Well done. Thanks for reminding us that a simple solution can solve a lot of problems for lower cost. Engineers sometimes get too wrapped up in looking at the newest devices, that we don't stop and "engineer" a easier design.
Nice, enjoy your articles.
BTW, if anyone tries to simulate the circuit of Figure 1, be aware it needs a long time-constant capacitive load (as mentioned in the article text but not shown in the schematic) to represent the FET gate. Otherwise, results can be strange. Had me scratching my head for a couple minutes. Interesting also to drive with a sine wave and monitor individual transistor current.
Well that you qualified the transformer circuit as appropriate to phase-shift systems, as it will get very strange when a significant deviation from a 50% duty cycle is attempted! But otherwise it's a great little circuit, and the observation about the benefit of a negative gate drive is pertinent.
The challenges of highside drive with variable duty cycles I hope will be addressed soon. I look forward to future installments.
In my past experience using this buffer in class D audio amplifiers I discovered an intriguing detail. Putting a 2k resistor between base and emitter of the transistors will considerably speed them up. Part of my research revealed that 2k is a nearly ideal number. The resistor is bootstrapped so it has minimal loading effect.
I have seen versions of the circuit in figure one providing tens of amps in different applications. It is not a new circuit, but it certainly would be potentially a very good choice in the application described. So thanks for reminding us again that an elegant solution does not have to be either complex or brand new. Sometimes simple works very well.
That's a commonly used and suggested technique, but generally we assume the transistors are somewhat less capable, forgetting that when acting as boosters to a wimpy driver, they're getting plenty of base drive, so only need a beta of 10 or so. Thanks, Robert!
The '2222 and '2907 are better dies than the '3906, etc. The common jellybean plastic versions are the 2n4401 and '4403, and their mmbt sot-23 counterparts. The Zetex parts are even better yet, but perhaps much better than needed, more expensive and not second sourced.
The Diodes, Inc. mmdt2227M is pretty cool - it packs the NPN and PNP transistors together in one sot-23-6 package. 5-cents each at Arrow.
Good article! I have used these circuits over the last 20 years with good results. For figure 3 just be careful of the "first pulse syndrome": when power is first applied the capacitors are at zero volts! ie: don't make the capacitors too large.
@bearchow: yes, a resistor connected across base-emitter (only one is required for the NPN/PNP buffer) can be useful.
For extra output current, I have used a resistor in the collector ckt of each of the voltage-follower transistors to then drive the BE junction of a complementary transistor (or drive the G-S in case of a MOSFET) with its collector (or drain) connected to the output (ie: the emitter-follower pair drive a common-emitter pair - this avoids the extra Vbe drop you would get with a second emitter-follower pair) - this is where that resistor across BE is really useful. (Sorry, I don't know how to upload a picture!) This arrangement generates heaps of drive current but need to be careful of stability.
Very good article, but misses a very important detail when describing the schematic in Figure 1: "a 50? source and was loaded with a 0.01 µF capacitor connected in series with a 1? resistor..." What is the value of the voltage Vcc? "...shows that the MMBT2222A is capable of sourcing almost 3A..." Yes, but capable at what voltage: 5V, 15V, 25V - it makes BIG difference!