LTC3703
applicaTions inForMaTion
Feedback Component Selection
If breadboard measurement is not practical, a SPICE
simulation can be used to generate approximate gain/
phase curves. Plug the expected capacitor, inductor and
MOSFET values into the following SPICE deck and gener-
Selecting the R and C values for a typical Type 2 or
Type 3 loop is a nontrivial task. The applications shown
in this data sheet show typical values, optimized for the
power components shown. They should give acceptable
performance with similar power components, but can be
way off if even one major power component is changed
significantly. Applications thatrequire optimized transient
response will require recalculation of the compensation
valuesspecificallyforthecircuitinquestion.Theunderlying
mathematics are complex, but the component values can
be calculated in a straightforward manner if we know the
gainandphaseofthemodulatoratthecrossoverfrequency.
ate an AC plot of V(V
)/V(COMP) in dB and phase of
OUT
V
in degrees. Refer to your SPICE manual for details
of how to generate this plot.
OUT
*3703 modulator gain/phase
*2003 Linear Technology
*this file written to run with PSpice 8.0
*may require modifications for other
SPICE simulators
*MOSFETs
rfet mod sw 0.02
*inductor
;MOSFET rdson
lext sw out1 10u
rl out1 out 0.015
*output cap
cout out out2 540u
resr out2 0 0.01
*3703 internals
emod mod 0 value
;inductor value
;inductor series R
Modulatorgainandphasecanbemeasureddirectlyfroma
breadboardorcanbesimulatediftheappropriateparasitic
values are known. Measurement will give more accurate
results, but simulation can often get close enough to give
a working system. To measure the modulator gain and
phase directly, wire up a breadboard with an LTC3703
and the actual MOSFETs, inductor and input and output
capacitors that the final design will use. This breadboard
should use appropriate construction techniques for high
speed analog circuitry: bypass capacitors located close
to the LTC3703, no long wires connecting components,
appropriately sized ground returns, etc. Wire the feedback
amplifier as a simple Type 1 loop, with a 10k resistor from
;capacitor value
;capacitor ESR
=
{57*v(comp)}
;3703multiplier
;ac stimulus
vstim comp 0 0 ac 1
.ac dec 100 1k 1meg
.probe
.end
With the gain/phase plot in hand, a loop crossover fre-
quency can be chosen. Usually the curves look something
like Figure 11. Choose the crossover frequency in the ris-
ing or flat parts of the phase curve, beyond the external
LC poles. Frequencies between 10kHz and 50kHz usually
work well. Note the gain (GAIN, in dB) and phase (PHASE,
in degrees) at this point. The desired feedback amplifier
gain will be –GAIN to make the loop gain at 0dB at this
frequency.Nowcalculatetheneededphaseboost,assum-
ing 60° as a target phase margin:
V
to FB and a 0.1µF feedback capacitor from COMP
OUT
to FB. Choose the bias resistor, R , as required to set the
B
desired output voltage. Disconnect R from ground and
B
connect it to a signal generator or to the source output
of a network analyzer to inject a test signal into the loop.
Measure the gain and phase from the COMP pin to the
outputnodeatthepositiveterminaloftheoutputcapacitor.
Make sure the analyzer’s input is AC coupled so that the
BOOST = –(PHASE + 30°)
If the required BOOST is less than 60°, a Type 2 loop can
be used successfully, saving two external components.
BOOST values greater than 60° usually require Type 3
loops for satisfactory performance.
DC voltages present at both the COMP and V
nodes
OUT
don’t corrupt the measurements or damage the analyzer.
3703fc
20
Linear Systems [ Linear Systems ]
