Often is not a matter of "ignoring" but just of accepting that is shared by convent!. TODAY we CAN ignore (near safely) FTs; YESTERDAY we was CONSTRAINED to "ignore" FTs because also criyng desperately Ancient Greek or Chinese Mandarin don't change anything with bipolars, which was low Ft devices and and so remained at that time...
Hi
Piercarlo
There is still a very big difference among amplifiers built with 2 MHz, 8 MHz and 30 MHz ft devices. Don't fool yourself. You want the higher ft devices.
Merry Christmas,
Bob
Among 2 and 8 MHz is sure. I'm not so sure going beyond 10 MHz. In my viewpoint when devices going faster usually trouble coming faster. Obviously is not so easy destabilizing single bipolar devices but usually output stages are composed by a double pair of cascaded devices and this may open a peep in the door that may entered by stability issues. Higher for higher is worth remembering that what we want are audio amplifiers, not VIDEO amplifiers.
Higher FT make sense just if we use as a clue for estimating that it's remain high enough over the whole output current span. Not too luckly, modern devices has been linearized only in current gain over output current range. But the "constant line" in frequency response over the same current span is regrettably again yet to came...
Merry Christmas again - that is going to end... but has been a very niceful day!
Piercarlo
The superiority of the higer fT devices aside, a good amplifier is made not by the transistors used but how they are used.
Look at the benchmark 50W/8ohm Blameless with a crusty old pair of MJ820/4502 power output transistors; ~0.001% THD at 1kHz and ~0.01% THD at 20kHz, with 30 or 34dB or so of loop gain at 20kHz.
You won't better that with a pair of MOSFET's.
Look at the benchmark 50W/8ohm Blameless with a crusty old pair of MJ820/4502 power output transistors; ~0.001% THD at 1kHz and ~0.01% THD at 20kHz, with 30 or 34dB or so of loop gain at 20kHz.
You won't better that with a pair of MOSFET's.
Last edited:
But class AB does require fast output transistors because it's by definition a mix of switching and linear amplifiers. The output stage is switching every time speaker current crosses zero amperes in either direction. Class G and class H have exactly the same requirement, but with additional switching points.
Traditionally these amplifiers require lots of feedback for linearisation, either global or disguised as local to keep some people pleased.
Has anyone seen the base current waveforms of the output devices in a class AB output stage while sourcing/sinking high current at anything but bass frequencies? They usually look horribly distorted, with turn on and turn off spikes and 25%-50% non-linearity. And I don't mean in simulation, I mean probing in prototypes.
But, if mixing linear with switching is the problem, why not just going full switching? Nowadays we have excellent switching devices. Open loop linearity is far better, like in class A, and not much negative feedback is applied (or can be applied, that's one of the problems, or advantages, or challenges... )
Traditionally these amplifiers require lots of feedback for linearisation, either global or disguised as local to keep some people pleased.
Has anyone seen the base current waveforms of the output devices in a class AB output stage while sourcing/sinking high current at anything but bass frequencies? They usually look horribly distorted, with turn on and turn off spikes and 25%-50% non-linearity. And I don't mean in simulation, I mean probing in prototypes.
But, if mixing linear with switching is the problem, why not just going full switching? Nowadays we have excellent switching devices. Open loop linearity is far better, like in class A, and not much negative feedback is applied (or can be applied, that's one of the problems, or advantages, or challenges...
)
Last edited:
Just as an aside, the amplifier you mention is a perfect example of designing by numbers only... now if only you would listen to it and compare to other designs/topologies
Whatever you say. The point I was making is that even old slow devices in a good design with modest global negative feedback giving ample phase margin can have demonstratably very low distortion; even at 20kHz.
Last edited:
The point I was making is that even old slow devices in a good design with modest global negative feedback giving ample phase margin can have demonstratably very low distortion; even at 20kHz.
modest global neg. feedback is not exactly the word..
have you measured d. self s blameless s open loop gain?..
.
Whatever you say. The point I was making is that even old slow devices in a good design with modest global negative feedback giving ample phase margin can have demonstratably very low distortion; even at 20kHz.
....i.e good things starts from good numbers, not from bad ones!
Hi
Piercarlo
You had never designed power amplifier and you know nothing about "problems" related in obtaining really low distortions from output stages. Is difficult also with your "superior japanese transistors". And it should be so also with rf devices in output stages...
Piercarlo
The NFB is modest at 20 Khz (although 34 dB at that frequencies may be yet a risky affair)
Hi
Piercarlo
If the loop gain at 20kHz is only 30dB (I just looked it up, 30dB it is) then you'll have plently of phase margin, and yes, the gnfb then is modest.
When it comes to crossover distortion there are several contributing mechanisms, and the switching speed of power transistors is only one of them - and the sky does not fall in if you choose to use 4MHz fT devices in a decent design.
Last edited:
Hi Glen,
I don't recall Self achieving 0.01% THD-20 with MJ802 devices (which I also don't think were 4 MHz ft). I'll have to go review what he did and what he achieved.
Is suspect if he did it, it was with his beloved CFP output stage, which is very picky with quiescent bias, and in which he actually starves the output transistors to achieve proper crossover.
In any case, I agree that a good design with 21193/4 devices can achieve 0.01% THD-20.
A generously-biased MOSFET amp can also achieve that number without EC, but it will have somewhat higher idle dissipation. Two pair of vertical MOSFETs, each pair biased at 150-200 mA, combined with a 1 MHz gain crossover frequency, will do it. Such a design will not suffer nearly as much trouble with thermal stability and distortions due to junction temperature variations inflicted by program material.
Cheers,
Bob
![THD100W[1].png](/community/data/attachments/150/150731-72d50ae7760ab70c573eaa65f426d2bc.jpg)
Perhaps somewhat tangential to this conversation, I'm looking for an ouput BJT, but it must be a TO-3 package, and it must be actually available new. I've narrowed it down to an 2ST2121/2ST5949 vs. MJ21193/MJ21194, based in part on information presented in this thread.
For my case, worst case load is 5.3 amps. If I examine the fT vs. Ic curve for a 2SC5949, which someone indicated is the same as the 2ST5949, the fT drops like a stone at Ic above 2.5A. If I extrapolate the curve to 5 amps, the fT is substantially less than the "slow" MJ21193 which is chugging along at 4.5-5.0 MHz fT.
My conclusion is that at moderate currents, ~1 amp, the 2ST5949 could be a better performer - the fT spec is way higher, but once loaded down the MJ21193 is better, perhaps indicative of better transient performance. I could resolve the problem with more devices, but I'm already at 3 pairs and heat sink area is effectively used up.
So, which to use?