building a synchronous buck converter: questions about mosfet switching

Discussion in 'General Electronics Chat' started by vortmax, May 19, 2014.

  1. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    I'm building a synchronous buck converter because I can't seem to find a commercial product that fits my requirements. Basically it needs to be able to work over the 2-16V range (4-5A) and be able to switch into bypass mode (100% duty cycle). This is for a battery powered application...so I want the regulator to keep the voltage on the rail at a particular level until the battery discharges under that level...at which point the regulator stops switching and just locks in the on state to allow the rail to follow the battery down. I've found plenty of external mosfet buck converters and gate drivers, but none of them like voltages under 5V or 100% duty cycle.

    I've come up with a design (see attached) that I believe should work, but I don't have much experience with designing this sort of thing, so I thought I'd ask the experts. Basically, it is a synchronous buck regulator using two N-channel mosfets. I'm using an MCU to read the output voltage on the rail and drive both mosfets with a PWM signal through a set of transistor-based totem drivers to keep switching speeds up.

    My concerns with this arrangement is in the synchronization between the high side and low side mosfet and resulting shoot-through. My original plan was to drive both mosfets with a single pwm output and just use a logic gate to invert the signal going to the low side mosfet, but I'm concerned that the latency introduced by that component will keep the low side mosfet on too long, resulting in shoot-through. With individual control (as I have it setup now), I can make sure that I break-before-make when switching, but I'm worried that might introduce too much latency on the other end and reduce efficiency. Then to top it all off...I do have some doubts about the stability of such a system in general.


    So anyway...do I seem to be on the right track, or am I completely out in left field? Any words of wisdom?
     
  2. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,288
    1,255
    I see a couple of problems:
    The switching NFET needs a voltage 8 or 10 volts or so above the source voltage not ground as you have in the schematic. The second is the inductor you have is very small - I think 1 Uh for the 5 amp one. This would require an extremely high frequency for it to work. Not something you might get with a micro. You might need to use BJT to run at the low voltages - not so efficient, but maybe the only way. Maybe someone else has some ideas. If you google daycounter boost converter calculator it will help you choose the values. (it works for buck or boost)
     
  3. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    Wait.... Isn't the source side of an n channel MOSFET the "low side" (e.g. the ground side)? That's how all of the examples I've seen online are. The MOSFETs I have picked out have a Vgs(th) of 1.5 V, so I figured they would be good down to a supply voltage of 2V or so

    As for the inductor, I actually copied the symbol from another schematic and haven't put much thought into its specific value (or the other passives for that matter). There is also an opamp/unity gain buffer missing on the feedback line
     
    Last edited: May 19, 2014
  4. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Several problems.

    One thing you need to keep in mind is that enhancement mode MOSFETs' conductivity (or lack thereof) is determined by the voltage on the gate terminal, when referenced to the source terminal, known as Vgs. If Vgs is below the threshold value, the MOSFET will be considered turned off, as the threshold value is 250 microamperes, and current that amount or less is considered trivial. You then need to look at the Vgs specifications for Rds(on); the lowest Vgs is specified at 4.5v to 2.6m Ohms. In between the threshold voltage (around Vgs=2v) and the low Vgs Rds(on) spec, the Rds(on) is usually considerably higher, which means you will dissipate more power as heat.

    If Vg = Vs, within the threshold voltage, the MOSFET will be considered turned off.
    If Vgs is >= an Rds(on) specification, it will be considered turned ON, HOWEVER...
    If Vgs>16v (Vgs>20v in many cases) the MOSFET you specified will be destroyed! If your low-side MOSFET's gate would have been able to be turned on, Vgs would have most likely exceeded 16v, and the MOSFET would have been destroyed. This happens instantly. The moment you exceed the Vgs specification, the MOSFET is toast.

    The highest voltage that you will be able to obtain at either gate is the Vcc of the microcontroller, less ~0.7v of the transistor voltage follower. You show the uC's Vcc as 3.3v, so 3.3v less 0.7v = 2.6v; which is barely above the MOSFET's threshold voltage. Your low-side MOSFET will only be able to pass a few mA's, and your high-side MOSFET won't do much better - if even as much. You're so close to the threshold value it's difficult to say without actually building it, which wouldn't be worth your time and money at this point.

    The MOSFET has a rating for Id of 90A, but that's a theoretical rating. If you try to actually operate a MOSFET with that package at that current constantly, you'll burn the legs right off of it, as the max package rating is 70 Amperes.

    The total gate charge is shown as 110nC typical; that's VERY high. This will make the MOSFETs very difficult to turn on and off; you will need very potent gate driver circuitry. What you show won't do the trick.

