Maker Pro
Maker Pro

A Current Balance Circuit for LED Strings in Parallel with Polyswitch Protection

Here is a picture I saved from http://adaptivepower.com/Resources/Documents/APS_App_Note_LED_Driver_Testing.pdf

Simulated LED load with adaptive power system, just thought it was interesting

LED_load_voltage_Blue_and_current_red.jpg
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
So much to answer on a phone! (Beware autocorrect errors).

Unbalanced LED current in parallel strings. Let's assume you have two strings of perfectly identical, perfectly matched LEDs on identical heatsinks in the same environment, but running quite warm. Initially they will draw the same current. Let's assume that some factor leads to one heatsink being slightly less efficient for an instant. This will lead to one set of LEDs (or even just one LED) getting slightly warmer. This will cause it's Vf to drop a little, which will unbalance the current in the to strings. This will cause the already slightly warmer string to draw more current and the slightly cooler one to draw less. The process will continue until some equilibrium is reached. This may be one string using 100% of the current from a constant current source, or the rate of heat transmitted to the environment starts to rise faster than the rate of heat increase (heatsinks transfer more energy as the differential between it's and ambient temperature increases).

It's not too hard to damage a LED. A high current pulse may well damage them or destroy them before a fuse blows. Heat is a great killer of LEDs. Have you seen LED strips with LEDs that are dimmer than others? Or LEDs that don't work? Or LEDs that don't come on immediately?

Most semiconductors fail short circuit, so a MOSFET failure will result in loss of current control. The current can rise to high value and can damage the LEDs before the fuse blows (if you have one). Regardless of whether or not you have a fuse, that string is no longer functional. If the one mosfet controls
Multiple strings then they will all go out. Keeping all the strings independent at the cost of a few cents seems reasonable (unless you're Chinese :)).

I'm not sure what your on about with power sharing and timers. It should not be an issue.
 
Unbalanced LED current in parallel strings. Let's assume you have two strings of perfectly identical, perfectly matched LEDs on identical heatsinks in the same environment, but running quite warm. Initially they will draw the same current. Let's assume that some factor leads to one heatsink being slightly less efficient for an instant. This will lead to one set of LEDs (or even just one LED) getting slightly warmer. This will cause it's Vf to drop a little, which will unbalance the current in the to strings. This will cause the already slightly warmer string to draw more current and the slightly cooler one to draw less. The process will continue until some equilibrium is reached. This may be one string using 100% of the current from a constant current source, or the rate of heat transmitted to the environment starts to rise faster than the rate of heat increase (heatsinks transfer more energy as the differential between it's and ambient temperature increases).

Sorry for putting that question on you, I totally agree with your scenario description.
My questions are out of interest in the general use of LED's and for specific need.

No invalidation, just playing the scenario.
There is a small variation in outcome if the string has separated LED's or they run parallel to their partner string sharing the same PCB and HeatSink. Not a real measure worth worrying about, but an effect.

This is my Scenario in General Current Mismatch:
Based on the Mismatch coming from the LED Characteristics. Your scenario is also valid.

Let’s say we started with known LED’s from the same batch with similar trends.

EG..
With 2 x 2 LED’s
Let’s say we have a Maximum of 10% for variation with about 33% chance of it occurring.
--
With 50 x 2 LED’s
To match the 2x2 scenario we need 50 LED’s in 1 string trending 10% different then the 50 opposite.

Of course, these numbers and the extreme comparison are not real.
My question is, does this basic logic work or make any difference if we design the balancing circuit or choose resistors with consideration for required correction?

Based on some base line assumptions and odds.
Maximum possible variation in % of Total voltage?
Example: 1 LED’s value.

Percent of correction you would like to achieve in the worst case scenario?
Example: 75% Correction.

Does this make sense?
Anybody want to contribute thoughts on real numbers you use for the possible variation and % correction?

Knowing if there is any logic in this scenario would help with discussion for balancing.
I would also like to know how you would apply balancing to our 151v strings within or outside PWM circuit?
------------------------------------

Most semiconductors fail short circuit, so a MOSFET failure will result in loss of current control. The current can rise to high value and can damage the LEDs before the fuse blows (if you have one). Regardless of whether or not you have a fuse, that string is no longer functional. If the one mosfet controls
Multiple strings then they will all go out. Keeping all the strings independent at the cost of a few cents seems reasonable (unless you're Chinese :)).

My question was in consideration for some type of spike with failure.
The fail short answers the question, so in this case we fall back to the maximum current of driver.

I'm not sure what your on about with power sharing and timers. It should not be an issue.

The driver is 150v and I was concerned about glitches such as occur in power from relay and Solenoid valves on the same power as 555. I will connect the 555 to a separate AC/DC supply, $10 or I may have one lying around.

I am looking at the possibility of a 100% switch at the 555/Mosfet interface.
if posible 100% would be default use, PWM would be used on rare occasion to experiment down to 50% duty or 500mA.

Perhaps just install the PWM and Mosfet if needed, makes sense :)
In which case any balancing components would have to be separate, perhaps a negative?

At some point I will be building this array and might not have internet for reference.
I am learning as much as I can now, please do not think you are being ignored.
I am listening very carefully to your advice and may not make up my mind for some time.
There are a lot of possibilities to leave open, and possibly 2 more of these arrays to build.

All Good?
 
