Cable Theory

[DonH50]

All In My Opinion:

1. While dielectric absorption in high-impedance low-level circuits could arguably be audible, I suspect it is of little consequence for nominal interconnects, let alone speaker cables. I'd have to think about it in tube circuits, but still it'd have to be a pretty bad cable. DA, dispersion and the like are pretty small effects...

2. The voice coil of a dynamic driver looks inductive, true, but electrostatics and some others look capacitive. And, of course, what the amp sees is the crossover network plus speakers, so you'd have to look at the phase as well as magnitude to estimate the dominate reactive component.

3. Once you know the element values, it should not be hard to model and measure the voltage across each driver. Amir's model should do. However, you'll need some measure of sonic efficiency (the electromechanical-to-acoustic transfer function) to translate that voltage to actual sonic output levels. Empirically, you could apply a series of test tones and measure the voltage at the input terminals and across each voice coil to derive the voltage transfer functions, then measure the output in an anechoic chamber to get the acoustic conversions. Could take a bit of work. I did it in college and a few times since but not recently.

4. Reactances can create resonators and thus can perhaps be a larger issue than pure resistance, particularly in terms of stability and high-Q peaks and nulls in frequency response. However, high resistance also contributes to the frequency response and causes poor speaker control (high damping factor, poor load pull, however you think of it) with resulting poor sound.

5. I have applied shield bias (d.c.) voltages as required for some low-level testing. Far below anything that makes sense to me for audibility (e.g. when measuring uV to fV levels).

FWIWFM - Don

 

[The Smokester]

All In My Opinion:

1. While dielectric absorption in high-impedance low-level circuits could arguably be audible, I suspect it is of little consequence for nominal interconnects, let alone speaker cables. I'd have to think about it in tube circuits, but still it'd have to be a pretty bad cable. DA, dispersion and the like are pretty small effects...

Any quantitative estimate on magnitude? I think it is commonly accepted that it is audible in capacitors in amps, is it not? (Usual qualifiers apply.)

2. The voice coil of a dynamic driver looks inductive, true, but electrostatics and some others look capacitive. And, of course, what the amp sees is the crossover network plus speakers, so you'd have to look at the phase as well as magnitude to estimate the dominate reactive component.

Right. Impedance is not just inductive. I know you know this but just as a basis to make sure my point is clear (also made in Tom's refered paper, BTW): There are several equivalent ways to represent impedance...One is as a cartesian vector. Capacitive and inductive components sum directly but with opposite signs and are orthogonal to the resistive component (in a cartesian representation). The total impedance is the vector some of resistive and reactive components. Adding capacitance and inductance to a circuit changes the reactance but doen't change the resistance. In the case of a speaker cable, the main effect of the cable may be due to its reactive component because the sound is created by a (almost) totally reactive component (I think, and I am talking about the voice coil). The cable and voice coil reactances will add directly together. Looking at the total impedance allows the resistive part to mask the real size of the effect. Ya gotta look at both and in the right places, too. So I am still not satisfied that the calculations on this thread are telling the full story.

3. Once you know the element values, it should not be hard to model and measure the voltage across each driver. Amir's model should do. However, you'll need some measure of sonic efficiency (the electromechanical-to-acoustic transfer function) to translate that voltage to actual sonic output levels. Empirically, you could apply a series of test tones and measure the voltage at the input terminals and across each voice coil to derive the voltage transfer functions, then measure the output in an anechoic chamber to get the acoustic conversions. Could take a bit of work. I did it in college and a few times since but not recently.

We are just interested in the relative magnitudes, aren't we? The "sonic efficiency" will be constant regardless of the cable and so would cancel out in a comparison. (Maybe I don't understand what you mean by sonic efficiency.)

4. Reactances can create resonators and thus can perhaps be a larger issue than pure resistance, particularly in terms of stability and high-Q peaks and nulls in frequency response. However, high resistance also contributes to the frequency response and causes poor speaker control (high damping factor, poor load pull, however you think of it) with resulting poor sound.

It is a complicated, and perhaps complex, system (even taking out the human factor).

5. I have applied shield bias (d.c.) voltages as required for some low-level testing. Far below anything that makes sense to me for audibility (e.g. when measuring uV to fV levels).

