Chronic said:
Only if the bandwidth and the phase margins are small relative to the
target audio bandwidth, which is unlikely.
Ahmmmm....
Lets say we are after a 50khz closed loop BW, i.e. so that there is
negligible loss at 20khz. Now suppose we are after silly distortion levels,
say , 0.005% at 20Khz. Typically this means large amounts of feedback, say a
minimum of 40db at 20khz, maybe up to 60db even. This means that we need
say, a loop unity gain bandwidth of say 5Mhz min. Now, if the Ft of a big,
high voltage high current, bipolar was say 50Mhz, which is a tad on the fast
side, its current gain would have dropped to 10, which aint so great.
Furthermore, it would be really pushing a 50Mhz ft transistor to get an
overall stable loop at 5Mhz., not forgetting that there will already be, by
design, a dominant pole rolloff, prior to the output stage.
Lets do some sums:
Cin of a bipolar ~= gm/2.pi.ft X re/RL, in emiter follower mode, i.e. Cin =
1/(2.pi.ft.RL)
For a 50Mhz bipolor this would be 800pf. at 4 ohms.
A mosfet, would be Cin ~= Cgs/(gm.RL), which at typically 1A/V and 600p,
would be Cin=125pf.
So, despite the much larger gm of a bipolar to back off its inherent large
Cbe in source follower mode, they still typically need much more high
frequency drive than mosfets. Furthermore, without additional buffering,
this larger capacitance kills the h.f gain of the class a main gain stage,
as already mentioned by Graham. Indeed, in the early 80s, such 50Mhz devices
were made from unobtainium.
There are a lot of other details, but I really don't have the time to go
into any more detailed technical design at the moment.
So... try putting full on voltage on a mosfet without a heatsink for a
while, then try that with a bipolar!.
And hopefully I haven't made any errors in my calcs;-)
Kevin Aylward
http://www.kevinaylward.co.uk/ee/index.html
