Take a converter that needs a Type III compensator. Change nothing but the output capacitor, from ceramic to an aluminium electrolytic of the same value, and it now needs only a Type II. Same inductor, same capacitance, same crossover frequency, same phase margin target. The type was never a property of the topology.

The rule of thumb says voltage-mode control needs a Type III and current mode can get by with a Type II. That is a fair description of what most people build. It is not what decides the question. The type is set by how much phase the plant has left at the crossover you asked for, and on a voltage-mode buck the output capacitor moves that number further than the control mode does.

This article comes before the compensator design series, which works a voltage-mode buck and its Type III through in detail. That series assumes the type is already settled. This one settles it, and the answer is not always the one the series goes on to build.

The rule, in one line

At the crossover frequency, the loop gain has magnitude one. Phase margin is whatever the loop phase is above $-180°$ there. Write that out:

$PM = 180° + \theta_{plant}(f_c) + \theta_{comp}(f_c)$

Rearrange for the only unknown you get to choose:

$\theta_{comp}(f_c) = PM - 180° - \theta_{plant}(f_c)$

That is the whole decision. Measure or model what the plant leaves you at the frequency you want to cross over, subtract, and you have the number the compensator has to supply.

Every type starts in debt. All three have a pole at the origin, because you want infinite DC gain and therefore zero steady-state error, and that pole costs a fixed $-90°$ at every frequency. Zeros are what buy phase back. So it is easier to work in the phase the zeros have to supply, which is the compensator's net contribution plus the $90°$ the integrator took:

$\theta_{boost} = \theta_{comp}(f_c) + 90° = PM - 90° - \theta_{plant}(f_c)$

This is the number to carry around. It is zero for a Type I, and every figure and table below is plotted in it.

Type	Poles and zeros	Boost it can supply	Use it when
I	1 pole, at the origin	$0°$	The plant has almost no lag left at crossover
II	2 poles (1 at origin), 1 zero	up to $90°$ , and rather less in practice	The plant is effectively single-pole at crossover
III	3 poles (1 at origin), 2 zeros	up to $180°$	The LC double pole is still intact at crossover

Fig. 1. Bode magnitude and phase of Type I, Type II and Type III compensators, showing that each added zero raises the phase the network can supply — Fig. 1. The three shapes. All start at $-90°$ from the pole at the origin. Each zero lifts the phase back up, and the peak of that lift is what you place at your crossover frequency.

Type I is a pure integrator

One pole, at the origin, and nothing else. Gain falls at 20 dB per decade forever and the phase sits at $-90°$ at every frequency.

That sounds useless for a switching converter, and for a voltage-mode buck it is. But it is the right answer more often than its reputation suggests. If the plant is already single-pole and you are content to cross over well below any resonance, a Type I gives you infinite DC gain, no steady-state error, and nothing to tune. On the power stage below, crossing over at 2 kHz with nothing but an integrator returns 72° of phase margin. Current-source drivers, LED strings, thermal loops and battery chargers all live here. The bandwidth is poor and the bandwidth is not the point.

Reach for it when the plant leaves you more than $-90°$ of phase at the crossover you want, which in practice means crossing over below every pole the plant has.

Type II adds one zero

A zero and a high-frequency pole on top of the integrator. The zero lifts the phase before crossover and the pole rolls the gain off after it, which keeps switching ripple out of the error amplifier.

This is the standard answer for a plant that looks single-pole where you cross over. Peak current-mode control is the usual case, because the inner current loop absorbs the inductor and leaves the output capacitor as the dominant pole. It is also the answer for a voltage-mode buck whose output capacitor has enough ESR to matter, which is the case this article is really about.

