Parallel diode circuits and current limiting

On parallel diode circuits... Several people ask "Can I drive a... with a...". Invariably, these people are "married to parts". By this, I mean that rather than design the circuit, then pick the parts, these people are picking the parts, then trying to design a circuit that uses these parts. People, that is backward.

If you ask me "Can I drive a... with a...", then I will tell you "Yes or no" based on Vforward and the maximum current for your diode. That does not mean the same thing as "Do you think this... is a good idea". That is not the correct question to ask anyway.

A reliable diode circuit requires current-limiting for each diode series. Without current limiting, a variety of problems affect the overall performance of the circuit. All of these problems relate to the fact that diodes vary, even diodes that are the same make and model. So if you have your heart set on a CXB3590 and you want to use several, they will not have the exact same Vforward at a given temperature. What's worse, they will have different temperature characteristics, so even if they started off with the same Vforward, unless your current is low, the devices will heat up, then the Vforward will vary. Finally, even if Vforward starts off the same, and even if these diodes have the same temperature characteristics, over time the diodes will change differently based on material properties, manufacturing tolerances, cosmic rays, you name it. Eventually, these diodes that started off exactly the same, will not be the same.

Who cares? As long as they light up I don't care what old guy says!

That is the problem. If the diodes start off similar enough, or if the current is very low compared to Imax for the diode, then they will probably light up initially. Sooner or later, they will not all light up. This will happen because the difference between Vforward is eventually sufficient that one or more diodes never light up. When this happens, any diode that did light up is taking all the current the power supply will provide.

The simplest case is two identical diodes in parallel. Let's suppose that Vcc is slightly greater than Vforward, and that Icc is equal to the maximum rated diode current Imax. Since there is two diodes in parallel, each diode should see half that current. Typically, 50% Imax is Itest for the diode in the manufacturer sheets, but Cree normally uses 2/3 Imax and that would be worse. Let's assume that due to current limiting for the entire circuit only Imax can flow.

If the Vforward for the two diodes is different more than about a volt, then the diode with lower Vforward will turn on while the other diode is off. It will literally take all of the supplied current. So that diode is on at 100%, and the other diode is off. Unless the diode is thermally able to sink the entire Imax for the amount of time it is on, it will eventually fail. When it fails, the other diode will light up, for a minute, then it too will get hot and fail because it is not thermally able to run at Imax for an extended period either.

If you design your parallel circuits so that each diode can withstand Icc indefinitely and can also dissipate the heat associated with operating at Icc, then you so over-designed your lamp. Your money would be better spent designing the lamp right, so that all the parts are running "in the sweet spot", not maxed-out or under-utilized.

In order to have current limiting in each parallel branch of a diode circuit, you need:
1) a driver per branch, or
2) a channel of an n-channel driver per branch, or
3) a current-limiting driver for the circuit, and then different current limiting per branch.

I think approach one is obvious. It is what most people try. The problem is that drivers are expensive and the reason that most people want to parallel diodes is that they think they can use only one driver for the circuit.

Approach two is not utilized enough in the DIY lamps I see. If you want parallel diodes, then get a two channel (or however many channel) driver. These will normally be cheaper for n=2 outputs, and normally much cheaper for n>2 outputs, vs. discrete driver per branch. The reason is that multi-channel drivers share a good bit of stuffs on the inside so there is your savings.

Approach three may be of interest to some DIY's because the idea is to use one driver for an entire diode circuit. This will provide good quality current limiting to the circuit and achieves this at one driver per lamp. Then, realize current limiting in a branch differently. For example, a suitable fuse would work. The idea is not to prevent the diode from running higher than its normal Idiode. The driver protecting the entire circuit is doing that when things are working OK. The purpose of the fuse is to current limit to Imax, NOT Idiode. That makes his job very easy, and very cheap to do. So if you want to run three of those Cree bad boys in parallel, and your driver can provide 5A, put a 2A fuse per branch plus your driver for the whole circuit and then no diode sees more than 2A and the average is the design current 5/3 A. If you want to do this, be sure to check the voltage rating to see it is more (considerably more is fine) than Vforward, and that the sustainable forward current is more than your desired diode current. These devices cost about $1, so the cost is the one driver plus three fuses, plus your time if you blow fuses and you are too cheap to buy fuse sockets. Hey a $1 fuse and a $1 fuse holder is better than a $40 COB or a $60 Meanwell driver.

Anyway, the correct question to ask when you want to use some specific diode and some driver is "How can I effectively use these ill-suited parts I selected mostly on the basis of stoner website posts to grow some weed". I'd buy a few fuses and use approach three.

Peace!

The attached picture shows the data I looked at when I thought about not having current limiting in each diode branch. "Nominal power factor" means the ratio of the design current to the rated maximum for the diode. For the CXB3590 at 72 VDC, this is 1.2A divided by 1.8A = 0.67. That number is not provided in the table but 0.65 is close enough. Then, by looking at the chart, we can see that if we parallel three, and one diode fails to light, the other two diodes see 0.98 of Imax. That is not good.


