FAB
"Yes I have already a copy of the patents. So it means that the circuit you have provided is an earlier version. I used one in its patent (for my own personal use). Anyway, the amp input circuit in the Perrot patent is also used in Halcro patent."
It was not an ealier version. Perrot made a kind of compilations of all differential stages he was aware of. He then measured them. No simulation at that time. His series of articles and his works are on the par with those of EF Taylor (not the same Taylor as of the schematics post 27) , Peter Baxandall and Douglas Self. (Note that all the three last authors published in Wireless World and Electronics World, which means how great has been this dying magazine).
You can get the copy of his works by goggling this :
"cd-rom de l'audiophile. 43 revues!! de 1977 à 1988".
Halcro did not use Perrot's patents, the circuit is a bit different.
Halcro or Peufeu (search "Peufeu" on this forum) use a Compound Pair, also named Complementary Feedback Pair, also named Sziklai pair. The collector load of the input transistor is a resistor connected to the emitter of the second transistor. Due to the low impedance provided by the following common base transistor, the bootstrap effect on this resistor is rather low.
In Halcro-Peufeu configuration, the two first transistors can be considered as one device, having higher gain and linearity, with an input node, a follower node and and an output node. The third transistor is simply common base connected.
In Perrot's circuit, the collector load of the input transistor is a constant current source. The open loop gain of the pair is higher and distorsion is lower than in Halcro-Peufeu. I made some simulations with TINA demo which clearly show this.
A fact overlooked by Perrot himself in his patented circuit is that the second transistor (which can be considered as a common emitter to the emitter of the input transistor and as a common collector to the collector of the input transistor) and the third transistor (common base) form a Rush pair, an other name for a differential series circuit. It is possible to replace this differential series circuit by a differential parallel circuit, better known a long tail pair differential circuit. Simulation shows there is no need of a CCS for the emitters node, a simple resistor will suffice.
~~~~~~~~ Forr
§§§
"Yes I have already a copy of the patents. So it means that the circuit you have provided is an earlier version. I used one in its patent (for my own personal use). Anyway, the amp input circuit in the Perrot patent is also used in Halcro patent."
It was not an ealier version. Perrot made a kind of compilations of all differential stages he was aware of. He then measured them. No simulation at that time. His series of articles and his works are on the par with those of EF Taylor (not the same Taylor as of the schematics post 27) , Peter Baxandall and Douglas Self. (Note that all the three last authors published in Wireless World and Electronics World, which means how great has been this dying magazine).
You can get the copy of his works by goggling this :
"cd-rom de l'audiophile. 43 revues!! de 1977 à 1988".
Halcro did not use Perrot's patents, the circuit is a bit different.
Halcro or Peufeu (search "Peufeu" on this forum) use a Compound Pair, also named Complementary Feedback Pair, also named Sziklai pair. The collector load of the input transistor is a resistor connected to the emitter of the second transistor. Due to the low impedance provided by the following common base transistor, the bootstrap effect on this resistor is rather low.
In Halcro-Peufeu configuration, the two first transistors can be considered as one device, having higher gain and linearity, with an input node, a follower node and and an output node. The third transistor is simply common base connected.
In Perrot's circuit, the collector load of the input transistor is a constant current source. The open loop gain of the pair is higher and distorsion is lower than in Halcro-Peufeu. I made some simulations with TINA demo which clearly show this.
A fact overlooked by Perrot himself in his patented circuit is that the second transistor (which can be considered as a common emitter to the emitter of the input transistor and as a common collector to the collector of the input transistor) and the third transistor (common base) form a Rush pair, an other name for a differential series circuit. It is possible to replace this differential series circuit by a differential parallel circuit, better known a long tail pair differential circuit. Simulation shows there is no need of a CCS for the emitters node, a simple resistor will suffice.
~~~~~~~~ Forr
§§§
constant curr constant volt
interesting idea and hot discussions here. hope sb will still follow on. I see such design has very high output impedance. Would anybody tell me, when amplifier is sth like current source, the speaker box will still work well or not? when spaker has little electric damper, can it work well?
interesting idea and hot discussions here. hope sb will still follow on. I see such design has very high output impedance. Would anybody tell me, when amplifier is sth like current source, the speaker box will still work well or not? when spaker has little electric damper, can it work well?
Delayed
Sorry for my much delayed answer. I was too busy to spend enough time on the forum to really look into your comments.
I think it is a mistake that my Q3 serves the same purpose as your Q3. My Q3 is a folded cascode error amplifier that happens to have its base connected to the output of the amplifier, but my Q1 is the actual emitter follower, i.e. the output stage that sets the output voltage.
Your Q3 is the actual emitter follower output. It is this Q3 that sets the output voltage.
So in my amplifier it is important that Q1 has a constant Vce and constant Ic for best performance (well, at least that's the goal for the moment). In your amplifier it is important that Q3 has a constant Ic (it has) and Vce (it has not). For that reason I mentioned adding a cascode to your circuit. I meant a PNP cascode in series with the collector of Q3 and the base of the cascode transistor connected to the output of the amplifier via a level shifter (e.g. some diodes). In this way the Vce of Q3 can be kept constant.
Vce modulation of Q2 is far less important.
