What I did NOT endorse was the wrong reflection coefficient. I see you've fixed that.
Hmm, your graph seems to show about 2 and change..<==edit..the model is showing one and change at 50%. What are the model parameters you used? length, L, C?
You mean, other than the wrong reflection coefficient. Did you put the wrong capacitance in the model?
I guess we can ignore the time delay you've presented, which is exactly what I've been saying.
The problem is, the speaker and output cable are ignoring you with respect to band limited, they conspire to alter delays based on impedance.
For LCR, of course. Unfortunately, the system also worries about delay, and humans can hear delay when it reaches some level.
Try just using L and R..You'll get garbage in comparison to reality.
jn
Last edited:
"albeit" for 1 uH cables??? What cable are you simulating? What's the capacitance?
I'm sure it amuses everybody...first, I don't recall anybody mentioning the DC resistance of the voice coil. And, second, how does one consider negative inductance? Remember the motional energy makes inductance measurements of a speaker really interesting at the lower frequencies. In free air, it gets crazy, locked VC shows the coil and iron with no additional energy storage.
When one chooses a 1uH cable and neglects capacitance, of course everything seems trivial.
How about modelling a real cable with the characteristic impedance of 300 ohms and the correct lumped parameters, and running the load from 1 to 60 ohms.
You know, duplicate what I posted perhaps a year ago.
jn
ps..OH, and btw. Thanks for your continued participation, it is much appreciated..
pps. Please put the parametric values, lengths, impedances on the plots. thanks
I've tried to make some ghetto low inductance cables, not for speakers, it is for a totally different stuff. For lowest inductance of course the best thing is two flat conductors but I didn't want too much capacitance either. Sort of a compromise.
So far the winners are :
- 10 conductor ribbon cable, with every other wire in parallel, measures around 2nH/cm, although theory says about 1nH/cm (probably influence of connectors on the 20cm cable that was measured)
- Two twisted magnet wires, 1.2mm2, about 2.5 nH/cm, but it's way too stiff
- Two stranded copper wires, 1.5mm2, stripped, put one wire in shrink tube, shrink, put other wire around, slip everything in a shrink tube, then shrink it : about 2.5nH/cm also, but much less stiff... copper braid in shrink tube should score well too.
Standard zipcord scored around 6-8 nH/cm...
So far the winners are :
- 10 conductor ribbon cable, with every other wire in parallel, measures around 2nH/cm, although theory says about 1nH/cm (probably influence of connectors on the 20cm cable that was measured)
- Two twisted magnet wires, 1.2mm2, about 2.5 nH/cm, but it's way too stiff
- Two stranded copper wires, 1.5mm2, stripped, put one wire in shrink tube, shrink, put other wire around, slip everything in a shrink tube, then shrink it : about 2.5nH/cm also, but much less stiff... copper braid in shrink tube should score well too.
Standard zipcord scored around 6-8 nH/cm...
Nope, didn't change any such thing and the 8R result is identical to the one I posted last time............besides 'reflection coefficient' is meaningless in the context of audio frequency risetimes.
No, I used 226pf total which is reasonable [Davis, Effects of Cable, Loudspeaker and Amplifier Interaction, 1991]. But actually, cable capacitance is irrelevant here, because it is shunted by source impedance and load impedance it has a very small effect. Results don't depend on capacitance - to quote Davis " The capacitive component of the cable is too small to have much influence at audible frequencies and is thus omitted from the model". However, for completeness I included a reasonable value anyway, but as Davis says it has negligible effect.
Yes - it's inaudible.
Try reading Davis 1991 or do some modelling with audioband risetime yourself, you'll find you are wrong.
Last edited:
Yes go ahead you should repeat that - but this time with a correct 25uS or so audioband risetime to fix the fundamental mistake.................
The truth will out.
Debating inaudible effect in audio? What a waste of time. 🙄
There are audible effects the members can be educated by debate. Try acoustic treatments.
Yes, the equation is LC = 1034 DC, with zip DC is about 4. L in nH per foot, C in pf per foot. So L and C are inversely related.
jn
Obviously, Davis has no concern with ITD, stereo imaging..
Perhaps you should find a source that includes the entire human hearing model.
I read Davis. He certainly has a broad brush, with some conceptual errors. However, I do not recall if those conceptual errors impact the current discussion.
Read him, saw stuff lacking.
Modelling? Been there, done that.
Hardware, actual measurements, been there, done that as well.
Got great results matching T-line both with coax and zip. Zip was lossier as well due to proximity crowding within the copper due to transition speed, it took even longer to actually settle. You know, copper resistance and all that.
Alas, I no longer have those BeO microwave resistors at my disposal, so I have to settle for metal film arrays.
jn
I note with pleasure, that your presented model with your favorite 25 uSec risetime introduced it's settling delay within 1 or 2 microseconds, and retained that following error out to the end of the slew.
And note with interest the "following error" as I very specifically stated as a consequence of the step function settling time.
Isn't it great to know that control system theory I presented works exactly as I said it did? Thanks for that confirmation.
Now that we agree that the high speed settling time error is the following error for the slower waveforms, the challenge is to determine if the variation in load causes sufficient variation in the following error such that it approaches human audibility. We've been bandying about 5 uSec as an arbitrary number, but in reality, researchers have demonstrated 1.2 uSec. I suspect that's really only justified in a lab setting with headphones, 2 to 5 is probably more realistic.
You said 1 uH and 226 pf, what kind of cable were you modelling and how long was it?
And what frequency was the 1uH measured at?
jn
Last edited:
I did mean to put a smiley after my text....
