Then in fairness, if people have decided that something can have an audible impact, surely it would be equally highly unlikely they will do open minded experiments too......? SY nailed this on the very early pages of this thread IMO.
The logical conclusion is that, to be meaningful, audibility experiments need to be conducted well, and with expectation bias in mind. Protaganists seem to know this and acknowledge it whilst getting on with embracing it IMO.
Well, if they are genuinely open minded they would approach it saying it may, or may not have an audible effect. If an audible effect occurs then determine whether the sound is improved or degraded by the change, or merely is "different" - there are ways of proceeding so that clear cut conclusions can be reached ... one is not comparing levels of "niceness", it's a process of troubleshooting - any other approach is pointless ...
Nope. Let's see what Terman actually has to say about it :
"Return Circuit of Two Parallel Round Wires.......
The self-inductance of a pair of parallel round wires carrying current in opposite direction, each having a length l, a diameter d, and spaced a distance D from one another (dimensions in inches), is
L = 0.01016 l ( 2.303 log10 2D/d - D/l - μrδ )
This neglects the inductance of the connecting link between the ends of the wires. As before δ is a skin effect factor that may be determined from Fig 16."
Then, for a parallel pair speaker cable of conductor radius 8mm, separated by 16.5mm, inductance can be calculated to be 116nH/ft or 383nH/m. μrδ works out at c 0.25 for audioband speaker cables independent of realistic conductor diameters.
So you are mistaken, JN. The cable I used in my simulation can be realistic and achievable, even Terman predicts within reason. Plastered all over the internet are alternate derivations and formulae which obtain smaller inductance still, BTW.
You've probably made an arithmetic mistake, JN, or tripped over unit conversion - and in measurement perhaps failed to ensure the two conductors are a return circuit where mutual inductance applies? This seems yet another hole beneath the waterline for your theory though, seems you're working with overstated inductance/length unfortunately.............
Last edited:
In fairness to JN, I should say the Terman text publishes an equation which differs from widely published and accepted expressions for inductance of two parallel wires in a return line - Terman producing results which are typically larger than common predictions.
The more common expression for inductance per length of a parallel 2 wire inc return current is of the form
(μ0/pi ) * ln [ (x-r)/r) ]
where μ0 is permeability of whatever seperates the wires, x is seperation and a is radius of round conductors. Whereas there are numerous derivations of equations in this form which appear correct IMO, I'm not aware of a derivation for the Terman equation.
Using this it's easy to obtain inductance c 100nH/ft for parallel wire lines. Whereas with the Terman prediction in fairness it's not.
Not that it matters in the scheme of things, just thought to point it out.
The more common expression for inductance per length of a parallel 2 wire inc return current is of the form
(μ0/pi ) * ln [ (x-r)/r) ]
where μ0 is permeability of whatever seperates the wires, x is seperation and a is radius of round conductors. Whereas there are numerous derivations of equations in this form which appear correct IMO, I'm not aware of a derivation for the Terman equation.
Using this it's easy to obtain inductance c 100nH/ft for parallel wire lines. Whereas with the Terman prediction in fairness it's not.
Not that it matters in the scheme of things, just thought to point it out.
We pretty much have kown that for decades. In the context of your post.
Yawn. 😉
Jan
Yup.. I've always loved that delta factor.
I early on spotted the difference between the two methods, so tested them for accuracy, with actual wires.
And the result is, Terman was THE most accurate. I was able to duplicate his equation results to within 4%. Other models FAIL for various reasons.
At no time will an actual set of round conductors dip to 100 nH per foot. I tried and tried to no avail, even using 1 mil kapton as a spacer, as well as using enamel magnet wire.
So you used 100, where did you get your data? Also, your prop velocity for the last cable was 93% lightspeed, so where did you get the data?
A twisted parallel pair cannot reach those velocities.
Can be? Have you measured a parallel pair set that had those values?
So if it's on the internet, it must be right?? Really?
I've played with quite a few of them, they fall apart like a cheap suit.
Ah, nice try. It's all measured.
Sometimes, it's simply a remeasure of Belden's or Times Microwave, sometimes its wire I actually made. But it's all measured.
Yes, he includes the magnetic field within the conductor, it doesn't have the 1/r falloff. It rises linearly from center. 30 nH per foot for a pair of wires.
He used actual measurements. appear correct?While multiple sources may appear to be correct, it's always worthwhile to actually make some wires and test them. Which I've done, and it confirms Terman.
Terman used actual measurements. I've confirmed them.
Have you built and tested your own wires, or have you tried measuring existing product from any manufacturer?
Accurate models are very important. Terman appears to be emperically derived, from actual measurements.
jn
Last edited:
Actually, you are incorrect.
Who is speaking of resonance? And, who said the impedance at the source is different from the impedance at the load.
As an add, what do you believe Terman considered audio? 300 hz to 3 Khz?
This century, what we think of as audio is significantly different..20 to 20K, amplitude and phase, and recently, consideration of Interaural time delays, to wit, the control of the delays to the tune of single digit microseconds. Essential if one wishes to establish a good and stable soundstage, totally useless if one is only concerned with listening to background out on the patio.
