PTC cause instability in SMPS

[Teun]

Hi,

Lately I've encountered some problems with the use of PTC fuses in combination with buck and buck-boost converters. It seems they can cause instabillity on driving the gates. Does anyone else ever come across this? And does anyone have an explanation for it?

Switched it with a regular SMD fast acting fuse and the whole thing becomes stable instantly.

[wraper]

Without a schematic this question is quite meaningless. PTC fuse generally will have higher resistance than a regular fuse.

[Berni]

I would also guess it might have something to do with your PTC fuse having too much resistance, causing the input voltage to be unstable. Perhaps just adding some extra input capacitance can calm it down.

[Mark]

I would suggest some research on the Middlebrook Stability Criterion. 

[Siwastaja]

Well any regulated output voltage switcher is a "negative resistance" device, i.e., when it sees input voltage drop, it increases input current, which causes more input voltage drop, increasing input current even more... and so on. When the voltage finally drops below Under Voltage Lockout (UVLO) limit, the switcher shuts down, voltage bounces back and here we go again with a relaxation oscillator, oscillating typically at some tens of Hz.

So you have to look at worst-case input current (not one calculated at the normal operating point) - i.e., calculate the input current at UVLO limit. Then pick a PTC polyfuse which has maximum resistance after ageing (for example, after 50 trips; exact rating depends on manufacturer) and calculate this doesn't cause excessive voltage drop.

Example calculation:
Buck outputs 10V 10A at 90% eff.
Input is 20V. Nominal current would be 1/0.9 * 10V/20V * 10A  = 5.6A
UVLO for the controller is set to 11V min, 12V typ, 13V max

Now assuming your 20V input normally goes as low as 15V.
You can't drop below the max UVLO (13V), otherwise the switcher could turn off.
So you can afford dropping 2V over the PTC (assuming other types of input resistance are negligible, i.e., good wiring and contacts).
At 13V, input current is 1/0.9 * 10V/13V * 10A = 8.5A.

R = U/I = 2V/8.5A = 0.235 Ohms

Then pick polyfuse with after trigging/ageing value of 0.235Ohms or less.

[Teun]

Ok I could see that. But it is a Buck-Boost with UVLO of 2.6V. The typical input voltage is a car battery (so about 13V is tested on a bench power supply). I don't see it drop that badly.

My case:
input: 12~14V
output: 13V, 3A | 16V, 3A

1/0.9*12V/13V*3A =3.08A
1/0.9*12V/16V*3A =2.5A

Max resistance listed for SMDC300F/24 = 0.072R

So say input current is doubled (6A trip current):
6A*0.072R = 0.432V drop over the fuse.

Even if the resistance would be 10 times higher I would think the effect would be negligible, since UVLO wouldn't be reached.
And the point is that the converter does work. It just starts shifting its frequency a bit when the PTC is used.

[Siwastaja]

Measure the input voltage after the PTC. Likely the PTC just causes excessive voltage drop and the circuit does not work properly at that voltage, nothing to do with the PTC itself.

Scope trace with one channel for input side of the PTC, another after the PTC would be interesting to see.

[ejeffrey]

Ok I could see that. But it is a Buck-Boost with UVLO of 2.6V. The typical input voltage is a car battery (so about 13V is tested on a bench power supply). I don't see it drop that badly.

My case:
input: 12~14V
output: 13V, 3A | 16V, 3A

1/0.9*12V/13V*3A =3.08A
1/0.9*12V/16V*3A =2.5A

Isn't that backwards?  i.e., it should be 16/12, which, assuming a 90% efficiency gives a worst case input current of around 4.5 amps.  You have a 3 amp fuse.  The trip current is the current where a brand new fuse is guaranteed to trip within some time, but it may trip for anything over 3A.  Each time it trips, but especially after the first trip, the cold resistance goes up, this decreases the trip point.

Next, is there any chance the supply operates at < 12 V for more than a fraction of a second?  If the supply drops to 9V the input current rises to 6A -- more than enough to trip the PTC.  You plan to use this on a car battery, but are testing it with a bench supply, what is the current limit on the supply?  If the 12V supply is current limiting (even during start up transient) and the voltage drops, the regulator will try to draw more current than you expect and that can cause unstable operation.  You should probably increase the ULVO threshold to something like 9-10V and set the PSU current limit to > 6A.  And use a larger PTC.

Finally, what is the operating temperature?  You also have to derate PTCs pretty seriously for operating temperature. At 60C the hold current on that fuse is only 2.2A.

This is the problem with PTCs.  The large trip/hold ratio, temperature derating, and aging means that you often have to way over-size them compared to what you think. For this application in an automotive environment where you need to handle a hot day in Phoenix or a cold morning in Minnesota you might need a fuse with a 6A hold rating that derates to 4.5A at the highest operating temperature.  If you are operating in the engine bay vs. the passenger compartment it might be even worse.  On a cold start that fuse might let 20 A through for several seconds, so your wiring all needs to handle that.  If the circuit has an upstream 10 A or 15 A fuse it might blow before your PTC rendering the PTC kind of pointless.

That said, I hate traditional SMD fuses.  They aren't as finicky as PTCs, but their cold resistance and aging are much worse than traditional fuses, plus when they blow you have to break out the soldering iron to fix them.  If you are going to us an SMD fuse, oversize it considerably and consider the chance that a transient fault could disable your device with no easy fix.  Can you fit a mini-ATO blade style fuse?

[Siwastaja]

Polyfuses are hard to design with, especially those physically small (SMT).

3A rated polyfuse may be completely unsuitable for a load guaranteed to never exceed 3A.

Trip it once and now the 3A fuse is a 2A fuse and trips for apparently no reason. Trip it 50 times more and now it's a 1.5A fuse. These numbers are examples. This is all in the datasheet.

Look at all the datasheet parameters carefully and if you have trouble understanding any of them, Google it or ask for more help, really.

I sometimes use polyfuses but really prefer the combination of active current limiting (by design, like Constant Current operation mode, or e-fuse style, trip using a comparator and reference) and fire protection by classical fuse. In such design, the classic fuse only blows when the active part of circuit fails due to a design mistake by me, or totally unexpected abuse such as overvoltage exceeding specifications written by me. In either case, semiconductors are already dead so the fuse doesn't need replacement, the whole card is swapped (or sometimes, repaired, replacing all blown parts).

[Teun]

Isn't that backwards?  i.e., it should be 16/12, which, assuming a 90% efficiency gives a worst case input current of around 4.5 amps.  You have a 3 amp fuse.  The trip current is the current where a brand new fuse is guaranteed to trip within some time, but it may trip for anything over 3A.  Each time it trips, but especially after the first trip, the cold resistance goes up, this decreases the trip point.

I assume you would divide output by input right? (16V out/ 12~13V in) I usually use 13V on the bench supply. Which has a 5A current limit. By my knowledge the supply never drops below 12V measured at the input of the switching MOSFETS (external switch design).