 |
 |
|
|
[An E-mail exchange. Reads from the bottom up.] Chuck: Line rejection is best achieved via simple output voltage feedback! This is a little counterintuitive, since one waits until the load voltage is actually disturbed by a line transient before doing anything about it. But more seemingly natural approaches turn out to have shortcomings. For example, feed forward techniques are nonlinear! If you remember your math, nonlinear equations are far more difficult to solve [superposition doesn't work, for one thing.] In practice, this means weird nonlinear solutions can exist. These show up as unexpected oscillations, unexpected responses to line and load transients--in short, unpredictable behavior. Unfortunately, Current Mode Programming works much the same. One is attempting to force the switch current to be a desired value, rather than allowing it to rise to be so. Though CMP can be linearly modeled (as can feed forward) it suffers from the duty cycle saturation problem I mentioned before: 1.) When a load transient occurs, the error voltage from the op-amp calls for a markedly different current than is currently flowing through the switcher's inductors. But the current in them can only rise so fast, due to di/dt = V/L. So the pwm hits Dmax or Dmin [possibly zero, or even unity in an unisolated Buck, for example.] But this means that the converter is unregulated, i.e. running open loop at a fixed duty cycle. Under these conditions CMP no longer reduces a switching topology to a first order system, and there is the danger of ringing = hi Q resonant behavior. 2.) When a line transient occurs, the effect is similar. If the input voltage suddenly rises, Dmin may allow the input current to rise above the current level called for by the error amp. This means that the switch turns off immediately after Dmin is reached, regardless of the exact control voltage from the error amp, and the converter is running open loop, at fixed duty cycle. The same is true if Vg falls suddenly. The di/dt may be too small to return the input inductor current to its required value. Thus D is fixed, and the converter returns to being a second order system. If it is not damped, it will ring. Perhaps the saving grace of voltage feedback is that the converter, with its input filter, yields a fourth order low pass filter even without feedback. Thus, if damped, little other than low frequency feedback is necessary, and a good loop produces remarkable results, to which I can testify myself! [whew!] So, the best answer I can give is that large banks of input or output capacitors are not necessary for good regulation with the boostbuck converter family using voltage feedback. The requisite performance comes from a well designed loop and shunt damping. [to shunt damp a converter, click on the Greek column icon at www.boostbuck.com and poke around 'til you find the right page;)] If lots of capacitance is required for hold up purposes, it can be used as the blocking cap Cd. If Cd becomes unreasonably large due to sloth on the part of the AC line or DC battery guys, push some of Cd into parallel with the energy transfer cap C. The esl of the electrolytics will mean that little of the switching current will pass through them anyway, and they will lower Ro1, so ro1 can be lowered. This improves efficiency when dropouts call on them to supply energy until Vg returns. Regards, Bill WIB, I guess what is mean is the output voltage to line voltage transfer function. Chuck -----Original Message----- From: WILLIAM BEHEN [mailto:billbehen@msn.com] Sent: Sunday, July 20, 2003 1:34 AM To: csamps01@harris.com Subject: Re: WEBSITE REQUEST Chuck: Only a partial response. When you say ripple rejection, do you mean line rejection? I am not sure if you mean 60 Hz (source) disturbances or load current variations, or triangular switching ripple. WIB Your Name: Chuck Sampson Company Name: Harris Corporation Address: 3680 Big Pine Road City: Melbourne State: FL Zip: 32934 Country: USA Phone: Fax: E-mail: csamps01@harris.com Comments: How do you meet the requirements for ripple rejection in a voltage mode Cuk? Doesn't current programming improve line rejection? I am using cpm in a small 1 Watt Cuk for a FET bias. The failure mode for cpm you describe on your web page seem to me to only occur when the switching frequency is less than 2 times the response of the converter. Why else would the current not rise up in time? Why does using cpm increase the amount of output capacitance? How does using VM control reduce output capacitance? I am sure you are correct about what you say, I just don't understand why. Thanks for the interesting web page. Chuck Sampson | Beef Eh! | Background | Tips | Engineering vs. Science | | Isolation of the Cuk Converter | Historical Perspective | | Welcome | 50K! | | Return Home | Safety First | Lesson in Ethics | The Consequences of Bad Design | Real World Applications | Deals | e-mail me! | |
||
 |
 |
