The TL model and the LCR model are identical, and produce identical results.
It's been shown here, you just prefer to not believe it.
Why does your "sim" not show the proper waveform with the 1 nSec rise? You've got a bad sim there.
jn
Thank you. Seems like a lifetime ago, but that's what I've been posting all along.............there's no TL effect in play in the audioband, lumped parameters are all one needs to consider.
Being identical is mutually exclusive with "no TL effect".
That wasn't even a good try.
So I repeat..The T line model demonstrates the exact same delay that an LCR model does.
What in heavens name did you do in your simulation? The fast waveform is all ragged.
Are you now trying to say that a 1 nanosecond risetime makes it through an inductor instantaneously as your bad simulation shows?
Try using real values for inductance and capacitance.
jn
No, the simulation is for a TL, and is correct. For 1nS rise time, TL and LCR models obviously differ in prediction and only the TL properly predicts reality. But for audioband risetimes, which are much slower, there is no difference between TL and LCR model predictions - both properly predict reality. This is simply because the cable does not behave as a TL with audioband risetime signals. Is this an audition for Groundhog Day II ?
Last edited:
Stop with the sillyness.
So then, you've designed a T-line with zero time from one end to another?
Give King Gustav my regards.
Or, you could sim correctly.
OH, I SEEE WHAT YOU DID!!!
Your simulation has the wrong reflection coefficient. You've modeled with the load z greater than the line z. No wonder.
Fix the reflection coefficient.
jn
Last edited:
No, that is obvious and the TL uses Scott's parameters which are reasonable. The simulation is correct, no surprises.
No, and the simulation is correct. Besides that's meaningless for reasons already set out. But what does it matter, we agree the TL model and the LCR model produce identical results in the audioband...............
All good sport and no hard feelings jneutron, and I'm done with this thread.
Last edited:
You have a good memory Tom....I had forgotten that scene, thanks for the refresh 😎.
Dinki Di doesn't smell too bad.....if you had to eat you probably could....Paul Hogan - PAL Dog Food
I'n not sure what Mad Max would make of this...
Dan.
The simulation is totally bogus.
The reflection when a signal hits a lower impedance load is negative..even wiki has that right.
The source is low z, the load is low z, the line is high. A 10 volt step will hit the load, reflect most back as a negative return step. When it hits the source, it reflects inverted again, sending the positive signal towards the load.
As Scott's earlier sim showed, the end result is that the load will continue to rise a bit at a time, each incremental rise a result of a smaller and smaller step hitting the load, getting smaller as the reflection coefficient will be close to negative 1, but not quite there.
Had you used the correct reflection coefficient, you would have simmed exactly that. Instead, you somehow ended up with a bad reflection coefficient, so the load signal has the appearance of a horrible overshoot which eventually decays to the final value. If you look very closely, that decaying envelope has very close to the same settling time that the correct reflection coefficient would have. If you also look closely, the timeframe for that settling is on the order of 3 to 5 uSec. What you called a geological time frame.
If the correct coefficient were used, the long slow 25 uSec ramp would also show a departure from ideal, that timing departure will be exactly the time difference as shown on the high speed settling. As I said, a tenet of control system theory is that the fastest a system can get to final value is as shown via the step response. When the system is driven at much slower speeds, this becomes less evident visually due to the timebases involved, but it is still there. In motion control theory, this is referred to as follower error. When I tune PID algorithms for my motion control stuff, I use this follower error both to determine performance, and to help tune the system better.
There has been no animosity here.
But I do wish you'd post the correct TL result after you fix the reflection coefficient.
jn
This also occurred to me once, although at the time I was "measuring" mp3 artifacts, subtracting the compressed file waveform from the uncompressed. Same general concept, though. Turns out the trouble is, even if we can see something on our Difference-O-Meter (and we always will, if we zoom in enough), we then still have to go back down the rabbit-hole to determine whether or not what we're looking at is indeed audible. Just because we can see it, doesn't mean we can hear it (to our perennial misfortune)...
-- Jim
Dunno about you guys, but I'm going to turn this damn computer off and go listen to some music.
Wait - the music is in the computer. Never mind...
-- Jim
That's the crux of it, isn't it? It is interesting (and reassuring) that the TL model and the lumped model (with enough lumps) converge on the same result. And that result is that a conductor like a speaker cable can exhibit a time delay which varies with the load impedance. That's interesting too, but in and of itself not very important, except that we also know that loudspeakers have an impedance that varies with frequency. Further, that Z vs. F might change dynamically, so the actual Z at a given F might vary depending on how much energy is present at other F's in the signal -- is this something that has been observed and measured? How big is this effect?