Who knew?
But using a low pass model to show that high speed (5 uSec) level time delays can't happen?? Circular argument. Self prophesy.
jn
That would be Terman then. Meaning the process of reflection.
That would be you then JN, implicit by insisting the cable acts as a meaningful transmission line........jneutron said:
Well no accurate inductance doesn't matter in this context, not within the tolerance of differences at issue. Latency arising is so small that it doesn't matter - this is self evident otherwise cable length would matter per se, but it doesn't.jneutron said:
I doubt Terman's stuff is empirical, but never seen a derivation so can't comment on conditions etc. However, if measurement based then conditions obviously matter as to applicability of any measurement/prediction. Since he produces no tables below 100kHz and has made pov clear as to irrelevnce at audio, who knows.
There are plenty of derivations of ln ((x-r)/r) equations, which are testable and verifiable and measurable at audio. And consistent with Davis and my own measurements, FWIW. No idea how or what you are measuring, JN. However But for reasons already set out there's no need for precision or splitting hairs here.
I'm confidebt he understood the difference between reflection and resonance.
You didn't read Davis, did you.
Self evident via a Low pass filter? Circular logic.
Nor I. However, his equation is dead nuts accurate with respect to actual measurement of wires.
I've verified the accuracy of his equation for parallel wires from contact using enameled magnet wire to 1.5 inches apart, from frequencies between 20 Hz and 500 Khz. 20 Hz being the area of 4% accuracy, the meter had difficulty distinguishing Ls from Rs due to the very small inductive reactance.
Measurements now? You've only talked about simulations and using the data of others..
I've not found any derivations which have the accuracy of Terman within the audio band.
Actual wires. Belden, Times Microwave, Andrews heliax, a few other vendors as well as wires I twisted, wires I braided, coax and triax I built.
It's not a case of splitting hairs.
I detailed exactly the characteristic impedance required to bring the latency down to clearly inaudible levels.. You cherry picked lumped values which are EXACTLY what I stated was needed to minimize the latency, then misconstrued the results.
Using a low pass model to prove high speed effects aren't there is still a circular argument.
And Davis disagrees with the use of the LR model on one page, and agrees on another. You spent the time with Terman, go back and read Davis.
jn
Frank, Dan, please guide me to any documented application outside of audio where wire directionality is a key element. AFAIK no scientific or medical measurement no matter how low level, sensitive, or in any frequency range has a documented benefit. We are not talking about which end to ground BTW.
Looks like you aren't aware of what bias controlled test is. It's your loss.
What is competent sound? Do you mean high fidelity?
If? You mean you still haven't read about the effects of audio cable lifters? 🙄 I'll post it here for you, it didn't make audible difference in double blind test. There, now you are better informed.
I've been quoting Davis, so you'd like to think so. Let's have another quotation for good measure:
"Expressions for transmission lines (such as characteristic impedance, impedance matching, reflections) do not fit audio applications, since the cable lengths involved are minute fractions of the shortest audio wavelength (about 16km at 20kHz in copper). This is discussed thoroughly in Greiner [1]-[3]
Therefore, cable and loudspeaker should be treated as lumped- circuit elements."
Davis also presents measurements of inductance/length using defined commercial instrumentation for 12 various cable types. They vary widely in value from c 150nH/m to c 1400nH/m, endorsing my choice of 333nH/m for simulation to illustrate negligible latency arising from 'nominal' cable types.
Greiner also presents measured and calculated similar inductance values to Davis for various cable types including 12 gauge zip cord achieving c 390nF/m with 0.4mm spacing.
However, as far as latency is concerned, even such wide variation in inductance, an order of magnitude, makes negligible difference to simulation of real loads. Perhaps there's 130pS latency from the cable with a nominally low inductance cable by my calculation. I can afford to be a long way off...... not that I seem to be.
To obtain the 5uS cable inductance latency you claim, JN, requires unrealistic cables, length and loads. Even 75m and 1R load might not get there. Furthermore it's not at all obvious that 5uS latency might even then be audible...........
Last edited:
Hi Scott.
I don't know of anything documented....I have not attempted to research the subject online, self experience only at this time.
For sure, in most applications wire/cable directionality would not be of concern.
In the turntable example I quoted, the two channels individually sounded same enough, but when listened to in stereo the difference was enough to side shift the stereo image slightly sideways, but not quite the same effect as panning which is of course a channel/channel amplitude alteration.
Flipping acoustic polarity of a system changes perceived distortion, image reality, and also perceived loudness (for sealed box loudspeakers)....as is to be expected.
The flipped one channel cable direction in the TT example caused a similar-ish effect to flipping acoustic polarity....one channel sounded subtlety louder than the other, and thereby skewed the image sideways, but the balance control did not cure the perceived imbalance.
The amp that I ran at the time had Stereo/Reverse/L mono/R mono/L+R mono switching.
I tried every diagnostic that I could think of including shorting the channels at the V15 III pins and shorting the channels at the TT output connections but still could not get correctly centered stereo image.
