Using diodes to reduce voltage by 1-2 V under a very high load (e.g., 20A)

[wraper]

I would say over all, you can parallel diodes for this purpose, but maybe use 10 ? or 20 % extra diodes to further prevent the possibility of what we haven't yet seen, that is, a diode blowing up. fair conclusion ?

Even with 50 diodes in parallel, you only need one diode with significantly lower forward voltage drop to cause the failure as it will take too much current on itself and eventually short out. I'd say that the more diodes, the worse, you'll more likely will get one that is significantly different from the crowd.

[CatalinaWOW]

I would say over all, you can parallel diodes for this purpose, but maybe use 10 ? or 20 % extra diodes to further prevent the possibility of what we haven't yet seen, that is, a diode blowing up. fair conclusion ?

Even with 50 diodes in parallel, you only need one diode with significantly lower forward voltage drop to cause the failure as it will take too much current on itself and eventually short out. I'd say that the more diodes, the worse, you'll more likely will get one that is significantly different from the crowd.

Remember, it could also fail open.  Maybe after initially shorting out.  And in this application a short may or may not be detected as a fault, and an open almost assuredly will not be detected.  So it "works".  Might even be entertaining.  If all the circumstances were "right" (or wrong depending on your worldview), you could see a diode fail every few minutes, with each failure dumping two transients into the system, one when it shorted, the other when it opened.  The interval between failures would tend to get shorter as each diode picks up more of the load.

This entertaining outcome would be rare and hard to duplicate.  As Chicken says, in a great many cases nothing exciting will happen at all.

[TheElectricChicken]

Even with 50 diodes in parallel, you only need one diode with significantly lower forward voltage drop to cause the failure as it will take too much current on itself and eventually short out. I'd say that the more diodes, the worse, you'll more likely will get one that is significantly different from the crowd.

If it was so easy to have such a fault why is it NOT showing up, ever? you tried to reproduce it, but it WORKED instead. Now you are saying that in order to have a problem, you FIRST need to have a faulty part, THEN that faulty part will cause further problems. I'm sure if you FIRST smashed hell out of the diodes with a hammer THEN they'd have a hard time working properly too.  What's the difference ? If you don't go in search of a unicorn diode or use a hammer, then diodes work fine.

[Chris C]

Now you are saying that in order to have a problem, you FIRST need to have a faulty part, THEN that faulty part will cause further problems.

I wouldn't consider a diode with an unusually low Vf a faulty part.  Under normal circumstances, low Vf would be a highly desirable characteristic.

I recall someone who needed a 1% tolerance resistor.  He didn't have any of those on hand, but he had lots of 5% resistors.  And figured that since tolerance is a range, at least some of them would be 1% or less.  All he had to do was test until he found one.

He was a bit surprised to find them all near -5% or +5%, with NOTHING in the middle.  Apparently the manufacturer was using the same idea.  They were all tested, and those with better tolerance were labeled and sold as such, at a higher price.  The automated sorting was presumably cheaper than running a separate manufacturing process to specifically produce precision resistors.

What if diodes undergo a similar sorting process?  With all the "unicorn" (low Vf) diodes removed and sold as a different part, then that would be awesome for running them in parallel.  And you might do so successfully, and repeatedly.  But suppose the manufacturer ends up with more low Vf diodes they can sell, and decides to temporarily skip the now unnecessary sorting; selling all the diodes regardless of spec as plain 400X series.  You try paralleling these, and what worked many times suddenly doesn't.

And it would be no one's fault but your own.  You broke the rules, using a part in a way that is technically questionable.  If it's for a hobby project, chances are it's no big deal; I've done my share of rule bending there.  But if it's for a product, and many products start blowing out in the field, you're in big trouble.

Then there's recommending something that's technically questionable to others.  As a general rule I try to avoid this.  On occasion I have not.  There was one case where I suggested something, for which there was some risk, but on average the reward outweighed it.  I explained all that based on my own personal experience, expecting it to scare most away, but surprisingly people tried it en masse.  Turns out that while the reward still outweighed the risk, the risk was much higher than I'd represented.  Fortunately most people didn't blame me for their losses, but I still feel responsible.

[SeanB]

That is done with diodes to select reverse voltage. You might have say 4 different diodes, all identical in all respects aside from reverse voltage. Tested during production ( after the die test to weed out those which were either short circuit or otherwise not worth packaging) and then binned according to reverse leakage at selected test voltages. Thus you will get the cheapest part at say 10V, a slightly more expensive one at 20V, a more expensive 30V and finally the premium 40V parts. During manufacture they might improve yield such that all devices pass 20v test after a while so the 10V and 20V parts now are equal, and still later on they might creep up so there are only 30V and 40V devices, which means the 3 lower parts are supplied from the basic lot with only a label change.

