I initially didn't see the small capacitor that appears to be in series with one of your transformer windings, what value is it? In my experience no series capacitor is needed, so long as your signal generator as a truly low DC offset when it's set for 0V.
Not all your pictures loaded when I first responded, the included screenshots help clarify what's going on. If your loop-gain measurements were made from the VSENS+ pin to the actual output, that's probably fine up to 1MHz or so, which is where your bode plot ends. For good high frequency results makes sure the scope probe GND lead is short (ideally two spring-tip probes, if you happen to have more than two hands), but I do think the results you got are usable.
I would like to emphasize that precise transformer gain and phase characteristics are not at all important to this kind of measurement. The transformer's only job is to transfer some of the signal generator stimulus to the 10Ω injection resistor with galvanic isolation. As long as whatever it spits out is mostly the same frequency as the input, it's good enough for this job. Ridley engineering and the like like to toot their own horn for making a very expensive and very nice isolation transformer claiming ridiculous gain and phase flatness, but those features do not help these measurements at all! These bode plots are purely ratiometric. The absolute magnitude or phase of the stimulus is not important, it just needs to be enough to get the resulting signals above the noise floor of the scope, and not too much to push the converter into some nonlinear response mode. The only actually good thing about theirs is the low frequency range, which is only useful if you are working with mains PFC converters and need to measure loop gains with 0db crossovers well below 60Hz.
Your transformer-only gain phase plots look promising, but the seemingly tiny series capacitor and the same wire colors for primary and secondary make it hard to judge just from a picture. If you can label each wire in that picture, and maybe sketch the exact schematic, noting which wires are twisted, and which is connected to the signal generator and the 10Ω resistor, then I could confirm for you.
Regarding the actual bode plot you got, it looks reasonable enough for a voltage-mode converter (as apposed to peak current-mode). I actually think with the polarity you have, 0° as shown on your bode plot corresponds to -180° in terms of loop gain. I think that because you can see a clear near 180° phase drop around 10kHz, which means that the phase is moving in the direction we expect when it hits the buck inductor + output capacitor resonance (as apposed to moving up 180°).
The reason why these bode plots often come out inverted is because feedback is negative. You get the same thing if you simulate a converters loop gain in LTspice. Try inverting one of your channels and the phase should flip, or at least that works on my RTB2004. Otherwise just go with what makes sense. First make sure you have the input and output specified correctly, which you do. The gain generally decreases to higher frequencies, which we expect from the low-pass nature of the LC filter. If you flipped input and output, you'd get a wacky gain curve. After you have that checked, you know that no buck converter operates with positive loop phase, the LC lowpass always adds some delay, and the error amp always just barely recovers enough phase margin to remain stable. So you can be sure that the phases in your measured bode plot are off by 180°. No big deal, invert a channel, or just measure phase margin to 0° instead of -180°, either are equivalent and fine. That being said, having the phase appear correctly is nice if you have to document these plots for you work. A final sanity check would be the basic load step response, a test you should do anyways, even though a bode plot is way more helpful for tuning.
All that said, at the operating point you made that bode plot at, it appears you have about 55° of phase margin at ~70kHz and 20dB of gain margin at ~300kHz, not too bad. Since that control IC has fixed internal compensation, the only way you might improve that would be to increase your output capacitance or inductance, which would keep a similar phase-vs-frequency curve round 50+kHz, but would decrease gain at all frequencies above the resonant point by lowering the LC low-pass resonant frequency. You can see that phase peaks around 50kHz, so ideally your 0dB crossover would be at 50kHz and not 70kHz, but honestly you're pretty close already.
Hopefully that makes sense, and happy testing!
Home
Help
Search
Login
Register
