Objectivist or Subjectivist? Give Me a Break

[j_j]

If they were not able to find any differences in speaker cables or interconnects this can be an indication that the method is not able to find them or they did not try enough to find them.

Sometime ago I tried to build my speaker cables using different very different types of cables and was surprised by the negative - I really managed to built some sounding very poor, although their RLC parameters would not affect frequency response more than .1 dB.

There is something wrong with what's being reported here, because at least one ABX test has identified #22 speaker cables at a fairly short length.

And, of course, .1dB, if it's broadband, is getting up to the level of audibility. You will not hear "loudness" differences, rather they will appear as "quality" differences.

Typically matching is required to .2dB, but all you need is .1 up and .1 down to create an audible situation, albeit one that does not sound like a loudness change, but rather a quality change.

 

[microstrip]

(...) But a thread on "how to measure", "hwat to measure" and "what does it mean" is interesting. But you're never, EVER going to get '1 number' as an answer.

I think that no one believes anymore in the '1 number' dream.

But should we believe that there is a small number of parameters, may be connected by simple relations, that describe sound quality and average audibility thresholds?

 

[j_j]

I think that no one believes anymore in the '1 number' dream.

But should we believe that there is a small number of parameters, may be connected by simple relations, that describe sound quality and average audibility thresholds?

Well, I've seen a variety of opinions, but this depends on what "numbers" mean. If you mean a spectral plot of noise floor, a spectral plot of distortion products when driven by an appropriate broadband signal at high level, ditto low level, polar patterns for speakers as well as direct and power response plots, including unwrapped and undelayed phase of speakers, variations in gain and response vs. level, maybe that's 6 sets of numbers... Or so.

 

[cnpope]

There is something wrong with what's being reported here, because at least one ABX test has identified #22 speaker cables at a fairly short length.

And, of course, .1dB, if it's broadband, is getting up to the level of audibility. You will not hear "loudness" differences, rather they will appear as "quality" differences.

Typically matching is required to .2dB, but all you need is .1 up and .1 down to create an audible situation, albeit one that does not sound like a loudness change, but rather a quality change.

The specific ABX tests that I was quoting, and about which I responded, did not, I think, involve any thin wires like #22, and they did not, as far as I was able to see, report any discernible effects for those that they did test. Of course, if one goes to thin enough wire there will be noticable effects. The tests I was discussing are reported at http://home.provide.net/~djcarlst/abx_data.htm.

If the discussions are not to go on round and round in circles forever, one first needs to eliminate the totally obvious effects, which are expected by the most elementary theoretical considerations. It is obvious that thin wires with a significant resistance compared to the impedances in the devices they are connecting will have manifest effects. The discussion really needs to focus on the situations where one has eliminated such cases, and where there is, supposedly, a conflict between theory and practice.

Microstrip: You said "If they were not able to find any differences in speaker cables or interconnects this can be an indication that the method is not able to find them or they did not try enough to find them."

The method they used was to listen to them. They were very specific about which cables they were comparing. It is hardly surprising that in blind testing the effects that can actually be discovered are much "coarser" than those that people claim to be able to hear under laxer conditions. This is merely a reflection of the fact that the human ear/brain system is influenced by many expectations, and in the absence of those cues it is much less able to discern real and verifiable differences.

Chris

 

[cnpope]

Sometime ago I tried to build my speaker cables using different very different types of cables and was surprised by the negative - I really managed to built some sounding very poor, although their RLC parameters would not affect frequency response more than .1 dB.

Did you compare them with double blind tests? If not, the results are not very compelling. In any case, apparently 0.1dB is around the margins of what can be detected, so that would not necessarily be surprising.

Chris

 

[microstrip]

hey

Did you compare them with double blind tests? If not, the results are not very compelling. In any case, apparently 0.1dB is around the margins of what can be detected, so that would not necessarily be surprising.

Chris

The sound of at less two of the pairs, one using tens of wires of Teflon insulated wirewrap cable and another using flat foils were so miserable that IMHO they did not need blind tests - surely this is just an anecdotal report without any value for you. IMHO differences of .1 dB can not change the sound from being great to average or even poor, YMMV.

 

[cnpope]

The sound of at less two of the pairs, one using tens of wires of Teflon insulated wirewrap cable and another using flat foils were so miserable that IMHO they did not need blind tests - surely this is just an anecdotal report without any value for you. IMHO differences of .1 dB can not change the sound from being great to average or even poor, YMMV.

