You have not published any kind of requirement specification for this buck converter. date with the schematic Input voltage range? Output voltage and tolerance? Output voltage ripple? Output current? Efficiency? Without at least these parameters it is not possible to make any informed comment about the choice or suitability of the components or the layout and it is not up to other people to try to infer information about your project from the components or the layout that you have presented. Some points that are immediately obvious: 1. The schematic symbol for Q1 is rubbish since it does not show the polarity of the device and does not represent the body diode in any meaningful way; 2. The pin mapping of Q1 from the schematic to the PCB footprint is wrong; 3. C8, C9, C11 and C12 are not connected and have been manually deleted from the PCB layout. Thesechanges will be overwritten if doing Update PCB.../Import Changes...; 4. As already stated, the PCB is not up to date with the schematic; 5.  6. The signals and supply on J2 and J3 have no ground return paths on adjacent pins on the connectors. This is a bad idea since it means that the loop area encompassed by the wires between these connectors and wherever they connect to and their associated ground or power supply return paths will be far from minimised and so may be prone to significant interference and crosstalk. This is especially bad news in any sort of switching PSU where there is a mixture of fast voltage switching edges and large currents being switched at high speeds.  7. For the same reasons it is unusual to place the controller chip at any distance from the output being monitored, which is what your PCB design appears to do; 8. It is not identified but it would seem that the net connected to J3 pin is to be the output voltage feedback point. This net has a lowpass filter formed by R6 and R7 in parallel and C10 to ground. This places a pole in the open loop response at about 2.3kHz. This will make stabilising the feedback loop more difficult since there is already a 2nd order pole due to the series 680uH inductor and the 200uF output smoothing capacitance, i.e. at around 432Hz 9. Prefixes in the schematic are incorrect and hence confusing.  10. Question: if an Arduino is being used to provode the loop compensation (as it is obviously not involved in generating the high speed PWM signalling) then why not use a dedicated Buck controller? What exactly does the Arduino bring to the party other than enormous scope for destroying the MOSFET every time you switch the circuit on until you get the compensation somewhere near right?

And to Andy's comments, the LTC1693 is really designed to drive N channel MOSFETs, not the P-channel you are using.  It probably will turn it on and off, but not optimally. I actually do use MCUs (ATTINY404s driving IRS2003s) in buck convertors all the time as it can be very effective, but if you want to use a processor you need the software tightly coupled to the hardware so I'd get rid of the clutter and design something much simpler.

@lukasryden, As MikeDB has sggested, for a digitally controlled SMPS I'd expect to see an MCU taking a feedback and maybe a current sense voltage (either for over current protection for a voltage mode or as the current sense for a current mode SMPS) via minimal RC filtering, straight in to an ADC input and then use the MCU to do the DSP for the loop compensation and the PWM generation and control so that it drives a PWM output straight into something like the gate driver chip that MikeDB refers to. If you want to include any sort of current sensing, it is generally easier and - in terms of robusteness and performance - safer to implement a Current Mode SMPS rather than voltage mode. Voltage mode looks attractive because it looks simpler but in fact they are quite fragile and easy to destroy by getting the compensatin a bit wrong. Without additional measures such as input voltage feedforward, they are also not very good in terms of line regulation and transient response compared to Current Mode converters. Current mode SMPS are easier to compensate but you have to beware the effects - and control - of sub-harmonic oscillations for duty cycles above about 50%. The attraction of digital SMPS control is that "it's just code" so it's "easy" to adjust and optimise. Unless, however, you have a fairly deep understanding of stability analysis in SMPSUs, you probably need to start with some code that is already written and has some IDE support for tailoring the compensation before you apply it to the real thing.

Actually I think you can blow up current mode convertors just as easily :-)    There are still (at least) two feedback loops and above 50% duty cycle it's almost impossible to keep these stable if you have varying loads or output voltage - which if it's Arduino controlled I assume is the case.   Also the current sense return path becomes a highly critical track - the slightest ring here and everything goes pop. Indeed if the original poster wants a wide variation in output voltage and current, a buck convertor is possibly the wrong place to start - look at something like a SEPIC or Cuk where the voltage is built from zero rather than dropped from Vin. One easy way to begin with SMPSs is to buy a few off the shelf two dollar Chinese ones, blow them up (you will !) and then work out why they blew up.  You can still use the inductors and heat sinks again.

