Diodes in parallel.

[Rick Law]

I've something that works in experimental setting.  I was thinking of applying it to something that I want to keep more permanently (and potentially more expensive), but I know that I am not knowledgeable enough to identify the draw backs.  If anyone can give me their thoughts on this, I appreciate it.

Seeing how schottky diode is used as "power source selector", I applied it to some of the experimental stuff I do...

I had something that I wanted to avoid outage when I need to pull the battery out to recharge, so as with the Arduino's, I decided to use a diode as power-source-selector.  I put two sets of batteries+diode in parallel to share/switch between batteries.  That worked.

I expanded it - to minimize the lost on diode, for each "battery+diode" set, I decided to use multiple (4) schottky diodes in parallel - I reasoned that multiple diodes in parallel would decrease the current through each diode hence keeping them closer to the (approx) 0.2v drop than the higher current 0.4v drop.  That also worked.

I've been thinking of applying that concept (power source switching/sharing and using multiple diodes to lower the voltage drop) to something more permanent (and perhaps more expensive).  That leads me to think: "What are the draw backs?",  "What if one of the diodes failed?", things like that.   

Your thoughts?   Thanks!

By the way: While I've been chewing on that thought for a while now and still hadn't done it, the post on ganging up PSU inspired me to ask since the same concept could apply with two PSU's - but I did not want to hijack that thread.

[SeanB]

Remember that diodes have leakage current, so multiple in parallel means more leakage current as well, which can, depending on the battery chemistry, be from mildly bad to FOOF level bad. If just pulling the battery out for a few seconds a easier thing is an ultracapacitor or other multi farad capacitor as a reservoir of power to hold up the unit, either in a string in series for higher voltage or just a 5V5 unit for a 5V rail.

[wraper]

"What are the draw backs?"

Connecting diodes in parallel is a bad idea (if it's dual diode within one package, usually acceptable). Current won't be shared equally. You should just use diode with higher current spec.

"What if one of the diodes failed?"

They usually fail short, so the whole thing will be shorted.

[Benta]

Connecting P-N diodes in parallel is a bad idea, but with Schottkys it will work. But the obvious question here is: why not just use a larger/higher current diode? The result will be exactly the same.

[wraper]

but with Schottkys it will work.

Proven to be very bad idea by failing TV PSUs. Schottky diodes as well as usual diodes have negative forward voltage temperature coefficient. Which means the hotter diode -> the lower forward voltage -> the higher current flows -> diode heats up even more, Vf drops even more, current increases. Thus imbalance only worsens.

[BrianHG]

Use something like this with a huge mosfet: https://www.diodes.com/products/power-management/ac-dc-converters/active-oring-controller/part/ZXGD3108N8

[BrianHG]

Also look at these diodes:
https://www.eevblog.com/forum/chat/take-a-look-at-these-new-field-effect-rectifiers-from-st-microelectronics/msg1428355/#msg1428355

https://www.st.com/en/diodes-and-rectifiers/field-effect-rectifiers.html?querycriteria=productId=SC1884

[Rick Law]

Connecting P-N diodes in parallel is a bad idea, but with Schottkys it will work. But the obvious question here is: why not just use a larger/higher current diode? The result will be exactly the same.

My experiment (with my limited stock of parts) requires very low current.  My ganging it was for lessening the voltage drop.  My 3A diodes proved to drop voltage more as compared to 3x1.5A diode at equal current testing to around 1A.  I lost my notebook, I have to retest that to see what the delta is again...

but with Schottkys it will work.

Proven to be very bad idea by failing TV PSUs. Schottky diodes as well as usual diodes have negative forward voltage temperature coefficient. Which means the hotter diode -> the lower forward voltage -> the higher current flows -> diode heats up even more, Vf drops even more, current increases. Thus imbalance only worsens.

I didn't know that would be a problem.

My experiment's typical current draw (about 1A) is well below the 1.5A rating for each diode.  The parallel 3x diode gang is to lower the current to each so as to lower voltage drop to my load and not aimed to increase max current to > 1.5A rated for each.  So this overheating should at worst case be same as using a single diode.

Thus, from my understanding of your reply, it would appear that I am okay as long as if my total current is not exceeding the rating for each individual diode.

Good that I asked. This issue rules it out for my other plan which would ganging for larger current.

Thanks for the info.

[wraper]

So from my understanding of your reply, it would appear that I am okay as long as if my total current is not exceeding the rating for each individual diode.

But then there is no point doing this. Imbalance would depend on how different diodes are, so it's a pure luck in picking diodes and each device built will be different. Also important how good is thermal coupling between diodes and if some diodes are not heated by something else. It's okayish when they are in single package as they are thermally coupled and likely closely matched. But when you start placing separate diodes on the PCB, it get's really bad. Even trace length and width to each diode starts matter a lot. It could be 20mV voltage drop on the track which causes 3 times difference in current through parallel diodes.

[wraper]

I even used thermal imager to look on several diodes connected in parallel. Not only there was huge random imbalance in heating, diodes in the center were vastly hotter. They were heated by nearby diodes which caused even more current flowing through them and heating even more while diodes on the sides were colder and current vastly reduced.

[Gregg]

If the batteries are the same chemistry why not just parallel them but fuse each battery with a fuse slightly larger than the load so that if one battery happens to fail short circuit, the fuse will just blow.  I do this with my shop alarm batteries after trying diode isolation and finding out I was just wasting energy and complicating things.  I put a small LED with a resistor across each fuse to give a quick visual if a fuse has blown.  To change batteries or isolate one for testing, just pull the fuse.

