Cellardoor:
"This waveform (start of) is not uncommon in music sequences - a drum hit at the end of a quiet passage generates a similar starting signal."
Hi Cellardoor,
well ,if captured by microphone, the waveform will not resemble that.
try to do FFT on your input signal.
Best regards,
Hartono
"This waveform (start of) is not uncommon in music sequences - a drum hit at the end of a quiet passage generates a similar starting signal."
Hi Cellardoor,
well ,if captured by microphone, the waveform will not resemble that.
try to do FFT on your input signal.
Best regards,
Hartono
Hi Johan,
You and Rodolfo write much I do not disagree with. So I am not arguing, merely discussing.
I've checked my #366 re your #384. Pure reproduction relates to (inductive) amplifiers, filters and chokes NOT superimposing their own waveform distortions when current drive or flow through them reverses due either to loudspeaker reactance or back-EMF.
Re your #383.
Electronic filtering is used with CD instead of passive to overcome phase effects, but SACD, DVD-A and new generation vinyl would never have arisen had the 16/44 of CD been adequate - which it is not, and we most certainly can't have fidelity if we test and cut all audio down to CD specification.
Hi Nikwal,
The reason an output choke is used in an amplifier is to let the output NFB node voltage slew before load current flows so that amplifier stages are not driven into an overload situation and oscillation due to the drive of the output stage being phase shifted with respect to the differentially sensed error when the output stage cannot quickly enough output sufficient load current without that choke. Thus a choke can alleviate the inadequacies that arise when a global loop often acts faster than the inner loops of a SS amplifier, and this relates to circuit design rather than SS devices themselves.
Hi Tony.
Nice work.
Re your 2nd conclusion in post#413.
The error arising re your 1st conclusion will be at the LS terminals wrt amplifier output terminals, thus any (and it is of higher frequency!) choke/cable error induced by a mid-bass driver and its crossover section will be superimposed upon the tweeter input !!!
There is a need to repeat this examination for waveform error spectrum with a 'composite LS system plus crossover' after the choke.
Cheers ......... Graham.
You and Rodolfo write much I do not disagree with. So I am not arguing, merely discussing.
I've checked my #366 re your #384. Pure reproduction relates to (inductive) amplifiers, filters and chokes NOT superimposing their own waveform distortions when current drive or flow through them reverses due either to loudspeaker reactance or back-EMF.
Re your #383.
Electronic filtering is used with CD instead of passive to overcome phase effects, but SACD, DVD-A and new generation vinyl would never have arisen had the 16/44 of CD been adequate - which it is not, and we most certainly can't have fidelity if we test and cut all audio down to CD specification.
Hi Nikwal,
The reason an output choke is used in an amplifier is to let the output NFB node voltage slew before load current flows so that amplifier stages are not driven into an overload situation and oscillation due to the drive of the output stage being phase shifted with respect to the differentially sensed error when the output stage cannot quickly enough output sufficient load current without that choke. Thus a choke can alleviate the inadequacies that arise when a global loop often acts faster than the inner loops of a SS amplifier, and this relates to circuit design rather than SS devices themselves.
Hi Tony.
Nice work.
Re your 2nd conclusion in post#413.
The error arising re your 1st conclusion will be at the LS terminals wrt amplifier output terminals, thus any (and it is of higher frequency!) choke/cable error induced by a mid-bass driver and its crossover section will be superimposed upon the tweeter input !!!
There is a need to repeat this examination for waveform error spectrum with a 'composite LS system plus crossover' after the choke.
Cheers ......... Graham.
Hi,
Graham explains it here in language I can understand. http://www.diyaudio.com/forums/showthread.php?postid=1213316#post1213316
This fits with (throwaway) comments made by other designers that went over my head due to a lack of understanding (yes, I'm a slow learner - steep learning curve was never in my vocabulary).
Graham explains it here in language I can understand. http://www.diyaudio.com/forums/showthread.php?postid=1213316#post1213316
This fits with (throwaway) comments made by other designers that went over my head due to a lack of understanding (yes, I'm a slow learner - steep learning curve was never in my vocabulary).
estuart said:
G’day Edmond.
