BJT vs MOSFET - taking the best of both

As my first posting I would like to ask this question:

I am planning to build my own Class A power amplifier, and with my current knowledge of electronics, I should be able to do such a thing. I -of course- want this amp to be better than best, and take the good things of all components... And so there begins the battle between MOSFET and BJT.

Both have advantages, as wel as disadvantages. But in some cases these (dis)advantages change. For example, a MOSFET would do good in a power stage, because it handles power better than a BJT (correct me if i'm wrong), while a BJT would do nice in a line stage.

Who can tell me when to use what device where and why? What exactly are the advantages of the one device over the other? When are those advantages best utilised, and when manifest the disadvantages themselves strongest?

I'm eager to know :)

Bouke
 
I'm not an expert, but...

Here's my take on things. I don't claim this is right or wrong, just my opinion. I would be delighted to hear other opinions.

There isn't a really good device for the output stage of a power amp. They all stink. :D

BJTs tend to be slow, they're subject to thermal runaway, and they have safe-operating-area (SOA) restrictions that often spells doom with highly reactive loads. Beta droop in most power BJTs is a major source of output stage non-linearity. Push-pull biasing is moderately tricky and the proper bias point is not widely known. But they can produce output voltages very close to the supply rails if properly applied, the Vbe-Ic relationship is very predictable and uniform across devices, and they tend to be cheap.

MOSFETs are fast - sometimes too fast. There's a lot of variation in threshold voltage in production devices, and it can be difficult to devise a push-pull biasing scheme that properly tracks threshold voltage changes with temperature. The gate oxide can be fragile (this is mostly a concern for handling prior to installation). Gate drive is primarily capacitive, and this can be a problem if you don't design for it. Rds(on) limits how close to the rails you can take the output voltage under heavy loads - and it increases as the devices warm up. But they don't have SOA problems, and don't play the thermal runaway game. Oh, and they tend to be more expensive than BJTs.

Both device types have the problem that complementary pairs tend not to be well matched.

For small signal gain stages, it's a case of horses for courses. BJTs can work very nicely in differential pairs for low impedance sources. JFETs are better for hi-Z sources, and for one-transistor gain stages if you don't mind (or are actively seeking) even-harmonic distortion. JFETs make excellent current source loads. I'm not up to speed on MOSFET noise behavior at these levels, but I imagine they would be fine for extremely high impedances - e.g. piezo transducers. MOS certainly seems to work fine in radio frequency gain stages.

At line levels it's your call. I lean towards BJT input op amps because I'm lazy, and because JFET and MOSFET op amps tend to have input stage shortcomings (poor common mode rejection, high input capacitance, etc.). If you were looking for major headroom, I imagine high-voltage MOS in line stages would be a good choice, but mostly because high-voltage BJTs are getting harder to find (SOA is a problem here too), and JFETs have never been common as high-voltage devices. The fact that MOSFETs are available in both enhancement and depletion mode types adds flexibility.

So what have I left out, and where am I wrong? There must be something bogus in this post.
 
If you want to build a pure class A amplifier, BJTs are more linear than MOSFETS if you use very large BJTs, having very high power handling capacity, and Ft greater than 10 Mhz, for instance 2SC3281/2SA1302 (Toshiba), which are specially designed for this purpose.

But BJTs which are designed for general industrial use (TIP 36...) are quite unusable for very high quality service, standard MOSFETS gives better results than such cheap devices, at the same cost.

In class AB amplifiers, it is very difficult to eliminate totally the crossover distortion , even with good BJTs, and MOSFETS gives better results, because of their "soft start" transfer characteristics.

P.Lacombe
 
I agree with Chucko's comparison. I would just add some more flavouring. What device you choose should be driven by what you want the device to do in the circuit topology you choose. Understanding what you want it to do is essential - at an electrical level.

Mosfets are transconductance devices: you must charge their relatively high input capacitances up and down in order to vary their drain current. They are majority carrier, resistive devices. Their transconductance is non-linearly related to drain current (approximately a square law). Outside hifi they tend to be used in high-speed switching applications, like switch mode psus, beacuse they have negligible turn-on and turn-off delays and do not suffer from secondary breakdown (tend to be reliable). Transconductance tends to decrease with temperature.

