Poor regulation / ripple in SMPSU Buck converters

[arcadeuk]

Hi everyone,

I am very much fluent in digital electronics, but I have to admit I am no expert when it comes to power supply design, so I could really do with some assistance please.

I  am trying to create a PSU that can convert 6.3VAC into 3.3VDC @ 1 - 1.5amp

I bought a bunch of ebay/China buck converter modules (more for experimentation with different chipsets so I can later integrate the buck convert of choice into my pcb)


So I moved my design onto the bench and used a Variac to generate the 6.3VAC for convenience. This is fed into a bridge rectifier, and a 15,000uF filter capacitor is placed in parallel with the bridge rectifier to give a smoothed DC output.

I hooked up the load tester and scope and applied load and the ripple varied as expected within reasonable limits with load variations. All fine so far!


I added a linear 3.3v regulator and connected the load to the output. The linear regulator worked as expected and provided a near enough perfectly smooth 3.3v. Again all fine.


I replaced the linear regulator with the first buck converter module and applied load. The peak-to-peak voltage on the output of the regulator is terrible, even with low (100ma) loads.

I wound up the variac to a higher voltage to give more headroom for the buck converter to operate, but still the same problem.

I tried a different buck converter, but still the same problem. I tried another, still the same! I tried adding a filter capacitor to the output of the buck converter, still the same!

So if we look at the waveforms, we can see I am getting 50hz ripple on the output of the buck converter. Why is this happening? If there is enough headroom in the smoothed DC from the bridge/capacitor than I didn't expect this ripple to occur.

Here is one example scenario on the scope (attached to this post)

Pink is the AC from the variac
Yellow is the smoothed DC (12.7vrms 960mv ripple)
Blue is the output of the buck converter (3.24vrms 800mv ripple!)

As you can see the ripple in the output lines up perfectly with the 50hz AC and the 50hz dipping of the capacitor during charging cycles. Why is this 50hz appearing on the output when there is so much headroom from the smoothed DC?

So as this is not what I am expecting to happen I must be doing something dumb. I would appreciate it if someone could point out what dumb thing I am doing

Thanks in advance

[arcadeuk]

I forgot to add the modules I have been using are based on LM2596 and MP1584 (and other One-Hung-Low variants)

[Phoenix]

Sounds like a problem with the power supply ripple rejection.

Things to try: Put a bigger capacitor on the rectifier output. Use a full bridge rectifier.

[exmadscientist]

Sounds like a problem with the power supply ripple rejection.

Things to try: Put a bigger capacitor on the rectifier output. Use a full bridge rectifier.

What, 15mF just isn't enough?

Sounds to me like you need to try some known-good switchers. For whatever reason, Chinese counterfeits of SMPS chips are everywhere, and would handily explain your results.

[Phoenix]

Sounds like a problem with the power supply ripple rejection.

Things to try: Put a bigger capacitor on the rectifier output. Use a full bridge rectifier.

What, 15mF just isn't enough?

Sounds to me like you need to try some known-good switchers. For whatever reason, Chinese counterfeits of SMPS chips are everywhere, and would handily explain your results.

I missed the 15mF, that is quite big. The rectified DC waveform still looks pretty bad so...

Another question - how are you connecting your oscilloscope? You say you're using a variac, do you have an isolation transformer too? This come to mind because a full bridge rectifier should have 100Hz ripple. But an incorrectly attached oscilloscope ground could cause 50Hz ripple to appear in the measurement.

[prasimix]

I'm not sure how much 15mF can help if you didn't add low ESR ceramic (one or more in parallel depending of required capacity and rated voltage) as close as possible to the buck power input.

[arcadeuk]

Another question - how are you connecting your oscilloscope? You say you're using a variac, do you have an isolation transformer too? This come to mind because a full bridge rectifier should have 100Hz ripple. But an incorrectly attached oscilloscope ground could cause 50Hz ripple to appear in the measurement.

