MOSFET Driver Supply Rail Problem

[Jepoy Baduria]

Hi all,

I am building a dc-dc boost converter with a UCC27254 (up to 5A peak current source/sink) to drive the low side mosfet. My input voltage is 20V at minimum, and switching frequency is 100khz. I am using an LM7815 linear regulator directly connected to the input voltage, to generate a fixed 15V to power the mosfet driver. I have attached the schematic of the open loop converter. The n-channel mosfet is an IRFB3077.

I just realized that the LM7815 cannot supply more than 2.2A (as the datasheet says), so I'm worried that the mosfet driver cannot source/sink the 5A it was supposed to, even though this is only transient current.

Should I place parallel LM7815s to increase the current capacity? Or should I just increase the bulk capacitance (C10)? I thought of simply replacing the regulator with an LM338 linear regulator (5A continuous, 12A transient), but this IC is too costly for my budget.



Any help is appreciated. Thanks!

 

[FvM]

The peak current will be exclusively sourced by filter capacitors. There should be no problem to supply the circuit through the 1A 7815 regulator. It might be reasonable to supplement the small 0.1 uF ceramic capacitors by 1 µF or larger ceramic capacitors for better peak current bypassing. The 100 µF electrolytic capacitor isn't particularly effective in this current and time range.

 

[Jepoy Baduria]

Why is the 100uF electrolytic capacitor not effective at the 100khz range?

 

[chuckey]

Its because the capacitor is made from two metalized films which are rolled up put into a can with the gel electrolyte. Because some of the current has to flow through this COILED up film, there is inductance, so at high frequencies it stops acting as a capacitor and more like an inductor. If they worked at all frequencies there would be no need to have the .1 MF caps. A different type of 100 MF might work OK, see the manufacturers data sheet. I would use a 1 MF polycarbonate cap.
Frank

 

[FvM]

Due to relative large ESR and possibly ESL.

 

[Jepoy Baduria]

I see. So the 1uF ceramic capacitors possibly have overall lower impedance at the frequencies concerned, due to lower parasitics. And this translates to better sourcing/sinking of very fast current transients.

But from the point of view of electrical charge that needs to be supplied per switching instant, how would you size the capacitance so that the power line ripple is an acceptable level? Inserting a larger capacitor than the 0.1uF ceramic is certainly intuitive, because this means more charge can be stored and supplied when the load demands it. Is there a predictive way to size how much capacitance is needed, given the specs of the application? Or is this more suitable for trial-and-error?

 

[Rajnaveen]

use 47uf parallel with .1uf ceramic . it will help you

 

[chuckey]

If you consider the circuit as +15V feeding via its internal impedance then the switch impedance of the UCC27254, then the 10 ohms feeding the G-S capacitance of Q1 (Cg).
The max charge current is 15v/10 ohms = 1.5A if every thing else is perfect. Say it decays to nothing in 4 X Cg X R 10 Ohms. This happens at the PRF. So the mean current can be found. As the max value is 1.5A and this is below the current rating of the UCC27254 and the LM 7815 their rating are OK. As a working approximation if the PSU time constant is 100 times the Cg X 10, then the ripple should be able to be ignored. The real problem is the Di/Dt and spurious inductance. I reckon its a good design! .
Frank