Dan, like a lot of people, here, I had deal with audio cables, in various situations, including PA systems, during 50+ years.
In each of the situations, being assured that the impedance adaptations have been properly set, I can think of a hundred ways to significantly improve the sound reproduction. In order: Loudspeakers, room acoustics, power amplification, source quality, ground loops in links, immunity to EMI RFI radiation. I doubt to have a life time long enough to never lean on "the sound of the cables" and the reflexions of microwaves in these.
The rule is very simple: work first on what matters more.
There is no point in accumulating layers of varnish on the body of a damaged vehicle.
[edit]In your last graph, my speaker's impedance is out of range: 6 Ohms. And the ones of my amplifiers too. 0.0xxx
Last edited:
No big deal, I have done the exactly the same professionally but for perhaps a decade less. I well understand and know the route to good sound that you suggest, my suggestion is icing on the cake after other things are taken care of and still worth trying. You have speakers with 'flat' Z ( 6R or so) out to 40kHz, but what about higher than 40kHz where RF pickup and RF line reflections happen ?. Terminating lines is one aspect of EMI/RFI immunity, I am suggesting this as retrofit for DIYers at essentially zero cost. IME this no cost 'tweak' usefully lowers subjective dynamic noise floor, why not use it ?.
Speaker....
Two Wire Line....
Dan.
Last edited:
I can push back at him calling my definition of timbre "wrong" in such a bullish way when all he offered is "no definition at all" which doesn't contradict what I posted - his posts are mainly just for argument sake, AFAICS
Sorry, but I treat people in the way they treat me - you have a great deal of patience for dealing with those who call you names & try to verbally insult you - I don't have that patience.
I said flat to 40KHz, because it was there I stopped the measurements. After this, the impedance is supposed to stay near the same.
Btw: As I consider the output line of an amp to be an input for the feedback, both my speakers wires and my speakers are shielded. (Overkill, and don't change anything according to my measurements). With the input filter of my amp set at 500kHz, what can I fear ?
As my high frequency driver is a compression chamber, not even HF entering in the moving coil sneaking into the air gap ;-)
Input side, I have a low pass filter set at 400kHz 12dB/oct. It is a diamond. One 6dB at the input, one 6dB between its two stages that prevent the typical instability of this topology in the same time..
To close this strange discussion, OMHO, Dadod has perfectly resumed the situation in a single sentence:
Last edited:
There are three ways to understand a signal in a cable:
1. Resolve the signal into a possibly infinite set of impulses, apply the impulses to the cable, follow the reflections, sum up the result.
2. Resolve the signal into a possibly infinite set of sine waves, apply the sine waves to the cable, follow the impedances, sum up the result.
3. Treat the cable as a lumped LCR network.
Done correctly, the first two methods will always give the same results. If the cable is short enough (or, equivalently, the signal frequency is low enough) then the third method will give a sufficiently good approximation for all practical purposes. JN seems to be saying that under circumstances where most people believe that '3' should be fine (e.g. domestic audio cables) we should actually use '1' because we can detect tiny time differences. PMA is saying that '3' is good enough if the signal is slow enough.
1. Resolve the signal into a possibly infinite set of impulses, apply the impulses to the cable, follow the reflections, sum up the result.
2. Resolve the signal into a possibly infinite set of sine waves, apply the sine waves to the cable, follow the impedances, sum up the result.
3. Treat the cable as a lumped LCR network.
Done correctly, the first two methods will always give the same results. If the cable is short enough (or, equivalently, the signal frequency is low enough) then the third method will give a sufficiently good approximation for all practical purposes. JN seems to be saying that under circumstances where most people believe that '3' should be fine (e.g. domestic audio cables) we should actually use '1' because we can detect tiny time differences. PMA is saying that '3' is good enough if the signal is slow enough.
One question I have about JN's theory is that our perception uses ITD for localization mostly at low frequencies i.e below 1KHz progressively using ILD for localization as frequency increases. AFAIk, JN's premise is that cable/speaker impedance mismatch will affect the power delivery (more correctly the settling time to full power) to the speaker differently for different frequencies & hence effect ITD across the frequencies? But if we limit the frequencies to those that perceptually use ITD i.e < 1KHz, does this affect his conclusions in any way?
Last edited:
I can only generate decent sine test signals up to 3 MHz which I’d say, given a shove and a kick, is the upper limit of ‘audio’.
I have never, ever seen any gunk on any interconnection at these frequencies looking with a 200 MHz BW scope and 50 ohm source Z and 2~10k load Z.
If I look at a digital signal (square wave, 10’s to 100’s of kHz) and the interconnect is not terminated then it’s bad - overshoot etc. But as soon as you terminate it, it’s beautiful again.
There are more important things to worry about in audio than speaker cables. As long as the cable pair is bundled together and of sufficient cross sectional area to ensure the total cable resistance is of the same order as the amplifier output impedance at LF we are good to go.