Now, if you are using the 10V part, relying on it breaking down at some point between 10 and 20V as cheap overvoltage protection, with a load that will survive 20V for a while, but now you are getting devices that will withstand at least 30V, and up to 45V ( your 40V part will be capable of more than 40V, probably over 50V on most parts before the leakage gets to the limit values), so now you find the equipment it is in is now failing from this. Vf also varies from unit to unit, even on parts next to each other on the reel. Not much, but you get a variation around a nominal value.

[wraper]

Even with 50 diodes in parallel, you only need one diode with significantly lower forward voltage drop to cause the failure as it will take too much current on itself and eventually short out. I'd say that the more diodes, the worse, you'll more likely will get one that is significantly different from the crowd.

If it was so easy to have such a fault why is it NOT showing up, ever? you tried to reproduce it, but it WORKED instead. Now you are saying that in order to have a problem, you FIRST need to have a faulty part, THEN that faulty part will cause further problems. I'm sure if you FIRST smashed hell out of the diodes with a hammer THEN they'd have a hard time working properly too.  What's the difference ? If you don't go in search of a unicorn diode or use a hammer, then diodes work fine.

WHAT THE HELL WORKED??? I had shown that the current through the diodes is very different and one of them heats up significantly more than others  . What the fail should I reproduce the hell??? Should it fail after a few minutes I was running the test? Have you made any real electronics ever? Overloaded parts don't fail immediately but wait to fail at customers site at the worst moment they could  . When pushing 6A through them, that diode which were the hottest, operated with current significantly over spec while others were underloaded. I could also figure out what current was flowing through each of them by running the diode individually and measuring the current needed to rich particular temperature, but I'm to lazy to do this now. Don't forget that those are just 6 diodes from the same tape. In real life, if you assemble like 100 devices with 6 diodes in parallel in each device, I'm sure that in some if them imbalance would be much worse than in this case.

[Chris C]

[SeanB],  I'd heard normal variations in reverse voltage preclude using a rectifier as a zener, but I had no idea they could be this large.  Very informative, thanks!

[TheElectricChicken]

WHAT THE HELL WORKED??? I had shown that the current through the diodes is very different and one of them heats up significantly more than others  . What the fail should I reproduce the hell??? Should it fail after a few minutes I was running the test?

I don't see what there is to be upset about. Sure, one gets hotter, use resistors and one has to be hottest, I doubt perfect even heating is an easy thing to get. The runaway effect was the goal, to show it, but I think it is rare, so rare that it is difficult to come across. Would resistors have one hotter than the rest without proving resistors have a runaway effect ?

End of the day, what I am saying is that yes, diodes can be used for voltage drop, to unload the heat from a regulator even at 20A. Nothing says they have to be parallel, but I'll say that if you use enough it will work, and using 20 % more than you need to in parallel is going to work and I would say that it will not fail in reality any more than other devices in the home which fail often even with 'perfect' engineering.

Have you made any real electronics ever?

what kind of question is this ? I posted pics in this thread of a recent effort of mine, which works fine, using diodes to drop voltage.

Overloaded parts don't fail immediately but wait to fail at customers site at the worst moment they could  .

even with standard engineering.

When pushing 6A through them, that diode which were the hottest, operated with current significantly over spec while others were underloaded. I could also figure out what current was flowing through each of them by running the diode individually and measuring the current needed to rich particular temperature, but I'm to lazy to do this now. Don't forget that those are just 6 diodes from the same tape. In real life, if you assemble like 100 devices with 6 diodes in parallel in each device, I'm sure that in some if them imbalance would be much worse than in this case.

Agreed on all points. I don't see that this proves that failure is imminent, more than any other design, I do see that the hotter a diode is, the shorter the life, same as LEDs, but if the power supply has 20% extra or doesn't run maxed out all day every day for 40 years, then it just works. Is it meant to go up into orbit and last forever, then sure, use the very best. Is it meant to work at home and get the job done today ? then use diodes that you have on hand and unload the regulator and save the regulator instead of the diodes..

I can't see how endlessly arguing to have me agree with your points when I agree with you ALREADY is going to help. Some people just won't take yes for an answer I have found out in life.

[c4757p]

It's not just running hotter. Do you even understand what positive feedback is?

[TheElectricChicken]

It's not just running hotter. Do you even understand what positive feedback is?

Yes, the hotter it gets the lower the voltage drop and the larger the current flowing through it and therefor it gets hotter still. A function that in the case of the diodes that we've tried, and the diodes on my desk are having to a degree that doesn't blow them up or make them go open.

The greenhouse effect and methane hydrate under the oceans is the one to worry about.

[c4757p]

Just warn me if you ever design a product I might end up buying...

[TheElectricChicken]

Just warn me if you ever design a product I might end up buying...

Why would it be different to any other product you buy. Are you the only person in the universe for whom an item has never broken down ?

[c4757p]

Certainly not! I've repaired tons of test equipment, in particular. Huge numbers of failures are due to manufacturers trying to get away with things like this to save a bit of money.

