Huh?
What does this have to do with harmonics? And reflections do to impedance mismatches in a transmission line?
What does this have to do with harmonics? And reflections do to impedance mismatches in a transmission line?
Last edited by a moderator:
When listening pleasure instead of making money is the sole reason to audition different cables, then proving with statistical valid output a subjectively perceived change in sound is the last thing one is interested in.
But mentioning positive experiences could stimulate others. Exchanging audio experiences is an important function of this very forum.
Only one or two guys tried my less the $25,- cable proposal, so far for prejudice !
Talking things into the ground is the easiest thing in the world.
Hans
Those who are trying to make money off of selling cables also audition different cables.
It can also be used by sellers to stimulate sales.
That depends on the type of experiences. Some are downright waste of time and money such as cryo-treating audio cables or cable lifters for the so called sound improvement.
So, how do you tell the difference between perception and supposed real audible difference?
And, how would that be verified?
Pavel,
what special winding techniques do you speak of? I was discussing with George how my toroidal winder could perform, but that had nothing to do with demonstrating anything in the audio band for me, just how to make a more accurate toroidal inductor. edit: the application I was winding toroids for needed 10 to 20 ppm matching inductive wise, so it behooved me (just had to say that) to make the winding as accurate and repeatable as possible.
John
Please reread your answers and see that you are not contributing in any sensible or constructive way.
I doubt if you are even interested in the subject of this forum.
Evenharmonics in sound reproduction are almost harmless or even agreeable up to a high level which can’t be said about unevenharmonics.
It seems you took the wrong harmonics in your alias.
Hans
I'd like to resurrect this discussion, maybe, with a few observations and questions. They're kind of random.
First, this paper shows another approach to making a direction coupler that might be more suitable for this task:
Dunsmore Coupler
I wonder what the best impedance would be for this.
Second, I took the LTspice files that Hans Polak offered for an update for Boris' cables and tried to apply a similar approach for Mogami W2972 speaker cable. Why this cable? A lot of people seem to like it and Mogami provides technical specifications suitable for this simulation.
Mogami W2972 Specs
Look! The capacitance changes with frequency! What is that about? It might be the actual effect of G, at least as the test instrument interprets its capacitance measurement.
Pressing on, I found that it was easy enough to tweak the values that Hans (can I call you Hans?) used to make this Mogami cable appear similar in terms of impedance to his design.
Of course, applying networks across the line short segments of the cable is an approximation of what an actual cable of the desired impedance is. You'd really need an infinite number of them to really make the cable the desired impedance. So, what is a reasonable and effective number of segments? I found that I could apply a Zobel at the end of a 2M cable and get almost the same results. But, is that the same?
Next, I tried to use a more complete transmission line model in the LTspice that simulates and actual lossy line. I found this in my investigations:
Lossy LTspice Transmission Line
Applying this, I got similar results to what Hans got with his lumped model, as you might expect. But, I had to try.
Then, I got wondering how much of the sonic effect Hans observed was due to the actual transmission lines effects jneutron posited and how much was due to the effects of a complex load at the end of an unmatched speaker cable on an amplifier sourcing power.
So, I just decided to use the lossy transmission line model for a 2M Mogami cable attached to a speaker impedance simulator circuit I found some years ago. (I can't remember the source, but I figured that the big thing was to have some changing reactive component to the impedance and a resistive component somewhere between 2 and 10 Ohms in the audio band. Above the audio band, I presumed - rightly or wrongly - that the impedance would just continue to rise up through the MHz region.)
That was interesting! An LTspice simulation of the impedance "seen" by the amplifier was OK up through a couple hundred KHz, but at a couple MHz rose to over 10K and then back down again. No wonder a Zobel network at the amplifier output is a good idea. These guys explain why:
Dennis Feucht on Oscillations
Chessman and Sokal on Oscillations
Adding a Zobel at the load end seems to smooth the impedance out, as you might expect.
Finally, I decided to see if I could really scare myself. So, I took a stability analysis I built in LTspice for a relatively common transistor based amplifier topology that uses negative feedback from the output back to the inverting input of the input stage. ADI supplies an example of the Tian Method for closed loop stability analysis in LTspice, so I used that:
The Tian Method
This lets you examine the closed loop gain and phase margins.
Well... When you have a simple resistive load, it's all pretty simple. When you add a pure capacitance in parallel with that resistor, it's still pretty simple, even though the margins change. More capacitance means lower margins. But, when you use a complex load at the end of a cable, not only do the margins change a lot, but the shape of the margin curves also change and they become rather lumpy. Imagine what that might do to the distortions.
