Propeller Head Plaza

Technical and scientific discussion of amps, cables and other topics.

Return to Propeller Head Plaza


Message Sort: Post Order or Asylum Reverse Threaded

Objective Measurement Results for Cables

68.222.71.211

Posted on September 30, 2007 at 21:16:49
Jon Risch
Bored Member

Posts: 6460
Joined: April 4, 2000
Contributor
  Since:
March 1, 1999
This is bound to be a popular thread, but I wanted to pass along the info, and let folks see what was out there.

At one of the online design sites related to one of the many electronics industry magazines that I get, a portion of a new book (one chapter?) has been posted in 6 parts, and several of the portions have measurement results showing diferences between cables, as well as input/output differences.

The series starts here:
Part 1
http://www.planetanalog.com/showArticle.jhtml?articleID=201802376
http://www.planetanalog.com/features/multimedia/showArticle.jhtml?articleID=201802376
OR AT
http://www.audiodesignline.com/showArticle.jhtml?articleID=201801792

Part 2
http://www.planetanalog.com/showArticle.jhtml?articleID=201803751
OR AT
http://www.audiodesignline.com/showArticle.jhtml?articleID=201803044

Part 3
http://www.planetanalog.com/showArticle.jhtml?articleID=201804426
OR AT
http://www.audiodesignline.com/showArticle.jhtml?articleID=201803044

Part 4
http://www.planetanalog.com/showArticle.jhtml?articleID=201806067
OR AT
http://www.audiodesignline.com/showArticle.jhtml?articleID=201805860

Part 5
http://www.planetanalog.com/showArticle.jhtml?articleID=201807797
OR AT
http://www.audiodesignline.com/howto/showArticle.jhtml?articleID=201807390

Part 6
http://www.planetanalog.com/showArticle.jhtml?articleID=202102592
OR AT
http://www.audiodesignline.com/showArticle.jhtml?articleID=202101992

If you have any trouble with the exact URL's not getting you to the article, try using one of the author's names as a seach term at one site or the other, this should pull up the series in a list.
Authors: Philip Newell and Keith Holland

Book: "Loudspeakers: For Music Recording and Reproduction", Focal Press
The 6 part article appaears to be from Chapter 6 of the book, titled:
Effects of Amplifiers and Cables

They reference research by folks at Cerwin Vega, and BTW, none of the measurement graphs they show in part 5 (where the really interesting measurements are shown) are in the referenced AES papers, as I have copies of both of those, and they do not contain the cable measurement graphs.

I have been keeping an eye out for articles by Alexander Voishvillo, formerly with Cerwin Vega, now with JBL Professional, because he has been repeatedly referencing my AES paper on Multitones, and has done quite a bit of work on Multitones himself. The AES papers cited by Newell and Holland, "Multitone Testing of Sound System Components " Some Results and Conclusions, Part 1: History and Theory" and " .... Part 2: Modeling and Application" by the Cerwin Vega engineering staff (4 engineers are stated as authors, including CV founder, Eugene Czerwinski) are interesting reading, but do not contain the measurement graphs in the Newell and Holland article/book.

Apparently only the book does.

Anyway, I hope that folks will find this article interesting, if anyone gets a cleaner copy of the graphs, I would appreciate a set.

My own impresion of the work presented is that the multitones may not have been fully optimixzed to avoid distortion product cover-up, but rather, chosen for convenience and the generation of a known pattern of potential distortion products (see the referenced AES papers by the CV crew for more details). Thus, the amount and number of distortion products shown may actually be lower than those actually present, due to partial cover-up of some of them.

For instance one very simple choice could have reduced some of the covered-up tones by NOT chosing a simplistic range of ten-to-one for the start and end points of the logrithmically spaced test tones, this alone would avoid the 1 kHz and 10 kHz tones from generating a whole sequence of "covered-up" distortion products, while not stepping on as many of the other tones present. As I outline in my paper, if the upper bound had been chosen to be say, 11.618 kHz instead of 10 kHz even, then almost all of the unneccessary cover-up could have been avoided. Oh, well.

It also looks like it to me that the distortion is a result of an interaction between the amplifier and the cables and loudspeaker load.
Many folks have been for years advocating that amps and cables should be tested with real-world loudspeaker loads, myself included, so as to be as realistic as possible, and to insure that back-EMF issues were included. Newell and Holland have done so.

