[Great Scott] should win an award for quickest explanation of a buck converter. Clocking in at five and a half minutes, the video clearly shows the operating principles behind the device.
It starts off with the question, what should you do if you want to drop a voltage? Many of us know that we can dim and brighten an LED using the PWM on an Arduino, but a closer inspection with an oscilloscope still shows 5V peaks that would be dangerous to a 3.3V circuit. He then adds an inductor and diode, this keeps the current from dropping too fast, but the PWM just isn’t switching fast enough to keep the coil energized.
A small modification to the Arduino’s code, and the PWM frequency is now in the kHz range. The voltage looks pretty good on the oscilloscope, but a filter cap gets it to look nice and smooth. Lastly, he shows how when the load changes the voltage out looks different. To fix this a voltage divider feeds back the information to the Arduino, letting it change the PWM duty to match the load.
In the last minute of the video he shows how to hook up off-the-shelf switching regulators, whose support components are now completely demystified as the basic principles are understood. Video after the break.
A buck converter takes your bucks and converts them into nothing.
Happens to me every day. I don’t really cares how it works, I just want it to stop.
You’re using the wrong part. You need to get the buck converter that outputs ponies.
that would be a bronco converter.
So inefficient, I prefer Stallion converters.
Well don’t buy your parts from the rodeo then…
Divorce then.
Excellent video! I only started playing with and understanding buck and boost converters a couple years ago and now they’re an indispensable part of my design portfolio.
My latest project was a GPS discipline board for the FE-5680A rubidium oscillator. There I use two buck converters – one to take the input power down to 15V regulated, and another to go from the 15V down to 5V. I use a TPS54331 for the 15V supply because it can go all the way up to 30 volts input (my board is spec’d for a maximum of 24V and has a 24V TVS on the front), and an AP1509 for the 5 volt supply. The AP1509 has fewer external components, but it’s maximum input voltage isn’t high enough for the 15V supply.
Inductor selection is an interesting trade-off. Higher inductance yields lower output ripple, but limits output current relative to physical size.
I’ve also had good results for a 1A 3.3V (from 5V) supply for an OCXO with an SC89Z. That supply yields less than 5 mV p-p of ripple in about a half a square inch of board space (and I wasn’t even trying hard).
Playing with these things has been more fun than I ever would have predicted.
One of my projects in the pipeline is a PoE splitter for the Raspberry Pi. That’s going to require a flyback topology, so that’ll be a new learning experience!
Have fun with the flyback. Hope you’ve got an application note or something to go off.
I do like PoE – it seems so elegant to have power & data on the same port.
(At risk of sticking my nose in…) the things that can really get you are:
1) Check your feedback very carefully before powering on :-)
2) Keep the leakage inductance down by making your layout tight.
3) The snubber on the primary is important so you don’t blow the main switch
4) You might need a snubber on the output diode. A simple R-C snubber can help a lot with noise.
http://youtu.be/QnUhjnbZ0T8
Mine is a little bigger ;)
http://youtu.be/68oQiPmZ5Aw
That is a lovely looking capacitor (:
I’m a Tesla coil guy myself but these are good fun too. I hope your project goes well. I had a boost converter a while back, it was free running as fast as it could switch and I used a neodymium magnet to adjust the frequency, making it free run faster and slower than it naturally would.
I admit I quite like the fact you can use this in reverse too.
Probably the easiest parts to use are from TopSwitch. I don’t use them cause they cost too much money.
Start with one of their parts, dunno if they have active clamp parts or not.
Also look up application notes for the ancient UC3842 or SG3825 series to learn how each circuit actually works.
Or the immortal MC34063…a nice app note that as well…
HAD,
Please move the report flag.
Fat fingered again
You typed the sentence, “Lastly, he shows how when the load changes the voltage out looks different.” I barely understand it. Was it supposed to be, “when the load changes out”?
“Lastly, he shows how when the load changes, the voltage out looks different.”
Let’s eat grandma!
Let’s eat, grandma!
Read “voltage out” as “output voltage”, perhaps?
WoW… 13k$ DSO should have earned sth. better than arseduino! Gimme ur scope and gtfo! I’ll make use of it.
I remember when stupid trolls at least spelled out their demeaning comments about things they know nothing about. Good times.
echo “bunny, pancake, head”
There is a mountain of high quality content on that channel. Even the production values are way above average.
that was an excellent video. it went fast, for me, but better fast than too slow ;) going to have to watch some sections a few more times.
thanks for this one; definitely worth seeing (and agreed; the video was above average for youtube) ;)
…
Why do Constant On Time regulators exist, when are they needed?
What is slope compensation? Why is it needed?
Know what’s cheaper than an arduino buck regulator? An MC33963A.
When are synchronous converters useful?
What is a type II and type III feedback circuit? Which one is for current mode control? Which one is for voltage mode control?
What is a hystereic converter?
When should you use a shielded inductor? What are it’s tradeoffs?
What steps are needed to convert a buck voltage regulator into a constant current regulator, like for driving LEDs?
Mc33963A = MC33063A
Have you got a copy of https://books.google.com.au/books?id=SRNxTY2I-8sC or would your recommend something else?
I don’t have that book. To be honest I designed power supplies a few years ago.
All of the reference books I had, including Marty Brown’s were full of errors.
The reason I point you to TopSwitch is that I believe there are a lot of canned designs from them with magnetics already design for many typical applications and voltage ranges.
TI also has lots of old unitrode reference material that is excellent and many canned designs, they usually begin with PMP#### unfortunately their E2E support forum is amateur and awful.
I’d say start with a working canned design, then read about how a simplistic UC3842 works implementing the same design topology. You will then have a working design and something to make some tweaks to.
Pro tip 1:
You will need to be careful with your oscilloscope grounding when measuring isolated supplies. Either use an isolated probe or only measure the primary side circuits with your scope ground clip on the primary ground. Only measure the secondary side circuits when the scope ground is on the secondary ground.
Pro tip 2: happy magnetics have a linear ramp up and down. Saturated magnetics look like a curve or plot of y=×^2. They will overheat and burn.
Pro tip 3: use good brand ferrite cores, like sendust or Micrometals, or risk thermal aging.
Pro tip 4: switching frequency sets the size requirment of the inductor and capacitors, but it also increases losses. FCC conducted emissions testing begins at 150kHz. So it’s popular to choose 133kHz so that your front end filtering will be easier to design on an offline AC to DC regulator.
Too bad books like this one are so expensive…
An Arduino is cheap enough, and using a premade part might be a learning experience, but it’s placing a black magic box in the center of the thing you are trying to learn about.
Having a driver’s license does not make you a mechanic.
How do you connect an arduino pwm pin to directly to the PMOS gate with pullup to +12V. Isn’t the arduino output limited to power voltage of +5V? I didn’t think there was a 12V rated open-collector output in an arduino. Either he used 5V input or it seems like an NPN or NMOS is needed to control the 12V gate.
I really like his videos, but they are just too fast for me to comprehend, unfortunately. I think it would be better as text.