MIC4950 Datasheet, PDF (13/18 Pages) Micrel Semiconductor – Hyper Speed Control 5A Buck Regulator

Micrel, Inc.
At the higher currents for which the MIC4950 is designed,
efficiency loss is dominated by MOSFET R_DSON and
inductor losses. Higher input supply voltages will increase
the gate-to-source threshold on the internal MOSFETs,
thereby reducing the internal R_DSON. This improves
efficiency by reducing DC losses in the device. All but the
inductor losses are inherent to the device. In this case,
inductor selection becomes increasingly critical in
efficiency calculations. As the inductors are reduced in
size, the DC resistance (DCR) can become quite
significant. The DCR losses can be calculated as in
Equation 4.
P_DCR

I
2
OUT
 DCR
Eq. 4
From that, the loss in efficiency due to inductor DCR and
core losses (P_CORE) can be calculated as in Equation 5.
Ef
f
iciencyLoss (%)


1




V_OUT
V_OUT  I_OUT
 I_OUT  P_DCR 
P_CORE


 100
Eq. 5
External Ripple Injection
The MIC4950 control loop is ripple-based, and relies on
an internal ripple injection network to generate enough
ripple amplitude at the FB pin when negligible output
voltage ripple is present. The internal ripple injection
network is typically sufficient when recommended R1-R2
and C_F values are used. The FB ripple amplitude should
fall in the 20mV to 100mV range.
If significantly lower divider resistors and/or higher C_F
values are used, the amount of internal ripple injection
may not be sufficient for stable operation. In this case,
external ripple injection is needed. This is accomplished
by connecting a series R_inj-C_inj circuit between the SW
and the FB pins, as shown in Figure 1.
MIC4950
Figure 1. External Ripple Injection
The injected ripple is calculated using Equation 6,
ΔV_FB(pp)

V_IN
 K_div
 D  (1- D)
1
f_SW  
Eq. 6
with K_div given by Equation 7
K _div
 ^R1//R2
R_inj  R1//R2
Eq. 7
and:
V_IN = Power stage input voltage
D = V_OUT/V_IN = Duty cycle
f_SW = Switching frequency
= (R1//R2//R_inj) × C_F
In Equations 6 and 7, it is assumed that the time constant
associated with C_F must be much greater than the
switching period:
1 _ T _{ 1}
f_SW   
Eq. 8
March 20, 2014
13
Revision 1.1