EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: dmwahl on March 17, 2016, 03:02:13 am
-
Can I use a h bridge driver to replace the high side diode in a boost converter with a FET instead? The controller is a UCC28C40? I realize I'd need to invert the input of the driver, but aside from that is there anything I'm not thinking of? Can't think why it wouldn't work, but I'd prefer to get my education before building it rather than after. The specific driver I'm looking at is the IR2104 (http://www.irf.com/product-info/datasheets/data/ir2104.pdf), with the output of U201 inverted into the driver input. LO would drive Q202 and HO would drive the FET replacing D201.
Ignore specific component values, I haven't figured out the exact values for everything yet. For those curious, the application is to boost the voltage from a 12V nominal solar panel up to ~28V to charge the battery in my lawn mower.
(http://)
-
I can't see why you would want to. Hope you will be at least operating the panel at power point to be efficient.
-
Why not use a synchronous controller. Its a rather delicate and unforgiving task. I did not seriously analyze your scheme but my initial gut response was that it would never work - at least not well.
Sent from my horrible mobile....
-
I'm trying to eek out as much power as possible from the panel, and yes, operating it at the mpp. It's not obvious from the schematic snippet, but I've included a digital pot to trim output voltage to get to mpp. If it's relatively simple to replace the high side diode with a FET and synchronous driver, then I'd like to try. If it's a definite no-go for some fundamental reason, then I won't bother trying, but I'd like to understand why rather than just going to a more expensive part. The schematic as shown does work, I've verified both in simulation and actual assembled boards.
I may still just use a synchronous controller, but they're more expensive and I'm familiar with the part shown.
-
In principle I don't see why not, it all depends on how much power you are losing in the Schottky diode. If you output current is one Amp or so I wouldn't bother. You've also got the additional problem of getting the dead time spot on.
As an aside, I recently prototyped an interleaved boost converter using an LM5032, TI app note AN-1820, Vin 18-36V, Vout 48V at 50W, measured efficiency 95%.
-
The IR2104 has a ~500ns deadtime built in, so in theory shouldn't that be enough? After the low side switches off and the high side has not yet turned on, there's still the body diode that will conduct until the switch turns on.
I'll look into the LM5032, I've also been reading up on the LM5122, but still want to understand what's going on here.
-
Half-bridge bootstrap driver with dead time insertion logic is just fine. IR2104 seems ok in this regard, and it would logically work, but it's horribly slow (rise&fall time); 520 ns deadtime is also overwhelming.
As you are trying to really maximize the efficiency (I might not bother with synchronous boost at 28V output), this slow driver increases switching losses too much. Modern bootstrap drivers can easily achieve 20 ns rise/fall and below. So get a similar part but with more modern specs.
Optimize the FET selection so that you minimize total losses, include (1)conduction, (2)switching and (3)gate drive losses to your analysis. Too big a FET is also a problem as it increases 2 and 3.
When you want to maximize efficiency, don't trust on the synchronous FET body diode during the short deadtime; not only it has higher Vf, but more importantly, it has poor recovery characteristics, causing extra switching loss. Try to parallel it with an ultrafast, soft recovery diode. You may want to experiment with different diodes. Body diode might be good enough, or might not. This parallel diode can be rated at typically one fifth of the actual average current because it conducts for a very short pulse, at miniscule duty cycle.
So, having both a good, really well selected diode, and a synchronous FET in parallel minimizes both switching and conduction losses.
Build prototypes. Build them on 4 layer PCBs, try to have the layout as final as possible, so that you can analyze actual PCB parasitics. EMI can be difficult in a boost converter, and when maximizing efficiency, you want to avoid adding an RC snubber. A properly selected diode in parallel with the FET may make the difference between a smoke-emitting converter and a working converter.
-
I would agree the 500+ns dead time is huge. I have a design in production right now based on the TPS43060 synchronous controller. It has a typical 65ns dead time and much higher current drivers to keep switching losses reasonable. TI has some sort of secret timing they call adaptive dead time but it is not clear how that is managed.
I suspect your plan would not blow up, but would not help you with efficiency either.
-
Is 4 layer really that important given the switching frequency is relatively low?
-
it's not the frequency, it's the rise time of the pulses that make a mess. I do all of mine in 4 layer to ensure I have an unbroken ground and the low=voltage loop and FB parts can stay clean. It is possible to do 2 layer for sure, but 4 layer offers options for much better performance.
-
It's not absolutely necessary but it doesn't cost that much more, and will make the design a lot easier and perform better, so why not. Fighting EMI is relatively important, and keeping unwanted inductance minimal is very important.
Switching frequency per se is irrelevant; edge rate is meaningful. As you are pushing efficiency, you are going to need fast switcing, so you are going to have the switch node (which doubles as a heatsinking copper pour) swinging at large dV/dt. Having an extra ground plane would be great.
Try to minimize the size. Layout is not trivial. Slower switching and RC snubbers both help, but reduce the efficiency.
4-layer is a often a good deal; you pay a bit extra, and get a better result with less effort.
It's a complex optimization task if you really go for efficiency. It's not only about EMI regulation, but parasitics and even minor design issues cause ringing and overshooting which can destroy mosfets easily. You cannot compensate by using higher voltage rated device because it would increase switching losses due to larger total gate charge.