A Buck Current-fed Push-Pull Converter / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

A Buck Current-fed Push-Pull Converter

previous 1 2 3

Move Thread

LAN_403

Goodchild

Wed Jan 15 2014, 03:25PM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

Finn, great project! I did have a question though, you mention that you fab your own PCBs. I noticed that you have vias on your board, how did you yet them to plate threw to the other side of the board for electrical connectivity?

I'm also glad to see you back into electronics, you are too good not to be!

All the best,
Eric

Finn Hammer

Wed Jan 15 2014, 08:06PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Thank you for your kind words, Eric

The vias are tiny copper tube rivets, which are supposed to work even without soldering them in, I donÂ´t trust it without solder though.

(picture 1,2)

I did a bit of work on the converter today. This first picture shows the gate waveform on the 2 push-pull transistors, and the overlap that is important in order to maintain a current path for the buck inductor.

(picture 3)

In the next picture, I have installed 10ohms resistors from buck inductor to drains of push-pull transistors, so that I can inspect switching behaviour under an ideal load. This test checks out as it should.

(picture 4)

In this picture, a zoom in on the switching period, reveals that the 2 switches share current during switching.

(picture 5)

At this point I feel it is time to attach a transformer to the converter, and why not use the one I want to use in the final project. The core is the big UU from Magnetics, which at 100kHz and 37.5 volts per turn are driven at 0.15 Tesla, which is a fair loss limited flux density.
The reason for using such an oversized core is the high volt/turn ratio, enabling a 10kV secondary winding with only 266 turns. The associated low winding capacitance should make it possible to create a 10kV transformer at 100kHz. Anyway, along these lines are my assumptions.
For now, however, I am satisfied with a transformer that will support 70volt in, and out, so a nice 1:1:1 winding stack, with 2 turns of copper strip was wound:

(picture 6,7,8,9)

First turn is prepared. Â Â Â Â  Â  Â  Mylar brought around.          First turn completed.         The finished article.

The secondary is sandwiched between push pull primarys, so I expected a nice, close to ideal transformer action.

Here in the next picture, the transformer has been wired into the circuit, and the load resistor is in place too:

(picture 10)

So what do I get from my effort:

(picture 11)

This is not the waveform I expected! I see transformer action, there is symmerty around x-axis, but the signals were supposed to show the buck inductor current, just like in (picture 4) above, but with the current waveforms flipped around x-axis.

In a desparate attempt to sort of harmonize the machanical ratio of buck inductor and transformer, I wound a quick and dirty tranny on an old gate transformer, a big one, and down at 12.5kHz, managed to get something that looked right, but it is not:

(picture 12,13)

I am transferring power, because the load resistor gets smoking hot, if I go for it, but why on earth does the big transformer not transfer the waveform as expected?

Anyway, the palace is still standing, so not all is bad.

Cheers, Finn Hammer

Thomas W

Wed Jan 15 2014, 08:46PM

Registered Member #3324 Joined: Sun Oct 17 2010, 06:57PM
Location:
Posts: 1276

Wow! very nice work, im curious, did you use any referances for calculating the transformer information, and if so, would you be willing to provide where you got the calculations?

Im seeing some great work here, with some luck i may get round to messing with some of these topologys now ive fixed my lovely 0-72V 0-20A SMPS power supply

Keep up with the good work!

Finn Hammer

Wed Jan 15 2014, 10:00PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

__=|(:3)-|--{__ wrote ...

Wow! very nice work, im curious, did you use any referances for calculating the transformer information, and if so, would you be willing to provide where you got the calculations?

Tom,
Yes, the ability to calculate number of turns on a transformer is an important one, so read the article linked below, it is all in there under transformer calculations. Be aware that it is in CGS units, so flux density is in gauss etc. Ferrites are are used over a vast range of frequencies, so different rules apply, with regard to flux density in the core. Whereas in an iron core for 50Hz operation, you would drive it close to saturation as you dare, this does only apply for ferrites untill 20kHz. Above 20kHz, the losses get so big that flux density has to be lowered in order to keep the core cool. So although the power ferrites saturate around 0.3-0.4Tesla, they are rarely driven beyond 0.15 T @ 100kHz, 0.1 T @ 300kHz and 0.05 t @ mHz. The phrase you will meet is that the ferrites are flux density limited up to 20kHz, beyond that they are loss limited.
i hope you will pick up on transformers, I know I have had a lot of satisfaction (and frustrations) working with them..