I do agree, we could do with a graph of what has the most effect.... I had to move the living room around for the missis the other day and now the stereo sounds ..... Bad so I wont be worrying about cable lifters for a while yet (i.e. never)😀
1.1uS to be precise. But even to obtain that requires a load impedance of 1R - which I used because that's what Scott used for reference and you endorse - not because I think it is representative of reality or real speakers. At 8R the delay is 0.13uS, more representative of nominal speaker loads perhaps.
The audible threshold in this thread keeps creeping downward as real delays are exposed as decreasingly small....... Even if so, 1.2uS delays don't arise from reasonable cables and loads.jneutron said:
You're not getting the meaningless of that question at all, are you ?jneutron said:
So let's see that re-run of your work but this time with a 25uS or so audioband risetime to fix the fundamental mistakes, you'll reach very different conclusions.................
What about the most critical part of all this in audio product, the listening, have you been there done that?
10u secs is the same as moving an ear 1/100 of an inch. Breathing will make more of a difference in ITD. And I'd bet some speakers move more than this with loud low notes.
And what is the characteristic impedance of the cable you modelled?
You may recall, early on I mentioned that bog zip runs 100, 150, 200 ohms impedance depending on insulation thickness. And that I recommend putting the cable impedance down into the realm of the speaker impedance. Recall also that I mentioned 4 zip cables in parallel, which will provide an impedance somewhere between 40 and 50 ohms.
So what do you use to discredit my assertion that 40 to 50 ohms cable z is a good thing? 66 ohms. 1 uH/226 pf.
Did you miss the fact that you've simmed exactly where I said provides the least amount of delay error?
Thanks for bolstering my assertion. I am buoyed by the fact that you were trying to disprove my assertion, and instead proved part of it..
I'm trying to get you to walk first...you can run later..😀
The 1.2 uSec number was learned back in 1972. I didn't want to give you a heart attack..
Ah, yes...1.1uSec perhaps "to be precise", but certainly not 1.2 uSec.
As opposed to your very clear establishment that the settling time does indeed exist, it is at 1.1uSec with a 66 ohm cable.
So, lets summarize..you state first that there is no cable prop delay that can reach into the audible regime...then you use a model which is predisposed towards your statement in that it cannot model the high speed.. and you use a low z cable as per my recommendation to minimize the delay, then state that because the delay is minimized, I must be incorrect that the delay can be large enough..
Circular logic. The last time I saw that kind of logic, it was a cable vendor.
How do you think I developed the test I previously posted? Duh.
My test, my speakers, my cables...my own time. And, most importantly, done with visual cues, with knowledge of what I listen for, in a completely controlled environment (controlled by the listener, me). So, as a data point, perfectly good. As a statement of fact of audibility, clearly and scientifically flawed.. That is why I said, if anybody else hears a difference, then there is a need for controlled listening tests. I cannot trust my own expectation bias.
Test equipment is a different story. There, I prefer a collaborator, an independent examination of measured results.
jn
Last edited:
Sigh...we need sticky.
As I previously stated, interchannel delays in the milliseconds can be changed by tightening a gluteus in the listening chair. Forget the head in a vice argument.
Also, nordmark found that dither actually increased our capability to 1.2uSec between 500 hz and IIRC, 12 Khz. Woofer motion, as you say, can certainly dither midrange content by that much, and more.
If the subject were easy, we'd all be rocket scientists.
jn
~1/100 of a foot so its closer to 1/8th inch
but again some perspective - even those boasting of R/L frequency matching in dynamic headphones or loudspeakers are proud of +/-1 dB matching between channels over frequency
a 1 dB peak or dip near a kHz is >10us group delay difference just from the phase slope associated with the linear system minimum phase response
I said""You said 1 uH and 226 pf, what kind of cable were you modelling and how long was it?""
Your response:
I fully understand what I am asking.
You used parameters which are not consistent with a cable made with parallel round conductors. I can duplicate those numbers, but it requires at least 3 zip style cables in parallel independently twisted to kill magnetic coupling.
To model an actual parallel conductor pair with that capacitance, you need to increase your inductance by a factor of (150/66) squared, or 5.16.
What does your simple model look like with 5.16 uH and 226 pf?
jn
Your response:
I fully understand what I am asking.
You used parameters which are not consistent with a cable made with parallel round conductors. I can duplicate those numbers, but it requires at least 3 zip style cables in parallel independently twisted to kill magnetic coupling.
To model an actual parallel conductor pair with that capacitance, you need to increase your inductance by a factor of (150/66) squared, or 5.16.
What does your simple model look like with 5.16 uH and 226 pf?
jn
Looks like you have not been there done that. I don't see any mention of cable lifters or stands.
You asked a question within the context of my statement (which you quoted) that I have modelled cable to load interactions using continuous modelling, microcap back in the day, using T-line modelling, building my own recursive analysis package on a pentium 150 screamer, and using actual hardware like amps, pulse generators, reed relays, into resistors and combo loads of various types.
So I answered it.
Within the context of cable lifters or stands, are you again saying that unless I've performed DBT's with them, you have ruled that I cannot post?
Really?
jn
Nope, you still fundamentally don't get it. There's a latency of 1.1uS for a 1uH cable and a 1 ohm load, when driven by an audioband risetime of 25uS. It has nothing to do with cable capacitance or rf characteristic impedance, depends only on cable inductance and load impedance. If the load is 8 ohms, latency is 0.13uS.
So let's see that re-run of your work but this time with a 25uS or so audioband risetime to fix the fundamental mistakes, when you do this you'll realise.................
- Status
- Not open for further replies.