Linear current regulator with pulsating 3.2A input current is bad? 3

[grigson]

Hello,

We are in Brisbane and are having big trouble (and angry customers) with a 6.8W flashing LED light which is a warning light…………

The linear regulator (LM350) keeps blowing up (its case is blown wide open).

Here is the schematic.

http://i48.tinypic.com/2cfxwqq.jpg

(the schematic uses LT1085 regulator because LTspice doesn’t have a model for LM350)
-also, as you can see, we’ve already put a 3R9 resistor in series with the LM350 to allow the LM350 to run cooler, but LM350 is still blowing up.


Here is the current in the 0R39 resistor (also same as input current)

http://i49.tinypic.com/35bt3c5.jpg

Here is the voltage on the 100uF output cap (the voltage that drives the LEDs)

http://i46.tinypic.com/5yg4ly.jpg

Here is the voltage at the “IN” terminal of the linear regulator

http://i46.tinypic.com/orj69s.jpg

Also:-
Vin = 28V
Power dissipated in LM350 is 6.2W.
Power dissipated in the 25W input resistor is 11W

The LEDs are Osram platinum dragon (LY W5SN)
LED datasheet:

http://www.thorlabs.com/Thorcat/21700/21732-M01.pdf

The LM350 is attached to a heatsink with an “LR20” insulating (also thermally conductive) spacer made by Denka:

Datasheet of thermally conducting material:

http://www.reipower.it/pdf/REI-Denka.pdf

LM350 Datasheet

http://www.ti.com/lit/ds/symlink/lm350-n.pdf

---------------------------
The LED flash pattern is

ON for 44ms
OFF for 10ms
ON for 44ms
OFF for 10ms
ON for 44ms
OFF for 298ms

…..then repeat endlessly
----------------------------

We originally get a contract company to design this for us but now they’ve disbanded.

The only thermal test data which we can find on this product is a thermal camera image of the PCB, which indicates that the surface of the LM350 was 123 degC.
-Unfortunately it doesn’t show what the ambient temperature was when this image was taken, or whether or not the PCB was enclosed in its enclosure (I suspect it wasn’t in its enclosure, as they wouldnt have been able to take the thermal image if it was)

This product is placed outside in the sunshine, and we wonder if the product’s amber diffusor is acting as a heat-trap, making the internal ambient far hotter than the external ambient temperature?


First of all, the the thing that concerns us is the flashing nature of this product.
…..So the LM350 linear regulator is conducting 3.2A pulses of current with the above flash pattern.

…..this means the “IN” terminal voltage of the LM350 is wildly going up and down as shown above.

….i am wondering if this is a harbinger of instability problems in this linear regulator and maybe linear regulators are just not meant for such pulsating current flow?

There are input and output caps connected up to the LM350 but they are not that physically near to it……..they are each about one-and-a-half inches of PCB track away from the LM350.

The heatsink on which the LM350 is mounted is not that big, and has no fins……its about 1mm thin metal………..the crazy thing is that the base of the product, on which the LM350 PCB + heatsink sits, is a huge aluminium bar and it’s unfortunate that the LM350s weren’t thermally coupled up to that in some way.




(Another point is that the LEDs are in parallel with no series equalising resistors, but we find that’s OK)


Anyway, I wonder if you believe that this ON/OFF operation of a linear regulator is a bad idea? (i.e. with pulsating 3.2A current)

 

[Comcokid]

A quick look at the data you've provided shows that you are pushing the LM350 near to beyond the datasheet limits in several areas.

1) The LM350 TI data sheet page 3 shows that the current limit at ~16 volts is 3A (the bottom of your input voltage swing), and at ~28V is 1.4 Amps (beyond the limit). Clearly during the turn-on time you are greatly exceeding the limits of the device.

Can you produce some with one or more LED strings removed as a trial to see if they last longer?

Can you produce some with two LM350s in parallel? Add another R1 on the second LM350 - this will make it close to an ideal arrangement for the two to current share.

2) If there was a measurement of a case picture of 123C, then the junction temperature is hotter and could be close to the limits.