Also, over at Texas Instruments, there are some good PDFÂ´s from Lloyd Dixon, look for *slupxxx.pdf:* there are quite a lot of them, and they are all you need.

Cheers, Finn Hammer

Thomas W

Wed Jan 15 2014, 11:10PM

Registered Member #3324 Joined: Sun Oct 17 2010, 06:57PM
Location:
Posts: 1276

Huge thanks for the info, Finn, i will read though the TI ones tommorow (just finished reading the Magnetics Inc one.

I intend to have a play around with this stuff tommorow, going to contact a local company that winds transformers and buy some ferrites + bobbins off him, mabye some copper strips too :)

Thanks again,
looking forward to see more awesome from Denmark!

Finn Hammer

Sat Jan 18 2014, 12:20PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

It is amazing how a lack of experience can make things look irrepairaple wrong, and then after much pondering, and methodic fault finding, the problem suddenly disappears without a trace.
Today I went trough one of these cycles. An abundance of noise kept an amplifier from behaving, and after dismantling the circuit, and putting it back together, the problem disappeared with out a clue, go figure.
The object of my focus was the current sense amplifier, if you look at the block diagram below, you see it at pin 7,8 and 9.

This amplifier is strapped across the current sense resistor on the back of the board, and this resistor is 0.1 ohms. This resistor sits in the ground return path of the converter, so for every ampere that is passed trough the load, 0.1 V is generated across it.

The output of this current sense amplifier is connected internally to a comparator that has the ability to terminate the gate signal to the buck switch, cycle by cycle. This happens when the output of the current sense amplifier rises to 3V, since this is the reference on the other input of the comparator. This feature enables the controller to perform cycle by cycle current limiting and thus provide protection of the buck transistor.

To make the amplifier reach 3 volts, a corresponding amplification factor has to be established, relating to the current that should trigger current limit, and since the voltage that is fed to the input is of negative polarity, the amplifier has to be configured as an inverting one.

With 0.1v/amp, and a (for now) desire to perform current limiting at 2A, the amplification should be 3/0.2=15 times.
This is acheived by using a 1K input resistor to the negative pin, grounding the positive pin, and strapping a 15K resistor from output pin to negative pin.

I have little experience with opamps, and frankly, I took for granted that the amplified output would trace the input, with complete fidelity, only with magnified amplitude. This is not so, compared to the input, the waveshape is distorted, but ok, the levels seem to match, so in the end things may work out. This is something that takes some getting used to, and perhaps I should try to breadboard a couple of different opamps to see how they perform. I made a precision rectifier once, and seem to recall that it takes some pretty expensive opamps to maintain fidelity on the output side, anyway:

In the little video I am going to show how the current sense amplifier works, in conjunction with the current sense resistor, and the current limit comparator, to terminate the output to the buck transistor. Another way to look at it is to say that the duty cycle is adjusted on the fly.

Interesting to see, how at first, only one of the cycles is adjusted, probably due to a slight difference in the size of the load resistors:

The bottom trace shows voltage at switching node, at 10V/div
The yellow trace at the top, shows inverted, the buck inductor curent at 1A/div

Finally, the cyan trace in the middle, shows output of current sense amplifier, at 1V/div.

Notice how, when it reaches 3 volts, the duty cycle is starting to reduce, and how this at first only happens at every second cycle, due to minor difference in load at each push pull transistor.

This is a wonderfull method to limit current, because it can go on forever without stressing any components, instead it relies on operation as usual, only with slightly altered timing.
This output of the current sense amplifier sets the final brute force protection in a worst case scenario, but the same signal is also used, at lower current levels, to perform average current limiting, in conjunction with the current error amplifier and the voltage error ampifier, the latter of which will be the focus for my next installment to this series.

Cheers, Finn Hammer

Finn Hammer

Tue Jan 21 2014, 09:47PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Today I closed the loop of the buck stage, and it worked.
It is the first time in my life I have ever closed a feedback loop, and it opens up to a lot of new possibilities, so I guess if I had been into the whole alcohol thing, it would have been Champagne!, but since I am not, I went to my Yoga class.

