Graham Maynard said:
Err, well yes, but I had to strain to hear it with headphones. The corresponding input was a typical Miles Davis muted trumpet riff. If you've heard these, you know that the sound is quite piercing. At the level for which I could just hear the error, the actual music input was quite loud.
New choke-induced high frequency components? Nonlinear? I don't know where you pulled this stuff out of. The simulated circuit was purely linear, so no new frequency components can be generated. If you're talking about scaling of the existing components, well sure, but everyone knows that linear circuits do that. No rocket science there. Maybe you're referring to some other circuit that you're not telling us about?
-22 dB is actually about 7.9 percent voltage. You're equating the scale factor of a linear system with a percentage distortion figure usually used to describe nonlinear circuits. Doing so is not at all useful and only adds confusion to the issue
I thought it would be reasonably obvious why I chose a scale factor of one here. Setting the scale factor to one preserves the amplitude relationship between the error signal and the input. If the scale factor were different from one, I wouldn't be able to determine the relative loudness of the error and the input. I can do so by having a playlist with two songs - the input and the error. By skipping back and forth between the two, I can get a good idea of the relative loudness of error and input. However, I can still exaggerate the error relative to the input if I want to by turning up the volume on the headphone amp when I switch from the input to the error. The headphone amp of the Benchmark DAC has plenty of extra gain for that. It should be very clear though, that scaling the error by anything other than one will lose the baseline of the error data - namely its true relative loudness when compared with the input.
Hi Andy,
Thanks for your discussion.
Please note, I asked for this investigation because there are those who categorically state that a series output choke has no effect upon audio.
It does; but attempting to prove it has led me into all manner of enquiry and challenge (???).
Linearity;- where output *continuously* varies in *direct proportion* to input.
Why is there this claim that a reactive network is said to respond linearly with dynamic waveforms ?
A reactive network can respond linearly only when there is a steady waveform, and only if the introduced waveform delay is acceptable at that single examination frequency !!!
__________________________________________
The alternating current flowing through, and thus the frequency related error voltage developed by, the series output choke (or a loudspeaker cable!) is directly related to composite loudspeaker current flow, not amplifier output voltage; ie. the choke error cannot be *linear* as compared to when that choke would be in series with another purely inductive load component !
___________________________________________
Composite loudspeaker currents change during every music cycle; this is due to delayed back-EMF development as capacitors store and release alternating charges and inductor fields attempt to generate appropriate back-EMFs. However, with changing input there is always a dynamically induced error component due to equilibrium never properly being able to develop. (Sub circuit resonances)
Where does crossover section delay come from ?
What happens to current change during that delay period ?
Current is not the same at the initiation of any music bandwidth limited change as it is after the first ninety degrees of dynamic waveform transit per single reactive component, and the components are serial; sub-sections are dynamically energised in a frequency relatedly serial manner, not synchronously, thus loudspeaker back-EMF becomes modified in time wrt input. (the same applies with bass driver/air-spring)
Thus the composite error voltage is not linearly related to input because of different sequential delays and different degrees of back-EMF development with frequency.
_____________________________________________
Has anyone checked for the variation in waveform delay through series impedance with frequency, due to a loading composite loudspeaker's inconstant back-EMF generation ?
The delay (error) is not constant, and yet the amplifier's driving voltage cannot change in time to accomodate it.
Time-frequency-amplitude. This is why the seemingly insignificant steady sinewave measured losses due to any series driving impedance do not equate to the surprisinly large dynamically induced changes in reproduction we actually hear. Time shifted (ie. non-linear) error components are generated !
______________________________________________
I still wonder whether 16/44.1 at the error detector is accurately capable of illustrating all the error.
______________________________________________
Cheers ......... Graham.
Thanks for your discussion.
Please note, I asked for this investigation because there are those who categorically state that a series output choke has no effect upon audio.
It does; but attempting to prove it has led me into all manner of enquiry and challenge (???).
Linearity;- where output *continuously* varies in *direct proportion* to input.
Why is there this claim that a reactive network is said to respond linearly with dynamic waveforms ?
A reactive network can respond linearly only when there is a steady waveform, and only if the introduced waveform delay is acceptable at that single examination frequency !!!
__________________________________________
The alternating current flowing through, and thus the frequency related error voltage developed by, the series output choke (or a loudspeaker cable!) is directly related to composite loudspeaker current flow, not amplifier output voltage; ie. the choke error cannot be *linear* as compared to when that choke would be in series with another purely inductive load component !
___________________________________________
Composite loudspeaker currents change during every music cycle; this is due to delayed back-EMF development as capacitors store and release alternating charges and inductor fields attempt to generate appropriate back-EMFs. However, with changing input there is always a dynamically induced error component due to equilibrium never properly being able to develop. (Sub circuit resonances)
Where does crossover section delay come from ?
What happens to current change during that delay period ?
Current is not the same at the initiation of any music bandwidth limited change as it is after the first ninety degrees of dynamic waveform transit per single reactive component, and the components are serial; sub-sections are dynamically energised in a frequency relatedly serial manner, not synchronously, thus loudspeaker back-EMF becomes modified in time wrt input. (the same applies with bass driver/air-spring)
Thus the composite error voltage is not linearly related to input because of different sequential delays and different degrees of back-EMF development with frequency.
