Electronics > Projects, Designs, and Technical Stuff
Ideal buck converter for high current
NiHaoMike:
--- Quote from: Pawelr98 on August 04, 2020, 01:56:29 pm ---The only problem I see is the possible kickback which can overvolt the capacitors.
--- End quote ---
It would take a very large inductor to approach the energy storage of the capacitors in question.
anishkgt:
--- Quote from: Siwastaja on August 04, 2020, 07:57:54 am ---Note that ending voltage of 8V means the average voltage is at 4V, which can be used for a rough loss calculation.
So while at 8V you could theoretize going without synch rectification, at 4V average it's almost a must.
Indeed, building it in single phase requires massive components and heatsinking, which easily blows your layout loop area up.
The more phases you use, the smaller the loops, and the easier cooling becomes; you can safely increase the distance between the phases (not going to extremes, of course, or you lose the benefits of shared DC input bus capacitance) to spread the heat.
These smaller transistors in smaller loops enables you to use higher frequency, which enables you to use smaller inductance values; making the off-the-shelf inductor selection current ratings go up!
As you have noticed, going below about 20A/phase gets you a wide selection of cheap off-the-shelf inductors.
Note you need a control IC that can do unlimited current-mode. Proper CC-CV would be optimal, but assuming you don't need good CC regulation, just pick one which limits current below the ratings of your components and doesn't go in a hickup or latched overcurrent modes.
I agree 4 phases would be the minimum. Some 2-phase control ICs have sync outputs phase-shifted by 90 degrees so you can built a 4-phase thing with two such chips. With more than 4 phases, the options for control ICs get more scarce.
--- End quote ---
I am looking at the ADP1850 (https://www.analog.com/media/en/technical-documentation/data-sheets/ADP1850.pdf) a dual phase buck-converter. 4 phase does not have gate drivers and the explanation is not very well documented so a beginner with buck-converters like find it hard to follow along. I could use ADP1850 in parallel or interleave, if its possible, to get what i want. I would just need 80A and CC at 100A a soft-start of couple milli seconds.
On page 16 Tss is calculated but does not specify the unit of time used. Can somebody please confirm it. On LTC3824 datasheet on page 9 they talk about Css calculation and the unit used is milliseconds so i assume its the same here as well but just not sure.
Jay_Diddy_B:
Hi,
I think that you will find the ADP1850 difficult to use in practice. It is a current mode controller, (which is good), but it measure the voltage drop across the bottom MOSFET to regulate the current.
This means:
1) the current limit will vary with the junction temperature of the MOSFETs
2) It make the layout extra hard, because you need to measure a small voltage in a circuit with high currents and rapidly changing currents.
I would suggest looking for a Linear Tech Controller that uses current sense resistors in series with the output.
Regards,
Jay_Diddy_B
anishkgt:
--- Quote from: Jay_Diddy_B on August 04, 2020, 02:26:18 pm ---Hi,
I think that you will find the ADP1850 difficult to use in practice. It is a current mode controller, (which is good), but it measure the voltage drop across the bottom MOSFET to regulate the current.
This means:
1) the current limit will vary with the junction temperature of the MOSFETs
2) It make the layout extra hard, because you need to measure a small voltage in a circuit with high currents and rapidly changing currents.
I would suggest looking for a Linear Tech Controller that uses current sense resistors in series with the output.
Regards,
Jay_Diddy_B
--- End quote ---
It has an option to use a Sense resistor as well, mentioned on page 17 under the heading of 'ACCURATE CURRENT-LIMIT SENSING'
Siwastaja:
Using the Rds(on) sensing isn't necessarily a bad idea; it automagically turns down the current if the MOSFET is heating up!
It isn't accurate, obviously, but you don't always need accurate charging current, just ensure power dissipation below maximum ratings. And Rds(on) sensing tracks that pretty well, if you have a transistor with higher Rds(on) or higher ambient temperature, it's doing the right thing for you. Careful with current sensing layout, of course. With a separate resistor, it's easier to do Kelvin sensing, but OTOH, separate resistor increases the loop area again.
ADP1850 doesn't have 90deg phase-shifted SYNC OUT - some ADI parts do have, I remember using one some time ago - so in order to have more phases than two, you would need an external clock generator. Any circuit which provides f_sw*2 with two outputs, 90 deg phase shift to each other would work. You could use them to drive two ADP1850's. ADP1850 internally adds 180 deg phase shift, so this would result in all four phases 0, 90, 180, 270 deg apart.
I would likely end up with 4 phases doing this, but well, do the calculation, including efficiency and price. My guesstimate is that with 4 phases, you can use the most popular commercial off-the-shelf parts for a simple BOM, but with 2 phases you may have trouble finding the inductors.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version