Problems with the AN57 document of Power Integrations

1 post / 0 new


The AN57 document gives an erroneous description of how to compensate a voltage mode flyback. AN57 suggests that a “type 2” compensator can be used to compensate a voltage mode flyback, but this is not correct. As stated on page 51 of “Power supply design volume 1: Control” by Dr Ridley, a type 3 compensator is needed for voltage mode converters operating in CCM. This is because of the severe phase dip which occurs due to the double pole of the power stage, and therefore two zero’s are needed  to offset this sharp phase drop. (there is only one zero in the type 2 compensator).

As Dr Ridley’s book states, the compensator needs to be a type 3 compensator, and one zero needs to be placed just above the power stage LC resonant frequency, and the other zero needs to be placed just below the power stage LC resonant frequency.

The “phase boost” network  mentioned in AN57 should surely always be used with any topswitch design. If it is not used, then one is deficient in the number of compensator zeros needed to compensate the power stage double pole. As you know, the “phase boost” network gives a zero at a frequency of  1/{2*pi*[RLED+RF4]*CF2}.

Also, the phase boost network also gives a pole at a frequency of 1/(2*pi*RF4*CF2)

In a typical voltage mode flyback, the compensator would require to have a pole placed at the ESR zero frequency, and another pole at about the RHPZ frequency.

Therefore, I conclude that the “phase boost” network of AN57 is always needed in a TOPswitch flyback design. If the “phase boost” network of AN57 is not used, then the compensation network is just a type 2 network, and that is insufficient to correctly compensate a voltage mode flyback. The only reason that the type 2 network  actually ever works with any topswitch flyback, is because as AN57 (page 12) confesses, there is a reliance on the fact that the flyback transformer will be wound with windings of such high ohmic resistance, that the power stage LC resonance is heavily damped, and therefore one can get away with using a type 2 network. This fact, coupled with the fact that the resonance frequency of LE*C is always made >500Hz, means that a type 2 network can be gotten away with in some cases.

 However, I would say that this is not reliable, and to improve the chances of stability in all cases, the type 3 compensator, involving use of the “phase boost” network (RF4 & CF2), should always be used with any topswitch flyback.

Another point, not a fault of AN57, but perhaps a loss of an advertising opportunity for Power Integrations, is that the advantageous method of feedback compensation of the TOPswitch means that the pole concerning the optocoupler base-collector region is totally insignificant in topswitch designs. This is a very good reason to use a TOPswitch. –Because the opto pole can be very variable even amongst the same batch of opto’s, but with a topswitch design, the feedback arrangement means that this has no effect on the feedback loop, since the opto pole is totally drowned out by the pole involving RF5 & CF3.

Another problem with AN57 is that it only discusses the use of the TL431 as the secondary side error amplifier (rather than eg using an actual discrete classical opamp integrator circuit.) The problem with the TL431 is that its cathode pin is also its power supply (Vdd) pin. This means that the RF3 resistor is limited in how high an ohmic value it can be before actually cutting off the supply current to the TL431 (TL431 supply current is typically about 1mA).

All of the schematics in AN57 are missing a resistor in parallel to the opto diode, which would be needed to prevent the TL431 bias current from passing through the opto-diode. However, such a resistor does not stop the problem that the TL431 bias current has to pass through RF3 without causing too high a voltage drop through RF3.For this reason, the TL431 circuit is limited, and the AN57 document  really needs augmenting to include the use of a discrete opamp error amplifier solution.

The problem with the “phase boost” network mentioned in AN57, is that even though it provides the type 3 response, the TL431 circuit given is not always good at giving the type 3 compensator. This is explained on page 297  of the book “Switch mode power supplies” by Christophe Basso. The problem with the “type 3” TL431 circuit given in  AN57 is that the RF3 resistor acts both in the gain definition and in the pole-zero position. Sometimes an RF3 value giving the right gain at the crossover frequency and enough bias in light load conditions can sometimes lead to an impossible solution. This can be solved by using a discrete opamp error amplifier solution. (of course, the TL431 can be used in such a manner, but not like in any of the AN57 app note schematics).

 Another point about AN57 is that it provides the equations so that someone can plot the graph for the type 2 bode plots. However, it doesn’t supply the equations for plotting of the type 3 bode plots.