Well, we really don't have access to sufficient information about your experiment to be able to form a useful judgment. However, I would be inclined to think that if one of them sounded "miserable" then there must have been something really badly wrong with it. But without more data, it is impossible to make useful conjectures.

Chris

 

[j_j]

The sound of at less two of the pairs, one using tens of wires of Teflon insulated wirewrap cable and another using flat foils were so miserable that IMHO they did not need blind tests - surely this is just an anecdotal report without any value for you. IMHO differences of .1 dB can not change the sound from being great to average or even poor, YMMV.

Well, you are saying two things. The first I'm not going to argue. You need a blind test for anything that isn't an obvious malfunction, and there really is no argument. Now, knowing the series 'R' would be nice, as well as knowing the speaker impedance as a function of frequency. I've seen some real gems there..

To the second, you may believe what you like, but that suggests that you have not listened to a test in which the only difference is a small level difference. So try it and then get back to us, ok?

 

[valkyrie]

I have pondered this very question - the audible effects of cable - and by implication the root of the subjectivist/objectivist debate for many years.

I HEAR CABLE DIFFERENCES. Sorry - generally I write off much of the audiophile creed as the mere mutterings of deluded hobbyists - middle-aged men with too much money in their pockets and too much time on their hands. But - REALLY - I can hear - even on my mid-fi rig - cable differences. Hard to characterize and even harder to understand. After all - wire is wire is wire.

And for the record - NO I do not hear power cords, cable lifts, fuses or any of the other absurdities that afflict our noble passion like dung flies at a spring picnic.

Recently I think I have come to anticipate what I believe it is we are hearing - and it is rather obvious if you reflect upon it. It is NOT RLC values. But it has to do with "settling time". Every L and every C has an associated discharge time - after all both inductors and capacitors are about "fields" being generated and discharged. I don't think any of us will argue that the substitution of a good film capacitor in a crossover network changes the sound of that network rather dramatically. Really - what is the film doing? It is changing the "settling time", i.e. the field discharge time of this storage device.

Similarly with cables - particularly speaker cables, less so with IC cables - a high current field propagates down a conductor and creates a substantial field - that much cannot be denied. That field then collapses, e.g. discharges, or dissipates back to the original conductor. This field is maintained both as a static and dynamic electric charge - for cables - ALL of them - have both capacitive and inductive qualities.

But this field does NOT discharge - in terms of time - equally in all cables. Hence conductor material, winding of the cable, layout of the conductors, insulation - ALL would make a difference in how the cable sounds. Because, given the rapidity of audio signals in the time domain, any remaining field (whatever its quality) would have an effect on the propagation of a new charge, i.e. signal as it attempts to propagate down the cable. If ou have a very slow cable, slow in the sense that it does not readily accept the collapsing field, then the cable will "buck" the new signal - hence the often repeated claims of muddiness, or extreme clarity that are associated with various cables.

Viola - we are testing for the WRONG aspects of cable science. We need to test for settling times, i.e. discharge rates.

Disclaimer: Normally I am of the "Phelonious Ponk" school of audio, i.e. much of what we call audiophilia is pure nonsense. This message does NOT amount to my admitting a change in this philosophical stance. However I can - or I strongly believe I can - easily hear differences in both speaker cable and with some interconnects (particularly the effect of KCAG, e.g. Kimber Silver cable interconnects).

Disclaimer #2: with the advent of remote control for my preamp (a modest Vincent SA-T8 pure tube unit) I was able to compare two IC's; an Audio Quest "balanced cable" (low budget) with an unbalanced Kimber "Hero". The result? NO AUDIBLE DIFFERENCE HAS EVER BEEN HEARD BY ANYBODY. But of course no audio memory was involved. I need to do this test again with the more pricey KCAG.

Disclaimer #3: I do not believe for even a second that ANY cable manufacturer is "testing" for the effects of settling time. I think they (cable manufacturers) are hardly more than alchemists using a set of heuristics gained by painful experience and lots of years awkwardly experimenting - but these clumsy, mostly medieval methods do occasionally result in a cable that does sound significantly different from most other cables. And sometimes even better.

 

[j_j]

Viola - we are testing for the WRONG aspects of cable science. We need to test for settling times, i.e. discharge rates.

