Problem with a buck converter: my error, or the device's

[Dextrine]

Hello everyone,

I am in the process of designing a buck converter.

This is the regulator I have chosen. http://www.aosmd.com/pdfs/datasheet/AOZ1242DI.pdf

The input will be 13V

output will be from 12 to 6 volts (will be adjusted via a thermistor on the FB voltage divider R2)

Now, the issue I am having is the following:

When there is no load on the output, the device works exactly as intended and perfectly with the parts I have chosen. However, when I put a load on it (DC fan 3.57W at 12V), it starts to intermittently cut the voltage to zero randomly and at a frequency of a few khz. Has anyone experienced this problem before with either another regulator or perhaps this particular one?

My best guess is that somehow the device thinks there is some sort of short since the application note lists "hiccuping" to be an indicator of a short. However, I don't think there is a short since it works fine with no load and the hiccuping occurs regardless of what type of load I use (fan or resistor)

I appreciate any help and thanks in advance.

 

[davenn]

The input will be 13V

with what current capability ?

However, when I put a load on it (DC fan 3.57W at 12V),

what type of fan ... brushed or something else ?

as an initial suggestion without all the info from you, if you are not supplying enough input current capability isn't high enough
then the converter may not be able to handle the higher start up current requirements of the fan

Dave

 

[CraigHB]

Usually you want input at least a couple volts above output to cover the drop-out of a buck converter. Maximum duty cycle for that part is 85% which means you need input voltage at least 15% greater than output voltage. Though that won't necessarily cause a problem other than low output voltage, it can if the controller is designed to do a restart when input voltage is too low. For an input voltage of 13V and an output voltage of 12V you would need a converter controller capable of at least 95% duty cycle.

Also the compensation network (Rc, Cc) is a pretty critical part of the design using that controller or any converter controller really. It provides the feedback compensation required as a result of the filter formed by the output capacitor and inductor. If those values are out of range the regulator can destabilize and do a restart at the frequency you are seeing. Showing a schematic indicating the parts you are using would be necessary to determine if those values are appropriate.

 

[Dextrine]

Thanks for responding Craig,
You know, that's the same thing I thought (about the input having to be higher than the output), but when the input is less than 15% higher than the output voltage, the converter does not do the "hiccuping" (I'm guessing because it's not working at all). I'll try increasing the input voltage (there is no issue increasing it up to at most 15.6V). I have included the schematic and this is the diode I'm using. http://www.diodes.com/_files/datasheets/ds30891.pdf

BUCK CONVERTER
http://4.1m.yt/GO-_r.png

 

[Dextrine]

with what current capability ?
what type of fan ... brushed or something else ?

as an initial suggestion without all the info from you, if you are not supplying enough input current capability isn't high enough
then the converter may not be able to handle the higher start up current requirements of the fan

Dave

Thanks for the reply Dave,

There is enough current to start the fan, because it does start and sometimes works smoothly for a little bit before it starts to hiccup. The startup current is between 1 and 2 A and the steady-state current is about .6A

 

[Dextrine]

I'm currently re-doing all my calculations and I think my loop compensation values may have been too high (at least for the resistor so far). I will keep this updated in case in the future someone else is stuck with a similar problem.

 

[rbelli1]

What is the inrush for that fan? Is your few kHz around 24?

What have you done with the thermistor? Is it at the end of a cable or with very short leads? Long leads on the feedback node can cause erroneous operation. A small capacitor across the top of the feedback divider (R1 in the datasheet) sometimes helps if you want longer leads. Also an RC snubber from the LX node to ground will reduce the output noise which may be interfering with the motor or feedback thermistor.

BoB

 

[Dextrine]

I haven't added the thermistor yet, just using fixed resistors right now. I managed to get a pretty consistent 12V out with 15V in. However, sometimes every like, 5-10 or so seconds, the output voltage will drop to zero briefly then return to 12V. I'm thinking it's not, nor was it ever, the hiccuping. I'm thinking I need to just improve my compensation loop. I will continue to update.

 

[Dextrine]

Sweet! I got it working perfectly with 12V output. Really honing in on the appropriate compensator loop values as well as a much smaller inductor (10uH) did the trick. Now, I have a different issue: there is some sort of strange noise I can hear that sounds like crackling, when I look at the oscope, my output seems to have some 5khz artifact which is what I believe I'm hearing. Any ideas on what could be causing this?

 

[rbelli1]

Try following these instructions for selecting an RC snubber:
https://www.fairchildsemi.com/application-notes/AN/AN-4162.pdf

Also an LC filter on the output could be an option if the snubber is insufficient.

