My Problem
I have designed and built a LM25085 circuit that nominally produces a 19V output. The design has about 100mV-150mV of ripple on the output voltage and uses the "reduced ripple configuration" for the feedback circuit. Under enough load to make sure the inductor is continuously conducting, the output voltage is higher than the designed voltage by 1.0 to 1.5V. However when there is no load, the output voltage is the correct value. My understanding of why this occurs is based on the fact that the LM25085 is regulating the output voltage based on the minimum value of the feedback voltage ripple, not the average value. Unfortunately I don't have a good idea of how to fix this problem.
My Understanding of what is causing the problem:
According to the LM25085 datasheet, the device uses a constant time on-pulse. The on pulse is produced when voltage on the feedback pin drops below 1.25V. The feedback pin requires 25mV of voltage ripple to produce stable switching output.
The datasheet proposes 3 ways of producing enough output ripple:
1. Minimum ripple configuration (lots of extra components)
2. Reduced ripple configuration (1-extra small cap on feedback voltage divider and at least 25mV ripple on output voltage)
3. Lowest cost configuration (enough ESR on output cap to produce 25mV on ripple on feedback even after being attenuated)
Lets assume I was using the "Lowest cost configuration". If I wanted a 2.5V output, the voltage divider would attenuate the output voltage by 1/2. The output voltage would need to have 50mV of ripple. This would mean the output voltage would be 2.5V when the ripple hit its minimum. However the average output voltage will be 25mV (50mV/2) higher than 2.5V.
A 25mV offset doesn't seem so bad, but the situation gets worse with more attenuation. Suppose I want a 25V output. The feedback divider will attenuate the signal by 1/20. The output voltage ripple will be 500mV, and the average on the output voltage will be 250mV higher than it should be. This will be the minimum offset, since the design should guarantee at least 25mV of ripple on the feed back.
An offset on the output doesn't seem so bad, given when the ripple voltage is twice that. But 0.5V on the output is slightly ridiculous, and the "minimum ripple configuration" or "reduced ripple configuration" seems more appropriate.
If a "reduced ripple configuration" there is a capacitor in parallel with feedback resistor. The capacitor acts like a highpass filter, to pass more of the output ripple back to the feedback pin. Thus you need only 25mV of output ripple no matter what the feedback attenuation is. I could be wrong, but, the feedback capacitor should change the average on the feedback voltage, but it will push apart the peak and valley of the voltage ripple. This implies that if the bottom of the feedback voltage ripple is -12.5mV below the average it will push up the output voltage by 12.5mV / attenuation. This would produce the same amount of output voltage offset as the "lowest cost configuration".
My Questions
First, is my understanding of the issue correct?.
Second, is there a good way to fix this issue at both light and heavy loads? I'm wondering if a "lowest ripple configuration" will help, or have the same issue as the two other configurations.