, it's not as good as a monolithic pair (two diodes on the same die).  Which also exist, though the datasheet may not say so, so it can be hard to tell, short of breaking the thing in half to see for yourself.

Adding ballast resistors, improves current sharing so that the limiting current can be much closer to 2.0 times.  This comes at the expense of voltage drop.

Current sharing also depends on diode type.  Some are better than others.  Typically, PN junction diodes have a strong negative tempco and low incremental resistance, so they parallel poorly.  Small schottky diodes (like BAT54) may parallel okay, on account of operating at relatively high current density (internal resistance is more significant, or even dominant).  SiC schottky are almost always in the PTC range, so can be quite safe to parallel.

The applied current matters.  For all semiconductor diodes (as far as I can think?), the ideal diode junction part of its response, always has a negative tempco; this is always in series with a resistive element (bulk semiconductor), which has a fixed resistance, and positive tempco.  At some current, the PTC will always come to dominate; but this point may be well beyond the continuous operating current of the diode, so it's not always useful.  Or, perhaps that's exactly the point, you're concerned with surges -- in which case it may not matter very much.  (For surge purposes, paralleling of diodes and TVSs is generally acceptable, if maybe not preferable.)

The exact current sharing factor can be calculated from forward voltage curves, thermal resistance, applied current, and ambient temperature; unfortunately I don't have a formula handy for that (if a closed-form solution exists), but fortunately that would exceed the scope of a beginner topic anyway so I'm not on the hook for that, right?...

Tim

[CaptDon]

Two diodes in pure uncompensated parallel is bad as shown in most of the data presented in those ST documents. The lower VF will conduct most of the current and then heats and carries more. Yes, yes, add compensating resistors, use twin diodes in one pack in parallel.....blah...blah...blah. It can be done but in most cases shouldn't. With diodes available into the thousands of amps where is the need to parallel a pair or more? ST says in their literature it is common in power electronics??? I have never serviced any unit built that way. I can assure you in the 4400hp locomotives where I design the power electronics I have never used diodes in parallel and specifically not on the alternator 3 phase output where we use diodes pucks about 4" in diameter. We do tons of low voltage LLC resonant converter crap and again I have never done and never seen diodes in parallel. Can it be done, sure...Should it be done, I can't think of one good reason. I do have many high voltage circuits where we use diodes in series, mostly 1N4007 type stuff and for many years used the parallel compensating resistors and small .001uf caps although these days the single piece slightly longer body high voltage diodes are more common and are a single package answer to what was a multi-package problem years ago.

[CaptDon]

I do agree with the poster that stated Schottky diodes with their usual PTC in parallel is do-able. Indeed that can work. Perhaps space saving with two diodes in on pack?? I doubt I would ever find the need two put a pair of SMT or thru-holes side by side but would opt for one that could do what I need it to do. The only reason I may consider a pair of schottky diodes in parallel would be to space them out a bit to spread the heat signature but I have not needed to do that yet.

[CaptDon]

re-read my own post, my english sucks.... on was supposed to be 'one' and two was supposed to be 'to'. Cheers mates!!

[archil]

my english sucks too,  but if you understand me, it is not to bad

[drussell]

re-read my own post, my english sucks.... on was supposed to be 'one' and two was supposed to be 'to'. Cheers mates!!

You can edit your previous posts with the "Modify" button, by the way...

As to the OP's question, as has been stated above, just like a standard bridge rectifier has four diodes in it for appropriate applications, dual diodes in both common-cathode and common-anode exist for when they are the appropriate choice for an application, which is generally not paralleling them.

If you need additional current handling capability it is generally a far better idea to just use a larger diode rather than trying to parallel them.

[Jwillis]

Some information on parallel diodes here.    Parallel Diodes
It can be done if you add a small resistance in series with each diode to balance the current .

[T3sl4co1l]

It can be done but in most cases shouldn't.

A better way to put it is: it can be done, but in most cases it isn't economical, because the increase in current capacity is less than proportional to the number of devices, and therefore the cost and capacitance go up much faster.

With diodes available into the thousands of amps where is the need to parallel a pair or more? ST says in their literature it is common in power electronics??? I have never serviced any unit built that way. I can assure you in the 4400hp locomotives where I design the power electronics I have never used diodes in parallel and specifically not on the alternator 3 phase output where we use diodes pucks about 4" in diameter.