Bipolars are current gain, minority carrier devices. Collector current is related to base current, often fairly linearly over a certain range of collector currents. The devices have input capacitance although much less than comparable FETs. Base-emitter voltage is non-linearly (exponentially) related to base current. Current gain tends to increase with temperature.

In my experience, bipolars are used most often in high-end power amps (with all due respect to Mr Pass :) ). I would speculate that this is because it is easier to achieve linear current gain, their transconductance is higher in the usual operating region, they enable greater phase margin with less devices because of their lower capacitances. Bipolar designs need more care wrt exceeding safe current operating limits and often have carefully designed current limiting circuitry. A great sounding amp can be made with either device provided you understand how they work and design circuitry around them to compensate for their foibles.
 
traderbam said:
Bipolars are current gain, minority carrier devices. Collector current is related to base current, often fairly linearly over a certain range of collector currents. The devices have input capacitance although much less than comparable FETs. Base-emitter voltage is non-linearly (exponentially) related to base current. Current gain tends to increase with temperature.

Lots of people seem to believe that BJTs are current-driven devices. Doug Self has convinced me that they are instead transconductance devices.

Self asserts that the relationship between Vbe and Ic is the main driver, and that Hfe (current gain, a.k.a. beta) is merely a side effect. He points to the fact that the exponential relationship Ic = e^(Vbe/k) is a very good approximation across several decades of collector current for the vast majority of silicon BJTs, and that there is little consistency between the beta curves of most BJTs - even of the same type!

This doesn't mean that you have to live with an exponential transfer curve; there are lots of tricks, using current mirrors and the like, to enforce a more linear relationship.

I agree with the rest of your comments, traderbam, but I didn't want to leave this assertion unchallenged.
 
Linearity of BJTs begins in less current than in FET(needs to be biased for high currents to be linear). BJTs have smaller tolerances than FETs. FETs don't have current noise (near to 0), so they are best for phono stages and similar low voltage applications. Fets have higher capacitances than BJTs, so they need more current for higher slew rates.
And it is not true that FETs are faster than BJTs. This could be true about 10 years ago.;)

Finally both are good for audio (FETs - Nelson Pass - great amps)(BJTs - Dieter Burmester, Mark Levinson and many others - also great amps). You have to know everything about device before you use it.
 
True, true... designing with mosfets requires a lot more trial and error in my experiences. I have successfully designed and built many full BJT amplifiers -- and getting mosfets to work well in the OPS instead of BJT's took about as long as the entire design.

BJT's are better for the first time designer/builder, much more forgiving when it comes to stability and biasing, ect...


DarkOne said:

Finally both are good for audio (FETs - Nelson Pass - great amps)(BJTs - Dieter Burmester, Mark Levinson and many others - also great amps). You have to know everything about device before you use it.
 
"Self asserts that the relationship between Vbe and Ic is the main driver, and that Hfe (current gain, a.k.a. beta) is merely a side effect. "

Isn't that a little like saying a resistor is a transconductance device since its current is related to its voltage? I guess it depends how you choose to look at it. I guess Mr. Self is saying that the exponential relationship between Vbe and Ic is more consistent across devices than hfe. Hideous isn't it? In many applications it is desirable to think of a bipolar having a small-signal gm of roughly 40Ic - such as when calculating the transconductance of a long-tailed pair. There are now a range of power BJT where the hfe is very constant with Ic and it may therefore be more useful to design around the hfe rather than the exponential gm. Horses for courses.
 
Here's just my opinion:

Have you ever considered IGBT devices?

And I don't know if someone of this forum has experience with room acoustics, but it's proved that the acoustic characteristics of the room are dominant at one or two meters distance of the speakers you're playing with (in an average room, not in "dead" rooms). This is because of the facts like reverberations, echo's, etc... So at this distance these characteristics are more important than the way your amp and speakers will play the music.
Of course it's nice to start with a proper signal, and maybe you'll see the difference between mosfets and BJTs on a scoop, but I wonder if anyone can hear the differences between a good mosfet amp and a good BJT amp!!! This is all called relativity, :) .

