• Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Does a constant off time sepic converter have a right_half_plane_zero?

Status
Not open for further replies.

Flyback

Well-Known Member
Hello,
Constant off time sepic converters seem very easy to make stable.

Take the attached PWM dimmed, 6W, 12VIN, 85VOUT sepic led driver which stabilises easily whatever compensation components are used.

So surely, continuous current mode, constant off time sepic converters (or boosts or buck-boosts) don't have a right half plane zero?

(ltspice simulation also attached....here's your challenge....try to make it go unstable by varying the feedback components, within some reason...bet you wont be able to)
 

Attachments

Flyback

Well-Known Member
certainly a neater, lower-component-count way than I did.
I like the use of the fet instead of the comparator and fet..........its got me thinking why the comparator is ever used, as Basso uses the comparator in his book for this purpose.
I see you chopped out the dimming fet.....of course, as you know, there is then more overshoot during pwm dimming switch on....however, the led current overshoot is not likely a problem.....though there is slightly more dissipation in the snubber resistor.

Stability of high duty cycle constant frequency led drivers is quite challenging....but with constant off time, its simple to do......I think there is no RHPZ in the constant-off-time converters.
I mean, the layman's description of RHPZ is that as the power demand of a converter suddenly increases, the off time gets constricted......but obviously that doesn't happen in constant off time converters, because the off time is, "constant".

So do you agree...no RHPZ in constant off time converters?
 

ronsimpson

Well-Known Member
Most Helpful Member
My bad memory agrees with you on RHPZ.
I do not like the overshoot. I started to fix that then thought why. Changing the current sense resistor by 2x will slow down how hard the supply pumps up the output capacitor at power up. This should help the overshoot.
 

MrAl

Well-Known Member
Most Helpful Member
Hi,

The right hand plane problems are a little easier to grasp based on a time domain analysis than a frequency domain analysis, even though they are usually talked about in terms of frequency response. This is because there is a fairly simple view in the time domain while the frequency domain view is more abstracted from the way we like to think about things.
The time domain view is very simple. We start with a basic idea like negative or positive feedback, and look at this in simple terms in the time domain.

Starting with negative feedback, when we increase the control signal which increases the error signal and we see the output increase and the feedback also increases. The feedback is negative so it tends to decrease the error signal which means the output does not continue to increase but reaches some limit.

With positive feedback, when we increase the control signal the error signal increases and that makes the output increase and so the feedback again increases, but now the feedback causes the error signal to increase so the output increases even more which may give rise to an output that increases without bound (unstable).

In a simple linear system we have a negative feedback signal so that when the output increases and the feedback increases the error signal decreases, and this bounds the output. So in short, output INCREASES, error signal DECREASES.

In a switching converter, we have to look at the pulse width to get an idea what is going on because the average pulse width determines the average output. But there is also the instantaneous view where we have to pay attention to one edge of the pulse.

In a buck converter, when the control signal increases the pulse width increases so the output increases and the feedback increases and so the error signal decreases so the output is bounded, just like in the normal linear circuit. But more important is that when the output decreases the feedback decreases and the error signal increases and then the pulse width becomes wider as the edge follows the error signal.

In a regular boost converter, the switch is turned on BEFORE the output increases because the output gets it's energy from the inductor not from the switch directly. The switch charges the inductor and when the switch turns OFF then the inductor supplies the energy to the output. So there is a delay in there where the output decreases, then the feedback decreases, then the error signal increases so the switch stays on longer, but because the switch only supplies energy to the inductor the output goes even lower during this time rather than climb higher as in the buck circuit, so the feedback responds to the output going low by first making the output go even lower, and that is positive feedback which is inherently unstable.

In one type of converter we see the output respond to the feedback immediately and it follows the error signal closely, while in the other type we see that the output does not follow the error signal but rather gets even farther from the required level for a time and that is the main problem. We have to find a way to fool the circuit into thinking that it is controlling the output in a more stable way.

So if we find a time in the cycle where the output goes low and the circuit through it's control makes the output go even lower, that means that there will be a right hand plane problem...that's short term positive feedback. The ideal situation is that when the output goes low the circuit tries to make the output go higher immediately. In the former case we have to find a way to adjust the feedback.
 
Status
Not open for further replies.

Latest threads

EE World Online Articles

Loading
Top