Ooooooh. SEPICs and the rest of the Cuk family converters. Getting way off topic but I have some interesting experience with those. Much of the published literature about them - with coupled inductors at any rate, including a few PhDs - is wrong and unfortumately gets parroted by most of the manufacturers in their datasheets and apps notes. Even the ones that get it somewhere near right seem to have done so by accident and with a rather surprised air about it. Here's just some of the stuff I did on the Cuk converter a while back: simple\_open\_loop\_Cuk\_ideal\_with\_ip\_ripp\_can\_160131\.7z at: [https://drive.google.com/file/d/0B2wonnsWWfTXRzU1X1hscXBMdnM/view?usp=sharing](https://drive.google.com/file/d/0B2wonnsWWfTXRzU1X1hscXBMdnM/view?usp=sharing) in: [https://drive.google.com/open?id=0B2wonnsWWfTXOTVmNDU2Y2MtYzJlZi00M2MyLWExZDAtM2JhY2NiNGI4ODBi](https://drive.google.com/open?id=0B2wonnsWWfTXOTVmNDU2Y2MtYzJlZi00M2MyLWExZDAtM2JhY2NiNGI4ODBi) I did find Feuchts papers very useful: Feucht\_papers\_on\_Cuk\_Based\_Converter\_Concepts\.7z And Ray Ridleys article on the SEPIC with Cuk's original notes appended to it is very good too. There's that and some more recent stuff he's added under Modeling (sic) here: [http://www.ridleyengineering.com/design-center-ridley-engineering.html](http://www.ridleyengineering.com/design-center-ridley-engineering.html) Deeply technical but well worth registering for.

Yes SMPSes are actually one of the most technical areas and a lot of rubbish - especially in IC 'application notes' - is repeated.  Only the original Unitrode ones and a few of their successor TI ones on series capacitor buck convertors are worth referring to.  The STM ones are good for actual PCB layout practices but then their evaluation board ignore them :-) Cuk also has become shall we say more controversial recently with articles such as [https://www.powerelectronics.com/power-management/step-down-dc-dc-converter-eliminates-ferrite-cores-50khz-enabling-power-supply-chip](https://www.powerelectronics.com/power-management/step-down-dc-dc-converter-eliminates-ferrite-cores-50khz-enabling-power-supply-chip) I actually use my own variant topology - or at least one I haven't seen elsewhere but as they say there's nothing new under the sun - which minimizes input ripple compared to the Cuk so I don't need excessive input filtering.

I realized my design had a lot of faults and I made changes right away after I posted this thread that's probably why the schematic looks like it does. Have some time off today so I'm going to post my renewed design later today. The biggest change is that I will likely use the Arduino due which makes the LTC 6992 obsolete and instead add a voltage regulator for a 5 volt source. The purpose of my project is to build a battery charger with bulk, absorption and float control Vin=24V Vout=12-14V Iout=2 A Vripple=0.1V Freq=100kHZ The driver is called ltc1693-5 accordingly to the datasheet its a P MOSFET driver. [https://www.analog.com/media/en/technical-documentation/data-sheets/16935f.pdf](https://www.analog.com/media/en/technical-documentation/data-sheets/16935f.pdf)

Ah ok - it was how you wired the LTC1693 that misled me.   In this case in theory pin 3 and 8 need to go to J1 pin 1 so that you are referenced to the Source of Q1.  However the max voltage on LTC1693 is only 13V so you would need to create a floating 'ground' for pin 4 at about 10 volts above input ground.  I still think you'd be better off with an IRS2003 which can handle 100s of volts :-)

Hmm, I think I misunderstood how the driver would work but if I connect pin 3,8 to a voltage divider and then to J1 pin 1 would that work?

