Your drawing shows the ground for the controllers far away from the end of the LED strip.
What are the physical distances involved between the stips and controller? between strips?

You want the ground of the controller connected to the ground of the LED strip, RIGHT WHERE IT CONNECTS TO THE STRIP.
Think of it as a star ground concept, the point where the LED strip begins should be the center of the star.

The LEDs need to see a clean digital signal that is referenced to the same voltage at the LED ground, not some wire 10 feet away, which will never be at the same potential.

Since these LED strips use high speed, single ended TTL signalling, where and how you connect the ground is EVERYTHING.
The inductance and resistance of even a few feet of wire with 20 A flowing will totally bugger any concept of ground.

Oh wow! Alright. Well, then.... should I add an additional switch that moves the ground path from the LED to the controller in lockstep with the data path? If I do that, should I remove the ground path that goes along with the 5v rail providing current for the strips or just have multiple ground references? I'm not sure how much shorter a ground wire parallel to the data line would end up being in the end.

 

Wouldn't a diode in series with each power supply work? All three grounds connected. A separate diode between each + output and the load.

Yes. (Sorry, Bob)

This is called droop-sharing. The diode acts as a ballast resistor and a reverse-current protector all in one. Because a diode has a V-I *curve* (as opposed to a brick wall type of straight line) response, multiple supplies can share current into a common load quite nicely if the outputs are tweaked to be nearly equal. The diodes also prevent control loop interactions (fighting), which is a probable cause or, or contributor to, the original problem.

Of course, I'm sure there are some supplies out there whose control loops are so finicky that this will not work, but my experience is that most normal industrial shoebox type supplies will. Lambda and others have app notes detailing this method.

Besides cleaning up the ground management as mentioned in several posts, diode-ORing the outputs will get the OP what I think he wants, redundant power to the LEDs. Thanks to decades of switching power supply design, there are lotsa high power dual-diode devices to reduce the body count and wiring.

An alternative is to replace the supplies with ones designed from the ground up to be shared. These would have a current share bus connection separate from the main power and ground connections, sometimes called 3rd wire current sharing.

ak

 

Yes. (Sorry, Bob)

This is called droop-sharing. The diode acts as a ballast resistor and a reverse-current protector all in one. Because a diode has a V-I *curve* (as opposed to a brick wall type of straight line) response, multiple supplies will share current into a common load if the outputs are tweaked to be nearly equal. The diodes also prevent control loop interactions (fighting), which is the cause of the original problem.

Besides cleaning up the ground management as mentioned in several posts, diode-ORing the outputs will get the OP what I think he wants, redundant power to the LEDs. Thanks to decades of switching power supply design, there are lotsa dual-diode devices to reduce the body count and wiring.

An alternative is to replace the supplies with ones designed from the ground up to be shared. These would have a current share bus connection separate from the main power and ground connections, sometimes called 3rd wire current sharing.

ak

Okay, so I should add some load balancing? That will require some modification to the work I've already done but if it's needed it's needed.

How do I go about choosing the right diode? I had considered adding load balancing via resistor. I calculated that adding a 6mOhm resistor across each of the three V+ leads would produce a 0.1v vdroop at the cost of 1.53watts dissipated at full 15.3amp load spread over the PSUs. The no-load voltage is currently at 5.10v and the full load voltage only drops to 5.06-7v worst case scenario (I plan to raise the voltage to 5.6v to save myself some amperage down the line. All calculations were done assuming 5.6v). The three PSUs are within 0.01v of each other.

I don't really know how much of a load increase I'm shooting for with the resistors/diode setup.

 

I give up.

AK: My suggestion that diodes will not do it was based on the fact that the voltages are different. If the voltages differ by even 100mV or so the higher voltage one will supply nearly all of the current. In fact , diides are used thus way to allow a higher voltage supply, eg. an AC adapter to override a battery supply when connected.

If the voltages are all the same and will remain so forever, then yes, diodes will allow equal current sharing, but then so will no diodes.

Why is everyone here ignoring the simple suggestion of powering 1/3 of the strips with each power supply? It makes no sense to me to do abything else inless there is a compelling reason.

It now sounds like the TS did this originally and it did not work because he did not connect the grounds together. The answer is to fix the grounding, not to unnecessarily cause other problems by paralleling the 3 supplies.

Bob

 

I give up.

AK: My suggestion that diodes will not do it was based on the fact that the voltages are different. If the voltages differ by even 100mV or so the higher voltage one will supply nearly all of the current. In fact , diides are used thus way to allow a higher voltage supply, eg. an AC adapter to override a battery supply when connected.

If the voltages are all the same and will remain so forever, then yes, diodes will allow equal current sharing, but then so will no diodes.

