|
|
|
FET (more exactly MOS-FET) class AB output stages topologically may be the same as class AB bipolar output stages are. Usually is a full complementary source follower, more occasional a compound-Sziklay like stage (that employ often a bipolar driver pair connected as emitter follower followed by a pair of power MOSFETs connected in common source fashion). Some early design have also employed quasi-complementary design.
MOS-FETs have many advantages over bipolar counterparts (mainly increased safe operating area, reduced dynamical current drive and practical absence of switching time in audio application which, with the aid of a reasonable amount of NFB, can improve effective power bandwidth up to frequencies not reachable even by the fastest power bipolars available) but they come with specific drawbacks that must be taken in account in designing and building MOS-FETs power amplifiers.
First of all should be demystified the old legend of "self-limiting" power dissipation and "self-sharing" current management when a bunch of these devices are connected in parallel. Those features were caracteristical of some early MOS-POWERs lateral devices (like 2SK135 and 2SJ50) produced by Toshiba (and now officially discontinued, although they are still produced - with other signatures - by second source suppliers, often born specifically to cope the audio market demands). More recent devices, as may be the IRF series from International Rectifiers, are in fact totally different devices which, being derivered from the same basic process used to fabricate power bipolars (only the step of fabricating gate structure is specific), share with bipolars the same hazard of thermal runaway what may drive them to a plain thermal blow-out.
All MOS-FETs however come with an other specific drawbacks which may harsh amplifier design, especially from the layout point of view. They, because of very high FT, simply don't tolerate layout not enough cured in their parasitics, that may somewhat turn these devices in a powerful (and self-destructive) Armstrong-like oscillators. MOS-FETs power devices *can't* be connected with wires to the driver circuit but should be closely wired on PCB to their driver stage which has, among his tasks, that of offering an adequate low driving impedance, low enough to damp the efficiency of internal feedback, which otherwise may lead the device into oscillation mode (especially, as told before, when parasitic strays, especially residual inductances, aren't appropriately reduced).
Despite of apparences, driving MOS-FET devices is not an easy task, especially when driven to high frequency and when modern IRF-like devices are employed (which exhibits awfully consistent internal capacitance - furthermore, differently from those boasted from early Toshiba devices, very non linear). Drivers, in order to proper charge and discharge internal capacitances of MOS devices, should provide a current drive practically proportional to the *speed* of signal applied.
Another drawback of driving MOS-FETs (especially when directly driven with no intermediate - and adequate - driver stage) is that VAS section see a load variable (sometime *heavily* variable) with frequency: not precisely a capacitive load but nearly so. This is due mainly to MOS-FETs consistent gate-source capacitances which, with increasing frequency of signal, "couple" the VAS output with the output load, which progressively short to ground the effective working impedance of VAS, so reducing the high frequency open loop gain of the system. If VAS is not appropriately design to make its intrinsic gain fairly independent from the loading impedance (a thing that require the embedding of some buffering and hence some extra active stage for doing it), the entire amplifier may exhibit very poor ("muddy") high frequency response, despite of the inherent high FTs of the devices employed in. This may be particularly evident with some "hybrid" design employing tubes in driving stage where, in spite of a *static* well behaviouring, may unexpectedly struck when a combination of high power/high speed power delivering is required by the load in response to the music content to be reproduced.
Other drawbacks of MOS-FETs are mainly the inexistence of real complementary pairs (especially from the transconductance and internal capacitances standpoints) and their undefined cross-over region which sometimes appear just "wobbly", with a kind of "double S" compounded transcharacteristics, which often lead the entire distortion amplifier to behave, when output power increase, in a fashion not too dissimilar from that boasted by the more usual (and more controllable) bipolar Darlington output stages.
Furthermore, biasing MOS-FETs output stage require, the same as in bipolar design, a Vbe multiplier circuit well thermally coupled with output devices just as a partial and minimal solution for avoiding thermal runaway. Effective and more satisfactorily solution however must employ dedicated circuits which, with the use of some feedback, stabilize the quiescent current of output stages. A good example of this solution exist in a dedicated integrated form, proposed by Linear Technology, with the IC LT1166.