How do you think they make them?   The pucks I think, are usually whole die/wafer, yeah.  Though I wouldn't be surprised if they use matched dies when convenient?  I haven't seen many from the inside unfortunately.

Modules however, I have seen plenty of, namely IGBT modules with co-pack diodes -- they tend to use smaller dies, lots of them in parallel.  Indeed, the same logic in this thread also applies to IGBTs.

Parallel IGBTs can be especially helpful when your options for discrete parts range from the cheap TO-247, up to somewhat pricier TO-268s, to much more expensive SOT-227s and up.  The 247s are so much cheaper, it's worth using them in parallel.  Even if it means needing six to obtain the capacity of four (i.e. a 1.5 sharing factor), you still save over a pair of SOT-227s or a module.

We do tons of low voltage LLC resonant converter crap and again I have never done and never seen diodes in parallel. Can it be done, sure...Should it be done, I can't think of one good reason. I do have many high voltage circuits where we use diodes in series, mostly 1N4007 type stuff and for many years used the parallel compensating resistors and small .001uf caps although these days the single piece slightly longer body high voltage diodes are more common and are a single package answer to what was a multi-package problem years ago.

Speaking of -- yet another option is to avoid it all together.  This is especially helpful for high frequency, low voltage converters: below some impedance (i.e. the ratio of peak switch voltage to peak switch current), you simply can't deal with stray inductance anymore, and the solution is to break it into multiple channels, sourcing current, wired in parallel.  Which can also be phased, to partially cancel out ripple currents.

The downside is, rather than beefing up just the inverter, you need to double most of the control circuit as well.  This can still be a worthwhile effort, for example going from 1 to 2-3 channels, the reduction in input and output bypass caps typically rewards both in terms of BOM cost and layout area.

Oh--incidentally, diodes in series, is very much a similar problem, just the reciprocal of it.  Reverse recovery has a runaway characteristic, so it is quite important to use matched diodes, and voltage-sharing methods (the capacitors mentioned) to keep them behaving.  And indeed, the HV parts are stacks of matched dies, which also track very well thermally on account of their being stacked.  So, when they're available in the ratings you need, they're great value.

Tim

[harerod]

In real life, with, let's say a Schottky BAT54A/BAT54C, something way more mundane might happen:
A certain design required a double diode BAT54A at one point and a single diode BAT54 at another.
To optimize the Bill Of Materials (BOM), a designer might choose to use the same component twice, thus reducing the number of different components to be purchased and handled. To avoid a dangling antenna, the designer would then connect the spare diode in parallel to the single diode.

[Tomorokoshi]

Somewhat outside of the scope of the immediate discussion here are cases where for redundancy reasons 2 or 3 Zener diodes are in parallel, with definitely no balancing resistor allowed.

[aheid]

I recently repaired a circuit where two plain diodes where in parallel for safety reasons in case one failed open.

They were acting as flyback/freewheeling diode in a 230V circuit, and due to the nature of the circuit, failing open would mean the possibility of a user being exposed to the inductive voltage spikes.

[drussell]

To optimize the Bill Of Materials (BOM), a designer might choose to use the same component twice, thus reducing the number of different components to be purchased and handled. To avoid a dangling antenna, the designer would then connect the spare diode in parallel to the single diode.

Somewhat outside of the scope of the immediate discussion here are cases where for redundancy reasons 2 or 3 Zener diodes are in parallel, with definitely no balancing resistor allowed.

I recently repaired a circuit where two plain diodes where in parallel for safety reasons in case one failed open.

They were acting as flyback/freewheeling diode in a 230V circuit, and due to the nature of the circuit, failing open would mean the possibility of a user being exposed to the inductive voltage spikes.

Obviously, paralleling diodes when you don't NEED to is fine, and as pointed out above, may well be desirable for reliability/redundancy, to optimize the BOM, or whatnot. 

In these cases, as long as one diode can handle the load, adding extras won't hurt anything.  It is when you need both of them to be able to handle the total current that you're likely to end up with problems unless you are very careful.  (Plus, it's usually not necessary anyway, as in most cases you can just choose a larger diode.)

[Siwastaja]

Diodes in double packages are matched to some degree, and thermally coupled, again to some degree.