So given all that, and given 5usec of ITD as a threshold of audibility for localization, does any of this matter? What would be the worst-case audible effect of this in loudspeaker sound? So far we have established that a loudspeaker with truly pathological impedance variations (I recall jn mentioning something about two orders of magnitude, is that right?) might cause a speaker cable to produce time-delay variations that are withing the limits of human-audible ITD -- but where does the ITD come from? For that to be meaningful we have to be talking about two (or more) loudspeakers, where the impedance exhibited by the two speakers was so different at the same frequency that one speaker was 5us behind the other. Even that would not be a problem if it was consistent; in order to have a "wandering" stereo image the impedance of the speakers would have to be changing wrt to each other over time, and changing dramatically (like by a couple of orders of magnitude).
And we're talking about loudspeakers, not headphones, so the sound from both speakers reaches both ears. In that circumstance I doubt the 5us threshold would hold -- I have no idea what the effect would be, but I suspect "none" is pretty close.
It's nice that jn has figured out that he can avoid this by the simple expedient of using a speaker cable with a lower characteristic impedance, and that he can do that by paralleling multiple pairs. If that helps him sleep better, then the world is a better place (maybe safer too given the stuff he works on). However, it's not something that I'm going to be concerned about. There are a lot of things I would like to improve in my stereo system before I start rewiring my active speakers (and how do I get the cable lifters inside the speakers?).
yes you have to compare to other much bigger issues that are much more variable, poorly controlled
multidriver loudspeakers radiating from different spatial location by 100s of microseconds sound propagation deltas?
individual driver's radiating surfaces having profiles >5 microseconds of sound propagation distances of 1.5 mm, 1/16"
most drivers breaking up, becoming multimode radiators over their upper octaves
bass radiators often move more than 1.5mm while covering up to a decade of frequencies
or the fact that varying Z load itself has an associated phase slope, group delay variations from Bode's Phase Integral
multidriver loudspeakers radiating from different spatial location by 100s of microseconds sound propagation deltas?
individual driver's radiating surfaces having profiles >5 microseconds of sound propagation distances of 1.5 mm, 1/16"
most drivers breaking up, becoming multimode radiators over their upper octaves
bass radiators often move more than 1.5mm while covering up to a decade of frequencies
or the fact that varying Z load itself has an associated phase slope, group delay variations from Bode's Phase Integral
Q: If it takes about 200 cells to model a T-line at 10MHz, about how many cells are needed to model that same line at 20kHz? Give both naive and better answers.
A1 (naive): Assuming constant line speed, then 10MHz/20kHz = 500. Number of cells required = 200/500. Approximately zero, but could use one cell if especially interested in accuracy or to satisfy a picky sceptic.
A2 (better): Line speed will be constant at most frequencies, but will slow at lower frequencies (becoming proportional to sqrt(freq)). Assume this change happens around 100kHz (although in reality it may be at a lower frequency so moving us closer to the naive answer). 10MHz/100kHz = 100, so factor of 100 needed. 100kHz/20kHz = 5 so factor of sqrt(5)=2.236 needed. Total factor is 223.6. 200/223.6 is about 1. One cell needed i.e. an RC low pass filter, with a bit of L mixed in with the R.
The 'settling time' seen in the T-line model is just a piecewise approximation to the exponential response of a first-order filter. Or, more strictly, the exponential response of the lumped version is an excellent approximation to the actual piecewise response. With a bandwidth-limited signal only the exponential response will be seen.
A1 (naive): Assuming constant line speed, then 10MHz/20kHz = 500. Number of cells required = 200/500. Approximately zero, but could use one cell if especially interested in accuracy or to satisfy a picky sceptic.
A2 (better): Line speed will be constant at most frequencies, but will slow at lower frequencies (becoming proportional to sqrt(freq)). Assume this change happens around 100kHz (although in reality it may be at a lower frequency so moving us closer to the naive answer). 10MHz/100kHz = 100, so factor of 100 needed. 100kHz/20kHz = 5 so factor of sqrt(5)=2.236 needed. Total factor is 223.6. 200/223.6 is about 1. One cell needed i.e. an RC low pass filter, with a bit of L mixed in with the R.
The 'settling time' seen in the T-line model is just a piecewise approximation to the exponential response of a first-order filter. Or, more strictly, the exponential response of the lumped version is an excellent approximation to the actual piecewise response. With a bandwidth-limited signal only the exponential response will be seen.
Precisely. The only consequence is that we should use identical speakers for both channels, and feed them with identical cables of the same length - but don't we already do that anyway? The lumped model tells us to do that!nezbleu said:
You have summarized in an excellent fashion, thank you.
Let's take a mono signal on two channels with loudspeakers.
If one channel is delayed by even milliseconds , we most likely will not hear anything, as we simply adjust without knowing, our head position. Nobody puts their head in a vice, and just twitching your glutes in the chair can change that..
If the midrange content only on one side is delayed, then the image of the midrange content will be shifted with respect to the balance of the content. The subject has a fixed source location to use as a judge for the shifted content.
