feedforward error correction in the Quad? - please go to
http://1.bp.blogspot.com/_6YZ4S4pLhUo/SXC0ZkrESrI/AAAAAAAAABE/VIzwkRcgEaA/s1600-h/Quad405-2.png
(source of this schematics: http://piercarlo-audioelettronica.blogspot.com/2009/01/quad-405-2-seconda-parte.html)
and to my own pdf schematic of Quad 405 in higher resolution.
http://1.bp.blogspot.com/_6YZ4S4pLhUo/SXC0ZkrESrI/AAAAAAAAABE/VIzwkRcgEaA/s1600-h/Quad405-2.png
(source of this schematics: http://piercarlo-audioelettronica.blogspot.com/2009/01/quad-405-2-seconda-parte.html)
and to my own pdf schematic of Quad 405 in higher resolution.
unfortunately not. But you can order original print there:
http://www.schaltungsdienst.de/kontakt0.html
info@schaltungsdienst.de
For directly compare the basic circuits I have introduce also the Quad 405 simplefield schematic in the pdf
If you read follow, you see, there are no real feedforward error correction in use by the Quad 405 circuit:
http://research.microsoft.com/pubs/102065/Malvar_CDA_WW1981.pdf
http://www.essex.ac.uk/csee/research/audio_lab/malcolmspubdocs/J27 DCDF IEE.pdf
http://www.audax.fr/forum/read.php?4,13537,13537,quote=1
Can you give my a "tutorial" how to order the schematics.
And how much will it cost.
If I need service manual from there, I call or I mail. So I cannot give you exactly navigation instructions about the right way for online manual order. I think, in Berlin they speak even your language. At first step I would ask about costs (mainly between 10 Euro and 25 Euro)
I don't think that's quite correct, there is some feedforward,
but the primary mechanism is feedback. The same thing
could be said of the Stasis amplifiers, although there the
feedback is local.
The feedforward path by the quad 405 is for me hard to find.
In general, for a feedforward distortion remover I need a feedforward canceller, i. e., a modulator, that produces exact the same harmonics for each fundamental signals than the unwanted harmonics from an amplifier's class A/B or B stage. Therefore I must know the exactly distortion character of this one.
The theoretical advantage (in opposite to NFB) is the not necessary short delay time, so I have no bandwidt limitation
But such kind of distortion reducing I have never see by a commercial audio amplifier product. But here are a detailled description of this one for radio frequency (RF) power amplifier
http://www.patentretriever.com/cgi-bin/GetURL.cgi
download US4,560,945
In general, for a feedforward distortion remover I need a feedforward canceller, i. e., a modulator, that produces exact the same harmonics for each fundamental signals than the unwanted harmonics from an amplifier's class A/B or B stage. Therefore I must know the exactly distortion character of this one.
The theoretical advantage (in opposite to NFB) is the not necessary short delay time, so I have no bandwidt limitation
But such kind of distortion reducing I have never see by a commercial audio amplifier product. But here are a detailled description of this one for radio frequency (RF) power amplifier
http://www.patentretriever.com/cgi-bin/GetURL.cgi
download US4,560,945
Look at the Renardson MJR-8, which explains the error feedforward in a IHMO very good way to understand:
http://www.angelfire.com/ab3/mjramp/amp8.html
Point is that by the use of feedback taken directly from the power stage output the distortion of that stage appears at the input of the power stage with opposite polarity (and greater amplitude) and thus can be passively summed into the output after the feedback point to cancel the remaining distortion to a large extent (theoretically to zero), at the cost of a slightly higher output impedance. The distortion of the output stage automatically (by the use of feedback) reflects itself into a signal at its input and thus need not be known beforehand.
The tricky part is to let the input and driver stage behave as linear as possible and to manage the frequency dependent gain of the input/driver stage. This gain needs to be precisely known and stable to correctly design the passive summing network.
- Klaus
http://www.angelfire.com/ab3/mjramp/amp8.html
Point is that by the use of feedback taken directly from the power stage output the distortion of that stage appears at the input of the power stage with opposite polarity (and greater amplitude) and thus can be passively summed into the output after the feedback point to cancel the remaining distortion to a large extent (theoretically to zero), at the cost of a slightly higher output impedance. The distortion of the output stage automatically (by the use of feedback) reflects itself into a signal at its input and thus need not be known beforehand.
The tricky part is to let the input and driver stage behave as linear as possible and to manage the frequency dependent gain of the input/driver stage. This gain needs to be precisely known and stable to correctly design the passive summing network.
- Klaus
Last edited:
KSTR:
thank you for this weblink. After read follow
Connecting a resistor, about 1k, at one end to the driver stage output, and at the other end to the output after the inductor will feed forward an opposite phase error signal, and will give output stage distortion nulling if the inductor has a small series resistance, about 0.1ohms, and is increased to about 3uH to compensate for falling driver stage gain at higher frequencies. With D applied to the load via 0.1ohms and -10,000D applied via 1k the distortion reaching the load will be zero.
