Not really an expert (obviously ) but thought paralleling LEDs had the potential of thermal runaways. Not sure how that would work on a single channel and one “LED strip” broke.
But LM3409 could also do 36V which I haven’t rejected yet. I still have 12x unpopulated PCBs of my LM3409 design which I suspect doesn’t require a redesign. Just some new values on the components.
Looked at my calcs for the 12V config, 150kHz, 90% efficiency sort-of. T_OFF 913.83E-09 ns
… That T_OFF is quite high and that’s due to me taking R_OFF straight of the datasheet (17k)…
Most (important) components have a 50V rating looking at my BOM…
Made all my calcs using LibreOffice and changing Vin/Vout mainly changes R_OFF and the inductor values.
But from what your saying maybe a Vin_max of 36V and still calculate for 24V output… When I quickly plug in some values I conclude that efficiency do goes up to 92% …
@thefreeman, what you wrote about not getting the Vin too close to the Vout got me thinking. The above alternative of going Vin ~36V and Vout 24V (i.e 3 drivers per Light of a total of 6x 12V XHP70.3HI’s) will get me into trouble on the inductor side. Both expensive and redesign of my existing LM3409 driver PCB.
Thinking about voltage drop. Voltage drop in the circuit, duty cycle limitation, battery->cable->lamp … and so on.
My preferred option of stacking 2x 12V LiPOFE4 battery for a 24V configuration…
Will eventually put me at that 12V draining the batteries closer to zero…
So until I get to 14% battery capacity left I will be at 25 - 1 = 24V at the “lamp level”.
Maybe not a good way of doing it??
Doing 36V instead would at least give me 36.5V by stacking three batteries and reducing one driver per lamp…
A buck/boost driver would have been better but more complicated,
Reading up on the LM3409 again since I kind of parked the project back in late October early November time-frame.
9.1.2 Operation Near Dropout
Because the power MOSFET is a PFET, the LM3409/09HV can be operated into dropout which occurs when the
input voltage is approximately equal to output voltage. Once the input voltage drops below the nominal output
voltage, the switch remains constantly on (D=1) causing the output voltage to decrease with the input voltage. In
normal operation, the average LED current is regulated to the peak current threshold minus half of the ripple. As
the converter goes into dropout, the LED current is exactly at the peak current threshold because it is no longer
switching.
It is indeed often recommended to not parallel LEDs, with two LEDs with a Vf difference the one with lower Vf will draw more current than the other, which means it will heats up more which in turn decreases its Vf, drawing even more current…etc. In practice it’s not really an issue with high power LEDs, there isn’t much Vf variation in the first place and unless overdriven near their limit there is a lot of room for some current imbalance, the Vf also increases substantially as current increases (with XHPs at least), counteracting heat Vf decrease. Many high power multi LED flashlights use parallel LEDs, an Acebeam X50 for example has 12XHP70.2 in parallel (6V). Quite a few LED bulbs I disassembled had a combination of series and parallel configurations, an XHP LED actually has 4 LEDs in a single package, wired in 4S (12V) or 2S2P (6V), though they might be binned to have less Vf variation and it’s probably better to series 2 6V XHP than parallel 2 12V ones. If it can simplify your circuit then IMO it’s a viable option.
I wonder why it needs such high inductance, 68uH is quite high and indeed it’s a problem since it needs to be huge for the current/saturation rating needed.
Your chart shows the rest voltage, the voltage of the battery under load will be lower, how much depends on the battery internal resistance, which comprise of the cell’s DCIR and battery management system (BMS : overvoltage/undervoltage/overcurrent protection) circuitry resistance if there is one.
For example a LiFePo4 cell like this one, has a voltage of ~3.3V with a 0.2A draw at 50% SOC, but it drops to 3V at 20A draw, which would be 12V in a 12V(4S) battery, and this is a high power density cell (very low DCIR).
Then there is the Vf at the desired current, which is going to be higher than the ”6V” or ”12V” nominal Vf, see this test of the XHP70.2, for example at 6V10A the actual Vf is 6.63V (13.26V at 12V5A).
