Inductor core loss not calculable?

[eem2am]

Hello,

This post is long but simple…….it just concerns redesign of a Switch Mode LED Driver and a potential problem brought about due to increased flux excursion in the new inductor.




Our design contractors have designed for us a Boost converter LED driver of the following spec:-

V(in) = 12V
I(out) = 450mA
(-but LEDs are in parallel so each individual LED current is 150mA)
V(out) max = 33V
Continuous conduction mode
Switching frequency = 983KHz.
PCB size = 11mm by 50mm
PCB = double sided.

PCB is potted into a cyclindrical moulding.

The LEDs are in series strings of 8 LEDs, and there are three of these in parallel. (24 LEDs in total)

The LEDs are Advanced Power TopLEDs (LB G6SP) by osram-os

LED datasheet:-
http://catalog.osram-os.com/catalog...Oid=000000020000fa02005e0023&act=showBookmark

LPS6225-103ML inductor datasheet
http://www.coilcraft.com/pdf_viewer/showpdf.cfm?f=pdf_store:lps6225.pdf



Each LED string has a 10R, 1206 resistor in series with it to equalise the parallel LED currents.
Unfortunately , each 1206 resistor dissipates 225mW even when the LED currents are perfectly matched…….if the LED currents become more mismatched then these resistors are going to be seriously over-stressed. (over-powered)

Therefore,

We wish to re-design this boost converter with all the LEDs in series, and throw away the series 10R resistors.

This means, the new spec becomes:
V(in) = 12V
I(out) = 150mA
V(out) max = 100V


(Obviously we must use a different control chip because the LT3477’s internal FET is only rated to 42V.)

Now the problem comes because we can not now use continuous conduction mode. This is because the duty cycle would be 0.88, and almost all control chips won’t provide a duty cycle that high. Some controllers will provide that high duty cycle, but their internal slope compensation is not sufficient to be able to handle that high duty cycle.
-We could use an external slope compensation network, but external slope compensation networks do not work well at high frequency (we need to use high switching frequency due to the small PCB size)
-We could use a voltage mode chip, but unfortunately there are very few of these about (there is UCC35705, but it has no on-chip FET driver, and no internal error amplifier).

Anyway, we therefore have decided to use a discontinuous mode boost converter to do it.

The current peak is obviously higher for DCM boost, and this means that the flux excursion will be greater in the inductor, ……….and since we have to operate at high frequency, we are worried about the inductor overheating.

We have decided to use Coilcraft XAL6060-103ME , 10uH inductor

XAL6060-103ME inductor datasheet:
http://www.coilcraft.com/pdf_viewer/showpdf.cfm?f=pdf_store:xal60xx.pdf


We have dropped the switching frequency down to 295 KHz, to try and keep the inductor cool (inductor core loss)…but as you can note, its still a high switching frequency.

However, the core loss calculator on the coilcraft website tells us that it cannot calculate our core loss because our inductor’s peak current (3.07A) is more than 50% above the inductor’s RMS current (1.62A).

The inductor datasheet does not tell us what type of ferrite is used, and there are no core loss curves, therefore we have no way of knowing what the core loss will be……we have only one option…..to build it and test it and measure inductor temperature……but this will be expensive if it fails.




Anyway, do you think that this XAL6060-103ME inductor will suffer massive core loss in our DCM Boost application?

 

[shrtrnd]

Contact Coilcraft and ask them about your application.
They've got some pretty sharp engineers there who know their products.
Everytime I've asked them something about one of their devices, they've got very
exact very specific information about their product.
I'm kind of surprised the XAL6060-103ME data sheet doesn't have what you need, but
the Coilcraft engineers will.
Good luck with the redesign.

 

[Electrobrains]

Hi eem2am

You have got an interesting project!

...Now the problem comes because we can not now use continuous conduction mode. This is because the duty cycle would be 0.88, and almost all control chips won’t provide a duty cycle that high. Some controllers will provide that high duty cycle, but their internal slope compensation is not sufficient to be able to handle that high duty cycle.
-We could use an external slope compensation network, but external slope compensation networks do not work well at high frequency (we need to use high switching frequency due to the small PCB size)
-We could use a voltage mode chip, but unfortunately there are very few of these about (there is UCC35705, but it has no on-chip FET driver, and no internal error amplifier).

Anyway, we therefore have decided to use a discontinuous mode boost converter to do it.

The current peak is obviously higher for DCM boost, and this means that the flux excursion will be greater in the inductor...

Some thoughts:
Why using discontinuous mode and the resulting over dimensioned inductor? You have only 150mA and (probably much less than) 100V output. Voltage ratio 100V/12V=8.3. That should theoretically allow for the use of a 1.5 to 2A rated inductor (not much considering voltage drops and efficiency factors).

I would use a current mode controller, running in continuous mode. I love the Unitrode 38 series. They can be used in many different applications. They are cheap, easy to use and very general.
(Without searching for newer types) I would use UC3843 (up to 450kHz) or UCC3803 (up to 1MHz) together with an external MOSFET and a low-cost inductor.
Those IC's can run duty cycles up to 100% (and if necessary, could easily be slope compensated).
I would use a normal step-up (boost) topology.

See circuit diagram! This is a minimum, "bare bone" diagram, that probably would need several added components (MOSFET gate resistor, feedback filter capacitors, input electrolytic cap etc.).
Important: The combined LED voltage must exceed the supply voltage (in this case 12V)!
I have not tested the circuit!

R1/C1 set the frequency, C2/R3 the system reaction speed and gain, R4 set the LED current, R2 set a max current on the input side, protecting T1.
I am not sure if C4 is really needed in this configuration, but I suppose it is. In any case C4 should be installed at a bit higher frequencies, to remove EMI/RFI disturbance!
Z1/R6 should at least be installed when testing (and strongly recommended always) for limitation of max output voltage. This circuit can produce very high and dangerous voltage if the output is left open. It would only be limited by the breakdown avalanche voltage of T1.

Theoretically, many more LED's could be put in series, but again, it's a matter of security. Be aware of that the voltage in this thread (100V) has to be well isolated!
EN60950 specifies that a voltage above 42VAC and 60VDC need isolation and protection against touch!
To be secure, assure that the air and creepage distance to any conducting part is at least 3mm (better 5.5mm)!

For everybody: Don't play with this circuit unless you have the full knowledge of what you are doing.