FWIWFM - Don

Yes. I got conned into buying a power cord like this one time. Has a cool blue light, though.

 

[c1ferrari]

...(e.g. when measuring uV to fV levels).

Okay, I'm cognizant of microvolts; however, what kind of unit is an "fV"

 

[microstrip]

Okay, I'm cognizant of microvolts; however, what kind of unit is an "fV"

femto = 10exp-15 . I would be very curious about how to make measurements to the fV level.

I have carried measurements of tens of femtoampere (10exp-14A resolution) and it was a nightmare!

 

[Speedskater]

femto = 10exp-15 . I would be very curious about how to make measurements to the fV level.
I have carried measurements of tens of femtoampere (10exp-14A resolution) and it was a nightmare!

Now with budget computer based test equipment it easier (make that less difficult). If you don't need to do the test real time, you can do what astronomy hobbyists did before budget CCD cameras were available. Using a web-cam they would take hundreds of consecutive photos and let the computer remove the noise.

 

[c1ferrari]

femto = 10exp-15 . I would be very curious about how to make measurements to the fV level.

I have carried measurements of tens of femtoampere (10exp-14A resolution) and it was a nightmare!

Thanks, mu!

 

[DonH50]

I like Tom's last line.

@Smokester -- Do you have an engineering/science background? I was going to quibble a bit but no point if you're an expert trying to make things clear for lay-folk (I understand the difficulty). Before responding further, and sticking my feet into my mouth too far, I'd like to know my audience.

microstrip, I found measuring to fV be a nightmare, even with a special meter and all the trimmings (including wicked-expensive cables). And, I too have (tried to) measure leakage currents in the fA range, another nightmare. Starts with a good screen room, or least a screen-box, then gets ugly very quickly. Microstrip, stripline, coplaner waveguides, hmmm...

Speedskater, I have a friend with something like $50k in his astronomy set-up (including the scope). Noise averaging is still required in CCDs due to dark current (small world; I actually worked with CCDs once upon a time, but on the signal-processing side rather than the imager side). I can't imagine an inexpensive PC system accomplishing reliable fV measurements (even uV would be a challenge); too many things to mess up the averaging. Too deep for here, I think.

My 0.000001 cents (one microcent) - Don

 

[The Smokester]

I like Tom's last line.

@Smokester -- Do you have an engineering/science background? I was going to quibble a bit but no point if you're an expert trying to make things clear for lay-folk (I understand the difficulty). Before responding further, and sticking my feet into my mouth too far, I'd like to know my audience... Don

Don, Yes. Thanks for asking. I have a Ph.D. in nuclear physics but am certainly not a practical electrical engineer. That means I have to go back to basic electromagnetism and derive everything from scratch to make sure I understand what people are saying.

Did you read the reference that Tom posted a page back? It is a good framework for the discussion, I think, because it exposes some of the moving parts (so to speak).

 

[FrantzM]

Hi
Now about cables in the femtovolt measuring domain .. How expensive are they? Would like to have an idea of size, construction and price... I am thinking that maybe , just maybe they may not be as expensive as my favorite whipping boys the $30K speaker cables .. Inquiring mind would like to know

 

[Speedskater]

Frantz, if you want to learn more about low level measurements and their problems and limits, do a quick read at about page 1-4.

http://www.keithley.com/knowledgecenter/knowledgecenter_pdf/LowLevMsHandbk_1.pdf

 

[Speedskater]

Also Ed Simon has been doing extremely low level distortion tests of audio interconnect cables.
See AudioXpress magazine #11/09 Distortion Meter By Ed Simon

Also some follow-up on the DIY forum. Somewhere in this 6000 plus post thread by simon7000.

http://www.diyaudio.com/forums/anal...curls-blowtorch-preamplifier-part-ii-591.html

 

[garylkoh]

We are discussing measurements in the fV and fA range. IMHO before diving into measuring cables at these levels, we need a better understanding of what we can hear and how it relates to these measurements. And then how far do we need to go in our measurements/estimates to be measuring something that is reliably audible. And finally, we can then figure out what nature of measurement corresponds to what we are hearing in wire characteristics (if we are hearing it at all).

