definition
Without knowing the Sumo 9 + I say that the term "CLASS A"
is just a name for a special kind of using a gain device.
Definition is that the signal current is superpositioned by a certain
DC-current in such a way that at peaks the current thru the device
is max double the DC current, at zero signal its the DC current
only and at the other peak the current thru the device is zero.
Everything over double bias or belo zero is not Class A per
definition.
Beware:
Definition has NOTHING to do with soundquality!
Uli
PS: Single ended or not makes no difference.
Without knowing the Sumo 9 + I say that the term "CLASS A"
is just a name for a special kind of using a gain device.
Definition is that the signal current is superpositioned by a certain
DC-current in such a way that at peaks the current thru the device
is max double the DC current, at zero signal its the DC current
only and at the other peak the current thru the device is zero.
Everything over double bias or belo zero is not Class A per
definition.
Beware:
Definition has NOTHING to do with soundquality!
Uli
PS: Single ended or not makes no difference.
Re: sumo
Not necessarily. Post #7 (mine) commented on the popular
disinction between "Pure" and "Other" forms of Class A.
If the gain devices maintain conduction through the operating
output cycle, then they would be said to be Class A, even if
only by technicality.
fscarpa58 said:
Not necessarily. Post #7 (mine) commented on the popular
disinction between "Pure" and "Other" forms of Class A.
If the gain devices maintain conduction through the operating
output cycle, then they would be said to be Class A, even if
only by technicality.
According to my experience - dynamic current sources like d'Agostino etc. do not bring sonic improvement compared to true class A with constant quiescent current. I would prefer well designed class AB to dynamic current source.
Please read the blue number below the green M6:It says 1.08 [A]
not 200 [mA]
By the way, Nelson once stated that the current source does not affect sound, so what is the problem if it has only 200mA, if the gain fet got 1A ?
Look at it Bernhard,
Although M6 never starves off, the current through R54 certainly will do. This is because M8 (almost) starves off at a certain point. You’re right that the bias current trough M6 linearises the transfer curve of M6 but this does nothing with the point where the current through R54 ceases. Something we don’t want. Note that all the current to the load needs to be supplied by M8. If M8 can’t, no feed for the load. The whole thing is highly non-linear without overall feedback, just as a class-AB
What counts in a class-A is that the transfer function of the output stage is linear by itself and free from even small discontinuities. BTW this is NOT a definition of class-A, a good tube PP class-AB can have this property too.
Cheers
Although M6 never starves off, the current through R54 certainly will do. This is because M8 (almost) starves off at a certain point. You’re right that the bias current trough M6 linearises the transfer curve of M6 but this does nothing with the point where the current through R54 ceases. Something we don’t want. Note that all the current to the load needs to be supplied by M8. If M8 can’t, no feed for the load. The whole thing is highly non-linear without overall feedback, just as a class-AB
What counts in a class-A is that the transfer function of the output stage is linear by itself and free from even small discontinuities. BTW this is NOT a definition of class-A, a good tube PP class-AB can have this property too.
Cheers
circlotron
o.k.
some effort to simplify the circuit…
but now look at the following output stage. It is a circlotron.
Each device has an idle current of 200ma. The total power dissipated
in absence of signal is roughly 9.4x10 = 94W.
The output power is 70 W/8ohm (360W dissipated) and 140W/4ohm (240W diss.) with distortion under 0.05 % as is, without global feedback.
Yes, naturally there is the current feedback alternatively
given by the two diode that maintains
positive the current in the output devices.
Maybe this is not a pure class A design but certainly
it is a clever, simple and elegant design. Very different
from those 80’s Japanese class A+,AA, etc.
The same figure shows the current in the output device
at max power.
I want to hear from Circlotron about this.
Regards
Federico
o.k.
some effort to simplify the circuit…
but now look at the following output stage. It is a circlotron.
Each device has an idle current of 200ma. The total power dissipated
in absence of signal is roughly 9.4x10 = 94W.
