I think maybe your dancing around part of it.
In acoustics anyway, one can't have a transmission line until it is a good fraction of a wavelength long. Sound is "like" an electrical signal in that it is two things in quadrature, a velocity and Pressure maxima separated by 90 degrees along the wave. With an electrical signal, one has a Voltage and magnetic field also in quadrature. In each case, it would seem that the Transmission line behavior would begin to stop as less and less of a wavelength was confined by the line (or a duct, or horn etc in acoustics).
The wave number reference or K also applies in the associated math for acoustics. For example in the radiation resistance curve.
In addition to the K issue, i would wonder how this looks when the source impedance isn't zero at all frequencies, and the bandwidth is limited to say an octave or two past what could possibly come from a signal source (excluding the requirement to reproduce a record tick precisely).
I am much more of a measurement person, a proper measurement usually trumps a computer model. Models are great but are only useful to the degree they match what you build, at least it's that way in the loudspeaker area.
Anyone have an oscilloscope with a differential input?
Try looking at the amplifier end of the cable subtracting the loudspeaker end of the cable. If your scope plug in has a summed out, feed that into a spectrum Analyzer and investigate what and how large it is, use music as the test signal and listen to the sum via headphones.
Best,
Tom
One camp thinks its profound and nobel worthy, the other a product of mistake and misapplication of undergrad physics. That image is basic stuff for any undergrad - its not that it isn't true, it's just not relevant here because the rise times involved are orders of magnitude faster than crops up in real audio programme material - and that changes everything. It's another misapplication of physical truisms IMO, which is about where the thread was about 200 posts ago !
Thank you - exactly !
I've done the scope bit, and found nothing to write home about except the expected small parasitic effects of cable resistance and inductance one might expect. No 5uS settle time, that's for sure ! Would be interesting to take up your suggestion to look at the spectrum and listen.Tom Danley said:
Again, you are incorrect. And again, I will state why.
The difficulty is in the measurement apparatus. Trying to measure a 5 or 10 uSec delay in a 500 hz give or take signal caused by swapping cables without disrupting the measurement is a real challenge within a low impedance circuit.
That is why I asked YOU what kind of resistor you were using to monitor current. Current viewing resistors will have as a base, an internal inductance, as well as a loop intercept area. Some, like flat wide film resistors, will even have the current path dependent on the slew rate of the change of current.
The internal inductance of course, causes the L dI/dt error.
The loop intercept area, or external inductance, will compromise the voltage pickup loop due the changing external magnetic field of the resistor.
Film resistors at speed, will shift the current centroid depending on the material parameters, this will cause a slew rate dependent inductance as well as change the external loop pickup of the measurement circuit.
Again, you've no understanding of the problem. I do recommend you find a good EE prof, I'd be happy to detail what is being discussed.
We agree. The problem is, what I speak of is exactly modelled by proper LCR models as well.
Again, wrong.
Who said I used an amp??
You need to look up a TEK 109. But that uses a source impedance of 50 ohms, so you have to build.
You can easily build a source using your own mercury wetted relay. But they are getting more difficult to find, especially reed style that you can run copper braid over. The other issue is the impedance of the storage media, lytics are useless at speed. MLCC caps are great for this, but keep the physical size down, and be very careful of cracks at the edges, as handling is the dominant failure cause.
It's also possible to use a simple fet switch or transistor, but you need to worry about how it handles overshoot with load>>line.
Of course, you can easily build a hot amp amp stage with no feedback, running a cable with load>line will break your margin all to heck.
The main concern is how much delay can be caused by the cable characteristics. The figure I posted a while ago shows how the load impedance can alter how fast the system can respond to the signal. If we keep the load Z tight, the settling time variation is pretty much moot.
The alternative is to put the cable Z in the middle of the load z range, and minimize the maximum delay.
For single channel operation, all of this is both academic and useless. Only for stereo channel operation where we rely on localization cues in the 2 to 5 uSEc range can this be an issue, and again, only if one channel's delay is different from the other in midrange frequencies. Drivers with high vc velocities will change the apparent inductance of the coil to higher frequencies because the energized coil is moving with respect to the gap iron.
The basic theory and practice is useful to me as I have long line (100 foot) runs driving coils which MUST be altered quickly as a feedback mechanism. As the physicists here dial in the machine, they will require faster and faster feedback, and at some point, I'll have to parallel up cables to lower the line's characteristic impedance. edit: I do not have the luxury of moving the electronics closer to the coils, so all I can do is reduce the settling time by dropping the line impedance.
I've been trying to get lucky to understand both the tline as well as the conversion to LRC, he will be stronger as a result... It would have been so easy to just convert the discussion to lumped element equivalents, but that loses certain aspects that are easily understood.