Settling time is precisely equal to frequency response. Frequency response (including phase) is the transform of impulse response, which is a superset of settling time. A long "settling time" in a minimumphase system means it has narrow bandwidth. In a non-minimum-phase system, it either means that, or it means there is some wild phase behavior going on.

 

[mep]

valkyrie-A very thought provoking response. I enjoyed it and you may well be onto something.

 

[valkyrie]

For the record

So - forgot to declare myself a subjectivist or an objectivist.

Really guys - I just don't know.

What I do know is that when I went to RMAF I had the occasion to listen to a Wilson (Sasha), ARC system playing some vinyl recordings of Simon and Garfunkel (music I knew well from my wild youth) - and my response was more or less to pick my jaw up from the floor - I have NEVER heard a sound reproduction system of such grace, such beauty, such incredible clarity and truthfulness in my entire life. I don't know how humans can conspire to make such an incredible reproduction system.

BUT - and it is a big butt - I have heard a great many other "high end" systems that were completely lifeless, dull, muffled, and just plain execrable - I mean a boom box would have been preferable. Included in this long list of uber-expensive "Duds" were a set of B&W 801s with a MacIntosh stack - horribly dull and lifeless - totally incoherent. A Martin-Logan set of Summits with some kind of expensive stack - thin, reedy and irritating. I heard the VTL room with Avalons at the same RMAF - dull, meaningless and non-communicative. And all of these were close - if not well over - $100,000.

Of course maybe it is about setup - maybe it is about the room - maybe it is about our internal mood at the time of hearing. But I really don't think so - I think the systems above just plain aren't up to the job.

I think that we - I hold a MS degree with over 30 years of actual engineering experience - are testing for the WRONG things. I believe that the aspects of a stereo system that make the most difference are NOT tested for - and further we don't even know HOW to test for them. We - engineers - tend to take the most prominent and easily understood aspects of a system and test those - and RLC values, while eminently testable have about as much meaning in terms of how a cable sounds as the barometric pressure that day.

Does such make me a subjectivist? Maybe - but at the same time I do NOT believe that any power cord makes one iota of difference in system sound. Neither do cable lifts, fuse replacement is just a raging absurdity - as are so many things in the so-called high end - Mpingo disks anybody? How about Tice clocks? Shakti stones? Or any of the other absolute nonsense about which our hobby is rife.

Our own Ethan Winer holds up a $20,000 power cord as an example of insanity - and I can only agree.

But you know what? Some "audiophile" systems - they are simply INCREDIBLE SOUNDING. They literally take your breath away.

and NO all systems do NOT sound the same and you cannot point to any absolute objective reason why they don't.

 

[valkyrie]

J_J,

I would suggest that your are "lumping" behavior. In a strictly mathematical sense, yes, variations in settling time should be observed as variations in the frequency domain. However such is a "gross" measurement. What if in fact the differences are on the order of 0.5 to 0.1 dB? And what if the spectra of these variations is rather narrow (say no more than 10 to 100 Hz). Will our "coarse" measurements of swept frequency response actually catch these deviations?

As for your comments in regards the transform of the impulse response - would you please go into a little more depth? I do not see, based on your single sentence, what it is you are trying to say (my failure more than likely). I DO understand the transform of the impulse response but not in regards an analog measurement - instead I understand such in terms of the Fourier Transform - multiplication in the frequency domain being the convolution in the time domain - and vice versa. But I am having a hard time translating your comments into a understanding on my end.

Thanks in advance.

 

[DonH50]

Long tails are also a problem when pole-splitting compensation is used and you don't exactly cancel the pole... Doublets are a huge problem in my (pulse) world.

 

[j_j]

J_J,

I would suggest that your are "lumping" behavior. In a strictly mathematical sense, yes, variations in settling time should be observed as variations in the frequency domain. However such is a "gross" measurement. What if in fact the differences are on the order of 0.5 to 0.1 dB? And what if the spectra of these variations is rather narrow (say no more than 10 to 100 Hz). Will our "coarse" measurements of swept frequency response actually catch these deviations?

If you did it right, yes. Of course, it's even easier to do this with an allpass sequence and deconvolution. Still, yes, it should and it will. There is no "gross" or "fine" here, if one shows the result, then the other should as well. If it doesn't, the measurement is bad.