I note the interchanging between phase shift and group delay, also by many other authors. Therefore, the so-called name "feed forward" and the difficult understanding descriptions for me. My interpretation of MJR-8 and Quad 405 is the parallel mode of two amps at one load. The class A stage drives the load about a serial resistor and the class B stage about an inductor (respective an low pass filter); by the Quad the class B stage is diagonal connected inside of a bridge network (cause additional capacitance between the classA output and the inverted input).
By a real feedforward error correction without harmonic modulator - the distorted signal needs a negative delay time. That is only possible in the Science Fiction world.
thank you for this weblink. After read follow
Connecting a resistor, about 1k, at one end to the driver stage output, and at the other end to the output after the inductor will feed forward an opposite phase error signal, and will give output stage distortion nulling if the inductor has a small series resistance, about 0.1ohms, and is increased to about 3uH to compensate for falling driver stage gain at higher frequencies. With D applied to the load via 0.1ohms and -10,000D applied via 1k the distortion reaching the load will be zero.
I note the interchanging between phase shift and group delay, also by many other authors. Therefore, the so-called name "feed forward" and the difficult understanding descriptions for me. My interpretation of MJR-8 and Quad 405 is the parallel mode of two amps at one load. The class A stage drives the load about a serial resistor and the class B stage about an inductor (respective an low pass filter); by the Quad the class B stage is diagonal connected inside of a bridge network (cause additional capacitance between the classA output and the inverted input).
By a real feedforward error correction without harmonic modulator - the distorted signal needs a negative delay time. That is only possible in the Science Fiction world.
Last edited:
tiefbassuebertr,
your interpretation is not quite accurate.
A global feedback loop is present as well, but it would not be able to do much about the hardly mastered switching and crossover distortion, as the current dumpers are biased in class C. The advantages of feeding forward an error signal to the output are: absence of closed loop, thus no instability and excess gain requirement and the output properties do not depend on the main amplifier`s input and are not defined by the main amplifier (in principle).
There must have been a great deal work (both theoretical and practical) behind the Quad amplifier.
your interpretation is not quite accurate.
A global feedback loop is present as well, but it would not be able to do much about the hardly mastered switching and crossover distortion, as the current dumpers are biased in class C. The advantages of feeding forward an error signal to the output are: absence of closed loop, thus no instability and excess gain requirement and the output properties do not depend on the main amplifier`s input and are not defined by the main amplifier (in principle).
There must have been a great deal work (both theoretical and practical) behind the Quad amplifier.
I don't think that's quite correct, there is some feedforward,
but the primary mechanism is feedback. The same thing
could be said of the Stasis amplifiers, although there the
feedback is local.
Why do you call your design "Stasis" when it just appears to be a compound transistor arrangement ??
regards
Trev
Because the Quad circuit is basically a sort of strange compound darlington arrangement as well, which has also a nice name (called Current Dumping).
The true current dumping is IMO the principle promoted by Sandman (Class S).
Btw.: The main trick in the MX-10000 is not the HCA non switching Class A circuit, it is definitely the feed forward.
Do we all agree, that at least in the MX-10000 a real feed forward signal is applied to the negative LSP terminal ?
IMO all approaches, which try to feed the error signal into the same point (the loudspeaker plus) parallel to the main signal, are always pseudo FF approaches, which actually use a kind of error feedback (like HEC e.g.)
Btw: The "HCA" circuit in the smaller models from Yamaha (MX-2000, MX-1000) unfortunately doesn't contain a non switching circuit nor true FF circuitry on the loudspeaker negative terminal, so if you want to listen to this special circuit, you have to buy you a MX-10000 or build one yourself.
Don't forget that this amp features 2ppm THD20k @6Ohms , so good luck with layout and wiring ...
Current Dumping was already taken.
Seriously, the first versions held the devices at nearly
constant voltage and constant current via cascoding and a
current bootstrap. This is not featured in the later schematics
you see on the web.
Hi tiefbassuebertr,
It´s mainly about voltages. Imagine a balanced resistor bridge, where in a relevant circuit R1/R2=R3/R4. For less power losses, two resistors are replaced with an inductor and a capacitor, giving the same impedance ratio, however, not making the examination easier: 1/C1/R2=R3/L4. The error cancellation relies on that balance (not ß).
No wonder you get confused by explanations like that and start talking about opposite phases, negative group delays and diagonal harmonic modulators.
It´s mainly about voltages. Imagine a balanced resistor bridge, where in a relevant circuit R1/R2=R3/R4. For less power losses, two resistors are replaced with an inductor and a capacitor, giving the same impedance ratio, however, not making the examination easier: 1/C1/R2=R3/L4. The error cancellation relies on that balance (not ß).