    You will need some kind of a charge pump device to get the gate voltage high enough to turn the MOSFETs on.
    You need a voltage level translator between the outputs of the uC and the MOSFET gates.
    You need a high current charge/discharge for the MOSFET gates.

    But you also need to describe your load to us, as we have no clue what you're trying to drive.
     
  5. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,288
    1,255
    Yes it is. But as it is now it is set up as a source follower, so the source will follow the gate minus the threshold voltage. So it doesn't switch it is just a current amplifier. You want it to turn hard on and off. You might be able to find a logic level PFET to put there then you could turn it on with a low. Or you could build a charge pump to get a higher voltage than your supply to drive the NFET. Read the data sheet carefully - the threshold voltage is where the FET just begins to turn on (usually only supplying 250ua.)
     
  6. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    Thanks, lots to digest there.

    You said the highest gate voltage I can get at the MOSFET is the vcc of the micro... I was under the impression that the transistor circuit would allow the micro to switch a higher voltage (vcc), therefore driving the MOSFET gate with that higher voltage.

    I originally was planning on using a p fet for the high side, but from what I was reading, it is less efficient.... Which is why all of the buck converters I've seen use an n fet on the high side. Is it enough to just replace the high side fet with a p fet?

    This circuit is to supply up to 4 amps of current to a pair of dc motors. The motors themselves are controlled by h bridge chips (drv8833) which handle current limiting, speed control, etc. The problem is, the divers have a max rating of 10V and my current motors want to live at around 6V. It's not hard to get battery voltage up above 6V (or 10 for that matter), so I need a way to limit it down. At the same time, once the battery voltage sags below 6V, I want to be able to disable regulation and just let the battery drive the bus directly
     
    Last edited: May 19, 2014
  7. AnalogKid

    Distinguished Member

    Aug 1, 2013
    4,535
    1,251
    The gate drive transistor pairs are complementary symmetry emitter followers, so the outputs will never be greater or even equal to the power rails. Even if the drive signals from the uC were perfect (0V and 3.3V), the gate signals would be 0.7V and 2.6V. That's sorta-off and not-very-on. If you invert the transistor drive pairs so they are common emitter saturating switches (PNP pulling up and NPN pulling down) the voltage swing will be much closer to 0V and 3.3V, but never higher and probably still not enough for minimum Rdson.

    One possiblity is to change the series FET to P-channel, but you still will be hampered by the 3.3V rail. Linear Tech and others make gate driver chips that solve this. They boost the Vcc for solid turn-on of the series N-ch FET, translate the drive signal, etc. Maybe they also provide dead-time management so the input can be a single PWM signal, but I don't remember that level of detail.

    ak
     
  8. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    Thanks, I just googled emitter follower and that side of things makes much more sense now.

    I looked into gate driver chips, but I couldn't find one with a drop out below 4 volts.
     
  9. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,288
    1,255
    Why not a 6 volt SLA battery or a 7.4 volt Lipo?:confused:

    You could switch to a logic level PFET and build a little driver for it and the NFET clamp. Add a diode from the inductor to ground so you can delay the turn on a bit for the NFET.
    Now would be a good time to figure out what frequency you want (or can) run at.
     
  10. crutschow

    Expert

    Mar 14, 2008
    13,006
    3,232
    When you resolve your other circuit problems you can use a non-overlapping clock driver (simulation shown below) to drive both MOSFETs to eliminate any shoot-through. It uses the propagation delay of the gates in the NOR latch to insure the non-overlap. A small amount of delay between the series FET turning OFF and the synchronous FET turning ON is okay since the body diode of the synchronous FET will conduct the inductor current until the FET turns on.

    NO Clock.gif
     
    shortbus likes this.
  11. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    I do have plans to use a 2 cell lipo, which does solve most of these problems. However I would also like to be able to use standard NiMH rechargeable batteries as well, which are much lower voltage.... But it is starting to look like that "feature" might not be worth the effort
     
  12. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    Thanks for all of your assistance. I did some serious reconsideration and redesign work and came up with a solution that I think will be more robust in the long run.

    I'm going to limit my batteries to 2 and 3 cell lipo's only, which gives me a minimum working voltage of 6V. I decided to ditch the synchronous plan and go with a non-synchronous external (P channel) FET controller (LM5085) instead. The regulator supports 100% duty cycle and also has a built in (adjustable) current limit. The current limit and output voltage are adjusted through feedback resistors, so I am going to use digipots to allow me to dynamically adjust them.
     
Loading...