I did succeed in scaring myself.
(If anybody is interested, I can supply the LTspice files.)
It's all really interesting and perplexing overall. Even if there is no effect from reflections in the unmatched loudspeaker cable - which I'm not suggesting, btw - there's lots of other changes that take place when you add a complex load that's made more complex by connecting that load with a speaker cable.
First, this paper shows another approach to making a direction coupler that might be more suitable for this task:
Dunsmore Coupler
I wonder what the best impedance would be for this.
Second, I took the LTspice files that Hans Polak offered for an update for Boris' cables and tried to apply a similar approach for Mogami W2972 speaker cable. Why this cable? A lot of people seem to like it and Mogami provides technical specifications suitable for this simulation.
Mogami W2972 Specs
Look! The capacitance changes with frequency! What is that about? It might be the actual effect of G, at least as the test instrument interprets its capacitance measurement.
Pressing on, I found that it was easy enough to tweak the values that Hans (can I call you Hans?) used to make this Mogami cable appear similar in terms of impedance to his design.
Of course, applying networks across the line short segments of the cable is an approximation of what an actual cable of the desired impedance is. You'd really need an infinite number of them to really make the cable the desired impedance. So, what is a reasonable and effective number of segments? I found that I could apply a Zobel at the end of a 2M cable and get almost the same results. But, is that the same?
Next, I tried to use a more complete transmission line model in the LTspice that simulates and actual lossy line. I found this in my investigations:
Lossy LTspice Transmission Line
Applying this, I got similar results to what Hans got with his lumped model, as you might expect. But, I had to try.
Then, I got wondering how much of the sonic effect Hans observed was due to the actual transmission lines effects jneutron posited and how much was due to the effects of a complex load at the end of an unmatched speaker cable on an amplifier sourcing power.
So, I just decided to use the lossy transmission line model for a 2M Mogami cable attached to a speaker impedance simulator circuit I found some years ago. (I can't remember the source, but I figured that the big thing was to have some changing reactive component to the impedance and a resistive component somewhere between 2 and 10 Ohms in the audio band. Above the audio band, I presumed - rightly or wrongly - that the impedance would just continue to rise up through the MHz region.)
That was interesting! An LTspice simulation of the impedance "seen" by the amplifier was OK up through a couple hundred KHz, but at a couple MHz rose to over 10K and then back down again. No wonder a Zobel network at the amplifier output is a good idea. These guys explain why:
Dennis Feucht on Oscillations
Chessman and Sokal on Oscillations
Adding a Zobel at the load end seems to smooth the impedance out, as you might expect.
Finally, I decided to see if I could really scare myself. So, I took a stability analysis I built in LTspice for a relatively common transistor based amplifier topology that uses negative feedback from the output back to the inverting input of the input stage. ADI supplies an example of the Tian Method for closed loop stability analysis in LTspice, so I used that:
The Tian Method
This lets you examine the closed loop gain and phase margins.
Well... When you have a simple resistive load, it's all pretty simple. When you add a pure capacitance in parallel with that resistor, it's still pretty simple, even though the margins change. More capacitance means lower margins. But, when you use a complex load at the end of a cable, not only do the margins change a lot, but the shape of the margin curves also change and they become rather lumpy. Imagine what that might do to the distortions.
I did succeed in scaring myself.
(If anybody is interested, I can supply the LTspice files.)
It's all really interesting and perplexing overall. Even if there is no effect from reflections in the unmatched loudspeaker cable - which I'm not suggesting, btw - there's lots of other changes that take place when you add a complex load that's made more complex by connecting that load with a speaker cable.
For speaker cable I would suggest trying Canare Star Quad. Its inexpensive and works pretty when wired up properly. https://www.canare.com/speakercable
Also, an interesting experiment would be to strip the rubber jacket off the speaker cable while carefully preserving all the internal wrapping materials and while maintaining the existing tight conductor geometry. Then Compare that to using the same cable with the rubber jacket sill on it. I wasn't expecting much difference 🙂
Another thing to try is put 100R noninductive power resistors across the speaker terminals. IIRC the resistors may have been made by Caddock(?) and come in something like TO-220 packages. Again IIRC this is something suggested by 1audio here in the forum some time ago. Believe he said it helped the sound even though they didn't know exactly why at the time. I can probably find the part number if anyone is interested in trying them.