This is just my way to analyse such problems, I don't want to be the anarchist here!!


Best regards,

HB.
 
hugobross said:
Here's just my opinion:

Have you ever considered IGBT devices?

Actually, I have... and already have sniffed this forum for info about that. I have learned that IGBT's aren't very suitable for the job. In Fact, they suffer from the disadvantages from BJT's and MOSFETs. High input capacitance, and nonlinearity. IGBT's were made for switching applications, and they perform perfectly there. I could use IGBT's in controlling biascurrents, i've heard the perform well in that too....

Reading all opinions posted so far, it seems that MOSFET's have most advantages over BJT's, and can be used best in DIY designs.

Does anyone know a little more about the application of JFETs instead of regular MOSFET's? And how about HEXFET's?

Bouke
 
It has been stated previously in this thread but I don't think that there has been enough emphasis placed on this point: MOSFETs are not linear devices.

As mentioned above by traderbam, the relationship between the applied voltage and the resultant drain current is approximately a square: Id = K*Vgs^2 (Where K is a constant). Therefore, theoretically, no matter how much feedback you apply you will be unable to reproduce the signal linearly.

You might now be saying 'Oh, this guy's a MOSFET hater' but that's not true. MOSFETs have their advantages: Little to no input current being one advantage, which would perhaps make them very useful as the input device of a line stage amplifier.

As for BJTs, they aren't linear devices either but they are more linear than MOSFETs. Note that not all BJTs are equal, some are more linear than others. If one takes care in both choosing appropriate BJTs and biasing in class-A, they stand a much better chance of amplifying the input signal linearly.

With that being said I believe that one can biuld a good amplifier using MOSFETs but could just as easily build an exceptional amplifier using BJTs.

Thanks,
Morrist
 
Nitpicking

"Isn't that a little like saying a resistor is a transconductance device since its current is related to its voltage?"

A resistor is simply "conductive", because the change in voltage across the resistor results in a change of current through the resisitor. Devices are described as "transconductive" only when the change that occurs happens in a different part of the circuit. A bjt IS a transconductive device because the change in Vbe results in a change in Ic, and Ic does not travel into or out of the base.

Also someone stated an equation in correctly.
It should read Ic = Is*e^(Vbe/Vt) where Vt is the thermal voltage (26mV at room temp for all trans) and Is is the saturation current(typically 10^-12 A to 10^-14 A). Is is device dependent.

Well, that was my first post. I hope I didn't **** anyone off.
 
traderbam said:
I guess Mr. Self is saying that the exponential relationship between Vbe and Ic is more consistent across devices than hfe. Hideous isn't it?

That's certainly the way I read it.

And no, I don't find it hideous. It's a challenge. If anyone ever finds a solid-state device (or electronic amplifying device, period) that is inherently linear, audiophiles everywhere will rejoice, and engineers everywhere will be out of work.

Until then, we have to make the best of what we've got.

There are ways to use the BJT's exponential Vbe->Ic relationship that are inherently linear. They're just not easy to apply unless you're a chip designer.
 
"Have you ever considered IGBT devices? "

Hugo,

IGBT are specially designed for switching purposes, at high currents and high frequency, not for linear amplification at audio frequencies. In my opinion it is impossible to obtain a clear sound with such devices, which are more suitable for cooking electrical appliance...

Regards, P.Lacombe.
 
Hint of the day!!

Take one VQ1000J from Vishay.
Use only 2 of the 4 MOSFET's
Set an idle current of 90mA and a VDS of 10V.
Then check the "Transfer Characteristics" in the datasheet.
Also check the values of Ciss, Coss, Crss.
Add some degeneration sourceresistance in the size of 10 - 20 Ohm.

I must say this is better than BJT and even nearly 100% matched.

The perfect one would be a ZVN3310 idling at .2 - .4 Amp and 25 Vds. but it cant handle the heat. :(

Sonny