I would be using an isolation transformer on the variac if I could find it, but I can't lol

All 3 channels of scope ground are connected to earth, the bridge rectifier negative is connected to earth, the buck regulator ground is connected to earth, so all measurements are relative to earth ground.

I initially was not measuring the AC waveform on the scope, I only added it in to see if the 50hz ripple on the buck converter output lined up perfectly with the positive phase of the mains AC.

The ripple is present, when only the output of the buck convert is being measured, and is visible on the scope (with buck converter ground tied to earth ground) or visible on my bench meters bargraph when measured as floating ground

[Phoenix]

I think this might be your problem. What country are you in? In Australia we have a CMEN system where the neutral is tied to earth. Since your oscilloscope is also tied to earth this creates a short circuit and can be quite dangerous for your oscilloscope or circuit under test.

Find an isolation transformer or use a 240V to ~12V step down transformer that's isolated. See if that fixes your problem.

[Phoenix]

I'm not sure how much 15mF can help if you didn't add low ESR ceramic (one or more in parallel depending of required capacity and rated voltage) as close as possible to the buck power input.

The ceramic cap you speak of won't help with 50Hz/100Hz ripple.

[arcadeuk]

I think this might be your problem. What country are you in? In Australia we have a CMEN system where the neutral is tied to earth. Since your oscilloscope is also tied to earth this creates a short circuit and can be quite dangerous for your oscilloscope or circuit under test.

Find an isolation transformer or use a 240V to ~12V step down transformer that's isolated. See if that fixes your problem.

I am in the UK,

I understand what you are saying, and I have been careful not to create a current path through the scope.

For the purpose of my original question, imagine I am not measuring the AC at all, I only did this as a visual aid. In this case, the buck converter ground will still be tied to earth, and the scope will s till be measuring the DC output relative to earth.

[Monkeh]

Do yourself a favour and find a transformer.

And for the love of noodles stop using earth as a current-carrying conductor.

[arcadeuk]

And for the love of noodles stop using earth as a current-carrying conductor.

I'm sure I am not, what makes you think I am?

[Monkeh]

And for the love of noodles stop using earth as a current-carrying conductor.

I'm sure I am not, what makes you think I am?

All 3 channels of scope ground are connected to earth, the bridge rectifier negative is connected to earth, the buck regulator ground is connected to earth, so all measurements are relative to earth ground.

Ponder that earth is connected to neutral (the same neutral you're using on one leg of your rectifier).

[arcadeuk]

Scope ground is connected to earth, so everything measured is relative to earth.

I have just picked up a Meanwell metal framed PSU on the bench, DC, negative is connected to frame ground, frame ground is connected to earth.  So this is normal right?

I work on pinball machines daily, all DC rails (of where there are many in a pinball machine, are referenced to earth)

I'm not arguing with you, but I don't understand your point. DC ground is always connected to earth right, unless the supply is specifically a floating supply?

[Monkeh]

Scope ground is connected to earth, so everything measured is relative to earth.

I have just picked up a Meanwell metal framed PSU on the bench, DC, negative is connected to frame ground, frame ground is connected to earth.  So this is normal right?

I work on pinball machines daily, all DC rails (of where there are many in a pinball machine, are referenced to earth)

I'm not arguing with you, but I don't understand your point. DC ground is always connected to earth right, unless the supply is specifically a floating supply?

But you're not isolated from a system which has an earth connection at a different point.

[Phoenix]

Scope ground is connected to earth, so everything measured is relative to earth.

I have just picked up a Meanwell metal framed PSU on the bench, DC, negative is connected to frame ground, frame ground is connected to earth.  So this is normal right?

I work on pinball machines daily, all DC rails (of where there are many in a pinball machine, are referenced to earth)

I'm not arguing with you, but I don't understand your point. DC ground is always connected to earth right, unless the supply is specifically a floating supply?

Your commercial power supplies have an isolated DC output (you'll find a small high frequency transformer or large mains frequency transformer inside them). This is OK to ground.

The supply you have constructed is not isolated from the mains. This is hazardous to ground (and hazardous to touch if wired incorrectly).