[bills]

Did Mr. Chicken ever indicate he would use diodes In  parallel (or series)  in a consumer product?
I think the point here is there are many ways to do things as a hobbyist, Right or wrong finding out is a learning experience .
I do agree that if you design a product you better be sure that it is safe,and semi idiot proof. 

[TheElectricChicken]

I do agree that if you design a product you better be sure that it is safe,and semi idiot proof.

That's like feeding antibiotics to chickens, it just makes stronger idiots., more resistant to the steps you take to defeat their foolishness. Same as herbicide resistant weeds. They just get stronger and take over the government.

[bills]

Here is one of my favorite quotes.
"I'am not saying go kill all the stupid people. I'am just saying let's remove all the warning labels and let the problem work it self out."  

[oldway]

Paralleling diodes is often used in bridge rectifiers for welding machines and battery chargers.
This let use low cost 35A press-fit diodes for high current output (100A and more).
But the diodes are mounted on the same dissipator, what reduces temperature differences and risks of thermal runaway.

[TheElectricChicken]

Paralleling diodes is often used in bridge rectifiers for welding machines and battery chargers.

how could they, that's terrible 

This let use low cost 35A press-fit diodes for high current output (100A and more).

  This makes me think old welders are an excellent source of high power diodes.   

But the diodes are mounted on the same dissipator, what reduces temperature differences and risks of thermal runaway.

Hmm, a bit like twisting leads together.

[madires]

Paralleling diodes is often used in bridge rectifiers for welding machines and battery chargers.
This let use low cost 35A press-fit diodes for high current output (100A and more).
But the diodes are mounted on the same dissipator, what reduces temperature differences and risks of thermal runaway.

Not by itself. The diodes need to be matched too to get a low unbalance in the load sharing. A few posts ago I've posted a link to a document with a sample calculation of the unbalance. Paralleling diodes for a low cost solution might be more expensive than most think. If a diode needs to be replaced one would have to find a matching one, i.e. buying a large bag and/or replacing all diodes with a new set of matching diodes.

[oldway]

The diodes need to be matched too to get a low unbalance in the load sharing.

I don't agree...These press-fit diodes are not matched, they are only current derated.
Sometime, the lead is crushed and drilled to reduce the section of the conductor and create a low-value resistor.
This reduced section of lead also serves as a fuse.

[TheElectricChicken]

These press-fit diodes are not matched, they are only current derated.
Sometime, the lead is crushed and drilled to reduce the section of the conductor and create a low-value resistor.
This reduced section of lead also serves as a fuse.

That's what my intuition tells me too. If you popped 30 1A diodes in parallel and ran 20A nothing would happen. If there was a fault with one of them, I wonder if it wouldn't just blow and then leave the rest just fine. It is often the case that faults can be burnt out. It's a practice I have heard of many times and I wouldn't be surprised if it did it itself in the scenario of parallel diodes. While one is shorted out, the rest are protected because of the low voltage across them and not carrying any current.

[madires]

The diodes need to be matched too to get a low unbalance in the load sharing.

I don't agree...These press-fit diodes are not matched, they are only current derated.

Actually it's the same Either you match the diodes or calculate the unbalance and derate the maximum current accordingly.

Sometime, the lead is crushed and drilled to reduce the section of the conductor and create a low-value resistor.
This reduced section of lead also serves as a fuse.

Classic series resistor for compensation

[wraper]

These press-fit diodes are not matched, they are only current derated.
Sometime, the lead is crushed and drilled to reduce the section of the conductor and create a low-value resistor.
This reduced section of lead also serves as a fuse.

That's what my intuition tells me too. If you popped 30 1A diodes in parallel and ran 20A nothing would happen. If there was a fault with one of them, I wonder if it wouldn't just blow and then leave the rest just fine. It is often the case that faults can be burnt out. It's a practice I have heard of many times and I wouldn't be surprised if it did it itself in the scenario of parallel diodes. While one is shorted out, the rest are protected because of the low voltage across them and not carrying any current.

Power semiconductors usually become short, not open. And all your circuit becomes busted.

[TheElectricChicken]

Power semiconductors usually become short, not open. And all your circuit becomes busted.

according to the encyclopedia of you ? no seriously, where do you get your guesses from ?

[retrolefty]

Power semiconductors usually become short, not open. And all your circuit becomes busted.

according to the encyclopedia of you ? no seriously, where do you get your guesses from ?

 Well it was true in the case of a large 7 megawatt variable speed drive at my refinery. The main power SCRs were assembled in series strings of 6 for a total of 144 SCRs for the 3 phase variable frequency output. They were designed to allow continuing running even with a shorted SCR in the string. The OEM stated that the SCRs if failed, it would be shorted and log the need for replacement at the next scheduled maintenace turnaround.

 Most failed open condition results from first shorting out and drawing enough current to 'burn open'. If one can maintain adequate heatsinking the short condition can be maintained.