Author Options:

BJT vs MOSFET? Answered

-- background:
In my perpetual pursuit for designing the best, cheapest, & best performing flyback (line transformer) driver, I've decided to try out some big BJTs, which appear to have higher voltage & current ratings @ considerably lower cost. In the past, I have tried MJE3055's, which work OK, and allow the generation of thin blue arcs from a 12V supply for a few minutes until the transistor dies due to high voltage kickback or overheating.

The FDP33N25 gives good results but is somewhat unreliable at 24V. So I decided to give these alluring "PHE13009" 400V 12A rated NPN BJT's a try. However driving the transistor adequately seems to be the problem. (I didn't realize these transistors would require like 5A base current w/ only HFe of 2!) It seems like almost all the "good" driver schematics utilize large $$$ FETs or even IGBTs, but almost every CRT, plasma globe, and ballast (SMPS) I took apart seem to prefer high power BJTs probably due to this exact cost difference. I was only able to get simalar performance to my 33N25 MOSFET when I stuck an additional TIP120 in as an additional darlington stage, which worked very nicely (white hot arcs) for about 1 second, then it popped! :( Since the collectors are tied together they are all exposed to >200V transients, I am sure that's what killed it. I substituted that transistor for another PHE13009 to see what would happen and with a third driving stage (2N2222) I could get somewhat acceptable results but I know I can do better. >:)

- My actual questions: 

* The GDP of this transistor is 40MHz, seems fast to me. (certainly faster than the 2MHz GDP of the 2N3055 which works well in my slayer exciter.) and MOSFETs have significant gate charge. ECE2630 glossed over transistors mentioning BJTs are faster, (small signal ones, anyway :P ) and some internet sources agree, but I am finding lots of sources saying the opposite! What's the deal? Which one "faster?" 

* How to traditional BJTs compare to IBGTs? Which are faster/better? The datasheet for this transistor explicitly mentions its use for "high frequency ballast and switch mode applications" which implies that it is well suited for my needs. It also includes a several inductor test circuits, but other than that the datasheet is pretty bare-bones.  I'm actually a little disappointed the HFe is so low (around 2-3) @ >10A. I may require more windings on my primary and a 48v supply. (not ideal)

* I know I will dissipate a little more heat due to the base current, and something I didn't consider was heat dissipated in the resistor biasing/controlling the base. (like 12V at several amps just to drive the damn thing!!) so it is worth the cost benefit of $0.5? Is there a configuration I could use that makes driving the transistor easier?


No schematic, how depressing.

The only advice I am going to give is to increase efficiency power the base through the primary like this.

Inverter 2.gif

This was my original schematic, I was hoping to be able to make a few changes to instead drive BJTs. (in fact, I think it would work if I removed the series capacitor from the gate drive and played with the value of the components.)


I see said the blind man.

Going by the data sheets they do have slow switching times for your application.

how are they slow? the PHE13009 has 40MHz GBP and 0.1uS fall time, which I thought was good enough compared to MOSFETs. The circuit you have I already know is going to have really bad rise times, given the nature of the oscillator (I assume you primarily showed it to show how 'they' drive 2N3055's, which are slower at 2MHz GDP.

Yea the schematic is just for the 2N3055 everything but the signal from the flip flop goes through the primary of the step up transformer, very efficient.

The datasheet I read had a 1.2 uS fall time but that happens with different sources.

Actually, I took a closer look at that configuration, now I think I understand what it's doing. It first pulls a small amount of base current from the supply through the weird PNP darlington configuration, then feeds current into the base of the 2N3055, then that yanks the voltage right to ground. Only problem is that it cannot ever reach ground because there has to be some current going into the emitter of the PNP darlington to maintain a current in the 2N3055 base, so I'd imagine the 2N3055 never saturates or turns on particularly hard. I'll simulate it and maybe try it out and see if it works! It certainly is something I have not entirely thought of.

This is the datasheet I am using:


At first glance, I assumed the driving transistors were in a darlington configuration used as an emitter follower, but I think it is actually using PNP transistors in the more typical common emitter mode of operation, so it has to pull base current out of the base to allow current flow through the 2N3055's. I wonder if it would be better to instead use a NPN-based darlington in the emitter follower config, so that the base current is included in driving the 2N3055's. Obviously a couple volts less than the supply rail is not going to be an issue given that the maximum base voltage is going to be around 2-4V.

I have heard from somewhere that if a transistor is allowed to saturate, then it takes longer to turn off, something to due with charge carriers and voodoo quantum physics. What's the deal here?

Thanks for the schematic! My primary reason for going with BJTs was that I thought it would be cheaper, but it appears as though I am going to need either a larger heatsink, or some power resistors like those 22 ohm 5W ones, which makes a simple push pull gate driver much more appealing in terms of cost. Nevertheless I will try to parallel a bunch of smaller 220 ohm resistors to see if I can get this to work. Thanks!


1 year ago

Also, I keep hearing this "saturation" term being thrown around. It seems like saturation region is the region to be in for switching applications, but that does not make sense because when a transistor is saturated, the dynamic impedance goes way up and there is significant Vds or Vce drop, which in my understanding is VERY inefficient. Wouldn't it be best to operate in the linear region with as hard of a gate drive as possible? To reduce the Vds or Vce as much as possible to ensure good performance?

Vce (sat) is the lowest it can be, its significantly higher than a FET. Until fairly recently, if you wanted high voltage switching, the BJT or IGBT was the only way to go.

No joke! I can't get a 2N3055 to have lower than 2V at 10A! Pathetic. The 100 mOhms on-state resistance of my chosen FET caused a maximum of 1V to develop across it. Guess FETs are the better choice for this application.

But to ask other questions, which are faster and why? BJTs, IGBTs, or FETs? Can I construct a IGBT from discrete components?

No, an IGBT is a specialist fabrication, like a proper thyristor or triac. Like I said, until recently you had little choice: for real power and high speed, it had to be a BJT. Fet speed is dominated by size. V channel designs are faster and have lower gate capacitance. Switching speed is determined by many parameters I can't remember the details of, 30 years out of University

I did once build an SCR out of discretes, but only tested it with low power. :) It had a harder time staying latched but it worked! So since the marginal cost of using these BJTs is now effectively zero, how could I achieve the 2A base drive with a 24V supply without dissipating too much power? Would a darlington configuration be appropriate? (to get the base drive down to maybe 700mA)

Darlngton drive will sacrifice the speed you need. Realsitically, you don't DO what you want with BJTs

I should use my vacuum tubes instead :)

I don't know how cheap, how simple and how effictive you wanna go....
But I do know my induction heater circuit works quite well with flybacks.
The mosftes I used are quite cheap too and perform well in the frequency range.
If you can accept 5-8 turns on the primary it should be just fine.
You can try lower numbers of windings until the mosfets fail due to the higher frequency.
Ready to go kits are available from china on fleabuy in case you don't want to design the circuit board too.

The 33N25 MOSFETs I have been using cost $1.36 in 10x quantity, while the BJT mentioned above is like 70 cents. But I guess they are in fact better suited for high voltage low current medium power applications. I messed around in LTspice using 2N3055's as substitutes and I think I have a circuit that should perform almost as well as MOSFETs when I build it, we'll see.