The $90°$ usually quoted for a Type II is a limit rather than a value. The zero and the pole are placed symmetrically about crossover by a factor $k$ , so the zero sits at $f_c/k$ and the pole at $k f_c$ [3]. That gives a boost of $2\arctan(k) - 90°$ , which reaches $90°$ only as $k \to \infty$ . At $k = 5$ it is $67°$ . At $k = 115$ it is $89°$ , and the pole is now sitting at $115 f_c$ . You cannot put it there. It has to sit at or below half the switching frequency, or it stops doing the job you added it for. Impose that and the ceiling that actually binds is

$\theta_{boost,max} = 2\arctan\left(\frac{f_{sw}}{2 f_c}\right) - 90°$

At the conventional crossover of one tenth the switching frequency, that is $67°$ , not $90°$ . The gap between the two matters, and I will come back to it with a part number in it.

Type III adds a second zero

Three poles, one at the origin, and two zeros. The second zero is there for one reason: an LC filter that is still resonating at crossover contributes up to $-180°$ , and one zero cannot buy that back.

This is the voltage-mode buck compensator, and it is what the rest of the series designs in detail. It is also the most expensive answer in components, in tuning effort and in sensitivity to tolerance, which Part 3 covers. If a Type II will do the job, use a Type II.

What actually decides it: the output capacitor

Here is the power stage from Part 1. A 24V to 5V converter, 100W, voltage mode, switching at 150 kHz. The output inductor is 4.7 µH with 20 mΩ of DCR, the output capacitance is 220 µF, and the load is 0.25 Ω. The LC resonance lands at 4.95 kHz. The design crosses over at 15 kHz and wants at least 55° of phase margin, so I will use 60° as the target throughout.

Now change only the output capacitor's ESR, keeping the capacitance at 220 µF:

Output capacitor	ESR	ESR zero	Plant phase at 15 kHz	Boost required	Type
Ceramic	5 mΩ	145 kHz	$-158.6°$	$+128.6°$	III
Polymer	20 mΩ	36 kHz	$-140.9°$	$+110.9°$	III
Aluminium electrolytic	50 mΩ	14.5 kHz	$-115.6°$	$+85.6°$	III
Aluminium electrolytic	150 mΩ	4.8 kHz	$-85.7°$	$+55.7°$	II

Fig. 2. Phase boost required from the compensator against output capacitor ESR, with the practical and asymptotic Type II ceilings marked — Fig. 2. The same converter across a range of output capacitor ESR. The black curve is what the compensator has to supply. Below the teal line a Type II can do it. Between the teal line and the grey one a Type II exists on paper and cannot be built.

The mechanism is the ESR zero. Every capacitor has one, at $f = 1/(2\pi R_{esr} C)$ , and it adds phase back to the plant. On a 5 mΩ ceramic it sits at 145 kHz, a decade above crossover, where it does nothing for you. Raise the ESR to 150 mΩ and it drops to 4.8 kHz, below crossover, where it cancels most of one LC pole's contribution. The plant arrives at 15 kHz with $-85.7°$ instead of $-158.6°$ , and a Type II is enough.

The worse capacitor makes the loop easier. That is not a typo and it is not an argument for fitting electrolytics. You pay for that phase in output ripple, in ESR loss, in volume and in a part whose ESR triples at $-40°C$ and rises again as it dries out. The point is that a component you chose for cost, ripple or footprint has quietly decided how many zeros your error amplifier needs, and most design flows never make that connection.

The 50 mΩ row is the interesting one

Look at it again. It needs $85.6°$ , which is under the $90°$ that a Type II is supposed to be able to supply. So a Type II should work.

It does not. Synthesising it gives $k = 25.9$ , which puts the zero at 580 Hz and the high-frequency pole at 388 kHz on a converter that switches at 150 kHz. The pole is nearly three times the switching frequency, which means it is not attenuating switching ripple at all, and no real error amplifier has the gain-bandwidth to place it there anyway.

That is the gap between the two lines in Fig. 2. Somewhere around 110 mΩ on this converter, a Type II stops being a design and starts being an equation with a solution. If you take one number from this article, take the practical ceiling rather than the textbook one.

And the crossover you asked for

The capacitor is half of it. The other half is a decision usually made before anyone thinks about compensation at all.

