You're on the right track.
t_t can't be much longer than, say, t_clk / 10 or 20, else so much time is spent transitioning that you're better off with a class C oscillator as such!
Typical times are under 1us, with under 50ns being typical for higher frequency and higher efficiency circuits.
Note that the output (drain) transition time is shorter than the total gate transition time, which helps a lot. Drain voltage is only transitioning during the Miller plateau (where Vgs moves very little), which is usually about half the total gate rise/fall time.
Note that assumes inductive switching, which is very typical. For a 4-switch converter, typically in any given cycle, half the transistors will be handling the transition, while the other half will be sitting around passively. This is because a synchronous rectifier doesn't have to deal with Miller effect -- which is not to say it's a slam dunk to drive, because the same phenomenon still arises, it's just behind a threshold. You need a low impedance gate drive, to firmly hold it off, otherwise when the main switching transistor turns on, some turn-on gets capacitively coupled to the sync rect gate and switching loss can go up massively.
Typical MCU output pins are similar to 74HC or LVC logic pins, having an equivalent output resistance around 30 ohms.
For t_t = 50ns, and using a ratio of 1 time constant, at 5V, this gives Qg max ~ 8.3nC (at 5V).
Oh, that introduces some other things --
1. Resistance. CMOS output pins are largely resistive, so it's not just that you're applying a current and stuff happens; it's more like a switch with resistance. So you get an RC time constant, rather than a constant slope and t = Q / I.
2. Time constants. Voltage swings 95% of the way after 3 time constants. But we don't need that much. Two time constants is 87% of the way, which will more or less include the Miller plateau in either rising or falling direction, and is a good figure for what you'll typically see at the gate waveform. Again, about half of that, or one time constant, is the edge speed you'll see on the drain/output, and is where most of the switching loss is dissipated. I wouldn't design based on one time constant, but you might consider this a worst-case scenario (so, 8.3nC really is the most you should ever consider, for this transition time).
3. Voltage. We need this to convert gate charge to effective gate capacitance. Q = V*C is the fundamental equation of the capacitor (or (dQ/dt ==) I = C * dV/dt is its more commonly seen derivative -- same fundamental thing
). The incremental capacitance changes with voltage (namely, it's small at low gate voltages and high drain voltages, massive during the Miller plateau due to feedback effect, then higher at high gate voltages and low drain voltages), but as long as we don't try to look too closely at the waveform (because of these bumps), the overall effect will be consistent with using this effective capacitance as the C in a regular linear RC time constant.
Some other tweaks:
- You have to use logic level FETs, obviously!
- There's no fundamental difference between logic and regular level types. It's just Vgs shifted down to be more convenient, but the gain is the same. Vgs(on) = 5V is about equivalent to a regular one at 7 or 8V. Which is to say, not very on -- you get what's in the datasheet (Rds(on), and usually Qg(tot) as well, guaranteed at 4.5V or so), but it's not exactly at peak performance. Keep this in mind for selection.
- The lower Vgs(th) means the Miller plateau falls at a much lower voltage, like 1 or 2V instead of 3 or 4V, where you have that much less current to pull through it. Logic-level FETs have poor turn-off characteristics for this reason.
- If you are switching low voltages, try to push towards 30V or lower Vds(max) parts, because they have more gain, and also less Rds(on) tempco. For a marginal case -- say a 12V power supply -- you might opt for an H-bridge of 20V parts, versus push-pull with 40V parts. Down at battery voltages, some quite excellent parts are available (there's a MOSFET with something like 300uohm Rds(on), Vds(max) = 8V and Vgs(on) = 1.8V, IIRC).
HTH,
Tim
Home
Help
Search
Login
Register