That'd be really nice. If it happens, I'd be happy to discuss it with them.
It'd be a good project for a masters level EE, I suspect not enough material for a thesis however. The t-line stuff I speak of is bog ScB stuff, and at that, only two to three weeks in.
jn
Last edited:
That will not work. As I said earlier, the difference between the current at each end will be exactly one transit time different, for reasonable cables that will be 20 to 50 nSec.
jn
Again, incorrect. edit: and no, it is not nobel worthy. It is the simple application of stuff you should have learned sophomore year in an EE program.
The delay is there independent of the risetime. A fast risetime is used to show the fastest response the system can have. As the risetime elongates, the system still has the delay, you just can't see it.
Again, you need to consult an EE prof, you do not understand the theory nor the practice.
jn
Last edited:
As I said, the cable end to end is different by one transit time.
You are using a ruler to measure color.
Please learn the topic, or at least find someone who you can ask.
jn
Cool, that's what I'll do then. I'll have to do it quick before the Budweiser effect leads us to playing with guns instead of talking about electronics, but I'll see what I can do.
🙂
That's one heck of a cookout you got there..😱
On the upside, I bet the neighbors don't complain about the music..
edit: OH, and don't mention cable lifters.. He'll instantly think bad things about you..
jn
Last edited:
That is not a typo but a technical error. Given the subject of the article perhaps it could be regarded as a serious technical error.jneutron said:
I think we can take the "bog standard two weeks into the standard E/M course" with a pinch of salt. I did 4 years of EM courses, starting with implicit relativity, then explicit relativity, then quantising (as a postgrad). I can't remember exactly where transmission lines first appeared, but it was certainly later than 2 weeks. Don't blow smoke into the eyes of those who have never studied EM at the required level and so may be unable to properly evaluate this discussion.
That I can agree with. A common sign of trip-ups is failing to apply the appropriate model: low frequency in this case.
Yes. True but irrelevant. A typical speaker cable length gets nowhere near a wavelength, unlike a continental-size AC grid.
As I understand him, jn believes that his TL model (and the equivalent 200 element lumped model) shows audio frequency behaviour which is sufficiently different from a single element lumped model that it might be audible.DrDyna said:
What else could it be ? Your argument just sank. See DF96 post about complex impedance for some insight into what really might be going on - but if you're really using a 50R pro sig gen, audio bandwidth, a short cable and a resistive load none of what you claim as to latency makes sense.jneutron said:
Ask him. I prefer giving him the benefit of the doubt, considering the difference between a capitol C and a capitol G. I proofed my article at least 7 or 8 times, and still found my typo's at 8.
It was taught in my EM course in the third lecture. Granted, courses may differ, but what is the difference between two weeks and 5 or 6. Either it was learned, or it wasn't.
""Those who have never studies EM at the required level"" should not be trying to discuss it, no? Of course, not understanding never stopped anyone from arguing a point on the internet.
The trip-up is that... anyone who needs to, can just substitute the LCR elements, and they will still arrive at the exact same results I've shown with the t-line model.
The real problem here is "one who never studied a topic" yet knee jerks in response..
We get it, you don't like t-line. Fine, substitute the LCR elements. You will arrive at the exact same conclusion that the t-line gives.
You understand incorrectly. Perhaps you can try again, with more effort?
jn
jn
Last edited:
Please slow down and actually read the posts.
Then start that post over..
I can wait.
jn
Did you actually attend it ? 😉
DF96 is right, the trip up is not applying appropriate low frequency, small wavenumber models.
This from one who does not read posts?
I talk about system settling time, and you try to find a delay from one end of a cable to the other???
You still do not understand. Please find a prof of E/M you can ask questions of.
Until you do, you will continue to waffle here.
I note with interest, you have not answered the simple questions on that jpeg.
Well, what are the answers? Show us some understanding on your part.
jn
A sufficiently large number of LCR elements must give the same result as a t-line. That is not disputed. The point at issue is that for sufficiently low frequencies the sufficiently large number is 1, not 200.jneutron said:
Nobody has said otherwise. Of course, it also has not been raised as a "point at issue".
I do note that nobody within the context of this discussion has even taken a specific L and C, and varied R to determine the settling time vs R.
And then change the L and C and repeat the R scan to see where the settling time minina is. Noting of course, that to match a cable, the L and C must meet LC = 1034 EDC, or about 4000 give or take.
Scott, IIRC, used 10 elements.
Cyril used 201 only because he was modelling out to 10 Mhz I believe. He had a distinct rf slant on some aspects of what he was doing, as he was troubleshoting amp/cable interactions that were causing hf burst oscillations.
jn
Last edited:
I think I have it somewhere, I'll look for it.
In the mean time, to keep you guys quiet, see attached 🙂
Jan
- Status
- Not open for further replies.