The output power is 70 W/8ohm (360W dissipated) and 140W/4ohm (240W diss.) with distortion under 0.05 % as is, without global feedback.
Yes, naturally there is the current feedback alternatively
given by the two diode that maintains
positive the current in the output devices.
Maybe this is not a pure class A design but certainly
it is a clever, simple and elegant design. Very different
from those 80’s Japanese class A+,AA, etc.
The same figure shows the current in the output device
at max power.
I want to hear from Circlotron about this.
Regards
Federico
Attachments
Bernhard said:This is witchcraft
djk said:
A bit of history: http://circlotron.tripod.com/
Do a search on this forum as well.
/Hugo
Certainly, it is a Sumo NINE (output stage)
The point is that on the J.B. patent the focus is on
the thermal stability produced by the particular path
of the control current
In my opinion another important point is the diode network.
Without that the output devices cannot work in class A with
a so low idle current.
Can someone explain to me how precisely it works?
Thanks
Federico
The point is that on the J.B. patent the focus is on
the thermal stability produced by the particular path
of the control current
In my opinion another important point is the diode network.
Without that the output devices cannot work in class A with
a so low idle current.
Can someone explain to me how precisely it works?
Thanks
Federico
"Can someone explain to me how precisely it works?"
It looks pretty clear, there is an idle current control, and a peak bias control.
IOW, dynamic bias. If one owned some really nasty low impedance speakers he could add another level shift circuit and make the bias a three-slope instead of a two-slope.
The NINE was rated at 70W/8R, 120W/4R in class A. It was built using the same chassis, sink, transformer (1.3KVA) as their 400W/4R Andromeda.
Notes for anyone wanting to build a NINE: the 2SC1831 has very, very high gain. The NINE+ went to a darlington connected driver/output. The LF353 was selected to run at ±20V, use a ±24V part like the OPA2604.
It looks pretty clear, there is an idle current control, and a peak bias control.
IOW, dynamic bias. If one owned some really nasty low impedance speakers he could add another level shift circuit and make the bias a three-slope instead of a two-slope.
The NINE was rated at 70W/8R, 120W/4R in class A. It was built using the same chassis, sink, transformer (1.3KVA) as their 400W/4R Andromeda.
Notes for anyone wanting to build a NINE: the 2SC1831 has very, very high gain. The NINE+ went to a darlington connected driver/output. The LF353 was selected to run at ±20V, use a ±24V part like the OPA2604.
> If one owned some really nasty low impedance speakers
> he could add another level shift circuit and make
> the bias a three-slope instead of a two-slope.
Excuse my ignorance,
I am very interested to this point, can you
exemplify this point ?
thanks in advance djk
Federico
> he could add another level shift circuit and make
> the bias a three-slope instead of a two-slope.
Excuse my ignorance,
I am very interested to this point, can you
exemplify this point ?
thanks in advance djk
Federico
"I am very interested to this point, can you
exemplify this point ?"
I'll give it a try.
First, view the amplfier as two halves: one for current gain, the other for voltage gain.
The voltage gain stage runs on ±20V on the NINE, the current gain stage on two floating 40V supplies.
The base current for the outputs must have a return path to the ±20V supplies otherwise no current will flow.
Ignoring the floating supplies for the moment, just consider the opamp. A common technique to force the output stage of an opamp into class A is to use a pull-down resistor to the negative supply. If we add a current follower we now have a simple single ended class A amplifier. The pull-down resistor will have to source all the current on the negative swing. It would be more efficent to use an active current source.
Let's say we build a pair of these single ended class A amplifiers and bridge them. We double the voltage swing, so we will need to double the current source too. The amplifier will now put out four times as much power, and will dissipate four times as much too.
Let's add the floating supplies now. For the time being we will ignore where the base curent comes from.
The amplifier is now a push-pull class AB amplifier. It only runs class A up to the point that the current souce supplies, it moves into class AB at higher currents.