Ic of my Q3 is only small compared to Ic of Q1; Q3 is only delivering base current to Q2, neglecting R1, wich is only there to speed up Q2. So Ic of Q1 is fairly constant, it changes only a few percent, depending on the hFE of Q2. The idle current of Q2 is also 2A, but Ic of Q2 will be modulated and can be between 0A and 4A for symmetrical signals, so there will be some base current modulation and consequently Ic modulation of Q3, but these changes will not have a very big influence on Vbe of Q3 (which is the same as Vce of Q1). Only if Ic2 is very close to 0 (near current clipping) then Vbe3 will become smaller.
The green graph in http://www.diyaudio.com/forums/showthread.php?postid=775791#post775791 shows that Ic1 is quite constant. The blue graph in the next post shows that Vce1 is fairly constant.
I'm not sure how it will sound; I did not build it yet, although I have an idea for a nice driver for this unity voltage gain follower. Maybe some day.
Its main advantage is the constant Vbe of Q1, irrespective of load current and voltage. This should give less distortion than a normal emitter follower.
Steven
Ultima Thule said:
Sorry for my much delayed answer. I was too busy to spend enough time on the forum to really look into your comments.
I think it is a mistake that my Q3 serves the same purpose as your Q3. My Q3 is a folded cascode error amplifier that happens to have its base connected to the output of the amplifier, but my Q1 is the actual emitter follower, i.e. the output stage that sets the output voltage.
Your Q3 is the actual emitter follower output. It is this Q3 that sets the output voltage.
So in my amplifier it is important that Q1 has a constant Vce and constant Ic for best performance (well, at least that's the goal for the moment). In your amplifier it is important that Q3 has a constant Ic (it has) and Vce (it has not). For that reason I mentioned adding a cascode to your circuit. I meant a PNP cascode in series with the collector of Q3 and the base of the cascode transistor connected to the output of the amplifier via a level shifter (e.g. some diodes). In this way the Vce of Q3 can be kept constant.
Vce modulation of Q2 is far less important.
Ultima Thule said:
But if I try to think about the currents going though Q1 I would interpret it like this way:
If we look at Q2 (especially with no emitter resistor) the Early effect will be very distinctive, said that what happens if we have high output voltage, this gives a decrease of the base current in most BJT's for a given current (which in our exercise is 2A), for low output voltage we have the opposite.
Now let's look at Q3 in your circuit, it's kind of shunting current passing by Q1 from Q3's emitter to base, but that's not much to talk about, however it shunts away current from emitter to collector by times hfe meaning there will for sure now not be any constant current through Q1, moreover the current modulation of Q3's base current needs also a small Vbe modulation meaning that the voltage is also not constant over Q1.
It's just thought through in my head so I can not be be fully sure but I guess you should see in your simulations something equal, for instance high output voltage = higher current through Q1(Q3 shunting less), and the opposite.
All this is in small quantities of course, but never the less I feel that Q1 is a "peace of dead meat", else if:
Ic of my Q3 is only small compared to Ic of Q1; Q3 is only delivering base current to Q2, neglecting R1, wich is only there to speed up Q2. So Ic of Q1 is fairly constant, it changes only a few percent, depending on the hFE of Q2. The idle current of Q2 is also 2A, but Ic of Q2 will be modulated and can be between 0A and 4A for symmetrical signals, so there will be some base current modulation and consequently Ic modulation of Q3, but these changes will not have a very big influence on Vbe of Q3 (which is the same as Vce of Q1). Only if Ic2 is very close to 0 (near current clipping) then Vbe3 will become smaller.
The green graph in http://www.diyaudio.com/forums/showthread.php?postid=775791#post775791 shows that Ic1 is quite constant. The blue graph in the next post shows that Vce1 is fairly constant.
Ultima Thule said:
I'm not sure I have yet grasped what is the main advantage with a circuit like yours(Q1), is it the constant base current need seen from the source (by means of fairly constant U/I over Q1 of course) that makes it more linear, or a bit something else?
If the main advantage is a more constant base current seen by the source then I could understand that might be the advantage, then, dose it have any merit's in reality?
But, on the other side, hope no one think I'm dissing, I do think this is an interesting discussion and brain exericse never the less!
I'm not sure how it will sound; I did not build it yet, although I have an idea for a nice driver for this unity voltage gain follower. Maybe some day.
Its main advantage is the constant Vbe of Q1, irrespective of load current and voltage. This should give less distortion than a normal emitter follower.
Steven
Re: constant curr constant volt
No, the output impedance is actually very low. If the output transistor Vbe can be kept constant and the transistor is driven from a low source impedance, the output impedance wil be near zero.
On this forum many threads deal with current driven loudspeakers and transconductance amplifiers. Look e.g. at the FirstWatt F1 from Nelson Pass.
Steven
lastguy said:
No, the output impedance is actually very low. If the output transistor Vbe can be kept constant and the transistor is driven from a low source impedance, the output impedance wil be near zero.
On this forum many threads deal with current driven loudspeakers and transconductance amplifiers. Look e.g. at the FirstWatt F1 from Nelson Pass.
Steven
Bump to get this back into currency, as it was just brought to my attention and has clear resemblance to linuxguru's buffer, which he has dubbed a cantilevered cascode buffer. See post 936 in the Discrete Opamp Open Design (he also alludes to this thread but does not provide the link).
I'm working on some variants, "unfortunately" for a client from whom I will need to get a go-ahead to publish on. That will probably be forthcoming, as he is contracting for only a single instrument at this point.
Brad Wood
I'm working on some variants, "unfortunately" for a client from whom I will need to get a go-ahead to publish on. That will probably be forthcoming, as he is contracting for only a single instrument at this point.
Brad Wood
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.