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

[Matie]

Power Supply:
Gophertc CPS-6011
60V 11A Adjustable C/V.

LED’s:
108x Cree XP-G3

LED Configuration:
6x 18 LEDs in Series.

XP-G3 LEDs = (2.74 Vf 150mA @25C) - (3.11 Vf 2000mA @25C)

Voltage Range = 49.32 to 55.98v
49.32v + .2 Transistor drop = 49.52v

BD139 Transistors have Maximum Collector-Emitter Voltage of 80.
Max Collector Current (DC) = 1.5A

The Minimum Voltage could be about 50v up to the Supplies Maximum of 60v!

Work out Voltage at V1
V1 = Resistor1 x 1.1A

Bipoloar NPN transistors 0.7v drop across.
0.7v. R1 = (0.7v / 1.05A) = 0.6666666666666667 ohms.

The voltage at the anode of D7 will be (2 * 0.7v) or 1.4v.

1mA down through diodes should bias them to the 0.7v on state.
0.5mA to bias each transistors well into saturation point, turned on.

6 Transistors = 3mA current
1mA through D7/D8 Diodes + 3mA for Transistors.

4mA of current required through Rr.

Rr can be calculated as (53.64v -1.4v Diode drop) / 0.0040A = 13.060 ohms.


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






Resettable Breakers

I have a 60V 11A adjustable Power Supply and wanted to safeguard against someone adjusting to 11A and blowing up Current Balancing Transistors.

Thermal Derating Curve
You will see below the Raychem RXE-065S1 trigger current changes 19% between 20 and 40C Ambient Temperatures.
The Littlefuse RKEF065 has very similar specs with tighter trigger Current variation of 13% over 20 to 40C Ambient Temperatures.

Time-to-trip
The other thing is time to trip in seconds, the RXE-065S1 is more like a Slow Burn Fuse at 5.3 Sec.

Which makes me think this might not be suitable for my purpose, but would protect from a slowly rising current?
The RKEF065 has a faster, 1 second trip time.
But would it trip before damage if someone dialed up the current on Power Supply.
Anybody know?
-

Here's some Specs for those who are interested, 40C Trip Currents are derived from graph - double check with data sheets before use.
-------


RXE-065S1 60V - Raychem

% of rated hold and trip current for Ambient Tempretures.
Thermal Derating Curve (81% of hold and trip @40C)

Hold Current (A) 0.65 @20C
Hold Current (A) 0.47 @40C

Trip Current (A) 1.3 @20C
Trip Current (A) 1.053 @40C

Time-to-trip 5.3 Sec
Time-to-trip with Fault Current of 10A = .1 Sec

Min Resistance 0.31 Ohm
Max Resistance 0.48 Ohm
---------------------------------

RKEF065 60V - TE Connectivity or Littlefuse

% of rated hold and trip current for Ambient Tempretures.
Thermal Derating Curve (87% of hold and trip @40C)

Hold Current (A) 0.65 @20C
Hold Current (A) 0.54 @40C

Trip Current (A) 1.30 @20C
Trip Current (A) 1.131 @40C

Time-to-trip 1.0 Sec
Time-to-trip with Fault Current of 10A = .09 Sec

Min Resistance 0.250 Ohm
Max Resistance 0.450 Ohm
---------------------------------

RKEF050 60V - TE Connectivity or Littlefuse

% of rated hold and trip current for Ambient Tempretures.
Thermal Derating Curve (87% of hold and trip @40C)

Hold Current (A) 0.50 @20C
Hold Current (A) 0.38 @40C

Trip Current (A) 1.00 @20C
Trip Current (A) 0.87 @40C

Time-to-trip 0.8 Sec
Time-to-trip with Fault Current of 10A = .06 Sec

Min Resistance 0.320 Ohm
Max Resistance 0.529 Ohm


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





Brief Explanation:
Basically we have 6 strings of 18 LED's each.
Each set of 18 LED's uses 1050mA and has a Current limiting Transistor.
The power supply is 11A max, hence it is capable of providing 1.83A per string of 18 LED's.
As a precaution, I am looking at protect each string from exceeding about 1.3A Current.

Questions:

Current Balancing Transistors?
Can anybody help with BD139 Transistors alternative?
I need 6 Transistors in a single package for the thermal connection properties!

PolySwitches?
Anybody using these Polyswitches, do you have a known Source Current reference to give some real world experience.
A device like RXE-065S1 would be perfect If the trip Current stayed within the 1.3 to 1.053 Amp variation (20 to 40C).

But when the Data sheet mentions Min Hold Current is 0.65A and Max Trip Current is 1.3A there is a very large gap between.
So the question is, how tight is the Trip Current in real use.