A mono signal however, has identical content on each channel, so any position or velocity dependent effects on Ls or Rs will by design be identical, assuming the speakers recieve the exact same content.
A beautiful test for cable induced delays involves a mono signal, two identical speakers, and 8 zip cords, say 18 awg.
Parallel 4 twisted pairs on each speaker such that the speaker sees reasonable cable impedance. Verify that the generated soundstage image is centered for all frequency content. No abberations to either side of the central image.
Next, take all four zips from one channel, split the zips, and twist all the positives together, and all the negatives together. Space the send and returns 8 inches apart for the length of the run to the amp. This maintains the identical resistance, but changes the cable impedance (as well as the LCR values) tremendously.
Now revisit the mono signal. Look for side shifting of any image content. Most of the image will still be in the exact center, you use that as the reference location (eliminating head positioning as a variable).
If I personally heard no change, so be it. If I personally heard a change, then I'd have to consider whether or not it's worth further research, as it could easily be my expectation bias. If I chose that path, I'd have to setup controlled testing.
For a stereo signal, it's a bit more complex. Here, the woofers on each channel may not track identically from content. Woofer spacial position will change path length for other frequency content, doppler rears it's ugly head, and variation in speaker Z (both dynamic inductance and eddy dissipation from flux dragging) makes the problem quite complex.
Unfortunately, worrying about cable to speaker interactions is very low on the list of things that keep me up at night.
The stuff I work on has no immediate safety implications. 10, 20 years from now, maybe.
I too, am not concerned w/r to my own stereo. I use bog standard plenum 14 awg from home depot in the walls, the 24 awg zip that came with the 5.1 system I bought a while ago, and I use 100 foot long landscape wires for the outdoor stuff. Wife, family, and company are far far higher on my list. Speakers are just background for me.
This analysis stuff is more for those who worry about such things with speakers..and for esoteric machines I play with during the day.
jn
Now you're talking...totally correct.
Unfortunately, that is an erroneous assumption.
While humans are incapable of hearing much over 20 Khz, they can hear ITD down between 2 to 5 uSec.
Anybody who has ever bought a scope probe back in the 80's knows the tradeoff between the equipment rise time and it's bandwidth, risetime being an exponentially decaying function. To retain ITD accuracy at the 2 uSec level implies a system interchannel bandwidth accuracy out to about about 500 Khz.
You are content to bludgeon the interchannel accuracy to about 20 Khz give or take.
Humans are far more sensitive to interchannel effects.
It is very important to consider what we are sensitive to, not pretend it does not exist.
jn
Channel gain/balance controls do not always correct replay problems...interchannel delay should be standard.
Dan.
Dan.
As I haven't changed my view I must have been correct all along! I am glad you have accepted that the effect, such as it is, can be explained at audio frequencies by a simple lumped model of the cable.jneutron said:
But we have two nearly identical channels. It would be quite unlikely for something which produces an in-channel effect around 5us to be so different between the channels that it can produce 2-5us channel difference. A typical listening room (or a typical listener) is likely to be far more asymmetric than this.
So the effect arise from the lumped (single cell) model of the cable (mild filtering) and will be inaudible under all except the most unusual conditions.
Now if we put one channel cable on blocks and left the other on the floor, would we notice the difference?
Because he craves attention. He has done this on other forum too and when others question his motive, he responds with disparaging remarks about how technically challenged they are or how good he is and how dare they question him... etc.
Your view has been that TL cannot be used at audio frequencies. Through the year or so we've had this discussion, I've patiently awaited your "proclamation" that they are identical and that you've said that all along.
The discussion has never been about the electronics producing such a delay.
So that sentence can be thrown out.
Which luckily, is not what the discussion has been about. Again, out goes that sentence.
No..a single cell is terribly asymmetric in terms of direction. What comes first, the inductor or the capacitor? By extending the number of cells, the cable model becomes more symmetrical.
I note with interest, you did not address what I'm actually discussing.
Personally, I find lifters and blocks of little concern or value, I do not think they work.
So, instead of actually discussion of a technical aspect, you choose to attack another? I also like how you inject "motive" into your blather, in a blatent attempt to try to make it a personal one instead of a technical.
Now that's useful. Why do you bother posting?
Lucky posted a totally incorrect simulation as "proof" of his point, which I explained had an error. My suspicion is that he accidentally put the capacitance in as nanofarads per foot instead of picofarads per foot. That would lower the TL impedance by a factor of 31.6, and if he used 8 ohms as the load, would explain the overshoot and oscillation he provided.
That's why I mentioned that earlier as a common error, we all make it one time or another.
jn
Last edited:
I have used 10pr and 20pr telephone IDF cables wired as paralleled twisted pairs for eons/decades.
Lower loop inductance, higher capacitance than standard fig 8, sure.
Better clarity and depth/positional imaging on every amp I have tried, Naim amplifiers excepted.
Dan.
Last edited:
- Status
- Not open for further replies.