Finally, I want to point out that this article was primarilly written from the stand point of professional monitoring and use, where certain things are more valued than for home playback, and that this POV does skew the importance various things are given. Those concerned with SOTA home playback would undoubtedly place more emphasis on absolute quality and every last bit of sonic clarity, as opposed to as much on reliability or longevity issues.

Happy reading!


Jon Risch

 

Hide full thread outline!
    ...
RE: Objective Measurement Results for Cables, posted on October 1, 2007 at 02:59:01
I only had time to skim it but it does look like a fun article thanks for the reference.


 

RE: Objective Measurement Results for Cables, posted on October 1, 2007 at 07:27:15
Looks interesting. Unfortunately without controlled listening tests it's impossible to know which of the effects they demonstrate are audible, or even which are the most important. What we really need is a series of listening tests that gradually vary one parameter of the cables at a time, thereby establishing audibility thresholds.

 

RE: Objective Measurement Results for Cables, posted on October 1, 2007 at 08:37:56
> Unfortunately without controlled listening tests it's impossible to know
> which of the effects they demonstrate are audible, or even which are the
> most important.

Not at all if the measurements are sufficient. For example, if the authors labelled the y axis for the plots that curiously have the label "dB" but no numbers with, say, 1.0x10-6, 2.0x10-6,... would the effect of the different cables be above or below the threshold of audibility?

The point of science is to replace having to perform an experiment to find out something with scientific knowledge which can be used to predict the outcome of said experiment. Your statement implies this cannot be done for cables at audio frequencies and many people including myself would disagree for those cases where the size of the change is either well below or well above the expected threshold of audibility. When it is near the threshold then there may be a case for experimentation.

 

Thanks for the effort, I'll hop right on it... next week... nt, posted on October 1, 2007 at 10:38:41
clarkjohnsen
Reviewer

Posts: 26841
Location: Massachusetts
Joined: May 5, 2000
d

 

Would this not imply that there should be one best design that is demonstrably superior? nt, posted on October 1, 2007 at 11:17:06
Norm
Reviewer

Posts: 30992
Joined: September 6, 2000
a

 

RE: Would this not imply that there should be one best design that is demonstrably superior? nt, posted on October 1, 2007 at 11:31:28
No. It would imply that normal designs under normal conditions are all the same in causing inaudible modifications to the signal (excluding the obvious).

If you were using the sound for purposes other than listening or the conditions were abnormal then there may be a case for considering the cable performance further. Or if you were developing marketing for the cable based on its actual performance but, then again, perhaps that would not be wise.

 

Oh...like this has been done for other audio DBTs?, posted on October 1, 2007 at 11:54:28
mkuller
Audiophile

Posts: 29842
Location: SF Bay Area
Joined: April 22, 2003
>What we really need is a series of listening tests that gradually vary one parameter of the cables at a time, thereby *establishing audibility thresholds*.>

Part of a test validation, perhaps.

You know, like in REAL science.

Do you know of any tests like this which have been done with other audio components?

Say, amplifiers?

 

RE: Objective Measurement Results for Cables, posted on October 1, 2007 at 13:31:16
bjh
Audiophile

Posts: 15528
Location: Toronto, Canada
Joined: November 22, 2003
It would be a promising development to see the "multitone test signal" make available for replication purposes; actual measurements would appear to be trivial.

One interesting aspect of the graphs in Part 5, i.e. Figures 6.7 and 6.8 is that the right-hand plots, those taken at the load, would appear (at least by superfical appearance) to represent a net sum increase in energy compared to the plots taken at the source; I'm assuming the y-axis (SPL-db), can be construed as a measure of energy.

Assuming this is the case (difference plots would help) my only guess for the apparent paradox would be that the additional enegry is Back EMF that is attenuated by the cable (to account for the lesser energy at the source).

 

"Your statement implies this cannot be done for cables at audio frequencies...", posted on October 1, 2007 at 14:32:04
No it doesn't - why would it imply that?

I think your point is simply that these thresholds have already been established, at least to a certain extent. I agree with that - but as you also pointed out the authors of those papers don't seem to have made much of an effort to match the measurements they made to any kind of audibility criteria, and that makes them not very useful.

Personally I doubt that these differences are audible except when the cables are either very long, very exotic (meaning high L, R, or C), or exposed to a high degree of RFI, but maybe there are exceptions.