If we parallel four, and one diode does not light, then we are at 0.87 Imax of the CXB3590 72 VDC (about 1.57 A) without any additional current limiting. That's OK. If the fourth diode does not light up, then the other three diodes can take that power until thermal considerations become important. Depending on why the fourth diode did not light, this kind of a design may give that diode time to fire. If so, then your lamp continues to work. If I wanted to use a fuse to protect each branch, then I would want 4 diodes and 4 fuses. The fuses should have an on current of at least 1.6A and a trip current of 1.8 A, or maybe a 2 A because that is all we are likely to find. Two amps is OK because even a slow blow 2A fuse will go before the Cree part will actually be damaged. Hopefully someone can tell us where we can buy these for a $1.

Running them in parallel and "drift" would only be an issue if you're running them towards the top of their limit. Or of course, if one fails.

It's just not that big of a deal otherwise.

i think mostly ppl will run driver per string with the meanwell HLG series if i understand this is option 1 which you recommend?

but peeps wanna run 2 or 3 cobs on a cheaper driver that they have already (option 2?) so you are suggesting it can be done as long as the current doesnt exceed the max current of the chip if/when 1 fails?

eg like a 5A driver with chips that have a max current of say 2A

so we could run 3 but if one blows the other 2 will be at 2.5A(above max) so better to run 4 so if 1 blows the 3 will be below max?
id think there could be issues with more than 1 not lighting up right away(during startup etc) or blowing out at some point

anyways if i understood correctly you are advising to run 4 in the scenario above right?

Howdy shrooms!

luvdemshrooms said:

Running them in parallel and "drift" would only be an issue if you're running them towards the top of their limit. Or of course, if one fails.

Click to expand...

For a parallel circuit, a design current close to Imax or a low number of branches makes this subject more important. More current means more lumens and improves the cost efficiency (lumens per dollar) of the lamp. Parallel design will become more important when more people realize there is a limit to how much voltage a constant current source can provide, but there is no limit on current.


I do not think a diode has to fail for the issue to be important to a lamp designer. Minor imbalances in parallel circuits will result in an imbalance of current flow. I used a diode failure as an example because it is easy to understand.

luvdemshrooms said:

It's just not that big of a deal otherwise.

Click to expand...

Thanks for your input.

PetFlora said:

Good questions. Not that I have a clue, which is why I use pre-made panels

Click to expand...

Everyone knows you need great soil. Some people prefer to make their own, some people prefer to buy Ocean Forest. If you place any value in your time, you are much better off buying Ocean Forest (it is a ripoff) rather than fooling around with a bunch of dirt. But some people spend tons of time and money assembling this collection of fish meal, chicken poop, shells, and god knows what else, and then they stir, bake, and mix, then let it set, etc. and they will swear that this dirt is the most awesome dirt since dinosaurs were doggies.

Uhm yeah what was the question?

Hey Dion. I appreciate your time!

Dion said:

i think mostly ppl will run driver per string with the meanwell HLG series if i understand this is option 1 which you recommend?

Click to expand...

Yes, one driver per series string is option one. This approach gives the best current regulation at great efficiency. The only bad thing is the cost of the driver. Meanwell gets great efficiency (0.95), thermally solid, and they are by no means a fragile part. This would be a highly recommended approach.

Dion said:

but peeps wanna run 2 or 3 cobs on a cheaper driver that they have already (option 2?) so you are suggesting it can be done as long as the current doesnt exceed the max current of the chip if/when 1 fails?

Click to expand...

Hmm. I doubt anyone will have a multi-channel LED driver laying around. Four smaller drivers should be cheaper than one big one. If the four smaller drivers are in the same chip, it could even be cheaper. For the IC designer, not us, the trade-off is the value in re-using internal components versus the thermal cost of having multiple FETs in one IC. If you have multiple diodes per circuit, then you would likely prefer a multi-channel driver because of cost. The selection of multi-channel drivers for applications using amps of current is not that great.

Dion said:

eg like a 5A driver with chips that have a max current of say 2A so we could run 3 but if one blows the other 2 will be at 2.5A(above max) so better to run 4 so if 1 blows the 3 will be below max?

Click to expand...

Exactly. That is a true statement. There are other reasons in addition to the diode failure issue, for example when you consider the power the current-sensing resistor in the driver has to sink. In other words, when one branch of a circuit carries more power, this extra power is dissipated in the current-sensing resistor, so you can loose a diode, then have thermal issues in another branch driver. Back at the end of the lamp 3 thread, if you look what happens in R2 and R3, these two resistors get all the extra current. That is not good. The same thing happens inside a meanwell or any other LED driver because they all work this way.

Dion said:

id think there could be issues with more than 1 not lighting up right away(during startup etc) or blowing out at some point

Click to expand...

I agree. What happens when one diode fails is more of a guideline than a hard rule. We both know if you hit a Cree diode with Imax for a minute, it is still fine. But I would need a good reason to violate the guideline. Sure, multiple diodes might fail, and they do not actually have to fail as you guessed. The key is to worry about what happens if the first fails. Then catastrophic failures are not nearly so likely as long as a small problem does not make things worse. If you just want the diode to light up, you don't have to care about this. Kinda like if you just want to get stoned, you don't need great weed. I want to help poor people build cheap lamps that rock. So think of my diode circuits as being covered with trichomes.