BOM optimization is funny, so is the market. A double diode part rated at the same current as a single diode could be cheaper. If the double diode is already in the design, even more obviously it would be used. If extra capacitance doesn't matter, it does no harm to parallel the second diode instead of use just one of them.

Given same diode, same batch, it's very very very unlikely there is no current sharing at all. Paralleling two diodes instead of using just one with the same rating will:
* Reduce the junction temperature
* Reduce Vf and hence, reduce power dissipation
* Spread the heat over larger area

Let's say there somehow is massive Vf variation between the two diodes. In this case, the one with lower Vf takes all the current, yes. But you are dissipating less power, and it's running cooler, than if you only had the higher Vf diode.

Just don't expect they share current very well.

For two diodes from the same batch (or a double package), I tend to assume I can go for some 20-30% higher current. Which is approximately how much I would derate anyway. It has happened that I had a diode in BOM already, run the math, and end up single diode being "just maybe acceptable", but a bit marginal. I.e., maybe it was a 4A diode, and I'm going to run 4A through it. Or I did the thermal math and get Tj=125 degC at worst-case ambient temp. Would like 115... So it works but it's marginal. So I put two of these diodes in parallel. Current sharing can be poor. If the utter miracle happens and current sharing is 0%-100%, it's still marginal. If one of them happens to have such exceptionally low Vf that it takes all the current, no harm done, it's also dissipating less than expected thanks to that low Vf, so it's better than marginal.

[madires]

I've seen diodes in parallel several times in SMPSUs (secondary side). Guess what a common fault of those SMPSUs is. Yep, a shorted diode.

[perieanuo]

What happens when two diodes are in parallel? if this is a bad idea, why exists shotkky diodes with common cathode? (i. e. to220 package).

diodes with common A or K are for rectifiers with median connection (see smps for computers, you have 2 identic windings in series in smtp transformer, median is GND and each winding is rectified by one diode, the common anode is "+" output)
minimise pcb tracks and pcb size

[Siwastaja]

Remember that LEDs are paralleled all the time in power LED modules without ballasting resistors despite the fact they are diodes as well, also have iffy Vf matching, and negative Vf tempco. They rely on the resistive component of the semiconductor itself + bond wires, and fairly good thermal matching through shared copper tab.

It's not the best thing to do, but doable.

[Circlotron]

I’ve seen setups where the current goes into the halfway point of a centre tapped inductor and each end goes to a diode. Okay for rectifying AC, no good for continuous DC. The inductor acts as an autotransformer and to whatever degree necessary provides an unbalanced voltage to the two diodes to force them to share current 50/50.

[TheUnnamedNewbie]

As has been said, it all depends on how well they are matched. Heck, almost every diode on a IC will be made up out of loads of tiny diodes put in parallel. Same applies to FETs, in the end I can only make a tiny FET on chip, and to get bigger ones I have to put them in parallel, with multiple fingers.

In the end, you can think of a single diode die as an infinite grid of infinitely tiny diodes in parallel. Usually they are so well thermally matched that the resistance of the interconnection and substrate itself gives enough negative feedback to compensate for the positive feedback of the Vf-vs-temperature curve. When they are not, you get hotspots, which is something that can happen in poorly designed devices, or when you throw on bad process control.

However, as others have said, it's usually not a great idea, as it can be more economical to simply get a bigger rated device (where the manufacturer might have done the device matching for you, in package).

[Gyro]

As has been said, it all depends on how well they are matched. Heck, almost every diode on a IC will be made up out of loads of tiny diodes put in parallel. Same applies to FETs, in the end I can only make a tiny FET on chip, and to get bigger ones I have to put them in parallel, with multiple fingers.

Although modern power FETs do suffer poor linear SOA due to hotspots as a result. Just saying.

[TheUnnamedNewbie]

As has been said, it all depends on how well they are matched. Heck, almost every diode on a IC will be made up out of loads of tiny diodes put in parallel. Same applies to FETs, in the end I can only make a tiny FET on chip, and to get bigger ones I have to put them in parallel, with multiple fingers.

Although modern power FETs do suffer poor linear SOA due to hotspots as a result. Just saying.

I'm no power-semiconductor guy, so I can't be sure, but isn't that more a matter of optimization? Most of those power FETs are targeted at switching applications, so linear-region SOA is not a primary concern in their design.