What happened to the natural phase shift (that should occur in every transistor due to it's internal C), but become disabled when the loop is closed by feedback to the differential?
The "open loop" operation of the amp cct when the global loop is open-loop is different than the "open loop" operation when the loop is closed by feedback? One can have their natural phase shift, the other one is "pressed" to have no phase shift by differential feedback, what happened here?
Is that really so? Is it wise to base engineering decisions on opinions and value judgements rather than on scientifically and statistically viable results? This is equivalent to saying that the ultimate goal of amplifier development is not building the technically best amplfier, but selling the best-sounding one (whatever best-sounding may mean). Which in itself is not bad, but not the same issue at all.
Cheers,
bk
Like TLF9999 said, the transistor is still the same, it doesn't know wheter it is used in common emitor, or common base, or in closed loop, or in open loop, it is operating without knowing it's position.
If we know what makes that same transistor can have less phase shift, maybe we can use that same cause for some good merit (in design stage), for example, to use the same transistor but having far less phase shift with some clever cct arrangement.
If we know what makes that same transistor can have less phase shift, maybe we can use that same cause for some good merit (in design stage), for example, to use the same transistor but having far less phase shift with some clever cct arrangement.
When I listen to to any piece of music or what ever I will hear it differently to everyone else - why because I've had a hearing test done and there is a chunk missing in my hearing spectrum between 2000 and 3000 Hz so that naturally will colour what I hear. Likewise my wifes left ear is affected more above 3000Hz and not her right so to her sound comes from a different direction.
My hearing cuts out at 11 KHz my wife at 13KHz and my daughter at 16KHz - we are ALL different so don't try and tell me this is better than that - to you it may be, but that is YOUR perception through your ears picking up the frequencies YOU can hear. Everyone else will hear something different.
David L
My hearing cuts out at 11 KHz my wife at 13KHz and my daughter at 16KHz - we are ALL different so don't try and tell me this is better than that - to you it may be, but that is YOUR perception through your ears picking up the frequencies YOU can hear. Everyone else will hear something different.
David L
Hi David,
What you say is true but not that important. If you all listen to something - whatever - and listen to the same thing reproduced, if it sounds the same to you then there is high fidelity. Same for each member of the listening party. Now if there is a defect that you can't hear, there is still high fidelity as far as you are concerned. If the entire party agrees to hearing what they heard live, chances are most, if not all would agree. My money is on your daughter, no stereotypes or preconceptions to deal with.
Your point of reference is you. Simple.
-Chris
What you say is true but not that important. If you all listen to something - whatever - and listen to the same thing reproduced, if it sounds the same to you then there is high fidelity. Same for each member of the listening party. Now if there is a defect that you can't hear, there is still high fidelity as far as you are concerned. If the entire party agrees to hearing what they heard live, chances are most, if not all would agree. My money is on your daughter, no stereotypes or preconceptions to deal with.
Your point of reference is you. Simple.
-Chris
I think this points to exactly what I was saying earlier. The problem with a THD+N analysis with an FFT is that some distortions, that we have real mechanisms for, simply appear as a higher noise floor. But they are not AIWN. These distortions contain correlated energy, and some of them are perceivable by the human hearing system.
This further points to a danger in the glib analysis of distortion mechanisms that says "the level is below the noise floor" (and hence is not perceivable.) Trouble is that the only noise floor that matters is the one that is defined by the limits of human perception. The noise floor seen in an FFT is not the same thing as the perceptual noise floor and it is quite reasonable for it to carry significant energy that is perceived by the ear. Indeed it may carry enough energy to contain quite obnoxious distortion products.
How about using FFT data and perform cross correlation analyses on the noise? This could pick up some signal if any.
In the peripheral of this subject, two interesting tidbits:
1- ppl have successfully created chaotic noise using standard analog op amp technology. Then they encoded a message in the noise. The receiver was a similar circuit fed with the noise signal. The (chaotic, quasi deterministic) noise generators were successfully synchronized and the signal decoded.
2- highly accurate identification of materials through surface laser scanning enables ID of specific materials from specific sources (such as product copyright protection needs). In practice authoirs show that you can scan any surface, say a credit card, and the random (=noisy) surface pattern of microscopic substructures creates a data stream oin the laser scanner. A similar piece of material scanned and the data cross correlated with the first one allows positive ID. In other words you can recover *significant* information from noise, either chaotic one, or truly randoim one.
Sources:
1 - Sync Chaos
2 -Everything has a fingerprint
In the peripheral of this subject, two interesting tidbits:
1- ppl have successfully created chaotic noise using standard analog op amp technology. Then they encoded a message in the noise. The receiver was a similar circuit fed with the noise signal. The (chaotic, quasi deterministic) noise generators were successfully synchronized and the signal decoded.