Let’s stop thrashing a non existent issue. This is audio, not a particle accelerator.
Pavel,
I already know of your work on this. However, you are stuck on some of the teachings we both received way back when.
As to documentation:
A reflection bridge will show everything I speak of. Just make one capable of audio frequencies.
Cyril Bateman designed one, built it, and provided photo's of actual cables driving speaker loads. He provided a complete analysis, and the bridge matched his results.
I have the paper, it has the design details, I was quite surprised.
I do not recall who sent me it, nor if it was published.
Jn
I already know of your work on this. However, you are stuck on some of the teachings we both received way back when.
As to documentation:
A reflection bridge will show everything I speak of. Just make one capable of audio frequencies.
Cyril Bateman designed one, built it, and provided photo's of actual cables driving speaker loads. He provided a complete analysis, and the bridge matched his results.
I have the paper, it has the design details, I was quite surprised.
I do not recall who sent me it, nor if it was published.
Jn
There are two issues here:
1. when a short cable is driven by a band-limited signal (e.g. music) does theory or experiment show a difference between the 'impulse' theory and the 'lumped LCR' theory?
2. does any difference matter?
The conventional understanding is No to 1, and somewhere between No and Who Cares to 2. JN is saying Yes to 1 and Maybe to 2.
1. when a short cable is driven by a band-limited signal (e.g. music) does theory or experiment show a difference between the 'impulse' theory and the 'lumped LCR' theory?
2. does any difference matter?
The conventional understanding is No to 1, and somewhere between No and Who Cares to 2. JN is saying Yes to 1 and Maybe to 2.
1) when input shorted ideal cable is driven by a BW limited step signal, you get sawtooth (triangular reflections) superimposed on exponential, e.g.. Depends.
Last edited:
Good point. For cables to affect localization, there has to be a difference. The only way we could discern any image shift is if the speaker dynamically changes it's impedance as a result of the power being fed to it.
I deal with this in my motion control work, which involves all types of motors, negative feedback, phase margin, all that hokey EE stuff. A stepper for example, even unpowered, has a magnetic field that fights rotation. The rotor inertia and that magnetic "spring" will have a resonant frequency. Powered, the spring constant changes and so does it's resonance frequency. What the motion control people miss is that if I try to rotate the shaft, the motor will fight me, changing the spring force. That changes the resonance of the rotor. In motion control, this means that during acceleration, the phase margin of the system changes. At rest, the motor has a set of time constants, but during acceleration, the constants change, raising the low pass frequency breakpoint. This changes the phase margin of the total PID loop. (Btw, Bode plots are used, and my motion systems are 6th order low pass so at some frequency, phase margin becomes zero. )
This allow me to dynamically increase the loop gain during acceleration past gains which will cause oscillation at rest. Also, for the devices I am fixing, it means that when the magnetic forces are high, I can use higher gains, but when I open the gap, I have to lower the gain.
Sorry for the verbiage. However, a speaker is a magnetic beast so the crux is important. As I said before, the speaker impedance is affected by cone position, cone velocity, and cone acceleration, as all these things affect the transfer function. And because it is stereo, the signals are different. Mono drive will show nothing w/r to image shift.
That said, if a driver is being pushed at high power at a bass frequency, the changing magnetic forces are altering the transfer function of the higher frequencies, meaning a change in the higher frequency impedance caused by the power drive at the lower frequency.
Read my response to Pavel.
We are taught that not because it's accurate, but because it's a rule of thumb. Using a smith chart in this regime is folly. But as you see, Pavel is "proving" it doesn't happen by showing both ends of the cable are the same. He is confusing the time it takes for the cable to charge with prop delay.
Build the bridge....pun intended... I will look for the drawings later today.So you cannot imagine a 6000 m long wave reflecting in a 6 m long cable?
Close. T-line theory informs me on how the mismatched cable charges, and it is absolutely equivalent to a distributed LCR model.
Jn
Last edited:
However, I should be more specific here. If you use a single RC to limit the BW of ideal step, you get sawtooth superimposed. Because there is a sudden break from zero line to exponential curve. If it is a double RC, which eliminates the first break from zero line to exponential, then there are no superimposed reflections. Audio is strictly band limited, so it makes no reflections. Period.
As for simulations, they are pretty close to the real world.
Last edited:
No, JN is saying they are fully equivalent (no to 1).
JN is also saying that if the change reaches into the ITD thresholds, it cannot be discounted, so possibly maybe to 2.
As Scott proved a long time ago, they are equivalent models. Anyone who glibly discount what I have been saying need only use the conventional LCR distributed model to arrive at the EXACT same results as the t-line.
So even if one doesn't agree with me, conventional analysis produces the exact same results.
Jn
Sure.
How did you do this shielding ?....coaxial/triaxial or shielded pair cable ?.....driver baskets grounded ?.....compensation networks across drivers or box input terminals ? etc , I'm interested to know what you have done and what you found please.
Dan.
- Status
- Not open for further replies.