Dion said:

anyways if i understood correctly you are advising to run 4 in the scenario above right?

Click to expand...

Yes. For Cree CXB3590, Idesign is 2/3 Imax. If I were going to use this part in a parallel circuit I would want no fewer than four branches in parallel so that imbalances between branches do not cause unstable operation. Two or three of the parts I would run in series. I might add to this idea in a future post.


Thank you so much for your comments Dion, it is always a pleasure!

If I wanted to use two or three CXB3590 in a lamp, I would definitely connect them in series. If I were considering four, then I might use series, or I might use parallel. This is how I would decide...


There are three ways to arrange four diodes and have balanced branches: series, parallel, or 2x2. Compare the cost of the drivers required for the series, parallel, and 2X2 cases... Cost of parallel 4 drivers (Vforward X Idesign) is..., cost of series 1 driver (4 x Vforward x Idesign) is... cost of parallel 2 strings of 2 driver (2 X 2 Vforward x Idesign)... prefer the series approach if the costs are similar... choose the cheapest alternative if costs are different.


Peace!

I messed up the first time I analyzed the n=4 case (the last post). I wanted to make sure I get this subject right, so I prepared the following table. I think this is the list of possible balanced diode circuits for n<10 diodes. By balanced, I mean the diodes are identical and the circuit is designed to provide the same diode current Idesign to each diode. This eliminates circuits like 2x 2/1, 2x3/1, 2x2/1/1, and 3x2/1. I thought about 2/2/2 a bit and decided that does not make sense! Eventually I'd like to draw each circuit so there is a standard list of circuit patterns with pictures.

Awesome thread. Everything laid right out. Really like that last pic. All you need to put up now is some nice labeled wiring diagram.Lol, seriously though great stuff. Gonna make it so myself and countless others have something to follow and reference.

DoubleTripleOG said:

Awesome thread. Everything laid right out. Really like that last pic. All you need to put up now is some nice labeled wiring diagram.Lol, seriously though great stuff. Gonna make it so myself and countless others have something to follow and reference.

Click to expand...

Hi! Thanks for the kind words. The subject of "all the combinations of circuits that..." is a fascinating one. In math, the related problem of "number of tree combinations that..." leads to the really big numbers. Fortunately, there are not that many combinations for n<10.

OldGuyInOregon, you have actually written a very detailed post which can actually be labelled as an article. Being an engineer myself, I can understand the pain of having to explain simple stuff like that to people. For designing a circuit, the first thing you need to know if your objective, the second thing is how do you want to achieve that objective and then comes the circuit design and then components. If you get components first and find out their current and voltage limitations later, it might not make as much sense.

Regards,
Jack Perez
http://blog.7pcb.com

Thanks jackperez, I qappreciate your kind remarks!

I have gone through your discussion and it is looking the pretty cool.
As per my observation that the most of the people run driver per string with the meanwell HLG series.
But peeps wanna run 2 or 3 cobs on a cheaper driver that they have already so according to it can be done as long as the current doesnt exceed the max current.
So we could run 3 but if one blows the other 2 will be at 2.5A(above max) so better to run 4 so if 1 blows the 3 will be below max?

thanks man! you've opened a new world to me!!

jackperez said:

OldGuyInOregon, you have actually written a very detailed post which can actually be labelled as an article. Being an engineer myself, I can understand the pain of having to explain simple stuff like that to people. For designing a circuit, the first thing you need to know if your objective, the second thing is how do you want to achieve that objective and then comes the circuit design and then components. If you get components first and find out their current and voltage limitations later, it might not make as much sense.

Regards,
Jack Perez
http://blog.7pcb.com

Click to expand...

OK so what about the really, really thick people (like me, and I'm talking tree trunks here!!)

I have read this through and understood most of it (I thought) but your table (Oldguy) seems to indicate the if you run 4 diodes in parallel you use 4 drivers (as in driver per string).
I thought that the idea was to have 4 diodes in parallel with 1 driver and each string would be fused below Imax for the diode to protect them in case of failure??

Dion said:

i think mostly ppl will run driver per string with the meanwell HLG series
but peeps wanna run 2 or 3 cobs on a cheaper driver that they have already so it can be done as long as the current doesnt exceed the max current
so we could run 3 but if one blows the other 2 will be at 2.5A(above max) so better to run 4 so if 1 blows the 3 will be below max?

Click to expand...

DevinaBove said:

I have gone through your discussion and it is looking the pretty cool.
As per my observation that the most of the people run driver per string with the meanwell HLG series.
But peeps wanna run 2 or 3 cobs on a cheaper driver that they have already so according to it can be done as long as the current doesnt exceed the max current.
So we could run 3 but if one blows the other 2 will be at 2.5A(above max) so better to run 4 so if 1 blows the 3 will be below max?

Click to expand...

wtf is this?

who r ya?

lol