Loop Compensation

Since the LM1572 uses current mode control, the loop re-

sponse does not involve the inductor. The error amplifier can

be modeled as a transconductance amplifier with a large

output impedance of 200k of resistance in parallel with 12pF

of output capacitance. In practical applications, the imped-

ance of the external compensation network from the Comp

pin to ground dominates completely, and the error amplifier

characteristics do not contribute any significant phase shift

to the loop. Therefore the error amplifier can for all practical

purposes be considered simply as a 2000µMhos of

transconductance, the loop phase/gain being determined

externally.

The simplest recommended compensation is a 3.3nF ca-

pacitor from comp pin to ground. This provides a pole at

240Hz. The overall loop then has a low frequency gain of

about 62dB at 1.5A, with a crossover at about 15kHz, and a

phase margin of about 33˚. A resistor may be added in series

with

this

capacitor

to

improve

the

loop

phase

margin/crossover frequency. Recommended values for this

are 1k to 1.5k. If the loop response needs to be further

improved, by increasing the value of this resistor, then a

small capacitor of about 470pF is required across the RC. Its

purpose is to limit the ripple on the Comp pin to within

100mV

behavior of the LM1572.

Loop compensation must be further validated by a bench

measurement, using standard Bode plot/spectrum analyzer

equipment. Step load transient response can also be tested

and should not reveal excessive ringing on the output of the

converter.

Tolerance of set Output Voltage

This section may be skipped altogether, unless the designer

wants to get a more precise understanding of the possible

variation or ’spread’ on the output voltage and how this can

be controlled better.

This ’basic resistive divider design equation’ seems to sug-

gest that R

output voltage. For example, referring to the Typical Applica-

tion circuit, where the values shown are R

2.21k, it may have been thought that using the following

values: R

acceptable. But the simple equation is just that: an ’ideal’

equation that unrealistically assumes zero current into or out

of the feedback pin. It can be easily shown that the effect of

any ’real’ current, flowing into the feedback pin for example,

is to raise the output voltage slightly from the ’ideal’ calcula-

tion. This is considered to be an output voltage ’error’, and

this needs to be understood and quantified.

Now, as mentioned, had the selection been: R

R

pation in the resistive divider by a factor of 10), it would also

have increased the error in the output voltage by almost the

same factor. A compromise can always be considered if

efficiency at light loads is a key concern, but first it must be

understood how to actually design the resistive divider for a

certain (maximum) error.

As can be seen from the Typical Performance curves and

tables of Electrical Characteristics for the LM1572, a current

of about 0.5µA (typical value) flows into the feedback pin at

regulation (I

thought appropriate to ’modify’ the basic resistive divider

equation by simply modeling this current in by an ’internal

resistor’ between feedback pin and ground. Its value would

be 2.42V/0.5µA = 4.84M

Ω . In fact this would have been

acceptable had the current been a constant. But as seen

from the Electrical Characteristics, this current can be as

high as 1.5µA. This would mean that this internal resistor can

actually be 3 times lower i.e. 4.84/3 = 1.6M

Ω. Therefore, it is

not the feedback pin current itself, but its variability that

poses the problem: a single fixed known value of feedback

pin current can be easily modeled, not a spread of values.

The most direct approach to the problem is as follows: to

continue to use the basic (’ideal’) resistive divider equation

as the basic design equation, but to also use the equation

which ’models in’ a 1.6M

Ω resistor as representing the worst

case error, then to compare these two equations to calculate

what the error is. Conclusions on how to reduce this error

would follow.

So proceeding in this manner, the worst-case resistive di-

vider equation (modified to include 1.6M

Ω resistor in parallel

with R

Comparing this with the basic equation provides the follow-

ing error equation

Note that the right hand side depends on R

output voltage directly). So for a given maximum allowed

error, first R

equation is invoked and used to calculate R

The effect of R

If R

’+’ direction as mentioned earlier, for the case of current

flowing into the pin. The error is reduced to 0.14% for the

value R

general it can be concluded that to restrict the error to within

0.25%, R

If R

basic resistive divider design equation is 4.286k for a 5V

output and is 1.462k for a 3.3V output.

There is an alternative way of stating the error, in terms of

current rather than resistance. Since if R

flowing through the resistive divider is 2.42/4k=0.6mA, there-

fore it can also be stated that the divider current should be

0.6mA or less. This will restrict the error (due to the feedback

pin current and its variation) to less than 0.25%. This ’thum-

brule’ does not depend on input or output conditions, and is

typical for most applications.

The other related problem is that to get an exact value for R

1

or R

one of the two resistors can of course be selected to be a

standard value, but the other value as calculated from the

equation, will more likely than not, not correspond to any

standard value. The so called ’EIA standard values’ are the

E6, E12, E24, E48, E96 and E192 series, listed in most

resistor catalogs. E12 for example has 12 values in every

decade of resistance. The reader can for example do a

search within a typical vendor’s index home page e.g. http://

www.vishay.com/ with the keyword E96 for the table of stan-

dard values.

www.national.com

14