Inductor in Boost Converter Getting Very Hot

[Muny]

Hi all...I've got a problem on my hands which has stumped me. I've designed a boost converter using parts I had lying around from other projects.  I've also modeled it in LTspice and it behaves in almost the exact same way as the model.  The problem is, however, that the inductor gets extremely hot.  According to the LTspice, the inductor is passing a current of 6.62 Arms. It is rated for 24 Arms at a temp rise of 20*C but the temperature has reached 100*C before I shut it off.  After some measurements, I have determined there are approximately 16W lost in the boost process.  This gives an efficiency of around 83.3% which is very typical as I understand.  The MOSFET and the diode do not rise more than 3*C, so I assume all of this power is being dissipated in the inductor.  Is there any concievable reason for this, or am I not understanding something about boost converters?

Schematic:

I wasn't able to find the exact model for the MOSFET (maybe this is the issue?) but I chose one with the closest characteristics.

My load is a 100W (@120V) light bulb.

The power supply I'm using is capable of 6A @ 32V, but it can handle surge currents...so I was also curious about what value, if any, I should use for the series resistance of that in the model.

The inductor: AGP4233-153ME http://www.mouser.com/ds/2/597/agp4233-774569.pdf

The MOSFET: FDL100N50F http://www.mouser.com/ds/2/149/FDL100N50F-1007460.pdf

The diode: STPSC20H065CTY http://www.st.com/content/ccc/resource/technical/document/datasheet/a4/98/dd/83/db/59/4a/9e/DM00123409.pdf/files/DM00123409.pdf/jcr:content/translations/en.DM00123409.pdf (both anodes are joined in the circuit)

The switching frequency is 17.96 KHz (I can raise this if need be, but I would not like to go any lower due to audible noise created by the inductor).

I have other inductors with higher inductance but somewhat lower current ratings, so I was hoping to be able to use the one I am now.  The ripple is ~1.15vp, so I may need to go higher unless there's another option like increasing the switching frequency?

If you need the LTspice model or any other measurements or information, I can most definitely supply it.  Any help would be much appreciated, thanks!

[chris_leyson]

Hi, had a look at the AGP4233-153ME data sheet and it wasn't very helpful and I got the feeling that this series of inductors might me intended for filtering applications, if that is the case then the ferrite is designed to be lossy i.e. a large B-H loop. You are well within the current rating for the inductor so I don't think current is the issue, however, the inductor could be saturating if you exceed the maximum volt time product. If you know the core area and number of turns you can work out the maximum volt time product for a given working flux density but none of this is quoted in the data sheet. If you've got a current probe check the inductor current for spiking at the end of the current ramp that will tell you if it's saturating. Alternatively use something like 0.1 ohm in the mosfet source and measure the current that way. Looks like it is either saturating or the ferrite is intended for filter aplications in which case it would run hotter but if I had to flip a coin so to speak I would go for saturation. A larger value inductor with more turns would give you a larger volt time product, I suspect the 15uH inductor you are using doesn't have very many turns so maybe start with something like 100uH. "Switchmode power supply design handbook" by Billings and Pressman is worth a read.

[Yansi]

It seems you just cobbled up a bunch of components and expect it to work. That aint gonna happen, not in this power electronics category, not at all.

Let's analyze it a little:  You need a 100W at 120V. Thats fine,  0.83A output current. Estimating the efficiency to 85% (just pulled out of my finger), the input power will be 120W, meaning input current will be 3.75A at 32V. Also fine. 

Assuming the converter shall work at CCM (continuous current mode through inductor), a duty cycle of about 80% is required.  Input voltage (32V) is applied for 80% of period of 17.92kH, that is for about 45us, the inductor current will rise for 96A.  Ninety six fucking amps! Thats your problem.  The inductor is extremely saturated. You be glad the mosfet didn't get blown into pieces.

So, to do it the correct way: Use higher switching frequency. Because we can and because 18kHz is hilariously slow for a 2017. So we choose 60kHz as the optimal value.
Now for a 30% current ripple through the inductor, the deltaI = 0.3 * 3.75A = 1.125A.
The duty cycle is calculated as  k = Uout / (Uin+Uout) for a boost converter in CCM, co we have the above mentioned 80%.
Using the stadard U = L * di/dt formula, we can calculate the reqiured inductor for this converter:
L = U * dt/di = U*k / (f*di) = 32 * 0.8 / 60000 / 1.125 = 380uH

Now you see, you have the inductance value quite substantialy out of reality. So either you have to get a proper inductor for it (standard value of 330uH will do absolutely fine, as the ripple current can usually be even 50% or more for ferrite core inductors)
or... you have to up your switching frequency VERY HIGH to be able to use your 15uH inductor, which I do NOT recommend, as it would require about 900kHz switching frequency (at 50% current ripple, and that possibly might overheat the inductor due to core loss as well) 

Bottomline: The above calculation of the boost converter is overly simplified. But it will get you at least some real component values.
Get yourself a 330uH inductor rated for 5Arms and at least 5A Isat, up the frequency to 60kHz and you'll be fine.

//EDIT: I've just noticed what crazy overkill you use for the mosfet and diode...  Just pure why...  200V mosfet rated for  10A like even the good old IRF640 has to do it, maybe you can select a more recent type with lower Rds, to keep the conductive power loss down. Diode rated for 200V 3A is more than enough. Select a one with low Qrr.

