1
\$\begingroup\$

I've been tinkering with switch mode power supplies recently and it seems there are various types of buck controller IC's some with a 'COMP' pin for frequency compensation and some without, and I was wondering what the major trade offs are to having or not having this input? I know that well designed frequency compensation can control your feedback gain and stability so does that mean that the chips without that pin are inherently unstable?

For example two chips I am looking at are the TPS54531 from TI and the MC34063A. The MC34063 is appealing because it's cheap and a buck converter design with an external switch for high amperage seems fairly straight forward. The TPS54531 requires frequency compensation which is a bit more complicated and the chip appears to be much more expensive.

Are there chips that are cheap that allow for frequency compensation or is it something that I don't even need to worry about?

Edit: I also found this application note for the MC34063A useful: https://www.ti.com/lit/an/slva252b/slva252b.pdf

\$\endgroup\$
9
  • \$\begingroup\$ The MC34063 is actually not from TI but from Onsemi. Ideally you'd also make the IC numbers links to the datasheets. \$\endgroup\$
    – Bimpelrekkie
    Commented Oct 30, 2017 at 15:00
  • \$\begingroup\$ Woops sorry about that, I've been reading too many datasheets. Give me a sec and I'll link both. \$\endgroup\$
    – Wired365
    Commented Oct 30, 2017 at 15:01
  • 1
    \$\begingroup\$ Some relevant reading. \$\endgroup\$
    – Jerry Coffin
    Commented Oct 30, 2017 at 15:15
  • 1
    \$\begingroup\$ Well, if the chip needs those 2 caps and a resistor, you cannot leave them out, if you do the circuit will become unstable. All buck converters have a loop. All these loops must be stable (obviously). To make the loop stable there might be an external frequency compensation network needed (example: TPS54531) or not (example: MC34063 and many others). It is similar to opamps, some need an external freq comp. capacitor, others don't. Some opamps only need it for unity-gain, others are unity-gain stable. \$\endgroup\$
    – Bimpelrekkie
    Commented Oct 30, 2017 at 15:34
  • 1
    \$\begingroup\$ One other point to keep in mind: you can cheat a bit. Faster reaction in the regulator lets you use a smaller filter cap. If you're building a one-off, you can cheat a bit on the design: over-compensate the regulator, and use a bigger filter cap to make up the difference. If you're doing huge production, then you get into the game of reducing compensation to the minimum, so you can also minimize the filter cap. \$\endgroup\$
    – Jerry Coffin
    Commented Oct 30, 2017 at 17:40

1 Answer 1

Reset to default
2
\$\begingroup\$

The fact that a compensation pin exists lets you tailor your compensator to meet certain design goals: a specific crossover frequency (the point at which the loop-gain magnitude crosses the 0-dB axis), phase and gain margins. You know that the frequency response of a switching converter is affected by parasitic elements like \$r_C\$ the capacitor equivalent series resistance (ESR) which introduces a zero for instance. So when then the compensation is internal (like with a 3-pin linear regulator for instance), then you may have conditions on these stray elements for which the stability is ensured or not. It is your role then to pick the right passive component and make sure its parasitics match the manufacturer recommendations.

On the other hand, if the comp pin is available, then you can choose the compensation strategy to neutralize the effects of the parasitics knowing that they will move between known boundaries during the converter lifetime. You can also tailor the transient response you can accept (fast response with overshoot then moderate phase margin, sluggish response but 0 overshoot with higher phase margin) by selecting where to place poles and zeros.

A lot of ICs integrate operational transconductance amplifiers (OTAs) for design reasons (small die area etc.) but I don't like them especially if you are about to implement a type-3 compensator (1 pole at origin, 2 zeros and 2 poles). As highlighted here, you see that depending on the division ratio between \$V_{out}\$ and \$V_{ref}\$ you are limited in spreading one of the pole/zero pair and cannot boost the phase as much as you would like to. Also, in an OTA, the transconductance \$g_m\$ enters the picture as well as the resistive ratio fixing \$V_{out}\$.

Regarding the MC34063, it is a hysteretic controller inherently instable and does not need to be compensated. It can be extremely noisy as the bunch recurrence can enter the audible range at high peak currents. I think it has been released by MOT after Signetics introduced the µA78S40: yes, some years ago : )

\$\endgroup\$
5
  • \$\begingroup\$ Nice answer, that helped clear up the confusion I was having with regards to the chips that don't have the compensation pin. Thanks for the heads up on the MC34063, I'm realizing that I may just need to go with a newer/more advanced chip to get a more stable output. \$\endgroup\$
    – Wired365
    Commented Oct 31, 2017 at 12:26
  • \$\begingroup\$ Glad if I could help! \$\endgroup\$
    – Verbal Kint
    Commented Oct 31, 2017 at 13:06
  • \$\begingroup\$ @VerbalKint can we have back the file you shared ? It seems that the link has expired \$\endgroup\$
    – AnalogMan
    Commented Nov 21 at 5:47
  • \$\begingroup\$ @AnalogMan, hello, I have updated the link but you can also have a look at my last book, 100% dedicated to loop control. \$\endgroup\$
    – Verbal Kint
    Commented Nov 21 at 7:11
  • \$\begingroup\$ Great, thank you @VerbalKint \$\endgroup\$
    – AnalogMan
    Commented Nov 21 at 11:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.