This is exactly where i was heading - you got that and jumped into the conversation... I hope more people do this. Anyway, you made a better point and clearly than I might have. FFT doesnt tell us enough. The averaging is a big deal.
Thx, Dick Marsh
I am interested in how an output handles itself when the lf content is running in one quadrant, and then suddenly has to traverse another, not necessarily through zero current/voltage. Bursting a 20Khz sine in quad 1 (+V/+I) such that the output stage flies into either 2 (-V/+I) or 4 (+V/-I) means the opposite pass elements really have to work.. and the entire stage has to respond to that. (I'd hate to think about some active element going into sat because of some silly little pulse...then taking time to recover..)
Where's the fun in that? I like sinx/x far more, it's cleaner.
jn
Last edited:
A pulse test which is as short as one pulse is also important... because music is composed of high crest factor signals that do not repeat continuously (other than musically).
Now we are getting closer to knowing what kinds of tests are relavent and might start to correltae better.
More input, please, from others on the sidelines ! All are welcome in discovery. new tests and new ways to test from existing equipment will show other views... might lead someone to have other insights to pursue.
Thx, Dick Marsh
I hope that more people don't get this confused about what a "harmonic" means or are as befuddled by what "average" means.
If you check back, I stated that I was wrong and apologized to Sy. He accepted such. Evidently you missed those comments.
This post was discussed quite a while ago, I won't repeat my comments. I would prefer to build a Zen rock garden in my front yard.
Two confusions here. If you could find an expression for random noise it would not be random - the clue is in the name! You can do an FFT of random noise, and back again, with no loss of information. You can do an FFT of any time series, and back again, with no loss of information.vacuphile said:
No, you have pointed out a shortcoming in your understanding of FFT in any application.vacuphile said:
I know this is an old post which has already been dealt with by others, but Dick seems to want to agree with it:
RNMarsh said:
You would need to switch on the sig gen long before you decided to do the sin x/x experiment!
The algebra of feedback works exactly for zero forward delay and reliably for negligible forward delay. Issues such as re-entrant distortion are predicted by the algebra, so are not phenomena caused by any delay. In almost any (possibly all?) audio amps the delay is negligible. As I said, a set of LP filters can superficially look like a delay but if they can be compensated away by an inverse filter than they are not a true delay. A real delay can only be compensated by a time machine - a suspension of causality.
For some reason lots of people seem to think that feedback works by the signal progressively getting further and further copies of itself merged back in. It doesn't, at least not for audio. It is pure algebra; effectively instantaneous. It is for each person to decide whether he believes this or not. Disbelief shows a lack of understanding about feedback.
The algebra of feedback works exactly for zero forward delay and reliably for negligible forward delay. Issues such as re-entrant distortion are predicted by the algebra, so are not phenomena caused by any delay. In almost any (possibly all?) audio amps the delay is negligible. As I said, a set of LP filters can superficially look like a delay but if they can be compensated away by an inverse filter than they are not a true delay. A real delay can only be compensated by a time machine - a suspension of causality.
For some reason lots of people seem to think that feedback works by the signal progressively getting further and further copies of itself merged back in. It doesn't, at least not for audio. It is pure algebra; effectively instantaneous. It is for each person to decide whether he believes this or not. Disbelief shows a lack of understanding about feedback.
Last edited:
Yes, let's stay focused on this fundamental point that I feel is important because it has resurfaced time and time again even by folks that should know better. It has been used for years by the no feedback lobby.
It was exactly the reason why I chimed in this discussion, because of annoying resurfacing of this fundamental misunderstanding that leads to wrong decisions. I hope the day will come when people instead of asking, "How much feedback is used in your gear", or "Is feedback used", will ask some more relevant questions.
Last edited:
Just to clarify, for the benefit of those that can do maths, consider the formula
1/(1-x) = 1 +x +x^2 +x^3 +x^4 . . .
This actually models, in a simple way, what happens with feedback. You can consider the infinite series on the RHS as being either a set of higher order distortion terms or a set of progressive interactions. However, they all happen at the same time. The algebraic model is represented by the LHS. The two sides are equal. There are no magic extra problems.
1/(1-x) = 1 +x +x^2 +x^3 +x^4 . . .
This actually models, in a simple way, what happens with feedback. You can consider the infinite series on the RHS as being either a set of higher order distortion terms or a set of progressive interactions. However, they all happen at the same time. The algebraic model is represented by the LHS. The two sides are equal. There are no magic extra problems.
Last edited:
Better read what I stated in earlier posts again. I never stated delay caused distortion. I stated that as the frequency increases, a phase shift takes place and eventually the negative feedback becomes positive. If it passes through or circles 1 + j0, then oscillatoin occurs. As the phase shift takes place, the distortion figure changes. Check Stereophile specs by JA as an example. But why would one design so close as to flirt with an oscillation to begin with?
The rest is not worth arguing about. My professor is dead. Have it your way.
Cheers.
1 = 1 + x + (-x) + x^2 + (-x^2) +x^3 + (-x^3) + x^4 + (-x^4) . . .looks more amusing
- Status
- Not open for further replies.