Fig. 3. Phase boost required against crossover frequency for a ceramic output capacitor, showing the LC resonance as the dividing line between Type II and Type III — Fig. 3. Ceramic output capacitor throughout, crossover frequency swept. Below the LC resonance the plant has barely started to roll off. Above it, the double pole is fully developed and you are buying back most of $180°$ .

Crossover	Plant phase	Boost required	Type II ceiling here	Type
2 kHz	$-17.8°$	$-12.2°$	$87°$	I is enough
5 kHz	$-85.1°$	$+55.1°$	$82°$	II
10 kHz	$-149.6°$	$+119.6°$	$75°$	III
15 kHz	$-158.6°$	$+128.6°$	$67°$	III
30 kHz	$-161.1°$	$+131.1°$	$46°$	III

Note the Type II ceiling falls as crossover rises, because the high-frequency pole is pinned to the switching frequency while $f_c$ moves toward it. The requirement climbs and the ceiling drops. The transition sits at 5.9 kHz, just above the 4.95 kHz LC resonance. Below the resonance the plant has hardly begun to roll off and almost anything will close the loop. Above it the double pole has done its full $-180°$ and you have to buy all of it back.

Part 1 chose 15 kHz for transient response, which is a perfectly good reason. Type III is the bill for that choice. Had it crossed over at 4 kHz the same converter would have needed a Type II, and the article would have said something different about voltage mode.

So where does "voltage mode needs Type III" come from?

It comes from the common case, and the common case is common for good reasons. Modern designs use ceramic output capacitors, because they are small and cheap and their ESR does not drift. Modern designs cross over as high as they dare, because that is what gives a fast load transient. Do both and you land in the top-left of Fig. 2, where Type III is the only option.

So the rule of thumb is a reasonable summary of what most people build. It is still a summary, and treating it as physics costs you a compensator you did not need on the designs that sit outside it. The honest version has three inputs rather than one: your control mode, your output capacitor, and the crossover you are willing to accept.

Work the arithmetic instead. Model the plant, read its phase at the crossover you want, subtract from $PM - 180°$ , and compare against what each type can actually deliver on your switching frequency. It is one line and it is right every time.

If the answer is Type III, Part 2 takes it from there and computes every component value on this same power stage.

If it is Type II, you have bought two fewer components, one fewer decade of pole-zero spacing to get wrong, and a network whose tolerance sensitivity is correspondingly lower. The placement follows the same $k$ factor with one zero instead of two, and I have deliberately not worked it through here. A Type II is most often reached from current-mode control, and current mode changes the plant as much as it changes the compensator, so that network belongs alongside the current-mode material rather than as an appendix to a voltage-mode example. It is the next thing I want to write.

Working out which compensator a converter needs means modelling the plant, sweeping the crossover, and checking a candidate network against the switching frequency rather than against an asymptote. That is three or four passes of the same arithmetic before you have placed a single component, and it is the kind of thing a design tool should do for you rather than leave in a spreadsheet. It is what I am building switchmode.io to handle: the plant model, the phase budget at crossover and the type decision falling out of the operating point instead of out of a rule of thumb.

References

[1] Texas Instruments, "Demystifying Type II and Type III Compensators Using Op-Amp and OTA for DC/DC Converters," Application Report SLVA662, Jul. 2014. [Online]. Available: https://www.ti.com/lit/an/slva662/slva662.pdf

[2] R. W. Erickson and D. Maksimović, Fundamentals of Power Electronics, 3rd ed. Cham, Switzerland: Springer, 2020.

[3] H. D. Venable, "The K Factor: A New Mathematical Tool for Stability Analysis and Synthesis," in Proc. Powercon 10, San Diego, CA, Mar. 1983.

[4] Analog Devices, "Modeling and Loop Compensation Design of Switching Mode Power Supplies," Application Note AN-149. [Online]. Available: https://www.analog.com/media/en/technical-documentation/application-notes/an-149.pdf