Last edited:

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
If LEDs are on the same heatsink it will act to reduce the temperature difference between them. The closer the temperatures of two strings can be maintained to each other, the smaller the potential drift in the sharing of current between them will be (assuming they are similar devices and we originally sharing current evenly). The best practical example is COB led arrays which may have 10 strings of 10 LEDs on a single device, all of them in close thermal contact.

If you don't have a low voltage power source you will need one for the 555. It won't run from 150V. And you will meet to have a common rail between them. Typically, and sensible in this case, would be to connect their negative terminals together as your common ground.

I can't make any comment about your scenarios because they really don't make much sense to me. My scenario was for (mythical) perfectly matched devices. The situation becomes worse and harder to predict as they move away from perfection.

It's better to design for stability than to hope for it based on an expected (or wished for) spread of characteristics.
 
If you don't have a low voltage power source you will need one for the 555. It won't run from 150V. And you will meet to have a common rail between them. Typically, and sensible in this case, would be to connect their negative terminals together as your common ground.

I can't make any comment about your scenarios because they really don't make much sense to me. My scenario was for (mythical) perfectly matched devices. The situation becomes worse and harder to predict as they move away from perfection.

It's better to design for stability than to hope for it based on an expected (or wished for) spread of characteristics.

No problems, I have a lot of experience with 555 and aware of its characteristics. Common Ground. Done.
-

Strings of 10 LED in parallel will suffer about 5 x the average imbalance as strings of 50 LED's because we work with larger numbers and the voltage of all LED's are averaged to make the total.
You are looking at the single LED, which is correct.
I am ignoring the single LED and looking at the string as a whole.

What you describe is a fault, something that should not exist with best practice, but could occur with a problem.
If the maximum current cannot exceeded the LED rating in design, it is a maintenance issue.
If runaway current can still take out LED's it is a design issue. But Best practice is also about balance efficiency.
-

From the position of this design, I just want to balance the LED's for improved performance.
So with normal balancing practice, are you comfortable with 2 strings at 150v.

If so, using what ever values you would like to suggest. what would you use?
-

Very grateful :)
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
No I'm not just talking about a single LED. I'm taking about chains. When I refer to something happening to one LED, the effect would be amplified if it happened to multiple LEDs.

I agree that it should not happen with best practice because best practice suggests you don't put strings in parallel.

Putting 2 strings in parallel might work for you, and it's no skin off my nose if you decide to do it. But short of having them well matched and sharing the same heatsink, don't expect me to tell you it's going to be OK.
 
Putting 2 strings in parallel might work for you, and it's no skin off my nose if you decide to do it. But short of having them well matched and sharing the same heatsink, don't expect me to tell you it's going to be OK.

Yes, they do share 50% each of the same heat-sink side by side.
I sensed there was some distaste for the design, glade you have been up front.

I would not expect you to say it is ok, but would appreciate your electronic knowledge to build balance circuit.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
OK, you said they were on a heatsink but I didn't pick up that they were on the same heatsink.

The closer the two strings are both thermally (i.e. the heat path between them is short) and electrically (i.e. variation of Vf with current and temperature) the closer the currents in the two strings will track each other.
 
OK, you said they were on a heatsink but I didn't pick up that they were on the same heatsink.

The closer the two strings are both thermally (i.e. the heat path between them is short) and electrically (i.e. variation of Vf with current and temperature) the closer the currents in the two strings will track each other.

The 6x2 SinkPAD have 1" LED spacing and good thermal conduction, these are clamped to Large Aluminum Heat Sinks chosen with about 50% head room, and also serving as structure for physically connecting to Aluminum frame.

The LED's (PCB's) will have an oversized heatsink based on passive conduction recommendations, and have fan blowing through each 3 PCB connected in rows. Heat is an important factor.
 
Here is the PCB Arrangement for 100 LED's.
2x 50 LED Strings in Parallel.

The 2 "LED strings" run either side of the 9 PCB's and equally share 9 common heat sinks for thermal balance. Each string is the same length and all LED's are directly opposite their partner in adjacent string.
Wires from the PCB Array to driver are short, high gauge and equal in length.

PCB_ARRAY_2x50_LED_149v_98_of_152v_Supply.jpg


Normally all PCB would be full, making 108 LED's in Total, and be powered individually by
9 Mean Well LDD-700H-W DC-DC Constant Current Step-Down LED drivers.

In this case I wish to add a Pulse Width Modulator and Mosfet to a Mean Well
HLG-320H-C2100
76~152V - 319.2W.

Note: You will notice the Centre PCB's above have some LED's missing.
This is to reduce the required voltage "by Data Sheets" to 149v which is 98% of the 152v Supply.
The Total Number of LED's can be reduced in even numbers, as required to match the Supply Voltage.

555 Pulse Width Modulator
ne555-pwm-led-dimmer-circuit.gif


Mosfet Output Circuit
Current2.jpg



----------------------------------------------------------------------------------------------------------------

Incorporating 150v Balancing Circuit?

I would like to incorporate a balancing circuit to improve performance.
Hope to get some assistance with the Math and required components for 150v 1050mA strings, or an alternative circuit choosing components to match above.

This would only be used as test equipment.
Those interested in a permanent LED system using 12 LED PCB could consider MW LDD Boost - Step Converters as a 12 LED PCB driver option which negates the Balancing issues.

LED_balancing_circuit_2_strings.jpg
 
Last edited:
Top