 

"Do you know of any tests like this which have been done with other audio components?", posted on October 1, 2007 at 14:33:46
I think it has been, at least partially - I know I've seen results for when THD in an amp becomes audible (several percent IIRC).

But I'm maybe missing the point of your post?

 

RE: "Your statement implies this cannot be done for cables at audio frequencies...", posted on October 1, 2007 at 14:57:54
> No it doesn't - why would it imply that?

Because you stated an experiment had to be performed to determine if the changes in cables was causing an audible change. I was trying to point out that a plot showing a variation of 0.000001 dB over the audible frequency range is sufficient to show an inaudible change in the signal.

> I agree with that - but as you also pointed out the authors of those
> papers don't seem to have made much of an effort to match the measurements
> they made to any kind of audibility criteria, and that makes them not very
> useful.

Yes but they are likely to be believers (I have not read the article yet) and this will be reflected in their words and how they go about things. So long as they provide honest data in a scientific manner this will not matter because they will be observing and reporting reality rather than just presenting their beliefs. Given their data one can compare it with standard audibility measurements from 50 years ago to see if it is well below, well above or somewhere near the threshold.

 

RE: Objective Measurement Results for Cables, posted on October 1, 2007 at 17:31:26
Jon Risch
Bored Member

Posts: 6460
Joined: April 4, 2000
Contributor
  Since:
March 1, 1999
Having some familiarity with interpreting multitone spectrum plots, I can shed some light on this.

The signal being measured is the electrical one, both at the amp and at the speaker end. The levels of the fundamentals of the test signal multitone (The ten frequencies that start at 100 Hz, and ending at 10 kHz with 8 more in the middle evenly spaced logarithmically, all of which are the ONLY lines shown that extend all the way to the top of the graph, and possibly beyond) would not change appreciably, energy that is at -40 dB down would only have the potential to change one of the fundamentals by approx. 0.1%, or approx. 1/10 of a dB.

What you noticed is the significant increase in distortion products, of which most are IM products rather than harmonics. If you look at fig.6.7, this shows the residual level of the various harmonics and IM products which are present with this particular test tone and spectrum analyzer, the pattern of which was predicted by modeling of that particular tone in one of the CV AES papers on multitones. The increase in distortion due to the cable and loudspeaker interacting with the amplifier, is the INCREASE above the levels in Fig. 6.7 that is shown in Fig. 6.8 and 6.9

What is significant, is that even right at the power amp, distortion is much higher with a real world loudspeaker load compared to a simple resistive load.

bjh wrote:
"Assuming this is the case (difference plots would help) my only guess for the apparent paradox would be that the additional enegry is Back EMF that is attenuated by the cable (to account for the lesser energy at the source). "

This additional distortion energy is the combination of back EMF and acoustically generated inter-driver and inter-speaker system crosstalk.

What is not clear in the Newell and Holland web article, and no information is given, is if the loudspeaker load was in isolation, that is, whether or not it had a stereo companion, or was being exposed to the test signal by itself, with no other loudspeakers playing nearby.

As I show at my web site, at:
http://www.geocities.com/jonrisch/biwiring3.htm
Fig. A, B and C ,
When another speaker is also playing, then a direct generated voltage is present, and this voltage ALSO gets injected into the amplifier outputs via the speaker cable. Since the other loudspeaker would be somewhere on the order of 6 to 12 feet away, the acoustic output will not reach the other loudspeaker till approx. 5.3 to 12.6 ms later, thus there is a significant time delay for the external stimulus being generated and reaching the power amp.

It is one thing for the distortion to be occurring at the same time as the original signal, this is bad enough, but with masking and various other aspects, it makes it harder to notice the distortion when the original tones are playing too. But when the distortion is a full 5 to 13 ms AFTER the signal, which may have ceased sounding for a moment, the distortion generated from the output of the other speaker will stand naked for that brief moment, making it MUCH more audible than when the original exciting signal was present.

This very aspect is what makes it ludicrous to try and simply look at the "thresholds of audibility" and declare something either hearable or not.

Of course, most objectivists like to limit the situation to load resistors and a silent room, but we don't listen to music in a vacuum with resistors.

In any case, it is highly likely that the measurements showing the multitone distortion represent just ONE loudspeaker being driven at a time; add in a stereo partner, or multiple surround sound and center channel speakers, and you have even more potential sources of intermodulation and increased levels of distortion.