[Rick Law]

So from my understanding of your reply, it would appear that I am okay as long as if my total current is not exceeding the rating for each individual diode.

But then there is no point doing this. Imbalance would depend on how different diodes are, so it's a pure luck in picking diodes and each device built will be different. Also important how good is thermal coupling between diodes and if some diodes are not headed by something else. It's okayish when they are in single package as they are thermally coupled and likely closely matched. But when you start placing separate diodes on the PCB, it get's really bad. Even trace length and width to each diode starts matter a lot. It could be 20mV voltage drop on the track which causes 3 times different the current through parallel diodes.

Got it - I understand that one.

My experiment happened to use the diodes from the same roll.  So that part is okay.  I should put some thought into how to get a shared heat sink or other ways to couple them thermo-wise.

Well, may be the experiment doesn't need them (< 1amp), but for sure if/when I do one for higher current.

If the batteries are the same chemistry why not just parallel them. ... ...

Experimental only.  So I am using batteries that I redeploy on something else regularly.

[wraper]

My experiment happened to use the diodes from the same roll.  So that part is okay.

Even when they are from the same batch, usually difference is huge.

[Zero999]

So from my understanding of your reply, it would appear that I am okay as long as if my total current is not exceeding the rating for each individual diode.

But then there is no point doing this.

Lower volt drop.

I can see what you're saying, but in practise it does work. The current sharing isn't perfect, but it's enough to give a lower voltage drop, than what one would expect using a single diode.

[wraper]

So from my understanding of your reply, it would appear that I am okay as long as if my total current is not exceeding the rating for each individual diode.

But then there is no point doing this.

Lower volt drop.

I can see what you're saying, but in practise it does work. The current sharing isn't perfect, but it's enough to give a lower voltage drop, than what one would expect using a single diode.

It's pretty crappy way to decrease the voltage drop to begin with. As BrianHG suggested, if you want low voltage drop, use MOSFETS.

[Zero999]

So from my understanding of your reply, it would appear that I am okay as long as if my total current is not exceeding the rating for each individual diode.

But then there is no point doing this.

Lower volt drop.

I can see what you're saying, but in practise it does work. The current sharing isn't perfect, but it's enough to give a lower voltage drop, than what one would expect using a single diode.

It's pretty crappy way to decrease the voltage drop to begin with. As BrianHG suggested, if you want low voltage drop, use MOSFETS.

It's much simpler than MOSFETs and I can assure you, it works quite well.

[wraper]

It's much simpler than MOSFETs and I can assure you, it works quite well.

Not any better than choosing beefier diode and not having any downsides.

[Zero999]

It's much simpler than MOSFETs and I can assure you, it works quite well.

Not any better than choosing beefier diode and not having any downsides.

Beefier diodes often work out more expensive, since it's often cheaper to buy a larger quantity of smaller diodes, than a single big one.

[sourcecharge]

zeners in parallel work.  multiple 1n4740 in parallel increases the total current capacity.

[ArthurDent]

As others have pointed out, using bigger/more diodes can create problems with reverse leakage/unequal current sharing.  Here is a good simple video that covers what you are trying to accomplish. A single cheap mosfet can work in most applications. Because mosfets are basically low value resistors when turned on they would be much easier to parallel without any added equalization circuitry if you needed really large currents.

[Zero999]

As others have pointed out, using bigger/more diodes can create problems with reverse leakage/unequal current sharing.  Here is a good simple video that covers what you are trying to accomplish. A single cheap mosfet can work in most applications. Because mosfets are basically low value resistors when turned on they would be much easier to parallel without any added equalization circuitry if you needed really large currents.

I can't watch the video at the moment.

MOSFETs are good for switching between several power supplies or reverse polarity protection, but they don't behave like diodes, without an external controller IC. If two slightly different voltage power supplies are connected together, via P-channel MOSFETs, then they'll just get shorted together, when the MOSFETs turn on.

In the prototype I worked on, lots of 1N5822 diodes were connected in parallel, which worked out much cheaper, than a more beefy diode and the package was more convenient for a quick lash-up, than an SMT package.

[wraper]

In the prototype I worked on, lots of 1N5822 diodes were connected in parallel, which worked out much cheaper, than a more beefy diode and the package was more convenient for a quick lash-up, than an SMT package.

Did you dare to use thermal imager?

[Zero999]

In the prototype I worked on, lots of 1N5822 diodes were connected in parallel, which worked out much cheaper, than a more beefy diode and the package was more convenient for a quick lash-up, than an SMT package.

Did you dare to use thermal imager?

No, there was no need to because none of the devices came anywhere near to overheating. Using multiple, smaller devices also spread the heat, over a larger area, which removed the need for a heat-sink. The voltage drop was significantly lower, with all the devices in parallel, than the equivalent big rectifier and was sufficient for the application, which only needed 10A continuously, 15A peak and 8×3A diodes gave ample headroom.

It simply isn't true that one diode hogs all of the current. It's true that current sharing is not perfect, but it's not all or nothing, like you're implying. This isn't the same as connecting multiple discharge tubes in parallel, which would indeed result in one device firing, leaving the others unused.

[chemelec]

When using diodes in Parallel it is recommended to put a Separate Series Resistor to each diode.
Resistor Values depend on actual current, but somewhat Low Value resistors.
These Resistors will help to even out differences between the Diodes.

[Zero999]

When using diodes in Parallel it is recommended to put a Separate Series Resistor to each diode.
Resistor Values depend on actual current, but somewhat Low Value resistors.
These Resistors will help to even out differences between the Diodes.

That's true, resistors will help to improve the current sharing, other than increased component count, the downsides are increased voltage loss and power dissipation, hence why I didn't even consider using them.