[chris_leyson]

Thanks Yansi, kudos  It's too late and I didn't feel like doing the math in my head. I'm surprised the fet didn't pop. 

[Muny]

Hi, had a look at the AGP4233-153ME data sheet and it wasn't very helpful and I got the feeling that this series of inductors might me intended for filtering applications, if that is the case then the ferrite is designed to be lossy i.e. a large B-H loop. You are well within the current rating for the inductor so I don't think current is the issue, however, the inductor could be saturating if you exceed the maximum volt time product. If you know the core area and number of turns you can work out the maximum volt time product for a given working flux density but none of this is quoted in the data sheet. If you've got a current probe check the inductor current for spiking at the end of the current ramp that will tell you if it's saturating. Alternatively use something like 0.1 ohm in the mosfet source and measure the current that way. Looks like it is either saturating or the ferrite is intended for filter aplications in which case it would run hotter but if I had to flip a coin so to speak I would go for saturation. A larger value inductor with more turns would give you a larger volt time product, I suspect the 15uH inductor you are using doesn't have very many turns so maybe start with something like 100uH. "Switchmode power supply design handbook" by Billings and Pressman is worth a read.

They advertise the product as a "Power Inductor," "designed for high current power supply applications."  They advertise another line of inductors as "Output power chokes," so I'm not sure what to make of what the former is intended for, but it sounds like it might not be the same.  I just checked, and yeah the inductor current is reaching almost 40A so it is not...quite... saturating.  Using a larger value inductor increases the duty cycle needed for a particular output voltage under load.  Towards the high end, the output voltage starts increasing exponentially so regulation wouldn't be easy due to a fixed step-resolution of the PWM input.  Thanks for the suggestion, I will see if I can get it somehow.

It seems you just cobbled up a bunch of components and expect it to work. That aint gonna happen, not in this power electronics category, not at all.

Let's analyze it a little:  You need a 100W at 120V. Thats fine,  0.83A output current. Estimating the efficiency to 85% (just pulled out of my finger), the input power will be 120W, meaning input current will be 3.75A at 32V. Also fine. 

Assuming the converter shall work at CCM (continuous current mode through inductor), a duty cycle of about 80% is required.  Input voltage (32V) is applied for 80% of period of 17.92kH, that is for about 45us, the inductor current will rise for 96A.  Ninety six fucking amps! Thats your problem.  The inductor is extremely saturated. You be glad the mosfet didn't get blown into pieces.

So, to do it the correct way: Use higher switching frequency. Because we can and because 18kHz is hilariously slow for a 2017. So we choose 60kHz as the optimal value.
Now for a 30% current ripple through the inductor, the deltaI = 0.3 * 3.75A = 1.125A.
The duty cycle is calculated as  k = Uout / (Uin+Uout) for a boost converter in CCM, co we have the above mentioned 80%.
Using the stadard U = L * di/dt formula, we can calculate the reqiured inductor for this converter:
L = U * dt/di = U*k / (f*di) = 32 * 0.8 / 60000 / 1.125 = 380uH

Now you see, you have the inductance value quite substantialy out of reality. So either you have to get a proper inductor for it (standard value of 330uH will do absolutely fine, as the ripple current can usually be even 50% or more for ferrite core inductors)
or... you have to up your switching frequency VERY HIGH to be able to use your 15uH inductor, which I do NOT recommend, as it would require about 900kHz switching frequency (at 50% current ripple, and that possibly might overheat the inductor due to core loss as well) 

Bottomline: The above calculation of the boost converter is overly simplified. But it will get you at least some real component values.
Get yourself a 330uH inductor rated for 5Arms and at least 5A Isat, up the frequency to 60kHz and you'll be fine.

//EDIT: I've just noticed what crazy overkill you use for the mosfet and diode...  Just pure why...  200V mosfet rated for  10A like even the good old IRF640 has to do it, maybe you can select a more recent type with lower Rds, to keep the conductive power loss down. Diode rated for 200V 3A is more than enough. Select a one with low Qrr.

Yep, that's about what happened 

Most of what you say is true, however it does not operate in continuous conduction mode. The current reaches less than 40A, though that is still high.  What are the benefits to having it operate in CCM?  Thanks for the run-through of the calculations, it should be useful.  My goal is to operate the converter between 100-300W of load.  Is this out of the realm of reality when using CCM?  It seems easy to find inductors rated for 330uH >=5Arms, but not for anything higher.

Like I said, this was thrown together with parts I had from past projects.  I will probably get more appropriate switching components in the future.

[Yansi]

it is out of reality to try designing this to work in DCM. The peak current will be very high, and the benefit of zero voltage switching and soft recovery of the diode is not worth it. You shall operate that in CCM, of course.

Also, operating in DCM means large flux swing in the core ---> higher power loss in there.  Not to mention you need a choke with very high saturating currents - usually aint cheap and it is hard to find a part with high Isat and significantly lower RMS rating.
The downside of CCM is that a bigger inductance is required. DCM is usually more efficient for low current. (it will work in DCM at light loads by itself, by the way)

If you have trouble finding suitable power inductors, what about designing your own? I never buy the "overpriced space grade whatever inductors" for such applications. A length of wire and a selection of different cores is the way to go, if you want to experiment with power conversion.

Note: 300W converter would require something like 100uH 10Arms  choke  @80kHz switching - which is I think very fine.  100uH 10Arms seems like a job for iron powder toroidal core #26 or better #52, size T150 or near that.