Zeta topology DC-DC converter unstable

Question

So I've tried to build a Zeta topology DC-DC converter based on a simple buck converter IC XL4016, which works perfectly in a regular buck topology. However, when hooked up in a Zeta topology as shown on the schematic below, the schematic becomes unstable.

If the output voltage is set using R1 potentiometer to, say, 9V, then everything is more or less fine at input voltages V1 < ~7V and V1 > ~12V except for +-0.5V changes in voltage with the input voltage. When the input voltage is close to the output voltage (V1 > 7V && V1 < 12V), the converter suddenly jumps to ~14-16V output voltage and "latches" in this state - no matter what the input voltage is (even if it becomes beyond the 7-12V range) or the potentiometer position, it stays at that output voltage until the power is cut off or the potentiometer is set all the way to minimum.

So, what could be the problem here both with the output voltage sudden rise and latching and with the output voltage a bit unstable (+-0.5V) outside the "death range"?

XL4016 datasheet: https://www.makerfabs.com/desfile/files/XL4016-Datasheet.pdf

P. S. It doesn't seem to be a problem with the maximum/minimum duty cycle: the duty cycle in zeta topology at input voltages close to the output ones becomes close to 50%, not 100% or 0%, and the XL4016 IC claims to be able to hadle the full 0%-100% duty cycle range.

I am using the Mastech HY3005 PSU for the input voltage, by the way.

The schematic works fine in Proteus simulation with the XL4016 IC changed to Lm2576-ADJ (I haven't found anywhere the XL4016 IC model)

^{simulate this circuit – Schematic created using CircuitLab}

Answer 1

Zeta is a wholly unsuitable configuration for these regulators.

For switching controllers, given suitable gate drive logic, and level shifting for the current sense if applicable, zeta topology is fine. (But more often, these signals will be common-ground, and so the more compact/efficient method is to choose a boost/flyback type controller and SEPIC topology -- which is simply zeta operated backwards, so has identical characteristics.) But these are regulators integrated on a common substrate, and that imposes additional restrictions.

The LM2576 limits output (SW node) voltage to -1 or above. The XL4016 limits output even more strictly, above -0.3V: which is especially perverse, given their table of suggested schottky diodes has minimum V_F circa 0.55V, I believe?

Whether simulation models implement these limits, isn't clear. Operation beyond specified ratings doesn't need to be modeled, so it may simply be whatever.

If you are saying it was tested and breadboarded, that's quite astonishing that it survived more than a volt at all. Most likely, as the flyback voltage exceeded 7V or so (typical V_EB breakdown), the switch stopped turning off reliably, and substrate charge injection may've increased current consumption, disturbed control circuitry nearby on-die, or induced CMOS latchup (well, it's not CMOS to be clear, but to the extent that similar SCR structures might exist on this old-school bipolar design, that is).

This would seem to explain the low voltage limits and peculiar behavior.

As mentioned in the other answer, C1 likely needs to be much larger. It is generally advantageous to use coupled inductors to reduce the ripple current seen by this part; coupled inductors are readily available these days in a broad range of ratings.

Answer 2

Problem fixed by significantly increasing the C1 to 47uF. Probably that's too much, but there is no way you can "overkill" it within reasonable limits. Adding an input 220uF filter capacitor reduced the voltage unstability greatly to +-0.05V.

Texas instrumets has some great materials regarging Zeta converters design:

• Designing DC/DC converters based on ZETA topology
• Power Topologies Handbook

While C1 value might seem to be unimportant (as I assumed firstly), it actually has a certain minimal capacity since the voltage drop on it might become too large. I have no idea, why it happens at duty cycles near 50%, however.

In the first link the C1 capacitor minimal value is shown on page 4 of the document. It is largely dependant on the frequency of your controller, maximum/minimum duty cycle and typical input and output voltages/currents, so this capacitor value varies from case to case.