We are now going to remove the current sources. The amplifier is now running in class B.
We are now going to restore the current sources. Without the floating supplies we had to run several amps though the current sources to get any real output. With the floating supplies we only have to run a pull-down resistor to the negative supply for enough base current for the outputs. With either the 2SC1831 or a darlington connected pair we are only talking a few mA of base current.
At this point the amplifier is short circuit poof, and never moves out of its class A point. Because the base current is limited by the pull-down resistor, the collector current is also limited. This is what I am calling the first slope.
The resistors to ground adjust the peak current the output stage can deliver in class B by making more base current available. The diodes shift the point away from zero, a short on the output of the amplifier with no signal voltage will cause no increase in collector current. A 'normal' amplfier with only a few millivolts of signal (or DC off-set) will try and drive 10A (or more) into a short. This amplifier will draw no more current that its class A bias point. This is what I am calling the second slope.
Another slope could be added to the class B side for more precise current limiting, although the amplifier is very rugged with ten TO3 outputs per channel. The output stage can soak up whatever the supply can provide until the mains fuse or breaker opens on a long duration short. Remember, you would have to be driving a large signal voltage to draw any more current than the class A bias.
Another slope could be added to the class A side for a dynamic bias. The addition of the floating supplies converts the amplfier from single ended to push-pull. This doubles the efficency of the amplifer, but allows it to be driven into class AB at very high currents. You could have it increase the bias at very high signal levels. Krell does this on some models. Would this sound any better? Who knows. My NINE sounds pretty good with the sleazy LF353 in it. Some day it will get a better opamp and some cap upgrades.
Note:
A FET output cannot be substituted without a different bias scheme. One could build a discrete FET opamp with a single long-tailed pair to drive both halves of the bridge, use the X feedback scheme, and dump the LF353.
exemplify this point ?"
I'll give it a try.
First, view the amplfier as two halves: one for current gain, the other for voltage gain.
The voltage gain stage runs on ±20V on the NINE, the current gain stage on two floating 40V supplies.
The base current for the outputs must have a return path to the ±20V supplies otherwise no current will flow.
Ignoring the floating supplies for the moment, just consider the opamp. A common technique to force the output stage of an opamp into class A is to use a pull-down resistor to the negative supply. If we add a current follower we now have a simple single ended class A amplifier. The pull-down resistor will have to source all the current on the negative swing. It would be more efficent to use an active current source.
Let's say we build a pair of these single ended class A amplifiers and bridge them. We double the voltage swing, so we will need to double the current source too. The amplifier will now put out four times as much power, and will dissipate four times as much too.
Let's add the floating supplies now. For the time being we will ignore where the base curent comes from.
The amplifier is now a push-pull class AB amplifier. It only runs class A up to the point that the current souce supplies, it moves into class AB at higher currents.
We are now going to remove the current sources. The amplifier is now running in class B.
We are now going to restore the current sources. Without the floating supplies we had to run several amps though the current sources to get any real output. With the floating supplies we only have to run a pull-down resistor to the negative supply for enough base current for the outputs. With either the 2SC1831 or a darlington connected pair we are only talking a few mA of base current.
At this point the amplifier is short circuit poof, and never moves out of its class A point. Because the base current is limited by the pull-down resistor, the collector current is also limited. This is what I am calling the first slope.
The resistors to ground adjust the peak current the output stage can deliver in class B by making more base current available. The diodes shift the point away from zero, a short on the output of the amplifier with no signal voltage will cause no increase in collector current. A 'normal' amplfier with only a few millivolts of signal (or DC off-set) will try and drive 10A (or more) into a short. This amplifier will draw no more current that its class A bias point. This is what I am calling the second slope.
Another slope could be added to the class B side for more precise current limiting, although the amplifier is very rugged with ten TO3 outputs per channel. The output stage can soak up whatever the supply can provide until the mains fuse or breaker opens on a long duration short. Remember, you would have to be driving a large signal voltage to draw any more current than the class A bias.