As we discussed in another thread starting here: http://www.whatsbestforum.com/showt...it-what-is-important-to-you&p=18387#post18387
..... we got some objectivists and some subjectivists to agree on phase accuracy down to almost 1deg. However, Amir's simulations proved that speaker cables effect phase by a factor of 10times less..... so we need to look at measurements other than phase.

As has been mentioned more than once in this thread, we cannot reliably hear a 0.1dB change in level. But at what reference level? Can we hear the difference between 18dB and 18.1dB but not 70dB and 70.1dB? What about 18dB of music with a 0.1dB injection of noise (distortion)? Or 70dB of music with 0.1dB injection of noise?

I found it pretty easy to hear the difference between dither and no dither when a 24bit signal is reduced to 16bit (and I think that it is well accepted in the recording industry that dither is needed when doing this bit-rate reduction). Since we are talking about dither below the LSB at 16bits, that's at -96dB. So, if the cable distorts the music signal at 96dB below the music signal, and at 1W we are talking about 2.66V, then 96dB below 2.66V would be 668fV (did I get my math right?)

That is assuming that we can hear something that is -96dB of the music signal...... which is below the threshold of hearing if we are talking about playback at 75dB normal listening level. The way to test this theory might be to take a 24-bit music file, inject white, grey, brown, blue and violet noise in at -96dB and then listen to it. If we can reliably tell what type of noise has been injected, then we may be able to agree that we can hear -96dB below normal listening level. If we can not, then raise the level of the noise signal, and find the threshold where 50% of us can hear it.

Then, we have a reliable level we need to be able to measure at.

Going back to cable theory, if I postulate that the cable "dithers" the music signal in somewhat the same way, and the level we need to measure at is at the levels (in pF and uH) that we are measuring, then we have reliably arrived at a point where some objectivists and some subjectivists will find common ground.

Reasonable?

It'll probably be harder to do than the DBT cable test, but maybe I can convince Marjorie of RMAF and Constantine of CAAS to include such a session so that we can get a reasonably large sample of test subjects.

 

[DonH50]

A few random comments:

1. P = E^2 / R => 1 W into 8 ohms is 2.828 Vrms or 8 Vpp. 1 lsb at 16 bits is 8/2^16 = 8/65536 = 122 uV. I agree that hearing something that far down is very unlikely...

2. Dither does a few things, among them helping to mask quantization noise (which tends to be more objectionable than white/pink/etc. noise) and reduce low-level nonlinearities in the ADC and DAC. Unless you have a perfect ADC/DAC, the noise floor and nonlinearities are (much) larger than theoretical. I believe the "nastier" sound of quantization noise compared to typical thermal/shot noise and influence (sound) of those low-level (though perhaps higher-order) nonlinearities are the primary reasons dither helps.

3. I am not sure how a cable "dithers" the sound but admit ignorance about what's in the "super" cables. Obviously any extra resistance adds noise, but I have a very tough time believe it contributes significantly. Perhaps the ones with various networks include resistors that add noise? Still hard to see it doing much at speaker levels...

4. Dielectric absorption can be measured but again I find it hard to believe it's a significant contributor sonically. In capacitors used in high-impedance (e.g. tube) circuits, maybe... If I recall correctly, it's in the uV region at audio frequencies.

5. I think it much more likely that the sonic influence of cables is related to their impedance working into the impedances of the components (player, preamp, amp, speaker, etc.) rather than effects measured at the uV level and below. Distributed RLCG causes various frequency peaking/ringing/etc. as well as dispersion (which is probably nil at audio frequencies) when presented with typical source and load impedances (which vary with frequency and perhaps amplitude, temperature, phases of the moon, etc.)

I have done measurements on cables and helped with as well as participated in numerous DBTs a long time ago. The results were enough to convince me that differences among cables can be heard in some situations, but were generally indicative of poor cables, or at least poor matching of cables to components. For example, for the vast majority of nominal cables and systems nobody could tell if it was nickle a foot heavy zip cord or $5/foot silver-tinned 000 copper going to the speakers. As for interconnects, there was virtually no difference fr the vast majority of cables and systems. Except in certain special cases of terrible choices, people could not tell the difference unless they looked at their wallets.

FWIWFM - Don