Question on < 100 KHz synchronous rectification buck converter

[BravoV]

I've done lots of step down circuits in the past using common buck converters like National (TI now) Simple Switcher families LM257X or even the jellybean switcher like MC34063.  Most of these converters are working at sub 100Khz range, and its relatively easy to work with especially they're available in through hole style, and also on pcb layout still quite forgiving compared to those > 500 KHz switcher. All this time I've been drooling at those high end > 1 Mhz step down using synchronous rectification controller, but as an enthusiast, I still don't have the gut to face the complexities involved especially on pcb layout and components placement.

Now the question, say I use a controller like MC34063 and coupled with mosfet driver (inverting and non-inverting) to drive two mosfets to form a synchronous rectification buck converter ? Is this workable or even remotely possible ?

Or to be precise, what direction I need to take to build such circuit ?

I'm aware the cost of building this might be significantly higher since at this low frequency, it will need a relatively big inductor with high inductance, and also huge filtering caps compared to > 1Mhz counterpart, but please ignore this in this discussion, since it will be only one of project and to learn this thing.

Btw, does anyone know if any ic manufacturers ever made a sub 100KHz step down IC controller using this synch rectification method ? I think I exhausted Google already on searching this, and also probably there is no economical value to make such product.

[slateraptor]

I'm aware the cost of building this might be significantly higher since at this low frequency, it will need a relatively big inductor with high inductance, and also huge filtering caps compared to > 1Mhz counterpart, but please ignore this in this discussion, since it will be only one of project and to learn this thing.

On the contrary, I suspect designing a sub-100kHz Buck topology as a 1-off project would be cheaper (admittedly, I've never built one up before, but hear me out ). As it turns out, MHz switchers are typically less efficient than their 500kHz-class counterparts, which in turn are typically less efficient than their 50kHz-class counterparts. What's gained by increasing switching frequency is a reduction in device and peripheral component size that can't be had any other practical way at a nominal cost of slightly reduced efficiency. There's definitely a point of diminishing returns, and the efficiency vs. frequency relationship is quite concave-down parabolic.

Pressman points out that ac switching losses account for a significant portion of lost efficiency. As a 1st-order approximation in a theoretically convenient scenario, and completely ignoring snubber losses, ac switching losses can be simplified down to



where V_DC is input voltage, I_O is output current, T is switching period, and T_s is switching transition time. So it's clear that the problem is not that ac loss is inversely proportional to switching period, but rather the reality that the T_s/T term becomes more significant in practical devices because rise/fall transitions can't happen fast enough to maintain static proportionality (think slew rate limitations) as a consequence of parasitics, device limitations and what have you other nuisances. And that's just ac switching losses...the real party doesn't start until precision synchronous rectification techniques come into the picture.

So it might superficially seem like the need for relatively large components values might cost you more initially, but I feel like the price you'll pay just to get it right at a much higher switching frequency will end up costing you more--time, monetary and headache wise. That being said, I'm almost motivated to try my hand at a 50kHz switcher attempt myself now.

[BravoV]

@Slateraptor, thanks for pointing those formula, yeah, I'm aware of it, infact thats one of the reason this thing sparked my curiosity. But with my "half baked" knowledge and skill in electronic as enthusiast, to be honest I'm lost and don't have any idea how to start from the scratch. 

So it might superficially seem like the need for relatively large components values might cost you more initially, but I feel like the price you'll pay just to get it right at a much higher switching frequency will end up costing you more--time, monetary and headache wise. That being said, I'm almost motivated to try my hand at a 50kHz switcher attempt myself now.

Really eager to see how do you plan to carry out this sub 100Khz sync. rect. converter, please keep me updated.

Actually personally I don't mind the size, again, its just I'm so curious and this solely purpose is to learn and see how this thing performs, if its so good, I might use it as a linear pre-regulator.

You can get drivers like the ISL6207.

http://www.intersil.com/products/deviceinfo.asp?pn=ISL6207

These might be marketed under half bridge drivers or high and lo-side drivers.

Hey Achmed, thanks, really interesting chip, currently just starting to learn and digest the datasheet.

Usually sync rectification is only worth it for higher output currents.

An alternative controoler which integrates pretty well everything is

http://ca.mouser.com/ProductDetail/ON-Semiconductor/NCP3020ADR2G/?qs=sGAEpiMZZMtinsmYoYdje0yXznloMszZ6WfOjTdsphg%3d

Its true about switching losses but as FET technology improves lower RDSON X Qg improves and smarter control strategys that reduces cross conduction and minimize body diode conduction it can sometimes be benificial in efficency and size to increase the switching frequency. Its always going to be some trade-offs.

This thing is working at 300Khz, any idea if there is similar chip that capable of running lower freq ?

[bfritz]

Be aware, that you will have to build in a "dead-time" to avoid shoot thru current.  Don't want the bottom side synchronous rectification switch to turn on, while the top is just starting to turn off.  It tends to let all the magic smoke out!   

One person commented that synchronous switchers are only "worth it" at high currents.  I agree that high currents will help things, but so does the output voltage.  If you are trying to regulate a 1.2V output voltage, having to rely on a diode on the low side kills efficiency real fast!

I think there are plenty of switchers in the marketplace that aren't in the nosebleed area for switching frequency.  I've also done layouts on MHz switchers, and don't find the rules to be any worse.  As a matter of fact, I think many times they are a bit easier, as you end up being able to choose smaller components, which makes the high current loops smaller, which makes the layout issues easier.

I also find that manufacturers are doing a much better job of writing data sheets that tell you what to do in the layout, and give great guidelines on how to achieve it.  I worked for a competing IC manufacturer to TI, but I liked the format of TI's datasheets in this area.  They give very clear guidelines.  Honestly, if you've understand how a switcher works, and what to watch for in the layout, I think you will be as effective with higher frequency switchers.

I do find a better selection of inductors in the 500KHz range, so you might want to look for a part at 500KHz, and another at 1MHz, and go through the loss calculations, to determine which does turn out better.  Some of the high frequency switchers have gotten tricky, and do things to recycle the energy from the switching of the FET's to other places in the circuit, to reduce the AC losses.  The data sheet for any part should give you information on how to calculate the losses and efficiency.  Note that many manufacturers have their own software that you can use to design a converter, and it will calculate efficiency for you.  Most of the manufacturers are doing a good job in this area.

[nukie]

Other than low rds-on mosfet, you also need low ciss for switching in the MHz range.