The RKEF065 has a faster, 1 second trip time but would it trip before damage if someone dialed up the current on Power Supply. Anybody know?

 

[Audioguru]

Cree and you do not know the range of forward voltage for the LEDs. Most might be minimum voltage or most might be maximum voltage then your tight design will not work because you cannot buy only "typical" ones.

You need to design with a high enough voltage that the LEDs and the current regulator work if all the LEDs have maximum forward voltage (Cree does not list it) and design so that the current regulator does not overheat if all the LEDs have minimum forward voltage (Cree does not list it).

How will you cool the tiny LEDs? How will you solder them?

 

[(*steve*)]

I think you should start at the beginning with respect to the current limit calculations.

Your calculations for base current are way off.

Your calculation of the the resistor value to provide this current is wrong. Or at least the value you get is out by a factor of maybe 1000.

Your method of current limiting seems to be based on minimum component count, which is not the best point to start from. Why not individually current limit each string?

You have not done any power dissipation calculations. How hot will those resistors and transistors get?

 

[Matie]

I think you should start at the beginning with respect to the current limit calculations.

Your calculations for base current are way off.

Your calculation of the the resistor value to provide this current is wrong. Or at least the value you get is out by a factor of maybe 1000.

Your method of current limiting send to be based on minimum component count, which is not the best point to start from. Why not individually current limit each string?

You have not done any power dissipation calculations. How hot will those resistors and transistors get?

-----------

Thank you steve,

It was suggested I should use an adjustable supply to experiment with different drive Current for LED's.

Quote: "Why not individually current limit each string?"
A/ That was the idea around the Transistors, at least to balance the current, but yes?
The Polyswitches were another idea from someone using them to protect LED's, but standard fuses seem like a better option if needed.

I do not really see a need to experiment and could purchase sets of LDD sizes, they are cheap and handy to have around!

I had a lot of concerns about this plan and calculations, so posted it in forums to get some feedback.
Maths has not been my strong point, so I got some help but still managed to stuff it up

I don't want to bother anyone for help with the correct maths as I think the original plan is best.
If anyone wants to broaden my knowledge, I am open to learn what I can.

If anyone is interested, this is the Sum and link to reference.

Reference: http://www.st.com/content/ccc/resou...df/jcr:content/translations/en.CD00145611.pdf

My plan is to use 9x Mean Well LDD Step Down Converters off a 48v supply.
1 LDD per 12 LED's.

Audioguru - Quote: "How will you cool the tiny LEDs? How will you solder them?"
A/The LED's are soldered on to 2x6 SinkPAD's with hotplate and clamped to a Heatsink!

 

[(*steve*)]

I would use something like this on each string, with the added protection of a 15V zener diode connected between the mosfet's gate and source.

R1 is a value which will drop 0.6V at the rated current, and R2 can be between 10k and 100k. Q2 is any general purpose NPN transistor.

Keeping each string independent means that if something bad happens to your current limiting only one string will bite the dust.

 

[Matie]

Thanks Steve,
If I can ask a few questions to understand this better.

What Mosfet would you use?
Is the Mosfet a more robust choice, I could look up a Data Sheet to learn more.
Assuming the Zenner after R2 just protects the Mosfet gate from to much voltage?

Q2 Pulls the Gate low, but it is not all clear to me.

From the position of application:
This differs from a Current Sharing Circuit, by being independent Current Limiter Circuits for each 18 LED string?
--

My Question is, could these individual circuits be used to adjust the current from a fixed supply.
I suspect a trim pot may not have the stability and reliability for job.

I had such a problem with resistance in a timing circuit and ended up soldering resistors into 12 position rottary switches, not to say this would be a good idea either.

I ask because the Gophertc CPS-6011 60V 11A Adjustable C/V has not been paid for!
The PayPal invoice could still be cancelled.
-

Not sure how or if the adjustment could be made, but it might negate the need of adjustable supply if it could?
An example might be a normal Mean Well driver for all strings, what ever the required division,
Or individual LDD-1000H/W per 12 LED PCB with adjustment Circuit.

Cheers

 

[Matie]

Cree and you do not know the range of forward voltage for the LEDs. Most might be minimum voltage or most might be maximum voltage then your tight design will not work because you cannot buy only "typical" ones.

You need to design with a high enough voltage that the LEDs and the current regulator work if all the LEDs have maximum forward voltage (Cree does not list it) and design so that the current regulator does not overheat if all the LEDs have minimum forward voltage (Cree does not list it).

I would like your opinion or experience with Cree LED consistence, is this the same across all Bins (S4)?

I began by looking for simplicity and arrived at a single driver.
HLG-320H-C2100
76~152V - 319.2W.
OPEN CIRCUIT VOLTAGE (max.) 156V

By running just 2 strings in Parallel and removing some LED’s from each to fit HLG-320H-C2100 voltage.

Product Characterisation Tool for S4 gives a forward voltage either side of Max Current target 1050mA.
1A... = 2.95fv 40C - 2.97fv @25C
1.1A = 2.97fv 40C - 2.99fv @25C

51 LED’s x 2.98v = 151.98v 1050mA per string.
Assumes the product variation is balanced through average over 51 LED’s.
As you describe, if the batch had a trend it is a problem.

What size tolerance do you build in to negate problem?
I have no problem changing the length of strings to match voltage but would like to know what you have experienced for a starting point?

LED's are soldered on to 2x6 SinkPAD's with hotplate and clamped to a Heatsink!

(*steve*)
This might not be what you had in mind with Current Limiter Cuircuit, but reducing the parallel strings to two might make changing resistors for different Currents within the supply voltage much more attractive?
As I understand it, I would be able to change the current for whole array by changing just 2 resistors.
The balancing should not be required over two long strings, some averaging of LED's will cover it.

Would the circuit be suitable for these voltage and load?

Cheers

 

[(*steve*)]

I would select a MOSFET based on the current, max voltage, and max power dissipation. It is not particularly critical.

Yes the zener is required to protect the gate when the supply voltage exceeds about 20V.

The transistor is turned on by the current sense resistor which lowers the gate voltage. This ensures the gate voltage is set to the value which keeps the transistor just turned on. This point is where the voltage dropes across the current sense resistor is about 0.7V.

This circuit is not intended as a variable current source. If you just want to trim the current to get all strings the same then you either select resistors or as additional large value resistors in parallel to increase the current for strings with lower current.

The most efficient way to make the circuit dim the LEDs is to use PWM. This is a new requirement...

Take a look at the LED resource to see a variant of the circuit I have you which can be used for PWM control.

The circuit is capable of handling long strings but ensure the MOSFET is rated for something in excess of the full supply voltage since failure of the MOSFET will take out your LEDs.

 

[Audioguru]

I have always used LEDs that are spec'd with minimum, typical and maximum forward voltage at various currents. Cree shows only typical voltages because they expect you to use constant current with plenty of extra voltage. The Bins show different colors, not voltages. Other companies like Lumileds have bins with different voltages.

I think each production run has a trend in forward voltage and most (typical) will have that trend. They all might be high voltage or low voltage.

 

[(*steve*)]

I think each production run has a trend in forward voltage and most (typical) will have that trend. They all might be high voltage or low voltage.

Can you quote figures that you've measured? I'm interested in the practical range of values that you have encountered.

I have 1000 cheap LEDs from a no-name manufacturer that are not guaranteed to be from the same batch. Maybe I should check a few and see what sort of distribution I get and compare it with the results you have seen?

 

[Matie]

The most efficient way to make the circuit dim the LEDs is to use PWM. This is a new requirement...

Take a look at the LED resource to see a variant of the circuit I have you which can be used for PWM control.

The circuit is capable of handling long strings but ensure the MOSFET is rated for something in excess of the full supply voltage since failure of the MOSFET will take out your LEDs.

I didn't know of the LED resource before, missed it some how.

So I found this link, PWM dimmer for LEDs http://www.reuk.co.uk/wordpress/lighting/led-dimmer-circuit/

I might have missed the variant of circuit, unless you refer to the 12v Pulse Width Modulator with NE555 Timer Oscillator.

I find the concept interesting, because this is how I thought LED's were driven from some memories of 35 years back. Had assumed this was how LED's could be brighter through higher peak voltage and current?
It would seem the new drivers are smooth which confused me because of previous thinking.

If I missed your reference can you point me straight, I will read the article to understand a bit more.

Thanks

 

[(*steve*)]

The LED resource is here. Section 3 is the part you're after. The circuits you've found are a reasonable source of the PWM to drive the constant current circuit. You probably don't want to drive the circuit I've given you at more than a few kHz and the choice of gate resistor is important to ensure the mosfet turns on and off promptly but without overloading the transistor used to reduce the gate voltage. A value of 1k is probably OK for driving the circuit at a couple of kHz.

 

[Matie]

The LED resource is here. Section 3 is the part you're after. The circuits you've found are a reasonable source of the PWM to drive the constant current circuit. You probably don't want to drive the circuit I've given you at more than a few kHz and the choice of gate resistor is important to ensure the mosfet turns on and off promptly but without overloading the transistor used to reduce the gate voltage. A value of 1k is probably OK for driving the circuit at a couple of kHz.

So a 555 oscillator connected to the final figure 3.5 in Section 3 - Driving a single high power LED.

I meant to mention that the LED's are rated at 2A which I am told is very conservative in Cree Specs.
This was another advantage in two long strings at 1050mA, if one string faults open the LED's are going to have no problems at 2.1A before being switched off and problem fixed.

So no problems with run away current, additionally only one half of each board need be in each string.
Not that it would mater, but the remaining LED's would also have double the Heatsink area.

 

[(*steve*)]

So a 555 oscillator connected to the final figure 3.5 in Section 3 - Driving a single high power LED.

yeah, figure 3.5, but driving your string of LEDs.

This was another advantage in two long strings at 1050mA, if one string faults open the LED's are going to have no problems at 2.1A before being switched off and problem fixed.

That's not a good way to go. One string may draw more current than the other if they are in parallel.

 

[Audioguru]

If you pulse an LED at a high speed with a very high current but have the average current low enough that the LED does not overheat then it is not super bright, instead it looks like its average current because our vision is slow and does not see the peak brightness of the pulse. That is how PWM dimming works.

 

[(*steve*)]

It's even worse than that. LED efficiency decreases with increasing current, so all other things being equal, for the same average forward current a LED will have a slightly lower average brightness as the duty cycle decreases.

 

[Matie]

yeah, figure 3.5, but driving your string of LEDs.
That's not a good way to go. One string may draw more current than the other if they are in parallel.

I would like to continue this discussion to ascertain where and to what degree it applies.
I do not have a strong conclusion either way and see a few possible caveats. Can we qualify this further?

Starting with Multiple batches of LED's with different trends, we have no control. So putting this aside!

I believe that it would be very difficult or require an extraordinary event to damage LED's in this design from thermal runaway, given the LED specification and the maximum available current.
Might be wrong, but lets start from this position and just look at the specific design for mismatch!

How would you describe the process that would lead to mismatched currents and how great could it get.

There seems to be some different answers for the actual application, perhaps within certain LED types?

if you take 100 LED's with Voltage Characteristics within a range around 3v, each lead contributes 1% of the total.
Unless you receive LED's from multiple batches the trend might be positive, negative or neutral but the net variation between 2 strings of 50 should be low just from the numbers involved.

In two strings each LED contributes 2% to an Average Total of around 150v, so the difference in voltage is divide by a larger sum. If you look at the same logic over a few LED's at low voltage the change can be a very high percentage of the total in comparison.

I am not looking to skimp on circuitry, but am wondering if the subject needs more nuance.
-------

PWM is a must do, if there are no problems it is to simple not to include.

You mention the Mosfet will take out LED's if it goes and should be well rated.
What is the process that would take the LED's out, is it likely and could it be fail safed with a standard fuse?

Have seen and used many 555 circuits, so familiar with modified oscillators and PWM.
Keen to apply the 555 PWM circuit to the 2 string single power supply described.

Removing LED's as required to match the end design is no problem at all.
My goal would be to work with little head room for the Maximum Power Utilization.
I will certainly understand and do what I can, but might need some assistance with the final component values.

Often in Timer/Drive circuits I would have problems with power sharing and just insert diodes and capacitors until it worked. If the circuit got to complex with need of separate power supply, it could warrant another path.

Do we need some consideration for the 555 and power sharing?

 

[Matie]

If you pulse an LED at a high speed with a very high current but have the average current low enough that the LED does not overheat then it is not super bright, instead it looks like its average current because our vision is slow and does not see the peak brightness of the pulse. That is how PWM dimming works.

Yes, that is my understanding.
The circuits I was looking at 35 years ago were all battery operated and utilized techniques to increase intensity from a low voltage source. So it is not getting more out then inputted, but I had no knowledge of LED's at the time.

 

[Matie]

It's even worse than that. LED efficiency decreases with increasing current, so all other things being equal, for the same average forward current a LED will have a slightly lower average brightness as the duty cycle decreases.

This is the case!

The increasing efficiency of LED's with lower currents is a good positive attribute for many application.
But achieving a specific minimum depth of intensity cannot always be meet by LED density.

So a positive feature has less value.
People will keep trying to push LED's in fields that they are not necessarily ideal, to push the efficiency envelope.

When you look at the Product Characterization for lm/watts the picture is not so clear for some applications

https://sheet.zoho.com/sheet/published.do?rid=0yszq8f05442075fd4554b4550b59d8ac6a03

 

[Audioguru]

I believe that if you buy a bunch of LEDs from an electronic distributor their forward voltage will be nearly the same either high, typical or low because they will be made about the same time and will be from the same batch. A few might be random from other older batches.

An LED can burn out much faster than a fuse so the LED protects the fuse.