Anyway, The voltage error amplifier is the topic of todays progressive step, but let me first recover what has been acheived up to now. The buck stage has been built and it functions to a point where I can controll the duty cycle, and thus the output voltage as a percentage of input voltage, and the 2 push pull switches work too. I have seen that the duty cycle is controlled by a positive voltage of 0-2.6V and that the highest voltage produces the highest duty cycle.

In order to controll the duty cycle with the voltage error amp, it is therefore necessary for this amp to start out with an output that is high, and this output should then decrease, when the controll voltage from the voltage divider across the output, rises above the voltage reference delivered from the chip.

In short, the voltage error amp should be configured as an inverting differential amplifier.

I have grown accustomed to being cautious before hooking something permanently up, so I breadboarded the intended layout to see if it functioned as desired, and it did. I have become fond of the small and inexpensive chinese powersupplies with digital controll, because they are so handy down on the bench, right next to the circuit. And did I mention that they are Inexpensive

With a 30k resistor from output to inverting input, where the voltage divider is also attached, the voltage out decreased from 2.6V to0V with an increase of Vin of only 0.14V. This means that the duty cycle would drop from full to zero if the voltage (at the measuring node) would overshoot with 0.14 volts.

This looked good, so I mounted the resistors on the board, turned up the voltage and saw that it worked.

In the video, you see the switching node voltage as the Magenta trace at the bottom, you will see it rise up as it follows the input voltage.
At the same time, the green trace above it shows the output voltage also rising.

At some point the output voltage stops rising, and at the same time, although the switch node voltage continues to rise, the duty cycle starts decreasing. This demonstrates, that the voltage error amplifier has taken over the controll of the duty cycle.

Enjoy:

So where to go from now? I guess I need to check the stability of the loop by applying step loads, and this is probably what I will doo next.
There is also the whole thing about current mode controll, which is supposed to take over controll from the voltage loop at some point when the supply approaches full power, but I can't quite grasp what it is good for.
Perhaps if I have the attention of the more proficient members of this society, they will step in with advice on the virtues of current mode controll?

Cheers, Finn Hammer

Thomas W

Tue Jan 21 2014, 10:19PM

Registered Member #3324 Joined: Sun Oct 17 2010, 06:57PM
Location:
Posts: 1276

Very nice, excellent work there Finn!

My buck-boost ICs/ PMW controllers and such arrived to day, so in the next week or so i may have a few pictures up in a thread.

Thanks for the insperation to get a move on with my own electronics!

Those powersupplys look rather nice, a bit too pricy for me, i think i could probibly, when i become more competent, build a few of those at a lower price each.

Keep up the good work!

TCW!

Finn Hammer

Sat Feb 15 2014, 11:36PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

At the present, the converter is able to meet a part of the design criteria, and that is putting out 31V across 4.7 ohm = 204W, close enough to the 200W I am aiming at.

In order to be able to measure this, I built this fan cooled resistive load:

It consists of, apart from the heatsink and fan, 4pcs. DSEI60-06 Ultrafast FRED diodes in a bridge, 7500ÂµF of smoothing caps and 4.7Ohms of load resistors, rated at 50W each. With this, I can load the supply continously at full output.

The transformer has 2 turns on both push pull windings, as well as the middle one, the output winding, and these windings are made of 10mm wide 0.3mm thick foil. I cannot think of any way to get better coupling between windings than that.

Still, I am getting some awfull turn off spikes from the transformer. as seen here at reduced input, just to make room for the full event on the scope:

That overshoot is 5 times the amplitude.

I can reduce it with a 10nF/10ohm snubber:

But this happens at the expense of a lot of heat: The top trace is the current trough the snubber, at this reduced power level the peaks are 1 amp, at full output, it amounts to 4 amps.

As measured with a tek type 134/P6021 current probe:

Transistors are running hot, snubbers are running hot, ferrite is getting warm, as expected, but I am not having the impression that this is an efficient supply. The buck stage is ok, but the push.pull stage is not. There is too much energy in those oscillations.
I can hardly believe that this is

The Magnitude of the turn off spike is baffeling me, and I hereby hope for help to solve it.

previous 1 2 3

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob