Tubes, FET's & BJT's - a few in-depth comparisons

[volki]

although i'm more or less familiar with tubes, JFET's and BJT's, i keep missing a few things when it comes to comparisons of orders of magnitude and such. so just to get it straight one time, maybe someone could fill me in with that? thanks a lot people :green:

A) what i'm rather sure about:
1. tubes are much more linear than BJT's.
2. tubes and JFET's have higher i/p and o/p Z and a higher dynamic range for i/p voltage
3. linearity of JFET's and tubes is dependent on optimum bias point, where for BJT's is mostly due to negative feedback
4. hierarchy of practical values of open loop voltage gain is BJT > pentode > triode > FET. relating to an amplification factor of 50-500 / 100-150 / 20-60 / 2-50.
5. triodes produce strong 2nd order harmonic, pentodes are more likely to produce 3rd order
6. tubes have a higher corner frequency than BJT's

B) what i'm not sure about:
1. does open loop gain change approx. inversely with distortion? meaning, statement A4 inverted, the hierarchy regarding linearity would be FET > triode > pentode > BJT?
2. somewhere i've read that JFET's are supposed to have a higher dynamic range than tubes i/p wise? which orders of magnitude can JFET's and tubes have for that, w/o NFB applied? for transistors, i read it's only about 1 mV for 1% THD...
3. do JFET's behave like pentodes distortion wise? meaning, they produce strong 3rd order harmonic? where do BJT's come in there?
4. regarding miller capacitance and such, i guess triodes have a higher corner freq than pentodes. what about FET's, and how much lower are BJT's? which orders of magnitudes are we talking about here?
5. what about electron transit times (for BJT's, too?)

of course some of these may be gross generalizations, there may be many exceptions of the rule, but just to get in the ball park. also, i'm aiming about the very propoerties of the components themselves, not about different kinds of circuits they can be used in. meaning no feedback unless mentioned, etc. and no power o/p applications, either.

thanks a bunch people :razz:

 

[PRR]

> some of these may be gross generalizations

Yes.

And you should also consider "practical" versus "ideal". The dynamic range of a VT can be quite large if you don't mind huge power waste. OTOH, BJTs usually give higher gain-bandwidth product than VTs, except since we can run BJT at very low power, we sometimes starve the highs right out, which in a VT is hardly worthwhile after you get it hot.

A lot of confusion can be found by looking at voltage gain. You should look at power gain too. You may not be able to use the ideal power gain without ideal transformers, but ultimately it IS all power.

However a VT's low-frequency power gain is "infinite", because its input impedance is infinite (or pretty darn high). In practice you are limited at the bottom by how high-Z a transformer you can wind, and at the high end by stray capacitances.

When you really need gain-bandwidth out of bottles, pentodes always win over triodes. Pentodes have much lower Miller (even at high voltage gain). You may have to load them with little 500 ohm loads and accept gains of less than 10 per stage to get many-MHz, but that is true in all wideband work.

Linearity is semi-simple. The perfectly linear device has the same Gm at any current. A naked BJT has a Gm (and gain) that rises directly proportional to emitter current. A standard VT's Gm varies very roughly as square-root of current. A JFET is similar. So if you bias each at 1mA and try to swing up to 2mA and down to 0mA, the BJT has over 25% THD, the VT around 10% THD.

BUT: a practical resistor does not vary with current enough to notice. If you stick a resistor under the cathode or emitter, you get a lower but more constant Gm. And at similar currents the BJT has much higher Gm than any VT. So when you stuff enough resistance under BJT to make its Gm similar to a VT, the Gm is VERY constant and linearity is excellent.

Pentodes only go 3rd harmonic at high levels. Low levels tend to be pure even-order. FETs also, except they don't have a point where odd harmonics rise slowly, they just flat-top.

> there may be many exceptions of the rule, but just to get in the ball park.

It is all about evading as many "rules" as possible, even if you have to pack-up the team and move to another ballpark.

And be very aware of the fact that excellent designs have been done with all three types of devices. Don't judge a book by its cover. Details often matter more than devices. And it is unusual for anybody to master even one device, let alone all three, so don't expect to ever understand it all.

 

[volki]

thanks prr - i had actually been expecting one of your valuable posts on this one :wink:

The dynamic range of a VT can be quite large if you don't mind huge power waste

.
talkin' low distortion still?

OTOH, BJTs usually give higher gain-bandwidth product than VTs, except since we can run BJT at very low power, we sometimes starve the highs right out, which in a VT is hardly worthwhile after you get it hot.

so you mean, in low- or line level circuits, there's not enough current around to get a decent gain-bandwidth product for BJT's in the first place, which forces you to go into squashing the voltage gain by NFB to extend practical bandwidth? but given enough power, a BJT's corner freq would exceed that of a VT?

If you stick a resistor under the cathode or emitter, you get a lower but more constant Gm. And at similar currents the BJT has much higher Gm than any VT. So when you stuff enough resistance under BJT to make its Gm similar to a VT, the Gm is VERY constant and linearity is excellent.

so obviously NFB (serial on this case, or parallel, too) is mainly used to reduce nonlinear distortion, not to extend bandwidth. the latter obviously is mostly done with op-amps, where the low pass fx of multiple stages add up.

A standard VT's Gm varies very roughly as square-root of current. A JFET is similar. So if you bias each at 1mA and try to swing up to 2mA and down to 0mA, the BJT has over 25% THD, the VT around 10% THD.

what about a comparison btwn VT and FET for low-level app's, e.g. a mic pre, or better still, the impedance converter (head amp) of a capacitor mic? the capacitor's power to drive the head amp's input being really low. - for class a w/no ac feedback: huge difference in max. i/p voltage swing and distortion characteristics?

also, to stay at the example of mic's - you often see pentodes wired a triodes, obviously because some pentodes have preferrable properties such as high input R, low stray capacitances, low leakage current, etc. what about distortion characteristics? do pentodes wired as triodes automatically take up all their gain characteristics, or do some pentode characteristics remain?

And be very aware of the fact that excellent designs have been done with all three types of devices.

that goes without saying :green: