important questions - think
Yeti,
Why does the linearity of the input Z matter?
Does the quasi-darlington really have any more local feedback than the darlington?
What makes you say the driver compensates for bandwidth in the quasi-darlington...are you implying that it doesn't in the darlington?
"And of course you can get closer to the rails." TRUE - but only by half a volt or so.
[Just for clarity, I'm talking about a proper quasi-darlington which includes two transistors, a resistor and a diode.]
Yeti,
Why does the linearity of the input Z matter?
Does the quasi-darlington really have any more local feedback than the darlington?
What makes you say the driver compensates for bandwidth in the quasi-darlington...are you implying that it doesn't in the darlington?
"And of course you can get closer to the rails." TRUE - but only by half a volt or so.
[Just for clarity, I'm talking about a proper quasi-darlington which includes two transistors, a resistor and a diode.]
The debate of quasi-complimentary Output stages vs True Complimentary Output stages has gone on as long as Transistors have been around. in the Early 1970's most Audio Amps used the All NPN aproach since at that time good PNP's were not available. Things are alot different now and Good PNP's are available. The Main problem with the All NPN output stage is that the NPN used on the Top half or Pos rail is an Emmiter Follower and indeed has a low output impedance, where as the lower NPN is a Comon Emmitter and not only has current gain but voltage gain as well. This can result in less stability than using Emmitter followers in Each half of the Circuit. quasi-complimentry output stages demand matched transistors or Crossover Distortion will result. True complimentry output stages should also have matched transistors and this often times requires that the designer hand select the Transistors and the result may not be what the transistor manufacture selects as a compliment. alot of times a larger NPN must be used that matches the PNP better that what the data sheet say is a compliment. Now one could do as Kevin sugested and use a Quasi-complimentary circuit in the Bridge mode and thus get the best of each. it all gets down to what sounds good to you. as far as opamps go early opamps also did the All npn output stages and some also sound quite good like the LT-1122. However most newer Opamps use True complimentary ouput stages since modern IC process can produce Both PNP's and NPN's with identical AC and DC specs.
"the NPN used on the Top half or Pos rail is an Emmiter Follower and indeed has a low output impedance, where as the lower NPN is a Comon Emmitter and not only has current gain but voltage gain as well. This can result in less stability than using Emmitter followers"
I agree with what you say, ppl, except this. I don't think an ideal emitter-follower has a low output impedance. It does not have a high output impedance either. It has current-gain and a transconductance. It's output Z depends on the source Z feeding it and which two points you are making the measurement across. When I think about it in this way I find the quasi is no different to the darlington.
How about looking at it this way: the darlingtons don't know whether they are in common-emitter or common-collector. This is because voltages are relative - so the BJT can't tell whether the output voltage is fixed and the collector voltage is flying up and down (as if it is the quasi) or vise-versa. All that matters is the relative voltages. A bit like Einstein's relativity idea. If all the relative voltages changes across the quasi are equal and opposite to those of the darlington then how can they tell which they are? Poor things. They don't know that they are in CC or CE configuration.
I'm labouring this point at the risk of causing a
because I hope it will lead to an "aHa!" with some readers of this and dispell some assumptions that are not appropriate. Unlearn and relearn. 
I agree with what you say, ppl, except this. I don't think an ideal emitter-follower has a low output impedance. It does not have a high output impedance either. It has current-gain and a transconductance. It's output Z depends on the source Z feeding it and which two points you are making the measurement across. When I think about it in this way I find the quasi is no different to the darlington.
How about looking at it this way: the darlingtons don't know whether they are in common-emitter or common-collector. This is because voltages are relative - so the BJT can't tell whether the output voltage is fixed and the collector voltage is flying up and down (as if it is the quasi) or vise-versa. All that matters is the relative voltages. A bit like Einstein's relativity idea. If all the relative voltages changes across the quasi are equal and opposite to those of the darlington then how can they tell which they are? Poor things. They don't know that they are in CC or CE configuration.
I'm labouring this point at the risk of causing a
because I hope it will lead to an "aHa!" with some readers of this and dispell some assumptions that are not appropriate. Unlearn and relearn.
traderbam> If one were to look at this in relative terms then your ausumptions would be Valid. However the DC voltages are not mearly just of the other polatity. Most often the all NPN stage will have quite different Drive Voltages on each half so as to Correctly Bias the PNP driving the lower NPN. Now i have heared some very good sounding Quasi-Complimentary Designs years ago I had a Heathkit AR-1500A stereo receiver that used a quasi circuit that looked like a copy of a circuit showen in the Latter pages of the RCA Transistor manual 1971 or so vintage. This was IMHO the Best sounding Stereo receiver of that era. The Harmon cardon Citation also used almost the same circuit and it also was well received in the Audio world at that time. However nowdays with the availability of quality PNP transistors I think that the True Complimentary Design is the way to go. Now if your talking Mosfets, A device i myself do not like as Output Transistors the situation is not so good as quality P Ch devices are still not as good as the N ch Equiv's. But for BJT's i like symmetry Nature is symmectrical. Beauty is also defined by People acording to reports i have read as requireing symmetry. A tree is symmectrical on both sides. What dose this have to do with Amp design Humm... Probaly nothing, But i would suspect that it has the same relationship.
Like i sed before this debate has gone on for Decades in Audio Circles and I gess will continue for quite some time. The true merot of any Audio circuit is how it sounds and more importent how it sounds to the persion using it. People have different views on what sounds good or what looks good for that matter and this is why we have so manny choices and this is a good thing as restricting one's options is only going to get in the way of someone not being able to get what thay want. Diversity is the spice of life and be it a certin Topology or a spicific Device each has attrabutes as well as Drawbacks and it is not the Topology nor evean what Devices a Designer uses to Construct a Circuit it is how well the Designer minimizes the drawbacks and hightens the advantages of the Overall design that will make or break a Design. Everyone has there own Bias on any aspect of Life and yes this includes somthing as subjective as what sounds good.
Like i sed before this debate has gone on for Decades in Audio Circles and I gess will continue for quite some time. The true merot of any Audio circuit is how it sounds and more importent how it sounds to the persion using it. People have different views on what sounds good or what looks good for that matter and this is why we have so manny choices and this is a good thing as restricting one's options is only going to get in the way of someone not being able to get what thay want. Diversity is the spice of life and be it a certin Topology or a spicific Device each has attrabutes as well as Drawbacks and it is not the Topology nor evean what Devices a Designer uses to Construct a Circuit it is how well the Designer minimizes the drawbacks and hightens the advantages of the Overall design that will make or break a Design. Everyone has there own Bias on any aspect of Life and yes this includes somthing as subjective as what sounds good.
The two main problems are the highly differing input impedance characteristics, as in quasi-comp OPSes, one driver/ops transistor is an EF pair, and one is usually a CF/Sziklai pair.
These two combinations have very very different input impedances - both from a resistive and reactive point of view.
This will cause the voltage gain stage to be loaded asymmetrically on either side of the output swing, putting it in a different part of its working region (more/less gain/transimpedance, etc.) - this will create some distortion.
Also, the gain linearity of the EF + CFP hybrid output stage around the crossover point is poorer than a properly implemented CFP NPN+PNP output stage. I can visualize the graphs from Self's book in my mind, but I don't have a copy of it on hand. If anyone here has a copy, they can look in his OPS chapter and see what I mean. I'll see if I can't find my copy from my local library and scan the relevant pages.
Also, since the EF and CFP output stages have very different stability characteristics, it can be trickier to get them perfectly stable. Also, one driver/OPS transistor pair that is ideal for an EF output stage may not be for a CFP output stage for stability reasons (make sure the driver has low Cob, if you can.)
It should also be noted that what we're talking about here are mere nitpicks - in theory, the Quasi-comp output stage is worse than either a CF or EF OPS, but in practice, it can achieve very low distortion levels (below 0.01%) - just not *as* low as one of Self's "blameless" CFP output stages.
If you decide to build one, it most certainly will not sound horrible - if you do it right, it will be excellent. Just know that even if you have 100% identical NPN output transistors, its linearity will still likely be poorer than a properly done CFP output stage with modern extended-beta transistors. (Toshiba, Sanken, OnSemi)
No harm whatsoever in trying, though.
I'll see if I can't post Self's thoughts on it sometime in the next few days.
These two combinations have very very different input impedances - both from a resistive and reactive point of view.
This will cause the voltage gain stage to be loaded asymmetrically on either side of the output swing, putting it in a different part of its working region (more/less gain/transimpedance, etc.) - this will create some distortion.
Also, the gain linearity of the EF + CFP hybrid output stage around the crossover point is poorer than a properly implemented CFP NPN+PNP output stage. I can visualize the graphs from Self's book in my mind, but I don't have a copy of it on hand. If anyone here has a copy, they can look in his OPS chapter and see what I mean. I'll see if I can't find my copy from my local library and scan the relevant pages.
Also, since the EF and CFP output stages have very different stability characteristics, it can be trickier to get them perfectly stable. Also, one driver/OPS transistor pair that is ideal for an EF output stage may not be for a CFP output stage for stability reasons (make sure the driver has low Cob, if you can.)
It should also be noted that what we're talking about here are mere nitpicks - in theory, the Quasi-comp output stage is worse than either a CF or EF OPS, but in practice, it can achieve very low distortion levels (below 0.01%) - just not *as* low as one of Self's "blameless" CFP output stages.
If you decide to build one, it most certainly will not sound horrible - if you do it right, it will be excellent. Just know that even if you have 100% identical NPN output transistors, its linearity will still likely be poorer than a properly done CFP output stage with modern extended-beta transistors. (Toshiba, Sanken, OnSemi)
No harm whatsoever in trying, though.
I'll see if I can't post Self's thoughts on it sometime in the next few days.
Mother Nature is ugly?
ppl,
I know what you are trying to say, and I too favour the symmetry in things. But the fact of the matter is that nature is not always symmetrical. It's a nuisance but it's true.
Electricity is an example of this. Charge carriers are electrons and they carry a negative charge. There are no "positrons" that carry a positive charge. If there were then PNPs would be much more like NPNs and there would be p-channel valves. Look at atoms: protons are the positively charge particles and they look nothing like electrons. If you study the physics of semiconductors the differences become clear.
I would love to find a pair of output transistors that are truely complementary. If you know some then please give the part numbers - I'll look them up and if you are right I'll eat my soldering iron (or something)
ppl,
I know what you are trying to say, and I too favour the symmetry in things. But the fact of the matter is that nature is not always symmetrical. It's a nuisance but it's true.
Electricity is an example of this. Charge carriers are electrons and they carry a negative charge. There are no "positrons" that carry a positive charge. If there were then PNPs would be much more like NPNs and there would be p-channel valves. Look at atoms: protons are the positively charge particles and they look nothing like electrons. If you study the physics of semiconductors the differences become clear.
I would love to find a pair of output transistors that are truely complementary. If you know some then please give the part numbers - I'll look them up and if you are right I'll eat my soldering iron (or something)
ThingyNess,
Your premise is that the input Z is "very very different". I simply do not agree that this is necessarily so. I don't understand what leads you to this conclusion.
I also don't agree with the stability differences that you mention. Why should one be less stable than the other? I do not understand.
Now, I can believe that using a CFP arrangement a la AKSA may improve low frequency linearity. However, this is not the point I'm arguing. I am arguing that a quasi-darlington will, in theory, make a better complement than a PNP darlington.
BAM
(Don't talk to me about Self's blameless designs )
Your premise is that the input Z is "very very different". I simply do not agree that this is necessarily so. I don't understand what leads you to this conclusion.
I also don't agree with the stability differences that you mention. Why should one be less stable than the other? I do not understand.
Now, I can believe that using a CFP arrangement a la AKSA may improve low frequency linearity. However, this is not the point I'm arguing. I am arguing that a quasi-darlington will, in theory, make a better complement than a PNP darlington.
BAM
(Don't talk to me about Self's blameless designs
)
Hi!
This is my first (of hopefully many) post to this forum, I am running a Quasi output 2SC3281 and 2SK216/SJ79 mosfets as drivers, input and voltageamp is a tube diff-stage. i came up with this idea about a year ago since i dont like mosfet OPS and wanted a simple, easy to drive BJT OPS. I have looked all over for something similar but have come up with zero so far!
/Anders
This is my first (of hopefully many) post to this forum, I am running a Quasi output 2SC3281 and 2SK216/SJ79 mosfets as drivers, input and voltageamp is a tube diff-stage. i came up with this idea about a year ago since i dont like mosfet OPS and wanted a simple, easy to drive BJT OPS. I have looked all over for something similar but have come up with zero so far!
/Anders
Folks,
Quasi versus full complementary - the beat goes on.
Observations:
1. One, the positive device, is in common collector, hence Zout = 26/mA flowing, Zin is beta(re + Re), and frequency response is excellent.
2. The other, negative device, is in common emitter, hence Zout is much greater, though we have a complicating factor since there is very powerful local negative feedback from the CFP connection.
3. CFPs, by their very nature, are highly susceptible to oscillation, and to achieve good damping it is often necessary to place a cap between collector and base of the driver. This cap has considerable effect on the slew rate of the driver, and robs vitality from the music, creating an asymmetric, distortive effect since slew rates differ markedly between the two output halves of the amplifier.
Any instability of any part of the output stage is immediately apparent in careful listening as a loss in resolution with attendant impairment of imaging.
The advantage of the fully complementary arrangement is the near identical slew rate of each half of the output stage; I think this is very important; we are artificially splitting the music in two, so the least we can do is treat each half the same way!
Incidentally, the CFP bipolar/mosfet configuration I extolled some time back is really only suitable for SE operation, certainly not push pull. Horses for courses.........
Cheers,
Hugh
www.aksaonline.com
Quasi versus full complementary - the beat goes on.
Observations:
1. One, the positive device, is in common collector, hence Zout = 26/mA flowing, Zin is beta(re + Re), and frequency response is excellent.
2. The other, negative device, is in common emitter, hence Zout is much greater, though we have a complicating factor since there is very powerful local negative feedback from the CFP connection.
3. CFPs, by their very nature, are highly susceptible to oscillation, and to achieve good damping it is often necessary to place a cap between collector and base of the driver. This cap has considerable effect on the slew rate of the driver, and robs vitality from the music, creating an asymmetric, distortive effect since slew rates differ markedly between the two output halves of the amplifier.
Any instability of any part of the output stage is immediately apparent in careful listening as a loss in resolution with attendant impairment of imaging.
The advantage of the fully complementary arrangement is the near identical slew rate of each half of the output stage; I think this is very important; we are artificially splitting the music in two, so the least we can do is treat each half the same way!
Incidentally, the CFP bipolar/mosfet configuration I extolled some time back is really only suitable for SE operation, certainly not push pull. Horses for courses.........
Cheers,
Hugh
www.aksaonline.com
Mirlo,
I don't have a drawing package so I'll describe it. The quasi has a pnp driver and an npn power transistor. You need a diode in parallel with a resistor connected between pnp emitter and npn collector. You need the same value resistor between npn base and emitter. The pnp collector connects to the npn base.
AKSA,
No, no, no on all counts! Do you have a circuit simulator? - this will make it clearer. I think you are focussing too much on the pin connections of the power transistors.
Anyhow, I've said all I can. Take it or leave it.
I don't have a drawing package so I'll describe it. The quasi has a pnp driver and an npn power transistor. You need a diode in parallel with a resistor connected between pnp emitter and npn collector. You need the same value resistor between npn base and emitter. The pnp collector connects to the npn base.
AKSA,
No, no, no on all counts!
Do you have a circuit simulator? - this will make it clearer. I think you are focussing too much on the pin connections of the power transistors.Anyhow, I've said all I can. Take it or leave it.
Got it! But what about this:
Suppose you replace the diode with an emitter follower:
I don't think there would be any downside, and the composite wouldn't carry the standing current in the PNP...
Or just omit the diode. Wouldn't the NPN have enough gain itself to make up for the loss of transconductance, of the (then degenerated) PNP with the possible advantage that you might improve stability, as with feedback resistors in current feedback op-amps?
-- mirlo
Suppose you replace the diode with an emitter follower:
I don't think there would be any downside, and the composite wouldn't carry the standing current in the PNP...
Or just omit the diode. Wouldn't the NPN have enough gain itself to make up for the loss of transconductance, of the (then degenerated) PNP with the possible advantage that you might improve stability, as with feedback resistors in current feedback op-amps?
-- mirlo
Attachments
Check out this Thred to see early schematics from the RCA Manual that showes Both types http://www.diyaudio.com/forums/showthread.php?threadid=6828 It should make this topic easy to understand.
Mirlo,
Your idea of using an npn instead of a diode is creative. Indeed, the pnp emitter current, which is also mostly npn base current, would be sourced from elsewhere than the output. The question is what benefit does this provide? Also, remember the idea is to make the quasi electrically equivalent to the darlington; the darlington driver current flows to the output.
As for removing the diode you must ask what you are trying to achieve. It's removal makes the quasi less like the darlington. And what makes you think it's removal improves stability? The non-diode and resistor form may have better performance in some respects, yes, but it deviates from the goal.
BAM
Your idea of using an npn instead of a diode is creative. Indeed, the pnp emitter current, which is also mostly npn base current, would be sourced from elsewhere than the output. The question is what benefit does this provide? Also, remember the idea is to make the quasi electrically equivalent to the darlington; the darlington driver current flows to the output.
As for removing the diode you must ask what you are trying to achieve. It's removal makes the quasi less like the darlington. And what makes you think it's removal improves stability? The non-diode and resistor form may have better performance in some respects, yes, but it deviates from the goal.
BAM
Darlingtons are somewhat nonideal at best: as emitter followers they have more nonlinearity than single devices.
I agree that if you are trying to match an NPN darlington on the positive rail with a fake PNP darlington on the negative rail, then you probably want the diode, for better symmetry.
But perhaps a better amplifier overall could eb made with composite devices on both rails. Now of course that means big PNPs on top, which is getting away from the point of this thread...
I think that removing the diode might improve stability by reducing the loop gain in the composite device, because it reduces the transconductance of the PNP part. If you look at the structure of a current feedback op-amp, you see that the composite PNP+NPN is very similar. I haven't done the analysis, but it is general practice to add resistors in the feedback loop of CFOAs for stability.
I agree that if you are trying to match an NPN darlington on the positive rail with a fake PNP darlington on the negative rail, then you probably want the diode, for better symmetry.
But perhaps a better amplifier overall could eb made with composite devices on both rails. Now of course that means big PNPs on top, which is getting away from the point of this thread...
I think that removing the diode might improve stability by reducing the loop gain in the composite device, because it reduces the transconductance of the PNP part. If you look at the structure of a current feedback op-amp, you see that the composite PNP+NPN is very similar. I haven't done the analysis, but it is general practice to add resistors in the feedback loop of CFOAs for stability.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.