2- highly accurate identification of materials through surface laser scanning enables ID of specific materials from specific sources (such as product copyright protection needs). In practice authoirs show that you can scan any surface, say a credit card, and the random (=noisy) surface pattern of microscopic substructures creates a data stream oin the laser scanner. A similar piece of material scanned and the data cross correlated with the first one allows positive ID. In other words you can recover *significant* information from noise, either chaotic one, or truly randoim one.
Sources:
1 - Sync Chaos
2 -Everything has a fingerprint
There is much fun to be had here.
If you take the FFT and elide the known components - those due to either the original signal or those belonging to the set of components modelled by harmonic distortion, then take an FFT of the rest you end up with a time domain signal that contains anything not part of either of those components. This you could actually listen to, and thus determine audible nature of the residual.
Is a difficult process, precise elision of a spectral peak is messy at best, and a peak with non-zero width not easy to interpret,
The other two examples bring to mind Andrew Chaikin's Algorithmic Information Content. It is always very instructive to look at any information stream in this light.
The "chaotic" analog process is probably not very random at all. The mere fact it can be sync'ed up at all is a good pointer. That is seems very random to the eye is a very bad guide. It is likely that its operation could be captured with a relatively small number of parameters. The space occupied the parameter set determines the entropy of the source, and thus the difficulty in cracking any such encryption. If you had some idea of the design of the chaotic system you could begin to construct models, and the application of Baysian analysis would quickly guide you to a parameter set that decoded the stream. Steve Wolfram's A New Kind of Science spends a great deal of time constructing discrete chaotic algorithms that have the appearance of random, or very complex, action, but are similarly simply the result of a simple paramaterisable process.
This is exactly analogous to what I have been talking about. In one domain a signal may appear random, but with a a suitable basis set, it can be revealed to be anything but. In audio the basis set is composed of a series of dynamic resonator bodies, not a harmonic series. Thus we should expect that a Fourier series will classify as noise some quite audible signals.
In Chaikin's world, the chaotic stream could be modelled with a reasonably short algorithm. The size of the needed algorithm is the metric of the information content of the stream, and thus its entropy.
A data stream that really does seem to be purely random is a bit of an enigma. If it is truly random there does not exist an algorithm shorter than the data stream that can code the data stream. Or another way of looking at it, more classically, there is no algorithm that can predict the next value, given all preceding values. The information content of such a stream of exactly zero. Yet it has the highest AIC! It is of course the perfect encryption key. And in the case of the scanned card for ID, it contains the perfect minimum of encrypted information - it is its mere presence that confirms identity. Since its AIC is maximum, every bit has maximum possible value in determining identity.
If you take the FFT and elide the known components - those due to either the original signal or those belonging to the set of components modelled by harmonic distortion, then take an FFT of the rest you end up with a time domain signal that contains anything not part of either of those components. This you could actually listen to, and thus determine audible nature of the residual.
Is a difficult process, precise elision of a spectral peak is messy at best, and a peak with non-zero width not easy to interpret,
The other two examples bring to mind Andrew Chaikin's Algorithmic Information Content. It is always very instructive to look at any information stream in this light.
The "chaotic" analog process is probably not very random at all. The mere fact it can be sync'ed up at all is a good pointer. That is seems very random to the eye is a very bad guide. It is likely that its operation could be captured with a relatively small number of parameters. The space occupied the parameter set determines the entropy of the source, and thus the difficulty in cracking any such encryption. If you had some idea of the design of the chaotic system you could begin to construct models, and the application of Baysian analysis would quickly guide you to a parameter set that decoded the stream. Steve Wolfram's A New Kind of Science spends a great deal of time constructing discrete chaotic algorithms that have the appearance of random, or very complex, action, but are similarly simply the result of a simple paramaterisable process.
This is exactly analogous to what I have been talking about. In one domain a signal may appear random, but with a a suitable basis set, it can be revealed to be anything but. In audio the basis set is composed of a series of dynamic resonator bodies, not a harmonic series. Thus we should expect that a Fourier series will classify as noise some quite audible signals.
In Chaikin's world, the chaotic stream could be modelled with a reasonably short algorithm. The size of the needed algorithm is the metric of the information content of the stream, and thus its entropy.
A data stream that really does seem to be purely random is a bit of an enigma. If it is truly random there does not exist an algorithm shorter than the data stream that can code the data stream. Or another way of looking at it, more classically, there is no algorithm that can predict the next value, given all preceding values. The information content of such a stream of exactly zero. Yet it has the highest AIC! It is of course the perfect encryption key. And in the case of the scanned card for ID, it contains the perfect minimum of encrypted information - it is its mere presence that confirms identity. Since its AIC is maximum, every bit has maximum possible value in determining identity.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.