Also, an interesting experiment would be to strip the rubber jacket off the speaker cable while carefully preserving all the internal wrapping materials and while maintaining the existing tight conductor geometry. Then Compare that to using the same cable with the rubber jacket sill on it. I wasn't expecting much difference 🙂
Another thing to try is put 100R noninductive power resistors across the speaker terminals. IIRC the resistors may have been made by Caddock(?) and come in something like TO-220 packages. Again IIRC this is something suggested by 1audio here in the forum some time ago. Believe he said it helped the sound even though they didn't know exactly why at the time. I can probably find the part number if anyone is interested in trying them.
Thanks.
My point in using the numbers of the Mogami cable was that, well, I could get the numbers from their data sheet. I could've bought some cable and measured it, but this was plain easier and took far less time.
I've been using Zobel networks at the speaker terminals for several years now, but never bothered to calculate what might be optimum. Then Hans' LTspice model caught my attention and I thought to try his approach. That led from one thing to another. "Another" has turned out to be a very deep rabbit's hole.
I only thought of sharing the information on the outside chance that somebody else might be interested or might have some insight into the questions I asked. And, to be annoying.
Based on your suggestion, I just tried comparing in simulation a 100 Ohm resistor in parallel with the simulated speaker impedance, as Demian suggested, to a Zobel with 10 Ohms in series with .01 uF. The 100 Ohm resistor certainly damped some of the resonant peaks above 1 MHz and was overall much cleaner. That's probably because the speaker cable has closer to a 100 Ohm impedance above 1 MHz than than the 1 Kohm plus the simulated speaker presents up there. (I'd have to measure what an actual loudspeaker has for impedance above 100 KHz. See? Rabbit's hole.) I'm not sure how that would affect the sound, but that's what the simulation shows with regard to impedance. It might affect amplifier stability in some way - see Dennis Feucht's article.
Then, I tried 100 Ohms in series with the .01 uF in the Zobel instead of 10 Ohms. That simulated pretty much the same as just using the 100 Ohm resistor by itself. The advantage of course is that you don't need a very high power resistor if you have a cap in series.
So, I still don't know what is optimum. But, working on it.
My point in using the numbers of the Mogami cable was that, well, I could get the numbers from their data sheet. I could've bought some cable and measured it, but this was plain easier and took far less time.
I've been using Zobel networks at the speaker terminals for several years now, but never bothered to calculate what might be optimum. Then Hans' LTspice model caught my attention and I thought to try his approach. That led from one thing to another. "Another" has turned out to be a very deep rabbit's hole.
I only thought of sharing the information on the outside chance that somebody else might be interested or might have some insight into the questions I asked. And, to be annoying.
Based on your suggestion, I just tried comparing in simulation a 100 Ohm resistor in parallel with the simulated speaker impedance, as Demian suggested, to a Zobel with 10 Ohms in series with .01 uF. The 100 Ohm resistor certainly damped some of the resonant peaks above 1 MHz and was overall much cleaner. That's probably because the speaker cable has closer to a 100 Ohm impedance above 1 MHz than than the 1 Kohm plus the simulated speaker presents up there. (I'd have to measure what an actual loudspeaker has for impedance above 100 KHz. See? Rabbit's hole.) I'm not sure how that would affect the sound, but that's what the simulation shows with regard to impedance. It might affect amplifier stability in some way - see Dennis Feucht's article.
Then, I tried 100 Ohms in series with the .01 uF in the Zobel instead of 10 Ohms. That simulated pretty much the same as just using the 100 Ohm resistor by itself. The advantage of course is that you don't need a very high power resistor if you have a cap in series.
So, I still don't know what is optimum. But, working on it.
It was in fact Cyril Bateman who searched for a reason why certain amps where blown when using cables with a low ohmic characteristic impedance.
He found that it was because of reflections within the LS cable when the speaker had a very high impedance, almost looking like an open end to the used cable.
This reflection was not captured by the amp's Zobel network and caused the amp to destruct itself in those cases, causing RF oscillations at ca 3.5Mhz for 5 meter cables.
Inserting a 100R resistor at the speaker terminals, plus parallel to this 100R a 33nF cap in series with a second resistor who's value together with the 100R should match the characteristic cable's impedance, fully cured all oscillation problems and prevented the amps from being destructed.
Hans
Still seems like it was Demian whom I recall posting here in the forum to the effect that even when an amp wasn't blowing up, the 100R resistor improved subjective SQ. If so, probably was in one of the old Blowtorch threads. Also believe Demian suggested a part number for the resistor. Still have two of those resistors somewhere around here from back then.
