my new prototype

[zilog]

The system will consist of this schematic: http://wintermute.csbnet.se/~zilog/power%20prototype%2012v%20smps.pdf paired with this DSP board: http://www.olimex.com/dev/images/DSP/TMS320-P28016-REV-A-sch.gif . As voltage reference goes, it will be wired to use 2.048V.

Not all details are worked out yet, but hopefully the main structure is set. I hope to control the switches using the PWM outputs from the DSP, and sense all feedback points using an AD-converter. Over-current protection will be handled by a comparator just for safety. I will be boosting the gate drive voltage to 17V to achieve better efficiency.

I will be using a digital input to sense whether or not the output is in balance, to know that I dont activate the balancing mosfets with too much imbalance during start-up, and to be able to keep the switches inactive when no rail pumping is present.

I dont know how much high-frequency content I can expect from the feedback points, so I am having some worries about aliasing from the AD-converter, hopefully I can sample fast enough not to notice any effects, alternatively only sample in quick bursts, letting the different parts of the smps be interleaved in time to avoid HF-ripple when taking feedback.

Am I way off in believing that I can take primary current feedback from two push-pull primaries using only one CT using Dmax = 0.9, and relying on symmetry, this should be possible, right? If I run into trouble with parasitic parallell inductance in the CT, I should be able to compensate for this in software, my main concern here is saturation.

Well, any opinions?

 

[zilog]

I have now wound the main transformer in the schematic, the core is a 3F3 ETD29, pictures from the construction are available here: http://wintermute.csbnet.se/~zilog/transformers/main/

Have I done this right, or have I made any mistakes other than not having very much clearance?

 

[Eva]

You will find that winding a 32mm OD toroid for 12V push-pull is much easier than winding an ETD29. Almost all commercial car-audio amplifiers use toroids. Bifilar symmetrical windings are particularly hard to achieve in an ETD29.

You need two current transformers in order to achieve proper balancing. A single CT will saturate with the slightest imbalance, it won't allow the DSP to detect and correct the problem.

 

[zilog]

You will find that winding a 32mm OD toroid for 12V push-pull is much easier than winding an ETD29. Almost all commercial car-audio amplifiers use toroids. Bifilar symmetrical windings are particularly hard to achieve in an ETD29.

You need two current transformers in order to achieve proper balancing. A single CT will saturate with the slightest imbalance, it won't allow the DSP to detect and correct the problem.

With maximum duty cycle of 0.9, shouldn't I be able to sustain a current imbalance of 10% without pushing the CT into saturation? I mean, during the off-time it's free to self-reset from the remaining imbalance flux.

If I skip the CT on the drains of the mosfets, could I successfully use one CT for the center-tap instead? Feels like all I need is to allow the CT secondary to swing to -20V or similar during the off-time in a Dmax=0.9 scenario. I have seen this done in http://www.national.com/ds/LM/LM5030.pdf , so it must be possible to do?

When it comes to the ETD29, could you please point out what I have done wrong? I want to learn. All but the outermost secondary are wound as tight as possible, the leakage-catch primaries are even wound in the same layer as the primaries they are supposed to couple with. It is only supposed to be able to handle an average output power of 100W, so I guess I have enough cooling area for the wires?

If I were to go for a TX 36/23/15 3F3 toroid (the largest I can buy over here), could you please show me how to wind this in an optimal way? I guess I should spread the copper over as large portion of the toroid as possible, and try to overlap primaries/secondaries everywhere to avoid leakage? Guess the hard part will be how to design the toroid in a way that allows easy PCB routing. I also need only one PCB-connection per winding leg as I want to stick CT:s on them...

 

[Eva]

See transformer pictures on this thread:

http://www.diyaudio.com/forums/showthread.php?s=&threadid=111566

Note that a bigger toroid core is not always better because the amount of primary turns required becomes too low.

You haven't done anything wrong with the ETD29, it's just that it's a true pain to wind and the result is not particularly good looking.

It's difficult to ensure proper current transformer reset in the way you want to use it. Try it but don't assume it to work from the beginning.

 

[zilog]

See transformer pictures on this thread:

http://www.diyaudio.com/forums/showthread.php?s=&threadid=111566

Note that a bigger toroid core is not always better because the amount of primary turns required becomes too low.

You haven't done anything wrong with the ETD29, it's just that it's a true pain to wind and the result is not particularly good looking.

It's difficult to ensure proper current transformer reset in the way you want to use it. Try it but don't assume it to work from the beginning.

Ah, still hope of life

What is the drawback of using too few primary turns? The air-gap for the PCB-to-toroid-turn becomes large in relation to the magnetic coupling to the core? Problems spreading the turns which worsens coupling to the secondary?

I guess I will need to write the software in a way to always skip pulses in pairs when over current is detected to avoid pushing the CT off center. I dont yet know how, but I am quite confident I can use software to keep track of flux imbalance in advance, so I dont need to react unreasonably fast.

I am still working on designing/winding the coupled output inductor. I have found that a T106-2 (which I already have a bunch of) can contain enough energy for 2x50uH windings at the current level I am interested in. The problem is that 50uH requires 2x61 turns of 0.8mm wire, and that demands 3 layers of winding..

Is this something doable with a toroid core? Do I gain anything by keeping the fast switching ends of the toroid apart to reduce capactace between both inductor windings? Feels like it would be easier to wind both windings bifilar in the same direction. The other alternative is to divide the toroid into two 180 degree halves, and stagger-wind each inductor in 3 layers, but that would worsen coupling between the inductors, but I will on the other hand be using active voltage balancing between the outputs, so do I really need that good coupling?

Any advice on what to do here? Reducing the inductance until I arrive at a low enough turn count doesnt feel like a viable option either since I then arrive at very high rms ripple in the output capacitors.