One way to improve the thermal transfer of TO-220 devices at high power is to NOT USE THE MOUNTING HOLE TO SECURE THE DEVICE! There is a older Motorola Application note on thermally mounting power devices that points out how screw-mounting ever-so-slightly warps the device which lessens the thermal transfer. I design electrical devices for aircraft, and when force to mount by this method production is required to use a torque screwdriver. I don't recall the exact torque range recommended for a screw on a TO-220 (research it - in may be in the Motorola app note) but it is almost less than finger tight! If you examine the design of power inverters for the consumer or commerical market - even super sloppy made inverters out of China - you will find that they skip using the screw hole and use either a clip to the plastic body or a screw-mounted bar over the plastic body of the device to get even pressure and maximize thermal conduction of the tab to the heatsink.

2) The TI data sheet indicates a input voltage range to 33 Volts. Since you indicate that the input is 28V this is obviously for some kind of vehicle flasher. Vehicular systems are subject to all sorts of nasty transients, some of which can greatly exceed the input voltage rating of the device. Research the term "load dump" or "co-inductance" and look for SAE specs on vehicular systems. Long input wires are inductors also. Long wires to the load are inductors. Quick switching could make these ring and produce fast transients - maybe.

You could increase the input-to-output differential of the LM350 slightly by adding a few diodes or a zener in the collector of the Q1-2904 transistor. The LEDs have 2.1 to 3.3V drop each - take 2.1 times 2 LEDs and .6 for the MURS120, and try a zener around 4.8V.

3) Spice is fine for roughing-in the initial schematic. But then the real-world takes over and it can be quite different.

 

[IRstuff]

> Is that REALLY a 0.39 ohm resistor shunt? That means you are already dumping 3A just across that resistor.

> According to the datasheet for the LM350, the output voltage programmed is supposed to be 1.25V(1+R2/R1). Your schematic has R2=470 ohm, and R1=0.39 ohm, which means that the LM350 is programmed for maximum output voltage, which is BAD.

> Your oscillscope trace presentation was nearly worthless, since I had to physically measure and scale the reading to get the voltage across the LEDs, which is ~7.6V, which from the datasheet puts the forward current at >1.5A per string, so with 6 chains in parallel, your circuit is attempting to dump ~10A from a regulator only guaranteed to supply 3A, not counting the 3A it's trying to dump across its shunt resistor.

> Your oscillscope trace further shows that the circuit completely violates even your duty cycle requirements, since the 2N3904 completely fails to completely discharge the voltage across the LEDs during the OFF periods. I'm guessing that the 0.39 ohm resistor is correct, since the time constant for the discharge is ~50ms, which divided by the 100 uF capacitance results in an apparent resistance of 500 ohm, the bulk of which is the 470 ohm programming resistor. I'm further guessing that your schlock contractor's simulation couldn't pull the output down fast enough with the resistor values they should have used, so they dropped them to get the transient response to a barely passable condition.

> You posted datasheets, yet apparently, no one in this process bothered to read them at all. Your LEDs have to be limited less than 4.8V for the double stack to keep the forward current under 0.5A per leg x 6 legs = 3A, which is the spec for the regulator. Your resistor values appear to be completely inappropriate for this design.

> I'm guessing that your "contract company" disbanded because they were completely incompetent, which may be coupled with the fact that your company was nearly equally guilty for not properly specifying the design. Thermal and ambient conditions are supposed to be in the design specification. Your raising that issue suggest that your company never bothered to tell the designers that their design had to work in a substantially harsher conditions than they thought. Of course, double shame on them for failing to do due diligence and asking what the operating conditions were supposed to be. Triple shame on your company for apparently failing to demand design reviews and to demand verification and qualification of the design.




TTFN
faq731-376

 

[grigson]

Thanks:

Comcokid,


..yes but surely the turn on time is a small , tiny interval, and will not be significant in heating the LM350?

When flash is on, the LM350 has 28 - (7.5+1.25 + 3.2*3.9) across it = 6.77V
When flash is off, the LM350 has nearly 28V across it ..but its not conducting current so surely this doesnt really matter?

We cannot easily remove leds as its a led module.
-Equal current sharing for LM350's looks like it needs exactly equal resistances at the output of each LM350, or better still an opamp feedback circuit.

I know what you mean about "load dump"....but load dump almost only ever occurs during bad servicing practice........to get a load dump you literally have to pull the battery terminal connectors off whilst the alternator is running...........only a fool would do that, and only a bigger fool would do it while any of the lights were switched on.

We have a front end 220uF capacitor and the 3R9 resistor before it, so i doubt we will see any transient activity......after all, a vehicle battery is an absoltely huge capacitance, and its voltage simply cannot be made to go more than 0.5V above nominal (unless its a faulty battery)



IRStuff:
Yes it is a 0R39 shunt.....but remember these leds are flashed, so the 3.2A only flows through it for about a third of the time because the flash duty is 0.3



I dont know where you get the 10A figure from, the LEDs used in the simulation are just three leds that give the 7.5V drop of the actual leds that are used in the real circuit (the simulator does not have the model for osram platinum dragin led)
The 470R is ismply to limit the current in the 2N3904 when it switches off the LM350

The 0R39 resistor means LM350 will regulate the current to 3.2A, which is the required current.

 

[OperaHouse]

I thought it was kinda novel approach switching the voltage below the LED conduction poin. Too bad you can fry an egg with it. I know everyone says to drive LED with constant current but that is only true with a pure series string. When strings are in parallel, ond LED shorting or opening can take out all the others. I would have use something like a LM2596 switching regulator that has a simple on/off pin. It has a bigger brother that will do 5A and 40V input. Run it at constant voltage a little above what LED needs for safety and add a little current regulation. This would be small and cool. I've bought complete boards like this on ebay from China foe only a couple bucks. Far less than what you paid I bet.

 

[LiteYear]

Hi grigson,

I don't think there's anything fundamentally wrong with pulsing a regulator, as long as you ensure the transitions don't exceed maximum values. The basic layout and component values of the circuit look fine. IRstuff must have had a rough day, and maybe missed that the regulator is being used in its current regulator configuration (hence the 0R38), not in the normal voltage regulator configuration. There's probably better and more efficient ways of doing the job, but that's not the issue at hand. The issue is why the regulator is popping.

I agree the RMS power dissipated by the regulator is about 6.3W, but even in theory it has peaks of over 21W for 44ms at a time with only 10ms rest in between. Given the short rest, it would not be wise to disregard the entire ~150ms on time as negligible. The component is rated for more then this, but that's still a lot of heat to get out of a small area. At those power levels you want to be very sure of your heatsinking. You'll see that without a heatsink the component will rise 35 deg/W, so you're heavily reliant on the heatsink to lower that temperature rise. As Comcokid suggests, even the securing method can defeat heat paste efforts. The placement of your (rather hot) input resistor could have an influence on the thermal performance too.

So it's possible that the heatsink is simply insufficient. You'd have to do some thermal calculations using the size of the heat sink, expected temperature differentials, and the specified thermal resistivities to find out.

Other than that, it's still possible that the circuit is producing unsatisfactory transients. The spice simulations look okay, but again, as Comcokid says, the real world is a-whole-nother world. I'm particularly concerned about the turn-on transient. Are the input and output caps you refer to the ones that appear in the schematic? Because those caps are probably too big to protect the regulator from internally generated transients. I'd at least have a 0.1uF cap right next to the input terminal. I'm also a little suspicious about the voltage on the input pin during turn on - the 220uF cap will hold the 28V for a couple of ms before it settles to the 15.5V operating voltage, and who knows how that interacts with the transient that will occur on the output pin.

I'm afraid I'm just musing here - this can only be solved by actually putting a CRO probe on the circuit itself and discovering the joys of theory/practice differences.

 

[MacGyverS2000]

Amusing, as always...

Dan - Owner

http://www.Hi-TecDesigns.com