Why is everyone here ignoring the simple suggestion of powering 1/3 of the strips with each power supply? It makes no sense to me to do abything else inless there is a compelling reason.

It now sounds like the TS did this originally and it did not work because he did not connect the grounds together. The answer is to fix the grounding, not to unnecessarily cause other problems by paralleling the 3 supplies.

Bob

Did you see my diagram? It is how everything is wired right now, and unless I'm misunderstanding you, it is exactly how you describe - everything split in thirds, commoned grounds. I'm still having problems, which is of course why I'm here. EDIT: Actually, my setup right now is simpler because I do not have the switches in the circuit for the sake of troubleshooting.

I have connected up 4 strips, and now the blinking has returned. I suspect this is happening because the loads aren't balanced and I'm running into a current limit on the first PSU. There are three assumptions with this hypothesis:

a) Commoning the DC grounds makes the PSUs run in parallel
b) A current limit condition on one of the PSUs would make ALL LEDs blink together
c) One of my PSUs is running into a current limit condition.

I don't really understand how a) could be the case, but I don't have the necessary experience to know otherwise. B) is reasonable enough, but c) is very confusing. The blinking happened when only 1 strip was connected to 1 PSU and the others had their grounds commoned (but had no load attached). Better matching the PSU output DC voltages solved this issue before, but it's back again with a heavier load attached. My amateur guess is that the greater load has caused even the miniscule discrepancy between PSU output voltages to become a problem.

I'm happy to begin work on a load balancing circuit, but I need some guidance. I don't know what Vdroop to shoot for with the resistor. I picked 0.1v, but that was mostly just an (uneducated) guess.

If I need resistors, what value? If I need diodes, which ones? If I need to figure that out on my own, fine, but how much Vdroop am I aiming for?

Thank you.

 

AK: My suggestion that diodes will not do it was based on the fact that the voltages are different. If the voltages differ by even 100mV or so the higher voltage one will supply nearly all of the current.

That is where we disagree. In my experience, casual output voltage adjustment with a 3-1/2 digit meter can achieve and hold 66/33 current sharing, more than enough to spread around the heat and significantly increase the long-term reliability of the supplies compared to 100/0 sharing. Yes, diode OR-ing can be configured as an all or nothing switch, but it also is capable of other actions.

And I'm not ignoring other solutions, just discussing the idea of one responder.

ak

 

Did you see my diagram? It is how everything is wired right now

I missed the big black magic marker line that connected the three power supply grounds

That should work. I don't see how one power supply could be overloaded when you have only 86W connected to a 100W power supply, unless they are well under spec. And one shutting down should will not effect the others with this circuit, unless it is the one feeding a single controller controlling all the strips.

So, with the switches out, how is it set up? One controller for each strip, or 4 separate controllers?

Do diagnose the problem, I would simplify it further.

I would first try taking out the repeater amplifier (actually it is a buffer) and wire it so the 4 separate controllers are uses. If there is a still a problem that affects all the strips, something very strange is going on, probably some incorrect wiring.

If that works okay, I would try connecting all of the data lines to one controller. It should be able to drive 6 inputs since they are likely CMOS. If that works, then it is the buffer or switching scheme that is causing the problem.

Bob

 

According to this paper, since I am powering a highly variable load, droop sharing is not an option for me at all.

https://www.cdiweb.com/Datasheets/power-partners/Load-Sharing-Considerations.pdf

Well...failing that...anyone have any other ideas? I suppose I can look at the circuit from the ground side like Bob has suggested. I will have to try to solve the ground loops without paralleling any of the supplies in the process (which I assume is happening when I tie all the DC grounds together. That explains the results I see).

But, I don't see how to achieve that when I have a switch that enables one controller to modify all 4 groups of LEDs simultaneously via a repeater/amplifier. It seems to me that no matter what, such a use case brings hardware attached to all three separated PSU circuits into THE SAME circuit thus producing group loops.

 

I missed the big black magic marker line that connected the three power supply grounds

That should work. I don't see how one power supply could be overloaded when you have only 86W connected to a 100W power supply, unless they are well under spec. And one shutting down should will not effect the others with this circuit, unless it is the one feeding a single controller controlling all the strips.

So, with the switches out, how is it set up? One controller for each strip, or 4 separate controllers?

Do diagnose the problem, I would simplify it further.

I would first try taking out the repeater amplifier (actually it is a buffer) and wire it so the 4 separate controllers are uses. If there is a still a problem that affects all the strips, something very strange is going on, probably some incorrect wiring.

If that works okay, I would try connecting all of the data lines to one controller. It should be able to drive 6 inputs since they are likely CMOS. If that works, then it is the buffer or switching scheme that is causing the problem.