_____________________________________________
Has anyone checked for the variation in waveform delay through series impedance with frequency, due to a loading composite loudspeaker's inconstant back-EMF generation ?
The delay (error) is not constant, and yet the amplifier's driving voltage cannot change in time to accomodate it.
Time-frequency-amplitude. This is why the seemingly insignificant steady sinewave measured losses due to any series driving impedance do not equate to the surprisinly large dynamically induced changes in reproduction we actually hear. Time shifted (ie. non-linear) error components are generated !
______________________________________________
I still wonder whether 16/44.1 at the error detector is accurately capable of illustrating all the error.
______________________________________________
Cheers ......... Graham.
Graham Maynard said:
Oh no! Impulse response and step response are also responses of the linear system
In case of linear system, response to every shape of the input signal can be calculated when the impulse response (or step response) is known.
You can also calculate your favorite first-cycle distortion ...
Hi Jorge,
I had realized the position of the res. but in this forum not there are theoretical certainties.
Thanks for the Norton formula, I use currents drivers only after several years, and not all the nonelementary mechanisms of the electronics are known me.
Rodolfo post #538 to share.
Graham post #540 to share.
Ciao
Mauro
I had realized the position of the res. but in this forum not there are theoretical certainties.
Thanks for the Norton formula, I use currents drivers only after several years, and not all the nonelementary mechanisms of the electronics are known me.
Rodolfo post #538 to share.
Graham post #540 to share.
Ciao
Mauro
Jorge said:
Jorge,
I am on the road, cannot check Hawksford paper. I understand that non-linear elements in the electrical-mechanical sustem may generate non-linear distortion. But, that is then a property of the system, and not of the back EMF, right? In other words, if it would be possible to remove the back EMF, the non-linear dist would still be there. And the back-EMF itself does not contribute to the back EMF? So, are we maybe comparing apples and car-tires?
Jan Didden
Graham Maynard said:
Graham,
Perhaps we come finally to the source of misunderstandings. In my understanding, a network whose properties do not vary with the signal is a linear network. A xover network, while complex, inductive, capacitive, generating time-varying responses, still is a linear network (in my view at least).
So, whatever you throw at it, the error voltage, however defined, will only have non-linear errors, no harmonic distortion, no IM.
So the output inductor of an amp that carries that dynamically varying load current, does generate an error voltage (assuming that this is meant to be the voltage across the inductor), which is linearly related to the input signal.
Jan Didden
I guess it is a matter of semantics, the basic relationship between voltage and current in an ideal inductor is well known and easily described, no nonlinearity is involved. Just as with back-EMF you use the term nonlinearity in a way I do not. A combination of ideal R's, L's, and C's simply cannot generate a signal at a frequency that is not contained in the excitation.
Graham Maynard said:
Graham:
Just to emphasize what has been said - particularly by Janneman.
A linear system is one that complies with:
1. The output for an input scaled by an arbitrary constant equals the output for the original signal scaled by the same amount.
2. The output for a composite (sum) of 2 or more signals equals the sum of the outputs corresponding to each individual input signal.
Customarily this two requisites are compacted in the definition:
"The output of a linear system for a linear combinations of inputs, is the linear combination of the individual signal outputs."
Where "linear combination" means the addition of an arbitrary number of items of the same nature, each one in time multiplied by a respective arbitrary constant.
Lc= a1.x1+a2.x2 ..... where a1 ... are constants and x1 ... are signals.
So for a linear system it holds:
A(a1.x1+a2.x2+...)=a1.A(x1)+a2.A(x2)+....
Where A is the system transfer function.
There is a curse to surface in sites like this, where a certain term as "linearity" has for some a very precisely defined meaning, and that can also be interpreted literally and unknowingly in a different fashion by others non familiar with the former more rigorous one.
The circuit simulated by Andy is linear by definition, and as such it behaves. Perhaps the term "linear distortion" is an unhappy one in that it evokes nonlinear behavior. What we mean here is a form of distortion none other that plain old frequency amplitude/phase deviation form flatness.
Rodolfo
P.S. Just to emphasize the last. A graphic equalizer set at any other position but 0dB in each and every band, is "nonlinear" (in the abused sense).
janneman said:
Jan
I relly cannot understand fully your question. From what I understood:
One can separate the two main components of speaker impedance: an ideal voltage generator and a distortion generator, both dependent on cone movement.
The ideal generator generates pure EMF in opposition to the driving source - so this is real back EMF. It's amplitude is small enough not to generate any significant distortion in a decent amplifier, as has been posted many times, starting with Hiraga.
The distortion component generator is a property of the speaker, it's amplitude is less than the EMF generator, and will matter even less to the amp.
So, back EMF as a factor that induces distortion in an amp is not significant.
Now, Hawksford papers are about how to design an amplifier to lower the intrinsec speaker distortion and not the other way around - that is, the speaker EMF generators still exists, but the outputs will be reduced by this amplifier.