Another slope could be added to the class A side for a dynamic bias. The addition of the floating supplies converts the amplfier from single ended to push-pull. This doubles the efficency of the amplifer, but allows it to be driven into class AB at very high currents. You could have it increase the bias at very high signal levels. Krell does this on some models. Would this sound any better? Who knows. My NINE sounds pretty good with the sleazy LF353 in it. Some day it will get a better opamp and some cap upgrades.
Note:
A FET output cannot be substituted without a different bias scheme. One could build a discrete FET opamp with a single long-tailed pair to drive both halves of the bridge, use the X feedback scheme, and dump the LF353.
Thank you very much
for your long answer djk,
I’ll have to study it for a while
a presto
federico
for your long answer djk,
I’ll have to study it for a while
a presto
federico
Thanks djk for the Sumo link, I found the matsushita recipe there and cooked my own little circuit...
Class AB Amp with 18V rails drives virtual ground of SE ClassA Amp with 6V rails and drives virtual ground of 15V op amp rails to allow higher output voltage of op amp.
RMS output is same like idle power consumption is 52W.
This time gain fet current is a sine and dynamic current source current is also a sine.
Idle current of ClassA amp is 1.5A and of Driver Amp is 180mA.
Again dynamic current source is not current sensed yet.
May be this is unpure again but it looks like I have to re-invent all that pseudo ClassA crap before returning to the golden calf.
And this time there could be the option to make a switch or relais that connects the SE ClassA amp to the 18V rails or 24V rails for cold winter... 😎
Also I find the circlotron very interesting...
Class AB Amp with 18V rails drives virtual ground of SE ClassA Amp with 6V rails and drives virtual ground of 15V op amp rails to allow higher output voltage of op amp.
RMS output is same like idle power consumption is 52W.
This time gain fet current is a sine and dynamic current source current is also a sine.
Idle current of ClassA amp is 1.5A and of Driver Amp is 180mA.
Again dynamic current source is not current sensed yet.
May be this is unpure again but it looks like I have to re-invent all that pseudo ClassA crap before returning to the golden calf.
And this time there could be the option to make a switch or relais that connects the SE ClassA amp to the 18V rails or 24V rails for cold winter... 😎
Also I find the circlotron very interesting...
Same circuit like above with as few changes as possible and pure ClassA has 165W dissipation @ idle.
I don't wanna bother you here with another madness schematic...
Could be possible to make a version that can be switched between both modes.
8 channels x 165W = 1320W
for a 4way active speaker stereo setup.
Was reading the xen-over threads today 🙂
Very interesting 🙂
But did you guys think about how all the xalephs you need for that, will heat up your room
I don't wanna bother you here with another madness schematic...
Could be possible to make a version that can be switched between both modes.
8 channels x 165W = 1320W
for a 4way active speaker stereo setup.Was reading the xen-over threads today 🙂
Very interesting 🙂
But did you guys think about how all the xalephs you need for that, will heat up your room
natural grown schematic
This is my latest circiut. I'm a very simple mind
I think it is much better than the first schematics of this thread.
The dynamic current sources of the ClassA amp part are current sensed now as it should be.
Idle current in ClassA amp is 1.67A , in driver amp 178mA , dissipation @ idle 60W, rms output power 50W.
Is this good & pure ClassA per definition now ?
All waveforms are sine. See picture.
Black = voltage across load.
Blue and red are currents through gainfet and current source.
This is my latest circiut. I'm a very simple mind
I think it is much better than the first schematics of this thread.
The dynamic current sources of the ClassA amp part are current sensed now as it should be.
Idle current in ClassA amp is 1.67A , in driver amp 178mA , dissipation @ idle 60W, rms output power 50W.
Is this good & pure ClassA per definition now ?
All waveforms are sine. See picture.
Black = voltage across load.
Blue and red are currents through gainfet and current source.
- Status
- Not open for further replies.