Consider what a long tail means on a minimum-phase signal. It means, specifically, that the bandwidth of the system is narrower. The relationship df * dt >= c where df is the frequency resolution and dt is the time resolution is, as they say, not just a law, it's the math

(n.b. there the 'c' may change depending on the standard of resolution you use, but is typically expressed either as .5 or 1 for pure Gaussians at sigma, and is worse for other shapes of signal or frequency response)

So, unless there is non-minimum-phase behavior (like the fellow after commented on compensation warned about) a longer impulse response is a narrower system bandwidth, and the two had better show exactly the same results. If not, you either have a bad measurement or a nonlinear system, and if you show nonlinearity in a wire at line level or speaker level, something is wrong.

The key of catching variations, using the simplest deconvolution techniques (see www.aes.org/sections/pnw/ppt.htm for the FFT class, recordings, and freeware) is to have a signal at least twice as long in time as you want resolution in Hz. It's not hard to do this, and signals of length 32k or 64k are not that hard to create. (see Johnston/Smirnov AES paper or Fezjo/Johnston/Bartlett/etc AES paper for various ways to do it) So, I most often use a signal that gives me .25 Hz resolution, give or take a bit.

You seem like you can probably digest the entire FFT tutorial, I'd recommend it, because you can also grab your own impulse response measurements if you want, there.

Remember to make cable measurements in-situ, so you see the effects of source and load impedence!

 

[j_j]

Long tails are also a problem when pole-splitting compensation is used and you don't exactly cancel the pole... Doublets are a huge problem in my (pulse) world.

Yeah, you have a non-minimum-phase world, I suspect, there. Transmission?

 

[j_j]

and NO all systems do NOT sound the same and you cannot point to any absolute objective reason why they don't.

I would dispute the second half. I'm pretty sure one could see differences, of course, what part mattered would be tougher.

 

[valkyrie]

J_J,

Thanks for your patience - the light came on. For everybody, at least everybody who is interested, what J_J is referring to are the same "impulse" response measurements that John Atkinson takes in his speaker measurements for Stereophile reviews.

Here is an example



We can get an idea of what the frequency response of the speaker/crossover combination by the way this plot behaves. While the time domain is demonstrated the response looks like there are some rings - oscillations - in the way this "lumped network" responds to a "step", which by definition contains an infinite band of frequencies. In fact I think this response indicates that the drivers are not in identical phase - one is negative, one is positive.

However how would we measure this for a cable? The graph depicted is measured via FFT of the resultant sound from the speaker drivers (some kind of microphone/ADC/computer combination) - if we are going to do this for a cable - and hope to get the "cable" response I would think it would be swamped by the information from the speaker drivers and speaker crossover.

Any comments? Do any cable designers use this type of analysis?

 

[DonH50]

I don't think I have any impulse plots, but there are a lot of threads about cables that include time and frequency domain plots in the Tech Talk Forum. Here's a link to my ToC (there are other contributors to that Forum as well): http://www.whatsbestforum.com/showthread.php?2829-Don-s-Tech-Series

As for me, I find some things are much easier to see on a frequency plot, others on a time plot. Yes theoretically they provide the same information, but when looking for time-domain ripples and reflections it is usually easier to use a TDR, whilst bandwidth and phase information is way easier to get from a VNA. YMMV - Don

 

[cnpope]

I have pondered this very question - the audible effects of cable - and by implication the root of the subjectivist/objectivist debate for many years.

Recently I think I have come to anticipate what I believe it is we are hearing - and it is rather obvious if you reflect upon it. It is NOT RLC values. But it has to do with "settling time". Every L and every C has an associated discharge time - after all both inductors and capacitors are about "fields" being generated and discharged. I don't think any of us will argue that the substitution of a good film capacitor in a crossover network changes the sound of that network rather dramatically. Really - what is the film doing? It is changing the "settling time", i.e. the field discharge time of this storage device.

Similarly with cables - particularly speaker cables, less so with IC cables - a high current field propagates down a conductor and creates a substantial field - that much cannot be denied. That field then collapses, e.g. discharges, or dissipates back to the original conductor. This field is maintained both as a static and dynamic electric charge - for cables - ALL of them - have both capacitive and inductive qualities.

It seems to me that what you are alluding to is all part and parcel of what is taken into account in the standard theory of transmission lines, as described by the "Telegrapher's Equation." The parameters in the equation are the resistance, capacitance and inductance per unit length, and also the conductance per unit length (dielectric leakage conductivity, which is almost certainly completely negligible in the context of audio cables). So knowledge of R, C and L should be sufficient for all practical purposes.

Chris