So even with the 100% duty cycle capable LM3409 you wont be able to drive your LEDs at your needed current with a x nominal battery voltage to x nominal Vf. With 24Vin, 1x12V or 3x6V LED would be the maximum, with 36Vin, 2x12V or 5x6V.
Really valuable input @thefreeman! The XLS is part of my links for the project. Just having a hard time getting my head back to the state it was last year…
Building a highbay lighting mast or a simple flashlight isn’t just hooking up power to the light source
Ok, I will continue investigating a higher input voltage in respect to the output voltage. And look at that inductor calculation to find a solution…
Note to self: Consider an output capacitor… Have two 1206 unpopulated output cap footprints in parallel prepared …
9.1.5 Buck Converters With Output Capacitors
A capacitor placed in parallel with the LED load can be used to reduce ΔiLED-PP while keeping the same average
current through both the inductor and the LED array.
With an output capacitor, the inductance can be lowered, making the magnetics smaller and less expensive.
Alternatively, the circuit can be run at lower frequency with the
same inductor value, improving the efficiency and increasing the maximum allowable average output voltage.
I see yes that makes sense, indeed without output capacitors it needs more inductance to keep the ripple current through the LED relatively low, with enough output capacitance you can afford to go up to 40~50% inductor ripple current (too high and there are too much AC/core losses).
Because TI doesn’t provide any calculator tools for the LM3409 I’d suggest downloading LTpowercad, opening an Async buck controller DCDC design (like LTC3894) and play with the components/values.
Here’s one I quickly played with if you want to start with it : async_buck test.ltpc - Pastebin.com , copy that in a text file and rename the extension to .ltpc
It will give you the efficiency, power losses (in each components) with models of real components, also the output voltage ripple, which in the saved file (200kHz, 36Vin, 26Vout 5Aout, 15uH, 2x1206 10uF) is less than 400mV, which looking at the XHP70.2 graph should be less than 0.5A LED ripple current if I’m not mistaken (2S 12V).
Another thing I wanted to mention is that the LM3409 can do analog dimming, this is a more efficient LED dimming method, you can still use PWM to control it, with a voltage divider and low pass filter to get a smoothed voltage to control the IADJ pin.
Thanks for the suggestion to use LTpowercad will have a look at it tomorrow!
The LM3409 datasheet has an example where they use output caps and provide the equations.
Seems mostly connected to frequency and ripple … And it’s a Co-Min that is being calculated so maybe not all that complicated. Not sure what oversizing that cap would lead to. Ti example simply goes from calculated 1.27uF to standard 2.2uF.
Don’t have any 2.2uF X7R 50V 1206 at my disposal. All my assortments boxes of cap’s is 0402, 0603 and usually 16V …
Since I order from Mouser I’m gonna need to hit £40 in total for my order. So throwing in a few different values to support a few potential configurations. Both support higher current and voltage. Probably should order some 6V Cree LEDs as well to maybe test three in series to get 18V for my 24V battery config…
My initial plan was to throw roughly ~50-60 Watt per LED before lumen/watt tapers off.
Did you mean high temperature? I read somewhere that almost half the capacitance could get lost. My board dosen’t appear to get very hot looking at it with my FLiR. But I did add your suggested cap!!
You live you learn! Great support @thefreeman truely awesome!
But if I step up from 35V → 50V that could potentially work as well.
I will read up on the topic later!
Searched for bias in LM3409 since I read somewhere about temp…
When selecting a ceramic capacitor, special attention must be paid to the operating conditions of the
application. Ceramic capacitors can lose one-half or more of their capacitance at their rated DC voltage bias and
also lose capacitance with extremes in temperature.
can you describe what you are trying to do?
light the ocean from the shore?
wouldn;t it blind you?
is it just to spot waves, then go out?
or to see while you are surfing?
) but thought paralleling LEDs had the potential of thermal runaways. Not sure how that would work on a single channel and one “LED strip” broke.
