..if you try you simulation complentary doubled then
you should not only find symmetric current sharing in the input stage.
You can also skip the current source in you output stage...
the output stage may run with 3..4mA...
Or first try Evas proposal with R6 175 Ohms...
should also end up in good balanced currents.
To All:
Have fun!
*leaving_soon_for_real_live_evening_activities*
you should not only find symmetric current sharing in the input stage.
You can also skip the current source in you output stage...
the output stage may run with 3..4mA...
Or first try Evas proposal with R6 175 Ohms...
should also end up in good balanced currents.
To All:
Have fun!
*leaving_soon_for_real_live_evening_activities*
millwood said:
Dude, you are going in circles,.
That's the single biggest flaw in this discussion, and the simulation was to show you that an ltp will not by itself automatcially balance leg current.
The LTP can not balance 100 % perfectelly...but with emiter resistors will be close enough...
Also the DC off set of the amp...will never also be 0 ...but close enough !
What is 100% correct???
Current mirrors are used for increased current drive for the VAS...because with CM there are no current lost in the colector load resistor of the LTP and all current from it, can be used for driving the VAS...millwood said:
Otherwise, people wouldn't be inventing current mirrors to balance leg current in LTPs,
I don't want to disturb the discussion about the input and VAS stage (only noting that the analysis in http://www.diyaudio.com/forums/showthread.php?s=&threadid=16796 seems be extensive and accurate), but add a point about the EF output stage.
In several places it is recommened not to tie the driver (and pre driver) transistors' emitters to the output node, but connect them with a single resistor additionally paralled by a 1uF capacitor.
Interestingly even the authors suggesting this in their "theory" chapters, don't (always) follow their own advice in the circuits presented in the "practical" chapters.
So can anybody shed light on this theory/practice split? Would the performance of the original poster's circuit be enhanced by observing the theoretical advice?
Regards,
Peter Jacobi
In several places it is recommened not to tie the driver (and pre driver) transistors' emitters to the output node, but connect them with a single resistor additionally paralled by a 1uF capacitor.
Interestingly even the authors suggesting this in their "theory" chapters, don't (always) follow their own advice in the circuits presented in the "practical" chapters.
So can anybody shed light on this theory/practice split? Would the performance of the original poster's circuit be enhanced by observing the theoretical advice?
Regards,
Peter Jacobi
Re: 150Wis it good?
The design looks similar to Marshal Leach's old design from the late 70's, but with more output stages. I built a similar amp
(~ 60W/channel) with materials available to me in 1979 - it is playing in the background as I write this.. The LEDs are a nice idea as voltage references for the current sources, as they are much stiffer than zener diodes of similar voltage, are lower noise (forward conduction vs. reverse avalanche), and pretty as well. Old-style deep red GaAsP LEDs have a voltage drop of about 1.6V, making them almost perfect for biasing silicon transistors. They are probably not easily available these days except on the surplus market, as they have been superseded by more efficient materials. Another good choice is the GaP standard green LEDs (yellowish green). They are readily available, and have almost exactly 2V drop when biased at 10ma. I don't know which LED the 150W design is using, but you can go through and calculate bias currents based on each voltage drop and choose what you think is best. In my design, I biased the input stage at about 1ma per side.
Following Doug Self's design logic, I would turn the resistors in the driver stages T10-13 into one resistor and place a 1uF capacitor across the resistor. This allows some reverse drive for the output capacitors, so that the output stage can slew faster and come out of clipping more quickly.
The design looks similar to Marshal Leach's old design from the late 70's, but with more output stages. I built a similar amp
(~ 60W/channel) with materials available to me in 1979 - it is playing in the background as I write this.. The LEDs are a nice idea as voltage references for the current sources, as they are much stiffer than zener diodes of similar voltage, are lower noise (forward conduction vs. reverse avalanche), and pretty as well. Old-style deep red GaAsP LEDs have a voltage drop of about 1.6V, making them almost perfect for biasing silicon transistors. They are probably not easily available these days except on the surplus market, as they have been superseded by more efficient materials. Another good choice is the GaP standard green LEDs (yellowish green). They are readily available, and have almost exactly 2V drop when biased at 10ma. I don't know which LED the 150W design is using, but you can go through and calculate bias currents based on each voltage drop and choose what you think is best. In my design, I biased the input stage at about 1ma per side.
Following Doug Self's design logic, I would turn the resistors in the driver stages T10-13 into one resistor and place a 1uF capacitor across the resistor. This allows some reverse drive for the output capacitors, so that the output stage can slew faster and come out of clipping more quickly.
Eva said:
I had no idea such a simple comment
would create such heated discussion.
Balancing a single ended LTP with a current sourced VAS
is easy, so much so a current mirror can be used in the LTP.
And whether the VAS is emitter degenerated or not doesn't
matter, a mirror or a changed resistor value can be used.
This is because the VAS current is fixed and the VAS Vbe
can be assumed.
(In the undegenerated VAS case you don't even
have to know what the VAS current is, within reason.)
For the double- ended 2 LTP and 2 VAS (back to back) VAS
current is not defined by a current source so the above
concepts are useless. The VAS can only be biased by the
LTP, the VAS TR's must be emitter degenerated and
current mirrors in the LTP's are out of the question.
(Without some other form of servo feedback)
The only question I think my comment did not
consider was if the overall feedback of the amplifier
added an extra constraint on the double-ended stage
such that it is actually easily designable.
I've been thinking hard and if an unbalanced LTP in one half
must be unbalanced in the other direction in the other half
then this is the constraint, but the feedback does not do this.
The Iq's can still be equal with unbalanced LTP's.
So what is the trick ?
sreten.
Tube_Dude said:
Bingo!!! that's it! And this is the condition of the first case in analisys..the 150 W amp!
Cheers!
Yes, but my problem is the interaction of trying to bias the
VAS and the balance of the input stage, I still can't see how
you can assume the input stage is balanced .......
Its quite easy to build an amplifier with an unbalanced LTP input,
it will still quite happily work, why is this not possible with the
balanced LTP topology ?
sreten.
sreten :
Is there another condition that provides equal output current in both upper and lower LTPs ?
No, this only ocurrs when both LTPs are balanced since upper and lower bases are tied together and current sources are of the same value
And the pair of VAS require equal input current to be balanced...
tube dude :
With ideal devices you could also end with leg currents unbalanced on the LTPs, the problem isn't in real transistors [as long as they are matched by pairs]
Is there another condition that provides equal output current in both upper and lower LTPs ?
No, this only ocurrs when both LTPs are balanced since upper and lower bases are tied together and current sources are of the same value
And the pair of VAS require equal input current to be balanced...
tube dude :
With ideal devices you could also end with leg currents unbalanced on the LTPs, the problem isn't in real transistors [as long as they are matched by pairs]
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.