Read Cyril Bateman’s paper and you will see that THD with and without this resistor within the audio band will differ for amps that do not blow up but have those RF oscillations nevertheless.
And this 100R resistor should be low inductance and so should the second resistor be in case of a low Char Imp cable.
And this 100R resistor should be low inductance and so should the second resistor be in case of a low Char Imp cable.
Here let me do a simple search for you https://www.diyaudio.com/community/threads/the-black-hole.349926/post-6597247
Thanks Bill. Still trying to figure out where in the threads he gave the part number for the recommended resistor. Presumably there might have been more than one post on the subject?
EDIT: Here we go: https://www.diyaudio.com/community/...rch-preamplifier-part-iii.318975/post-5805680
"it did seem to have the desired effect of a different (better) sound pretty much everywhere it was used."
EDIT: Here we go: https://www.diyaudio.com/community/...rch-preamplifier-part-iii.318975/post-5805680
"it did seem to have the desired effect of a different (better) sound pretty much everywhere it was used."
That was 25+ years ago! Caddock has TO220 power resistors that are pretty pure resistive until you get into real RF. 25W should be more than enough with any speaker and program material listened to by humans. Also 25W into 100 Ohms is 300+ watts into 8 Ohms.
The rationale was that a zip cord construction will be around 100 Ohms Z as are most twisted pairs. At RF what matters is the next section. What is after that you don't have much impact on. Most speakers are inductive at ultrasonic frequencies so a resistor that matches the transmission line should make sure the amp is not driving a reactive load. I did not want to add a cap in series to not add another frequency selective aspect to the network.
The trick with a series cap also seems to make a difference on power cords.
The rationale was that a zip cord construction will be around 100 Ohms Z as are most twisted pairs. At RF what matters is the next section. What is after that you don't have much impact on. Most speakers are inductive at ultrasonic frequencies so a resistor that matches the transmission line should make sure the amp is not driving a reactive load. I did not want to add a cap in series to not add another frequency selective aspect to the network.
The trick with a series cap also seems to make a difference on power cords.
There's really two different topics that have been addressed here. Or, at least, argued about.
One is jneutron's theory that reflections in a speaker cable tend to mess up the ITD characteristics. I'm not sure an actual conclusion was ever reached on that.
The other is that the high frequency impedance presented to an the amplifier output can and often does cause bad behavior in the amplifier causing distortions of different kinds. That one has been measured and can be simulated as well. The solution to that is to make certain that the transmission line, the speaker cable, has a decent termination at the speaker end that is close to the RF impedance of the cable. That pretty much eliminates virtual short circuits at some frequencies and open circuits at others. Neither help an amplifier work well. Since the cable impedance at lower frequencies in the audio band tends to rise a lot because of the effects of the shunt conductance, "G", it's hard to terminate things well there. (Same issue as with the transmission line ITD dilemma.)
Two points on that.
If the goal is to terminate the line properly at high frequencies, which I'll arbitrarily say is 1 MHz, putting a cap in series with the resistive termination at the speaker has no downside that I can find. Most any resonance caused by the interaction of that capacitor with the (probable) rising impedance of the speaker and crossover network probably is damped by the series resistance. But, the capacitor does restrict the power dissipation to a low level in the audio band. I don't think that anybody really drives their loudspeakers at 40 Vrms at 20 KHz, at least for long, so a half Watt resistor will do.
Now, about that audio band impedance...
Today I tried to measure the RF impedance of some commercial speaker cable a friend gave to me. This cable has two pairs of wires in a quadrifilar format, much like the Mogami and Canare cables. Nothing really unusual about that. But, the so-called "negative" pair is insulated with what I think is a kind of carbon filled plastic, which they call semi-conductive. I got flaky results in my measurements and will try them again over the next couple of days. The "1/8 wavelength" unterminated transmission line test gave me results I didn't expect. (Like 15 or so Ohms)
But, wouldn't an insulation like that dramatically move G away from 0? That may confuse the measurement and it certainly confuses me. It also might make for speaker cables with lower impedance within the audio band, too. It makes me wonder about the propagation delay associated with the "negative" conductors versus the other ones with more conventional insulation. What might that do?
As for power cables, most every measurement of home power systems I've seen published has shown that the impedance at the outlet goes from a very low value down around 50/60 Hz (the cable resistance) up to a value somewhere around 50 Ohms or so at high frequencies (a couple hundred KHz) where it settles out. Ralph Morrison has a graph of this in one of his books, if I recall.
