LTC6902

14

6902f

APPLICATIO S I FOR ATIO

generator is driven by the master oscillator frequency, not

the output frequency. This gives some design flexibility in

setting. When making the choice, usually the faster mas-

ter oscillator is the better choice. This is especially true

when the main goal is to lower peak radiated or conducted

signal levels measured during regulatory agency testing.

Regulatory testing is done with strictly specified band-

widths and conditions. Modulating faster than the test

bandwidth or as close to the bandwidth as possible gives

the lowest readings. The optimal modulating rate is not as

straightforward when the goal is to lower radiated signal

levels interfering with other circuitry in the system. The

modulation rate will have to be evaluated with the specific

system conditions to determine the optimal rate. Depend-

ing on the specific frequency synchronization method a

switching regulator employs, the modulation rate must

be within the synchronization capability of the regulator.

Many switching regulators use a phase-locked loop (PLL)

for synchronization. For these parts, the PLL loop filter

should be designed to have sufficient capture range and

bandwidth.

Even when running the LTC6902 at the maximum modu-

lation rate, the frequency hopping transitions are slowed

by the part’s servo loop. The frequency transitions are

slowed by a 25kHz lowpass. This is an important feature

when driving a switching regulator. The switching regula-

tor is itself a servo loop with a bandwidth typically on the

order of 1/10, but can vary from 1/50 to 1/2 of the

operating frequency. When the input clock frequency’s

transition is within the bandwidth of the switching regula-

tor, the regulator’s output stays in regulation. If the tran-

sition is too sharp, beyond the bandwidth of the switching

regulator, the regulator’s output will experience a sharp

jump and then settle back into regulation. If the bandwidth

of the switching regulator is sufficiently high, beyond

25kHz, then there will not be any regulation issues.

One aspect of the output voltage that will change is the

output ripple voltage. Every switching regulator has some

output ripple at the clock frequency. For most switching

regulator designs with fixed MOSFETs, fixed inductor,

fixed capacitors, the amount of ripple will vary some with

the regulator’s operating frequency (the main exception

being hysteresis architecture regulators). An increase in

frequency results in lower ripple and a frequency decrease

gives more ripple. This is true for static frequencies or

dynamic frequency modulated systems. If the modulating

signal was a triangle wave, the regulator’s output would

have a ripple that is amplitude modulated by the triangle

wave. This repetitive signal on the power supply could

cause system problems by mixing with other desired

signals and giving a distorted output. Depending on the

inductor design and triangle wave frequency, it may even

result in an audible noise. The LTC6902 uses a pseudoran-

dom noise-like modulating signal. This results in the

regulator’s output ripple being modulated by the wideband

pseudorandom noise-like signal. On an oscilloscope, it

looks essentially noise-like of even amplitude. The signal

is broadband and any mixing issues are minimized. Addi-

tionally, the pseudorandom signal repeats at such a low

rate that it is well below the audible range.

The LTC6902 directly drives many switching regulators.

The LTC6902 with the spread spectrum frequency modu-

lation results in improved EMC performance. If the band-

width of the switching regulator is sufficient, not a difficult

requirement in most cases, the regulator’s regulation,

efficiency and load response are maintained while peak

electromagnetic radiation (or conduction) is reduced.

Output ripple may be somewhat increased, but its behav-

ior is very much like noise and its system impact is benign.