I agree. Also, the measured loss in amplitude is actually caused by the LF attenuator made by the series input resistor and the resistor returning Q1's base to ground. Slightly attenuating the input signal is hardly something to worry about - all you have to do is to tweak the volume control a bit to compensate. I’m not sure how this counts as sonic degradation.
Q1 simply cannot be looked upon as a common-base amplifier in operation and I think most people don't realise just how high an impedance the base of Q1 appears to the signal source in an amplifier topology such as this. NFB forces the base of Q2 to follow the base of Q1, so in effect, a bootstrapping action sends the impedance as seen looking into the base of Q1 sky high.
The only real thing to be concerned about with respect to a series input resistor is the effect of Q1's miller capacitance. Fortunately, in most applications, a value of a few kilo-ohms it totally negligible. If you’re worried about it though, all you need to do is to cascode Q1. A sensibly proportioned series input resistor to limit RF in conjunction with a bypass cap to ground is a silly thing to omit.
Cheers,
Glen
G.Kleinschmidt said:G’day Edmond.
I agree. Also, the measured loss in amplitude is actually caused by the LF attenuator made by the series input resistor and the resistor returning Q1's base to ground. Slightly attenuating the input signal is hardly something to worry about - all you have to do is to tweak the volume control a bit to compensate. I’m not sure how this counts as sonic degradation.
Q1 simply cannot be looked upon as a common-base amplifier in operation and I think most people don't realise just how high an impedance the base of Q1 appears to the signal source in an amplifier topology such as this. NFB forces the base of Q2 to follow the base of Q1, so in effect, a bootstrapping action sends the impedance as seen looking into the base of Q1 sky high.
The only real thing to be concerned about with respect to a series input resistor is the effect of Q1's miller capacitance. Fortunately, in most applications, a value of a few kilo-ohms it totally negligible. If you’re worried about it though, all you need to do is to cascode Q1. A sensibly proportioned series input resistor to limit RF in conjunction with a bypass cap to ground is a silly thing to omit.
Cheers,
Glen
True!
However, I prefer a lower input series resistor, say 100R. With 2K (spiced) THD20 rises from 3ppb to 10ppb. But removing the input cap completely give rise to some instability, due to a rather special topology of my amp.
Cheers,
Graham Maynard said:
Graham,
Just to keep the record straight:
I was NOT advocating 'prior modification of an AF signal'. I was advocating using an AF signal in the first place. Your suggested signal needs to go through a filter that makes it an AF signal rather than a signal that contains frequencies from here to eternity.
Jan Didden
Hi Glen, ( also Edmond by your "ditto" )
You asked >> But why, for heaven's sake, calling this a part of the sensing network? <<
Because it is !!!
"For every action there is an equal but opposite reaction" this being one of the first fundamental rules I was taught in science.
Thus, when you check the damping characteristics of a NFB amplifier via the output terminal, you can see the effect an input filter has !
Current flowing through an input filter is not constant, but modified by inner loop response wrt global response, so although NFB will attempt to minimise the voltage difference between the differential bases, this does not mean we can assume that the voltage at the first base remains an identical copy of the source voltage waveform at the input socket with simple group delay, because the group delay and thus base voltage become modified by the base current flowing in response to NFB attempting to control the inner loop characteristics !!!
Cheers ........ Graham.
You asked >> But why, for heaven's sake, calling this a part of the sensing network? <<
Because it is !!!
"For every action there is an equal but opposite reaction" this being one of the first fundamental rules I was taught in science.
Thus, when you check the damping characteristics of a NFB amplifier via the output terminal, you can see the effect an input filter has !
Current flowing through an input filter is not constant, but modified by inner loop response wrt global response, so although NFB will attempt to minimise the voltage difference between the differential bases, this does not mean we can assume that the voltage at the first base remains an identical copy of the source voltage waveform at the input socket with simple group delay, because the group delay and thus base voltage become modified by the base current flowing in response to NFB attempting to control the inner loop characteristics !!!
Cheers ........ Graham.
Hi Graham,Graham Maynard said:
I did NOT ask: "But why, for heaven's sake, calling this a part of the sensing network?"
Rather: "But why, for heaven's sake, calling this a part of the sensing network?" Notice the Italics.
