Hi,
I'm not sure about Ft, one would presume the higher the better,
as long as the output stage is not prone to self oscillation.
There are a number of interesting issues hinted at but not explored
in Self's amplifier design books, which when combined might make
an interesting amplifier.
The main points of the basic amplifier.
1) Degenerated PNP LTP input, current source for the tail with correct decoupling.
2) NPN "darlington" VAS, current source for the tail with correct decoupling.
3) Accurate and stable biasing
3) Two stage output DEF, or CFP, with 0.1R value emitter resistors
The interesting points :
1) Optimum input degeneration.
2) Cascoding the input to the VAS, to use as a reference point
for the compensation scheme and consequently -ve PSRR.
3) Optimum 2 pole compensation.
4) (how to make slew rates more symmetric)
5) 2 transistor bias looks appealling.
5) The double EF stage is shown to have the lowest first watt distortion.
6) Multiple output device pairs, say 3, all with 0.1R, will have lower
distortion and as the current triples you get 3 x more "class A"
whilst maintaining optimum bias.
7) My suggestion : changing the drivers to CFP might be an idea,
only needs two small signal transistors and swapping the drivers,
the resulting triple stage biasing is the same as the double EF,
and consequently the distortion profile should be the same except
for the high current 3rd order stuff, which should improve.
(would this topology be prone to oscillation ?)
The boring points (IMO) :
1) Getting the best possible noise performance, it should be good.
2) Getting the best possible THD performance, it should be good.
(What matters is getting the best distortion spectra at all power levels)
3) Getting the best possible PSRR, it shoulds be good and symmetric.
/sreten.
I'm not sure about Ft, one would presume the higher the better,
as long as the output stage is not prone to self oscillation.
There are a number of interesting issues hinted at but not explored
in Self's amplifier design books, which when combined might make
an interesting amplifier.
The main points of the basic amplifier.
1) Degenerated PNP LTP input, current source for the tail with correct decoupling.
2) NPN "darlington" VAS, current source for the tail with correct decoupling.
3) Accurate and stable biasing
3) Two stage output DEF, or CFP, with 0.1R value emitter resistors
The interesting points :
1) Optimum input degeneration.
2) Cascoding the input to the VAS, to use as a reference point
for the compensation scheme and consequently -ve PSRR.
3) Optimum 2 pole compensation.
4) (how to make slew rates more symmetric)
5) 2 transistor bias looks appealling.
5) The double EF stage is shown to have the lowest first watt distortion.
6) Multiple output device pairs, say 3, all with 0.1R, will have lower
distortion and as the current triples you get 3 x more "class A"
whilst maintaining optimum bias.
7) My suggestion : changing the drivers to CFP might be an idea,
only needs two small signal transistors and swapping the drivers,
the resulting triple stage biasing is the same as the double EF,
and consequently the distortion profile should be the same except
for the high current 3rd order stuff, which should improve.
(would this topology be prone to oscillation ?)
The boring points (IMO) :
1) Getting the best possible noise performance, it should be good.
2) Getting the best possible THD performance, it should be good.
(What matters is getting the best distortion spectra at all power levels)
3) Getting the best possible PSRR, it shoulds be good and symmetric.
/sreten.
Christer said:
Actually, I think this is one of the most sensible answers to the question. I am afraid the advice is not that simple to follow though.
Otherwise we have the excellent component called voltage controlled voltage source. Unfortunately the manufacturers still seem to have problems putting it on the market.
I prefer power controlled power source. Both extremes are bad.
As the result, I accept loss of half of microphone and speaker powers on maximally linear I/O resistances (my greatest secret!). No need for special components, just optimal usage of what we have already.
Now... Thank you gentlemen for interesting discussions, I have to make a break: if I don't finish my current prototype of Hi-end PA active speakers before a folk festival I loose a lot. 2 weeks only left...
sreten said:
6) Multiple output device pairs, say 3, all with 0.1R, will have lower
distortion and as the current triples you get 3 x more "class A"
whilst maintaining optimum bias.
7) My suggestion : changing the drivers to CFP might be an idea,
only needs two small signal transistors and swapping the drivers,
the resulting triple stage biasing is the same as the double EF,
and consequently the distortion profile should be the same except
for the high current 3rd order stuff, which should improve.
(would this topology be prone to oscillation ?)
My RMI-FC100 amplifier have 3 pairs with 0.1R Re and CFP drivers.
The only method to keep OPS away from self oscillating is to bias CFP drivers very high.
The trick is to not allow drivers to go into class B where instability and oscillation appear near crossover point.
http://www.diyaudio.com/forums/showthread.php?postid=1349109#post1349109
Hi,
There seems to be a previous misunderstanding of D.Self.
For Class B/aB/AB crossover distortion is the dominant mechanism.
It is not the only mechanism, but can be the only one that matters.
Class A has no c/o distortion but each stage is not perfectly linear.
As Self shows the input stage driving the Vas creates hf distortion,
3rd order, the Vas ~ constant 2nd order, and the output stage for
class A "wingspread" diagrams are not flat either, mainly 2nd/3rd.
Self has not attempted to identify the limiting areas for class A,
i.e. the relative levels of each stage, but suspects the input.
It should suffice to say it is at a very low level, mainly below
the noise floor except at h.f., it has low high harmonic content,
unlike non - class A, and with standard single pole compensation
a rising distortion with frequency level is inevitable, for the
standard topology this is also true open loop.
Double pole compensation .... for class A ?
no from what I just found out on the forumhere.... AB bjt type amplifiers while playing low (normal listening level) output power there transistors do not turn off! so no crossover distortion!! only when you begin to play or use up the watts does it turn into class AB and the devices do turn off
Lawrence
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.