BoB

 

[CraigHB]

I've built a few converters for lower voltage and higher currents and have yet to need a snubber on them. You can get excessive ringing using high di/dt rates with switching MOSFETs, but it's generally not a problem for converters, it can be though. Of course for controllers with on-chip switches there's no option there. The di/dt rates are set by design. For those controllers you can still put a snubber on the "switch" output, but I've yet to have a need to do that. It's something you can try anyway, but it can come at a cost to efficiency.

Most of the time issues with converters come down to loop compensation. Ideally you want to tune these things using a frequency analyzer to plot frequency response of the actual circuit. For a hobbyist without access to expensive equipment you kind of have to wing it with trial and error. Typically you'll see formula in data sheets to arrive at component values which gets you close, but they don't account for the parasitic characteristics of your components and PCB layout. For that you need measurements on the bench to get real world numbers.

Current mode controllers that only require a type II compensation network are a lot easier to tune. I've done one voltage mode converter requiring a type III compensation network and that was pretty difficult to tune without actually measuring frequency response on the bench. That was very time consuming doing the trial and error method with all those poles and zeros. I'll probably not do that again sticking with current mode controllers. Sometimes you need maximal efficiency for higher currents. When current mode controllers can use inductor DCR current sensing they're just as efficient as voltage mode controllers.

Less than ideal PCB layout can give you trouble with parasitic impedances, but unless you do something unusual it's generally not a problem I've run into myself. Data sheets just about always provide recommendations for that so it's just a matter of following them.

 

[Dextrine]

Thanks for your really helpful reply Craig.

One thing I've noticed with my circuit is that this audible crackling sound is completely removed when I touch the FB pin with anything conductive (a probe, a random piece of metal, some wire).

I guessed that perhaps there is some capacitance needed between the FB pin and ground and so I tried 10uF, 2uF, 1uF, .1uF, 500pF, and 22pF.

At 2uF and above, the noise is completely gone, HOWEVER my output voltage oscillates much more (unnoticeable unless looked at on the oscope) The other, smaller values, seemed to not have any effect at all except that when I had them in, touching the FB pin with the probe no longer got rid of the noise.

I've tried increasing the capacitance in my RC compensation loop in order to get higher frequency response at near DC levels, but it doesn't get rid of the noise nor even attenuate it.

Any ideas on what this could be? The fact that touching it with anything conductive fixes the noise issue seems odd.

Thanks in advance.

 

[rbelli1]

touching it with anything conductive fixes the noise issue

I have seen that before. It was a layout issue. Try to move the feedback node and boost node away from the switch node as much as possible.

BoB

 

[Dextrine]

I have seen that before. It was a layout issue. Try to move the feedback node and boost node away from the switch node as much as possible.

BoB

I have my input capacitor kind of far away from my chip and the rest of the components. Do you think that might be an issue?

 

[rbelli1]

How far is "kind of far"? It could be a problem. The path from the input cap to the output cap/inductor is a very high current path. Extra impedance (both resistance and inductance) can be a problem.

BoB

 

[Dextrine]

The input capacitor is about 1.5 cm away from the converter. I've been reading some application notes and it seems like not having this loop as short as possible will generate some EMI, not sure if it would generate noise in the audible spectrum, but I'll try shortening this loop and seeing what happens.

 

[CraigHB]

There's actually some sensitivity to input capacitor placement on the PCB for buck converters. Again check data sheet recommendations for PCB layout. Buck converters are most noisy electrically on the input side and least noisy on the output side. Boost converters are the opposite, most noisy on the output side and least noisy on the input side. You should use an array of MLCC caps or a higher value electrolytic cap in parallel with a lower value MLCC cap close to the controller as possible to minimize electrical noise on the input side.

It's rather indeterminate what is going on that is making a difference when you touch a probe to the circuit. It sounds like the inductor is buzzing which is not that uncommon. There are mechanical forces within the inductor due to the expansion and contraction of the magnetic field. Unshielded, non-potted inductors have the most tendency towards vibrational noise. A shielded, potted inductor like the Vishay IHLP series will eliminate mechanical noise mostly since coils are not able to vibrate against the core nor is the core able to vibrate against the PCB. Also the core in a cheap inductor can resonate mechanically creating vibrational noise.

At low output currents the converter can go into DCM (discontinuous conduction mode) versus CCM (continuous conduction mode). In DCM the converter may stop and start at some frequency depending on load. It really depends on how the controller is designed to handle DCM, there's various techniques employed. This cycling can lead to inductor noise.

Some controllers offer a forced CCM pin option which removes the possibility of noise resulting from DCM, but it comes at a cost to light load efficiency. Otherwise if the converter is stopping and starting at a high frequency for whatever reason the inductor can buzz (typically caused by instability or overload). Hiccup overload protection is most noisy mechanically. Some converter controllers have a pin option for pulse by pulse current limiting which can reduce mechanical noise in overload situations.