Bob

About your comment about the single power supply, that's exactly what I thought as well.

With the switches out of the circuit (the repeater is also removed), one controller is dedicated to one group of LED strips and that controller is connected to the same PSU as the LEDs it's controlling. I only have 3 groups attached right now:

PSU1 gets two LED strips and the controller for them called Group 1. PSU2 gets two LED strips and two controllers, one for each, called Group 2 and Group 3. The rest are disconnected.

When this setup is run at max with their grounds commoned, eventually all the LEDs will "glitch" blink together and the controller modifying two strips will become unresponsive to the second strip (and for some reason the last two or so LEDs in the first one, go figure). Removing the data leads from the other controllers allows the glitched group to recover.

With the grounds not commoned, everything works as expected, but there will be loops once the switches are back in circuit.

 

And the matched voltages will track each other over time and load and temperature variation to better than 100mV? I doubt it.

Here is the result of using diodes or resistors to share two 5V power supplies with 100mV difference in voltage.

 

What kind of test equipment do you have?

Bob

 

With the switches out of the circuit (the repeater is also removed), one controller is dedicated to one group of LED strips and that controller is connected to the same PSU as the LEDs it's controlling. I only have 3 groups attached right now:

Wait a minute. If I am understanding you, there is no connection between the two groups except the common ground, and they still both blink at the same time? Impossible (or at least nearly so.)

If this is the case, I think you have a wiring error.

Bob

 

What kind of test equipment do you have?

Bob

Just a digital MM. I wish I had a current clamp...but sadly...

Wait a minute. If I am understanding you, there is no connection between the two groups except the common ground, and they still both blink at the same time? Impossible (or at least nearly so.)

If this is the case, I think you have a wiring error.

Bob

I know! That's what I thought! I don't think I do. I mean, probably, but I don't can't see where. The circuit is as simple as I can make it. PSU --> Ctrl --> LEDs for both PSUs (I'm only using two for test purposes right now).

Also, very important, I was wrong about the controller glitching. That does NOT seem to be related to commoning the grounds OR the connection status of the other data cables as I stated. This is a separate issue. Maybe it has to do with attenuated signal strength due to the fact that the controller is pushing two light bars. I don't know. It seems to happen primarily when the light bars heat up significantly.

The blinking issue when grounds are commoned is still a thing, though. Near as I can tell, they do not blink when DC V- between both PSUs is NOT connected. And, yeah. They all wink in unison even though half of them are on separate PSUs. Super weird.

 

I guess the next thing to determine is if it is the power supplies that are blinking out or the controllers.

Is it possible it is the AC supply that is blinking out?

Do the power supplies. Have any status indicators?

Can you monitor a power supply voltage and see what it does when the lights go out?

Bob

 

How about trying driving all 4 strips off one contoller?

Bob

 

Alright. I'll try both of those things and post results back when I get them. Thanks!

 

I guess the next thing to determine is if it is the power supplies that are blinking out or the controllers.

Is it possible it is the AC supply that is blinking out?

Do the power supplies. Have any status indicators?

Can you monitor a power supply voltage and see what it does when the lights go out?

Bob

So, in the process of trying to find out the information you were asking, I learned something that may explain everything.

I think there is a temperature-related short in one of my LED strips.

When this strip gets hot, all but the first 4 LEDs stop responding. This strip is in front of another and that one stops responding too. Once this happens, the only way to get response back is to shut off the system, let it cool, and turn it back on. The other LED strips don't have this problem, which makes sense because this strip was connected in line with another one that I accidentally fried by not being careful enough with my mm probes (in my defense, the contact pads are extraordinarily small and the LEDs are tightly packed). I thought this strip came away unscathed because it worked for a while, but apparently not.

I'm currently in the process of rebuilding my circuit and we'll see if some of these very strange problems disappear. I won't be able to replace the damaged strips until tomorrow, but I should be able to test 2/3s of the system and that should be enough.

I'll let you know what happens. Thanks again for the help!

 

Does not explain why the strips on the other power suppy blinked out.

Bob

 

Does not explain why the strips on the other power suppy blinked out.

Bob

Oh? Well, crap. I did take a reading from the PSU powering the misbehaving strip once it started blinking. I didn't detect any drop in voltage from the PSU output.

I'll run some more tests once I finish getting this wiring installed.

 

And the matched voltages will track each other over time and load and temperature variation to better than 100mV? I doubt it.
Here is the result of using diodes or resistors to share two 5V power supplies with 100mV difference in voltage.

That looks like an excellent example of how simulations do not always indicate real-world behavior.

ak