@lukasryden, You can play with the LTC1693 in LTspice. :)

good idea :)

I should have said that: simple\_open\_loop\_Cuk\_ideal\_with\_ip\_ripp\_can\_160131\.7z at: [https://drive.google.com/file/d/0B2wonnsWWfTXRzU1X1hscXBMdnM/view?usp=sharing](https://drive.google.com/file/d/0B2wonnsWWfTXRzU1X1hscXBMdnM/view?usp=sharing) in: [https://drive.google.com/open?id=0B2wonnsWWfTXOTVmNDU2Y2MtYzJlZi00M2MyLWExZDAtM2JhY2NiNGI4ODBi](https://drive.google.com/open?id=0B2wonnsWWfTXOTVmNDU2Y2MtYzJlZi00M2MyLWExZDAtM2JhY2NiNGI4ODBi) is a 7zip archive of an LTspice simulation and plot file. :)

@lukasryden No - pin 3 and 8 need to go directly to J1 and a negative voltage regulator drives pin 4.  All in all it's the wrong device for your application

simulate it in LTspice and yeah it's the wrong device :( I will look into IRS2003s and how I will connect it.

@lukasryden Or STM L638x if you can get hold of them easier.  Careful - they aren't quite pin compatible - low side input is not inverted and ST parts have the internal diode.

I was wondering if I have to invert the gate output of the irs2003s? Or just use a BJT to switch the PMOS but I believe I read somewhere that's not the best for higher frequency switching can't really remember

Ah no - change to using an NMOS fet as well.     Drain to the input battery and source to the coil

and power PMOSes are just about ok for audio - nothing more :-)


That should give you some clues.  Replace Q4 with a reverse diode if you prefer.   PSUREQ and /PSUREQ aren't exact complements - there's a programmable dead time in them as the one built into the IRS2003 isn't long enough for my application, but you can probably tie them together.

Oh and if you are wondering what HV1 and HV2 are for - that's where you put a resistor during development which you gradually reduce in value to zero.  Saves a lot of dead transistors :-)

I simplified the design by removing the driver and just replacing it with a transistor. Spacing everything out on the PCB since space is not a concern due to that this will be more of an evaluation board. R9, R10, and C13 is an optional filter from ina240 datasheet that I may or may not use.C21 and C18 is also optional depending on how input and output looks. the snubber values I just choose something arbitrary. So is this utter crap or will it work? [https://easyeda\.com/editor\#id=9bcb2dceed33437491767f0aabf9c4d3\|ea7bb088c0e44e2586a4640e5ae62fd6](https://easyeda.com/editor#id=9bcb2dceed33437491767f0aabf9c4d3|ea7bb088c0e44e2586a4640e5ae62fd6)

I haven't had time to run a sim of your open collector PMOS driver but I'm pretty sure that it will be too slow. It may be able to pull down fast enough but the only thing pulling current back outp of the PMOS gate capacitance is the 10k resistor. Even a very basic calculation of the gate capacitance of 1100pF * 10k says that the turn off time is of the order of 10us. No way are you going to switch at your target 100kHz! Do as Mike suggested and use a (bootstrapped or internally diode pumped) high side NMOSFET gate driver and an NMOSFET. The saving on the cost of the MOSFET will probably pay for the driver chip. Don't try to cut corners with power supplies, especially not switch mode power supplies: you could end up - and I mean this literally - with a faceful of shrapnel.

I think Andy's trying to be polite - you'll be lucky if the MOSFET turns off at all through 10k so it will go into some sort of latch up and die ! Also the transistor driving it's gate will eventually pop as the peak current spikes will be way in excess of it's capabilities. Sorry but SMPSes are never easy even after you've been designing them for 45 years.

Thinking back over dinner to before we had MOSFET drivers, we used to use this circuit arrangement to turn MOSFETs off.  You could add this to your circuit.  Peak transient current of about 1A is fine, compared with the 3-4A you need for the turn-on bipolar.  Obviously swap the polarity for your circuit. 

Ok, will try some simulations this weekend

If you want to try simulations in EasyEDA then the Simulation Tutorial is essential reading. Otherwise, carry on with LTspice. :)