Thus these measurements may not fully represent the complete and full extent of just how much distortion is actually present, it is likely that the actual distortion would be even greater than that shown.

Jon Risch

 

"This very aspect is what makes it ludicrous to try and simply look at the "thresholds of audibility"..., posted on October 1, 2007 at 18:08:32
"This very aspect is what makes it ludicrous to try and simply look at the "thresholds of audibility" and declare something either hearable or not."

You mean because the audibility threshold depends on the material being played?

We're used to that - absolute polarity, for example, is relatively easy to hear with test tones, but there is only a single blind test result I know of showing its audibility for music (and that was a very particular passage involving mostly percussion). Same thing with many other such effects. A well designed test takes that into account by playing a variety of musical material as well as test tones.

It's annoying, but that's life. There are no easy answers.

 

RE: Objective Measurement Results for Cables, posted on October 1, 2007 at 18:49:16
Jon Risch
Bored Member

Posts: 6460
Joined: April 4, 2000
Contributor
  Since:
March 1, 1999
andy19191 wrote:
"Not at all if the measurements are sufficient. For example, if the authors labelled the y axis for the plots that curiously have the label "dB" but no numbers with, say, 1.0x10-6, 2.0x10-6,... "

As far as I can see, the graphs are labeled, with a visible range of 50 dB, and the level of the distortion products showing as high as "82 dB" or so.

After looking at the original AES article from CV engineers with the test signal modeled there, as well as actual measurements of some drivers in that same paper, the levels of the distortion products shown in the Newell and Howard paper implies a level that would have the primary tones at a level of 120-130 dB on that scale, meaning that the higher level distortion products are at -38 to -48 dB. My experience with reading and interpreting multitones, as well as the calculated modeling and actual measurements provided by the CV engineers provides a pretty solid base for this kind of assessment.

Note that many of these distortion products would be very aharmonic IM products, very plainly audible because they are not part of a harmonic series, but rather, at odd and unharmonicaly related frequencies. There are also more than just one of them, over a dozen clustered at levels within 10 dB of the highest ones. A veritable cacophony of odd ball distortion tones much harder to ignore than a simple 2nd or 3rd harmonic distortion product.


Jon Risch

 

Is it not possible that other measures are needed especially if people hear a difference? nt, posted on October 1, 2007 at 19:38:48
Norm
Reviewer

Posts: 30992
Joined: September 6, 2000
a

 

RE: Is it not possible that other measures are needed especially if people hear a difference? nt, posted on October 1, 2007 at 23:58:24
If people are perceiving a difference due to what is going on between the ears rather than impinging on the ears then it would depend on what you are trying to achieve. Many people involved around here are trying to sell/believe in magic cables and so the answer is obviously yes. A few are only interested in getting cables good enough to avoid audibly distorting what is impinging on the ears and so if this is small enough then no.

 

RE: Objective Measurement Results for Cables, posted on October 2, 2007 at 01:58:00
> As far as I can see, the graphs are labeled, with a visible range of 50
> dB, and the level of the distortion products showing as high as "82 dB" or
> so.

I was referring to the later unlabelled graphs and in the context of my post the point was that measured minute differences are inaudible and the text was not overly reliable.

Concerning the figures you are referring to, I cannot offer anything without knowing what was measured, how and under what conditions. Without this one cannot work what is going on in terms of basic physics and hence determine if some magical property of cables has been discovered, a measurement misinterpretation/mistake revealed or whatever.

Anyway, if the CV engineers have provided a model for the effect then the physics is presumably understood. What is it?

 

What figure number (from the article) are you referring to? nt, posted on October 2, 2007 at 05:36:24
bjh
Audiophile

Posts: 15528
Location: Toronto, Canada
Joined: November 22, 2003
.

 

The key is "if people are imaging differences." What if they are not?, posted on October 2, 2007 at 06:16:50
Norm
Reviewer

Posts: 30992
Joined: September 6, 2000
I doubt seriously whether measurements cover all attributes of cables. If all the measurements cover are inaudible, I am certainly not convinced. I think science would advance more were studies to focus on finding the measurable characteristics of cables that most people prefer versus those they dislike.

 

RE: The key is "if people are imaging differences." What if they are not?, posted on October 2, 2007 at 07:27:50
> The key is "if people are imaging differences." What if they are not?