The feedback signal is sensed at the output and only at the output. Of course, the impedance of the input filter (or output impedance of the pre-amp, I don't care) does have an effect on the NFB. Suppose, for example, that the input is completely left open, (only for AC of course), then we have (almost) no NFB at all. But this doesn't mean that the NFB is sensed at the input filter, or the latter is part of the sensing network.
Cheers,
estuart said:
True!
However, I prefer a lower input series resistor, say 100R. With 2K (spiced) THD20 rises from 3ppb to 10ppb. But removing the input cap completely give rise to some instability, due to a rather special topology of my amp.
Cheers,
Oh, har-har.......from 3 parts per billion to 10 parts per billion.....
Me thinks that confirms, to a degree, my claim that a resistor of a few k has a rather neglible effect on an amplifiers sonic performance.....
Cheers,
Glen
Graham Maynard said:
Hi Graham.
Of course we can't assume that the voltage at the base of Q2 will be exactly the same as that at Q1. Basic negative feedback theory tells us why - a signal+error voltage needs to be developed between the two bases.
However, when we properly analyse the effects of such things we can reach rational conclusions about their influence on the operation of the circuit.
Cheers,
Glen
G.Kleinschmidt said:
Oh, har-har.......from 3 parts per billion to 10 parts per billion.....
Me thinks that confirms, to a degree, my claim that a resistor of a few k has a rather neglible effect on an amplifiers sonic performance.....
Cheers,
Glen
But these ppb figurers hold only for a completely bootstrapped input stage (including the LTP current sources and mirrors).
Cheers,
Hi,
most of us recognise that a standard LTP circuit works in two parts.
The non-inverting input transistor amplifies in common emitter configuration.
The other half (the inverting input transistor) operates in common base configuration receiving it's input from the emitter of the first transistor.
I had overlooked the fact the the LTP can and MUST operate in the opposite direction in exactly the same way. An input into the inverting input from the NFB loop swaps the configurations between the two transistors around.
The non-inverting input transistor operates in common base mode when an input signal is fed into the inverting transistor base. Simple symmetry demands as much.
Are Estuart & GK saying that something else is at work here?
Or are you disputing that the quality of the common base does not enter into the NFB amplification process?
Let's stop this arguing and agree the facts so that us less knowledgeable amateurs can understand the electronics that are at play.
most of us recognise that a standard LTP circuit works in two parts.
The non-inverting input transistor amplifies in common emitter configuration.
The other half (the inverting input transistor) operates in common base configuration receiving it's input from the emitter of the first transistor.
I had overlooked the fact the the LTP can and MUST operate in the opposite direction in exactly the same way. An input into the inverting input from the NFB loop swaps the configurations between the two transistors around.
The non-inverting input transistor operates in common base mode when an input signal is fed into the inverting transistor base. Simple symmetry demands as much.
Are Estuart & GK saying that something else is at work here?
Or are you disputing that the quality of the common base does not enter into the NFB amplification process?
Let's stop this arguing and agree the facts so that us less knowledgeable amateurs can understand the electronics that are at play.
Hi Andrew,
You are perfectly right regarding the feedback path. Of course, the non-inverting input transistor operates in common base mode. Therefor, as I said before, the (non-inverting) input has to be terminated with a sufficiently low impedance, in particular at HF. BTW, it just happens that an input filter does this job, as it acts like a ground return path at HF.
Are Estuart & GK saying that something else is at work here?
YES! And I mean this: Although an input filter does influence the behavior of the NFB loop, it is NOT a part of the sensing network. The same applies to Cdom for example, it has an (enormous) impact on the NFB loop, but, also in this case, it is not a part of the (sensing) NFB loop.
Cheers,
You are perfectly right regarding the feedback path. Of course, the non-inverting input transistor operates in common base mode. Therefor, as I said before, the (non-inverting) input has to be terminated with a sufficiently low impedance, in particular at HF. BTW, it just happens that an input filter does this job, as it acts like a ground return path at HF.
Are Estuart & GK saying that something else is at work here?
YES! And I mean this: Although an input filter does influence the behavior of the NFB loop, it is NOT a part of the sensing network. The same applies to Cdom for example, it has an (enormous) impact on the NFB loop, but, also in this case, it is not a part of the (sensing) NFB loop.
Cheers,
- Status
- Not open for further replies.