1
\$\begingroup\$

I'm making a buck converter, this is the circuit I've:

enter image description here

The average voltage at the output is ok, but I'm getting lot of noise at all nodes of the circuit:

I'm using as a main source a 30V 5A bench power supply, and linear regulators 7815 and 7805 to power the circuit.

This is the voltage at the output: The average of the signal is 15V and I'm getting an overshoot of 2vp I'm using a 10x probe.

enter image description here

This is at the output of the 7805, here I'm getting about 1.5Vp of noise

enter image description here

And this one at the output of the 7805, here the noise is about 1vp enter image description here

I'm using Manhattan wiring for the power stage:

starting with the input capacitors, 2*479u + 1*0.33u

enter image description here

the control stage is in a breadboard, wires from IR2110 to the transistor gate and source are about 6cm each.

Finally the load is a 8 ohm resistor.

This is the complete circuit:

enter image description here

Edit:

That's how I'm measuring the voltage at the output of the 15v regulator.

enter image description here

enter image description here

\$\endgroup\$
  • \$\begingroup\$ For your oscilloscope probes, are your leads very short? I can't tell from the pictures, but the wiring looks a bit of a mess... Not a crack on your wiring, we all get it, just an observation. Noise from a source can actually radiate to your ground probe lead and the longer your probe, the worse it is. I have seen this exact same thing. The best thing to to is to use an active probe that measures right at the correct point. Shorten your measurement loop as much as possible. The problem might just disappear. It has happened to me... \$\endgroup\$ – slightlynybbled Apr 21 '16 at 1:21
  • \$\begingroup\$ The probe is about 1.5m \$\endgroup\$ – Luis Ramon Ramirez Rodriguez Apr 21 '16 at 1:25
  • \$\begingroup\$ The probe is a shielded 1.5m or there are dangling wires extended 1.5m that - then - connect to the probe? \$\endgroup\$ – slightlynybbled Apr 21 '16 at 1:30
  • \$\begingroup\$ it's shielded . \$\endgroup\$ – Luis Ramon Ramirez Rodriguez Apr 21 '16 at 1:34
  • 1
    \$\begingroup\$ Acceptable? Depends on the application. If you can deal with it, I can :). This is very characteristic of long ground leads on the probe, I still strongly suspect that the issue is a measurement issue. Food for thought. \$\endgroup\$ – slightlynybbled Apr 21 '16 at 2:23
5
\$\begingroup\$

You will likely encounter issues besides this, you cannot realistically prototype 125KHz switching circuits on a breadboard, even if the high current loop is kept small. Other parasitic inductances and lack of a ground return plane from current loops to the controller and the power stage are still too large. 6cm of free standing wire is a much larger loop than even 6cm of wire directly adjacent to its return path on a PCB, and 6cm is already much too far away for the controlled to be from the switching elements.

Please adjust your expectations as even if this ringing problem is solved, you will not get very good results or performance, though you can probably get it working more or less.

That said, the specific problem at hand is due to your input and output capacitors. Namely, you only have 0.33uF of input and output capacitance.

I have not missed the pairs of electrolytic capacitors on the input and output. Electrolytic capacitors are only capacitors at low frequencies. By 100KHz, none the less 125KHz, they're just glorified resistors and have high ESR at that. And your ringing looks like it is in the MHz. The electrolytic shave too much inductance and too much ESR to do anything at all. Indeed, they are almost certainly making it worse.

125KHz is not the best frequency to be switching at. It's this bad spot where electrolytics are of little use, but still so slow that you need a substantial amount of ceramic capacitance as close as possible to the input switches and output sids of the inductor as possible. Many many uF worth, as this is the only kind of capacitance that will do much good. Ceramic capacitance.

Also, given your large loops and high power inductor, the huge gate resistor is probably making things worse. I would recommend removing it entirely, and placing a low value resistor in series with the bootstrap capacitor instead. 10 or less ohms is a good starting value. This will be more effective but not be prone to gate oscillations like a series hard resistor. Not that your issue is those oscillations per se.

Also measuring buck converters is hard. You really need to measure directly across the terminals of your chunky array of ceramic capacitors on the input and output. The high current loops can be creating a voltage gradient on your ground ("ground bounce") and depending on where your ground paths and measurement paths are, it can couple a significant amount of the switch mode ringing into your measurement. And the input and output, also depending on if their ground return paths have to pass through sections of ground carrying high currents.

Finally, layout is by far the most important part of switching coveters. It plays more of a role than even component selection or circuit changes. You will not have good results and encounter mysterious and frustrating problems if you aren't very careful about that aspect. Using copper clad boards is definitely good, but tighten up those loops more and eliminate any breadboard and long free standing wires and you'll start to see better results.

Maxim has a great app note with a lot of very useful plug and play equations on this, check out https://www.maximintegrated.com/en/app-notes/index.mvp/id/986 for help selecting input and output capacitors.

\$\endgroup\$
  • \$\begingroup\$ Yes, I will make it in a PCB. The magnetic core and UC and using work fine between 100 kHz and 160 kHz, could change the switching frequency in that range make it any better than working at 125kHz? \$\endgroup\$ – Luis Ramon Ramirez Rodriguez Apr 24 '16 at 16:50
2
\$\begingroup\$

If your wiring matches what you have on the schematic, it's all wrong. You're connecting 15V regulator, 5V regulator and 30V supply together!

Second, SMPS circuits have fast/strong ripple currents so your ground wiring has to be up to scratch. I can't tell from the PCB but you need to connect grounds from both sides of the inductor with short ground wires together and to the SMPS controller GND as well.

Third, SMPS circuits ring. This is normal. If you need to get rid of high-frequency ringing noise, you need to look into snubbers. Here's one guide by Ridley Scott: http://www.ridleyengineering.com/component/content/article/15-dc/127-71-designing-snubbers-for-nonisolated-converters.html?showall=1&limitstart=

Most people wouldn't probably bother with the second snubber across the high side mosfet. Here's application note by Fairchild if you want more thorough treatise: https://www.fairchildsemi.com/application-notes/AN/AN-4162.pdf

As pointed above, first you have to make sure your oscilloscope probe is not causing issues. See example in the fairchild application note of what we mean by "short" ground lead. Since this is a low voltage circuit you can get away by soldering a wire to the GND point and pressing it by finger

\$\endgroup\$
  • \$\begingroup\$ About the regulators, what should I do if I want the converter to work from a single power supply? I don't want to have two power supplies in the power stage, that would mean to have three if I want to isolate the uC? \$\endgroup\$ – Luis Ramon Ramirez Rodriguez Apr 21 '16 at 11:20
  • \$\begingroup\$ @LuisRamonRamirezRodriguez Use regulators, just don't tie the outputs together. You can run VCC and VDD both from 12V so you just need one regulator. You can actually generate 15V from 24V supply if that's easier to procure. If you don't want any regulators, you could set the supply voltage to 18V which can run everything. You obviously need to recalculate the PWM ratio for that. \$\endgroup\$ – Barleyman Apr 21 '16 at 11:28
  • \$\begingroup\$ Ok, thanks. Just to point I'm not connecting the regulator outputs together, the 78xx family has three pins and the symbol I'm using just two. \$\endgroup\$ – Luis Ramon Ramirez Rodriguez Apr 21 '16 at 11:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.