The negative side is that it does not work with standard design speakers.
So yes, we may quite well be comparing apples to car tires.
Hi Jan,
Aha ! It has been a differently held understanding of the term 'linear' that has caused an inability for us to appreciate each others views.
But L-C 'linear' in time ?
Only with steady waveforms, and if you examine each in time isolation.
Hi Rodolpho,
What you relate is very clearly written, but it does not suggest the same outcome that some other contributors here defend;
ie. they claim that the choke does not cause tweeter waveform distortion because the error voltage it develops is entirely linear with amplifier output.
By your reasoning ( and I am sure others here agree ) there can be a split of sum total energy between tweeter and mid/bass sub-circuits at the series output choke or cable end.
In other words, if the mid/bass sub-circuit, or if the choke/common cable, were to be removed, the tweeter voltage would be different.
Yet the tweeter voltage due to direct amplifier connection must be correct because it has not been modulated by co-connected mid/bass sub-circuit resonances which, via their current flow, develop a choke/cable potential that is not continuously linear with overall amplifier output.
As you say - plain old frequency amplitude/phase deviation; but with dynamic waveform change this equates to an initial non-linearity in time.
Cheers .......... Graham.
Aha ! It has been a differently held understanding of the term 'linear' that has caused an inability for us to appreciate each others views.
But L-C 'linear' in time ?
Only with steady waveforms, and if you examine each in time isolation.
Hi Rodolpho,
What you relate is very clearly written, but it does not suggest the same outcome that some other contributors here defend;
ie. they claim that the choke does not cause tweeter waveform distortion because the error voltage it develops is entirely linear with amplifier output.
By your reasoning ( and I am sure others here agree ) there can be a split of sum total energy between tweeter and mid/bass sub-circuits at the series output choke or cable end.
In other words, if the mid/bass sub-circuit, or if the choke/common cable, were to be removed, the tweeter voltage would be different.
Yet the tweeter voltage due to direct amplifier connection must be correct because it has not been modulated by co-connected mid/bass sub-circuit resonances which, via their current flow, develop a choke/cable potential that is not continuously linear with overall amplifier output.
As you say - plain old frequency amplitude/phase deviation; but with dynamic waveform change this equates to an initial non-linearity in time.
Cheers .......... Graham.
Dear Janneman
you wrote:
I think you miss one point:
the open loop output impedance is non-zero, thus highly reactive, (although linear) load does affect the phase accuracy of the signal which is fed back.
cheers
Postscriptum
and yes, xover+speaker+acoustic room LRC model is linear circuit by definition
you wrote:
I think you miss one point:
the open loop output impedance is non-zero, thus highly reactive, (although linear) load does affect the phase accuracy of the signal which is fed back.
cheers
Postscriptum
and yes, xover+speaker+acoustic room LRC model is linear circuit by definition
darkfenriz said:
Yes, agreed, and that phase shift IS linear distortion. It does NOT generate harmonic components, or IMD components.
The term linear distortion is commonly used to mean deviations from the desired (normally flat) freq and phase response. This is the type of distortion generated by linear components, like a cap or an inductor. Except for overload of saturation, linear components do not change depending on the signal. A cap value does not change depending on the signal amplitude. This is basically what Rodolfo above showed mathematically. The response of a linear component to a composite singnal is equal to the combined response of the signals if they were used separately.
Non-linear distortion is caused by components whose properties change with the signal, like the gain of a transistor is dependent on the signal amplitude. So, the output of a transistor stage from a composite signal is NOT the same as the combination (vector sum) of outputs of the individual signals.
So, as I see it, back-EMF and other speaker properties like vc inductance, mech resonance, all contribute in some way to the linear distortion like non-linear freq characteristic and non-linear group delay. (To a first approximation; I know that for instance the 'springiness' of the cone suspension is somewhat dependent on the cone position, but that is a much lower secondary effect).
Now back to the subject, I have the impression that Graham talks about what I call linear distortion. And as such, I agree with him, but at the same time that then means that what he posits are phenomenons that are well known for many years, and I am lost as to what the significance here is. I do not mean that negatively, it is just how I see it.
Jan Didden
Graham Maynard said:
Getting closer...
Hi Graham,
Thanks for the further precision. So, the question now appears to be, how can a network of linear components (and the cable, output inductor etc are linear components) cause a non-linear response as I understand non-linear responses, the ones that cause harmonic components and IM components. Are we on the same sheet of music now?
With your expression "with dynamic waveform change this equates to an initial non-linearity in time", do you mean my view of non-linearities, and if not what is your meaning then of these "non-linearities"?
Jan Didden
Hi Jan
I thought you would go a step further, maybe I wasn't clear enough.
I meant to say:
Thank you for interesting exchange of views.
I thought you would go a step further, maybe I wasn't clear enough.
I meant to say:
..., which must result in some extra group delays and an increase in non-linear distortions like TIM.
Does not generate non-linear distortion ITSELF, but can affect the work of acive non-linear devices.
Thank you for interesting exchange of views.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.