Who said they were imagining differences? Go look at the McGurk effect video. Is that your imagination or is that a real effect caused by your brain? Now map that onto the situation for cables.

 

RE: Objective Measurement Results for Cables, posted on October 2, 2007 at 09:13:43
morricab
Audiophile

Posts: 6548
Location: switzerland
Joined: April 1, 2005
"What is significant, is that even right at the power amp, distortion is much higher with a real world loudspeaker load compared to a simple resistive load.
"

Back EMF from the speaker Jon? Now if this is being fedback into the amp and the amp has a high amount of negative feedback then you could possibly get this signal reamplified and who knows what you get then. I believe this kind of potential distortion was first pointed out by Otala in the late 70s.

 

RE: Would this not imply that there should be one best design that is demonstrably superior? nt, posted on October 2, 2007 at 09:15:11
morricab
Audiophile

Posts: 6548
Location: switzerland
Joined: April 1, 2005
"If you were using the sound for purposes other than listening or the conditions were abnormal then there may be a case for considering the cable performance further"

And you base this conclusion on what?? Do you know the relevant thresholds for hearing these measureable differences??

 

Guess THD is the only validated audible amp difference then...(nt), posted on October 2, 2007 at 10:50:31
mkuller
Audiophile

Posts: 29842
Location: SF Bay Area
Joined: April 22, 2003
(nt)

 

Answer really wanted..., posted on October 2, 2007 at 17:06:06
Jon Risch
Bored Member

Posts: 6460
Joined: April 4, 2000
Contributor
  Since:
March 1, 1999
...or is this an 'a priori' rejection?

andy19191 wrote:
"I was referring to the later unlabelled graphs and in the context of my post the point was that measured minute differences are inaudible and the text was not overly reliable."

Which unlabeled graphs?

Minute differences? I wouldn't think that distortion that is at -38 to -48 dB is 'minute', especially since there are more than one distortion product showing at or near those levels. If there were just 4 products at that level, then the total distortion would be in the range of approx. -32 to -42 dB, hardly something that one could call 'minute' or wish away as if it were insignificant.

And what in the world does "text not overly reliable" mean? Is this just another way of saying "I don't believe these people, and refuse to even entertain their data or arguments? How scientific is that?

andy19191 wrote:
"Anyway, if the CV engineers have provided a model for the effect then the physics is presumably understood. What is it?"

They modeled the amount of distortion that would occur with a particular multitone pattern due to the air and due to the horn placed on the driver (not including the driver's own distortions), and then compared that to an actual measurement of a typical driver, and an experimental one they were working on. Other work has modeled what the distortion pattern would look like with a certain (very low) amount of distortion, such as with a power amp. Finally, the measurements in the Newell and Howard paper show what the distortion levels are for a load resistor, and this provides a residual distortion level that is very similar to that of the previously modeled low amount of distortion.

No one has yet proposed a model for the distortion created by the amplifier/cable/loudspeaker combinations (that I am aware of). I would suspect that some one at CV or JBL might be working on it, and I have a few ideas myself, as well as having tried taking my own measurements of similar and related conditions, and seeing levels of distortion higher than load resistors and very short lengths of cable/wire.

Keep in mind that the length of the speaker cable under test in the Newell and Howard paper is 6M, or almost 20 feet.


Jon Risch

 

RE: Answer really wanted..., posted on October 4, 2007 at 05:09:42
> Answer really wanted or is this an 'a priori' rejection?

I have an interest in how the performance of audiophile devices are presented. It is not a passionate one but it is real enough to read the odd paper when time permits.

> Which unlabeled graphs?

The unlabelled ones later in the article.

> And what in the world does "text not overly reliable" mean?

It means trying to make the data fit the belief rather than the other way round. Audiophiles performing experiments almost always do this and, so long as they honest, it usually does no harm in the scientific sense.

> They modeled the amount of distortion that would occur with a particular
> multitone pattern due to the air and due to the horn placed on the
> driver[...]

The modelling you describe does not include a model for the cable?

 

Another imaginative Objectivist!, posted on October 4, 2007 at 14:28:12
bjh
Audiophile

Posts: 15528
Location: Toronto, Canada
Joined: November 22, 2003
RBG with his imagined DBTs, you with your unlabelled graphs... you fellas are so creative!

 

Page processed in 0.023 seconds.