That approximates to each leg of the power line being like a 25 uH inductor in parallel with 25 Ohms of resistance. Gee, just like official LISN networks.
So, it makes sense that if you terminate the AC mains at high frequencies in something like the characteristic impedance of the line, at minimum you reduce standing wave peaks in the overall mesh at high frequencies. It also helps with AC line filters that often have actual peaks in the response at a few tens of KHz when driven from the actual AC mains source impedance rather than 50 Ohms across the band. Or, think about what putting a suitably rated cap between line and neutral might do.
It all gets very complicated when you move beyond the easy first order description of things. None of this is quantum physics - ok, everything is really quantum physics at some level - but often the details are not looked at closely. This is how you blow up space shuttles. Thankfully, amateurs like me don't design space shuttles.
One is jneutron's theory that reflections in a speaker cable tend to mess up the ITD characteristics. I'm not sure an actual conclusion was ever reached on that.
The other is that the high frequency impedance presented to an the amplifier output can and often does cause bad behavior in the amplifier causing distortions of different kinds. That one has been measured and can be simulated as well. The solution to that is to make certain that the transmission line, the speaker cable, has a decent termination at the speaker end that is close to the RF impedance of the cable. That pretty much eliminates virtual short circuits at some frequencies and open circuits at others. Neither help an amplifier work well. Since the cable impedance at lower frequencies in the audio band tends to rise a lot because of the effects of the shunt conductance, "G", it's hard to terminate things well there. (Same issue as with the transmission line ITD dilemma.)
Two points on that.
If the goal is to terminate the line properly at high frequencies, which I'll arbitrarily say is 1 MHz, putting a cap in series with the resistive termination at the speaker has no downside that I can find. Most any resonance caused by the interaction of that capacitor with the (probable) rising impedance of the speaker and crossover network probably is damped by the series resistance. But, the capacitor does restrict the power dissipation to a low level in the audio band. I don't think that anybody really drives their loudspeakers at 40 Vrms at 20 KHz, at least for long, so a half Watt resistor will do.
Now, about that audio band impedance...
Today I tried to measure the RF impedance of some commercial speaker cable a friend gave to me. This cable has two pairs of wires in a quadrifilar format, much like the Mogami and Canare cables. Nothing really unusual about that. But, the so-called "negative" pair is insulated with what I think is a kind of carbon filled plastic, which they call semi-conductive. I got flaky results in my measurements and will try them again over the next couple of days. The "1/8 wavelength" unterminated transmission line test gave me results I didn't expect. (Like 15 or so Ohms)
But, wouldn't an insulation like that dramatically move G away from 0? That may confuse the measurement and it certainly confuses me. It also might make for speaker cables with lower impedance within the audio band, too. It makes me wonder about the propagation delay associated with the "negative" conductors versus the other ones with more conventional insulation. What might that do?
As for power cables, most every measurement of home power systems I've seen published has shown that the impedance at the outlet goes from a very low value down around 50/60 Hz (the cable resistance) up to a value somewhere around 50 Ohms or so at high frequencies (a couple hundred KHz) where it settles out. Ralph Morrison has a graph of this in one of his books, if I recall.
That approximates to each leg of the power line being like a 25 uH inductor in parallel with 25 Ohms of resistance. Gee, just like official LISN networks.
So, it makes sense that if you terminate the AC mains at high frequencies in something like the characteristic impedance of the line, at minimum you reduce standing wave peaks in the overall mesh at high frequencies. It also helps with AC line filters that often have actual peaks in the response at a few tens of KHz when driven from the actual AC mains source impedance rather than 50 Ohms across the band. Or, think about what putting a suitably rated cap between line and neutral might do.
It all gets very complicated when you move beyond the easy first order description of things. None of this is quantum physics - ok, everything is really quantum physics at some level - but often the details are not looked at closely. This is how you blow up space shuttles. Thankfully, amateurs like me don't design space shuttles.
IME and in the experience of others there is likely more to speaker cable than what is being discussed here. Its possible to have two pieces if nearly identical speaker cable sound different from each other. Depending on the exact physical difference, the difference in sound might not be much though. Some people can still reliably recognize the difference. For one example, I already suggested stripping the rubber jacket off of one set of originally identical cables, while leaving stripped set otherwise fully intact. Of course that could have some effect on Z0, but that needn't be the only thing affected. As CG noted, "It all gets very complicated when you move beyond the easy first order description of things." Considering only the effects of Z0 could be considered a first order model of sorts.