The straightforward reason for that is that the brain has less work to do, trying to filter out the musical performance from all the distortion/noise artifacts. If a NR device reduces the contribution of unwanteds produced solely within the playback system, then the brain works more effectively on separating the crud in the source signal waveform.
It's all about adjusting the workload for the brain ...
Bybee devices appear to sometimes take a little signal with whatever they are fixing. It has always annoyed me, but usually they work to the better.
Hi Scott,
Hi fas42,
-Chris
I do. It was a moment of weakness I tell ya!
Hi fas42,
It's not the feedback that is getting slow. It's the circuit driven to a point where it doesn't function well. You got things a little backwards there and fingered the wrong guy! Once you take a circuit beyond its design limits, any conclusions are invalid.
-Chris
HI Stuart, snup,
They are great for reference as well. There is a lot of really good information buried in those pages. Jan works hard to put it all together.
-Chris
So can I. The Webzine is so useful that I buy them, even though I can't really comfortably afford them. I'll continue to do so.
They are great for reference as well. There is a lot of really good information buried in those pages. Jan works hard to put it all together.
-Chris
I know about the hard work. And lately Marsh's headamp was the nearly trigger. It is basically lazyness, that I havent subscribed. I havent had time for this hobby for the last 2 years. just picking up speed again. I havent read a mag in years.
Chris, I'm saying exactly what you're saying - the circuitry, minus the feedback loop, is being driven to a point where it misbehaves, by the feedback - the feedback loop if intelligently engineered will have minimal delay, it's not the problem unless poorly integrated into the actual physical circuit.
Design limits? What are they - take a 20kHz signal, feed it in into a lower impedance load, give the circuit a realistic power supply, with all the parasitics, throw in a bit of mains noise and distortion for extra authenticity, ramp up the volume a bit - and watch the behaviour fall off a cliff. I was able to get a reasonable circuit posted on this forum to immediately generate 100x worse distortion then nominally sim'ed to achieve, by doing this.
Last edited:
And I'll bet a lot of it, unless it is indeed skirting the edges of gross overload, is that most of the strange behavior is due to simulator artifacts. One learns to distrust simulators, and in some cases massage them to see a change in the results. As an example, a very common phenomenon is making step sizes too large, which can provoke numerical oscillations. If you see what appears to be thus, look at the waveform and see if the half-period of the putative oscillation is equal to the step size. Another sign of untrustworthiness is when Fourier tells you that a whole series of harmonics are of nearly equal amplitude---an almost sure sign that there are long time constants lurking, and you need to increase the sim time significantly. You can set things up to look at the last bits, rather than demand a huge amount of plotting.
Of course, best of all: breadboard the thing and see if there is a correspondence between simulations and reality.
Yes, but the real world awaits ... speakers with unpleasant impedance curves, real mains power being fed into real power supplies that have been fitted to your circuit, and someone enthusiastically seeing "what she'll do" ...
I believe in stress testing, immediately, to highlight the weaknesses - no point in attempting to refine something that has an intrinsic less than optimum area in its design, in my book; if that area is rethought at the first instance, then a far more robust solution is highly likely to be the result.
I believe in stress testing, immediately, to highlight the weaknesses - no point in attempting to refine something that has an intrinsic less than optimum area in its design, in my book; if that area is rethought at the first instance, then a far more robust solution is highly likely to be the result.
snup, that doesn't make sense. Why a new army truck, say, can do its job no matter what, is because it's put through hell before being accepted by the defence force - then, it always behaves as desired, no surprises.
No. It's just the realistic result of the components being pushed beyond their normal "comfort" zone - typically, having voltages and currents slewing at rates beyond "nice" levels; if the models of the components have realistic parasitics included then the glitching behaviours will happen, they can't not do so.
If one wants to see if the simulator itself is misbehaving then just use ramping sinusoidal waveforms to drive the circuit - a discontinuity of behaviour at some point, because of software bugs, should be pretty obvious.
I was about to say......
but then i thought better of it ...3.2.1....
Why would you use a faulty program for your file-processing. I don't quite see the benefits.
but then i thought better of it ...3.2.1....
Why would you use a faulty program for your file-processing. I don't quite see the benefits.
Hi fas42,
Feedback cannot speed anything up. If you could open the feedback loop and feed the appropriate signal amplitude in at the frequencies you are talking about, the circuit would misbehave - badly. The feedback loop is not an active circuit, it can't force anything too fast. What it can show you are stages saturating in between, slew rate limiting and things like that.
If you are talking about feedback saturating a clipped circuit, that is unrealistic too. As soon as any stage saturates, feedback is defeated and the normal transfer characteristic no longer applies. The only difference with feedback applied is that now the input stage knows what the output is doing and it attempts to correct the problem. It will overshoot, guarantied. But the feedback network didn't speed anything up. The presence of feedback only enabled corrective action in a circuit that wasn't working to begin with. You can't blame feedback for this.
I don't simulate anything. Every single thing I can talk about has been running on my bench and measured. Therefore, I can not comment on simulator problems.
-Chris
Feedback cannot speed anything up. If you could open the feedback loop and feed the appropriate signal amplitude in at the frequencies you are talking about, the circuit would misbehave - badly. The feedback loop is not an active circuit, it can't force anything too fast. What it can show you are stages saturating in between, slew rate limiting and things like that.
If you are talking about feedback saturating a clipped circuit, that is unrealistic too. As soon as any stage saturates, feedback is defeated and the normal transfer characteristic no longer applies. The only difference with feedback applied is that now the input stage knows what the output is doing and it attempts to correct the problem. It will overshoot, guarantied. But the feedback network didn't speed anything up. The presence of feedback only enabled corrective action in a circuit that wasn't working to begin with. You can't blame feedback for this.
I don't simulate anything. Every single thing I can talk about has been running on my bench and measured. Therefore, I can not comment on simulator problems.
-Chris
Chris, I think I'll pull out that circuit I was playing with, and take a few screenshots of what I'm talking about, make it easier to point at the behaviours.
Key in all this is that the corrective action is effectively "forcing" the circuit to work at very high speed, at the precise instant of the distortion being worse - the whole circuit is slewing like crazy at that moment to counter the detected non-linearity at the output, which will be more or less successful, depending upon everything.
Feedback will be brilliant, if the whole of the circuit, including power supplies, is engineered to do the job - I've done a number of rounds of preliminary exploring of ideas where I'm aiming for predictable, "correct" behaviour up to 200kHz - this gives me excellent results, and a very significant safety margin in behaviour.
Key in all this is that the corrective action is effectively "forcing" the circuit to work at very high speed, at the precise instant of the distortion being worse - the whole circuit is slewing like crazy at that moment to counter the detected non-linearity at the output, which will be more or less successful, depending upon everything.
Feedback will be brilliant, if the whole of the circuit, including power supplies, is engineered to do the job - I've done a number of rounds of preliminary exploring of ideas where I'm aiming for predictable, "correct" behaviour up to 200kHz - this gives me excellent results, and a very significant safety margin in behaviour.
- Status
- Not open for further replies.