I dropped into a friend's hifi store one day and he (Chris) said to me....'check this out'.
He played a known music passage and we both agreed that all was good.
He then flipped the direction of one interconnect and......the image skewed sideways.
WTF....and then the penny clanged, this is the problem with my turntable !.
When I fitted the cable to my TT, I stripped both ends and soldered these to the TT output connections.
I then cut the looped cable and fitted the RCA's....I was aware at the time that directions were reversed but discounted this as of no consequence because I had been taught/bludgeoned at college that ''cables/wires do not have directional properties.....and don't you forget it''.
When I got home I went straight to my TT armed with soldering iron and corrected the cable direction issue.
This corrected the image centering problem that had dogged me for months and suddenly all was well.
This experience later inspired me to make purposely non-directional interconnects.
The result was truly 3D sound, with sounds coming from miles through the curtains, and from miles behind the couch when playing natural sound recordings.
Sufficiently realistic that friends looked for rear speakers behind and to the sides of the couch !.
Music is an asymmetric waveform and this may be the key.....typical static testing methods run as perfect as is possible symmetric waveforms and rectification/averaging and therefore may not reveal minor/micro wire/cable asymmetries.
Thinking outside the square, maybe current direction very subtly changes noise spectrum including vlf or 1/f noise spectrum....dunno yet.
Dan.
You keep asserting this. It is untrue. The single cell LCR model fully models everything that can happen to band-limited audio on a short cable, including a 5us time delay. The single cell model assumes low frequency, not long (or short) time delays.jneutron said:
If it happens to audio it is not a "high speed effect". If it only happens to fast pulses then it is not an audio effect.
Are you sure it's -mu*delta?
I have +mu*delta.
You mean, like these?
page 464:
Page 465
Hmm..internally inconsistent
Yup. Now take a close look at what the 12 cables actually are. First, the single pairs (except 8)
1. #3awg at 1100 nH per meter.
2. #7awg at 1200 nH per meter
3. #5awg at 1400 nH per meter
7. #12awg zip at 800 nH per meter.
8. CVT (lord only knows) at 750 nH per meter
12. #18awg zip at 1100 nH per meter.
Note that the single pairs, #1, 2, 3, 7, and 12 all exceed 800 nH per meter.#8 is some weird combo.
4. #10awg, 6 conductors spiralled over a plastic core. 300 nH per meter
5. 19 awg, 16 wires woven into a flat cable. 230 nH per meter
6. 26 awg, 32 twisted pairs ~150 nH per meter
9. #19 by 8 wires, 500 nH per meter
10. #23awg, 8 wires braided ~270 nH per meter
11. #28 awg, 36 pairs ~.180 nH per meter
The only normal cables there are the #12 zip at 800 nH per meter and the #18 zip at 1100 nH per meter.
So again, what are you modelling? Pick and choose to suit a preconceived notion?
As I recall, you previously said 130 nSec.
And also, as I recall, you showed 1.1 uSec with a ambiguously low inductance cable model.
Actually , if you use 4 meters of a real #12 awg zip cord as measured by Davis (800 nH per meter), what are you going to get?? It's a rhetorical question, as you refuse to use real numbers in a sim..
Finally, you 've said something that is accurate.
Oh, btw..a little tidbit from Davis, page 461 (first page)
Hmm..what about 4 ohm speakers?
The takeaway? If you want to discount the latency, choose a set of parameters which I already stated reduce the latency to a low value.
jn
Last edited:
Ooops, my typo in the post should be an addition '+' not subtraction '-' ..., didn't do that in the calc though.
Just picking some kind of median, plausible, nominal inductance/length value for a decent cable. It doesn't much matter because the result tolerates an order of magnitude spread in cable inductance anyway, latency is so small. If it upsets you, just tell yourself I chose a somewhat short but plausible decent cable, after all it's total inductance that matters, not inductance/length per se.jneutron said:
Yes, 130nS into 8R nominal load. That is the same as 1.1uS into 1R, ie the load you suggested - 1R is not my idea and IMO not representative but included for consistency and hopefully permits comparison with your regime.jneutron said:
Into 8R, about 400nS I'd estimate.........and scaling to suit whatever result you might like if you choose a low enough load, say 3.5uS at 1R. It's just a LP LR filter after all.jneutron said:
As ever, curious to know how |Z| ever dips below voice coil R.........as DF96 posts it's unrepresentative to consider purely resistive loads in any event.
Last edited:
Ah, ok..You've run 4 orders of magnitude variation in you statements, so I was wondering.
It's only an LP filter if that's the model you choose. Remember, looking for single digit uSec variance or delay using a LP filter model is circular reasoning, not scientific at all. Nor accurate.
As for dipping below DCR, ask Otala and Huttunen, Davis referenced them with that statement of "as low as 1.2 ohms".
Now, most importantly..run a 200 element LCR model using the correct numbers. See how the distributed capacitance works... And, note how absolutely accurate the t-line is when compared with a real model. Don't worry, we've seen that already thanks to Scott.
Then look at that "3.5 uSec" number again.
jn
- Status
- Not open for further replies.