Zilch's AK Design Collaborative - Econowave Speaker

[tubino]

Y'see, if we can measure it, it's science and we can understand it. If we can't, it's voodoo, like Monster Cables at $100 per foot. Trust me; it works! Just hand over your cash and sign here.

Tubino: I've read it several times. Greg Timbers saying, "Trust me".

I don't think you're being fair here. No one disputes that the CC battery changes the potential of the caps so that the AC signal (within p-p range) will never drop below positive. If you want evidence that there is a quantifiable difference in caps wrt signal behavior between signals crossing zero and signals that don't, fine, but it's not a JBL problem or an audio problem, really. It's a signal processing problem that has surely been addressed to the nth degree by academic engineers and industry engineers, and published in engineering journals. (I work in an academic environment, so I would look in searchable databases of refereed journals.) A reference librarian in a good library could track down articles in minutes.

After which you would be able to explain how, when, where there is a difference in signal behavior.

At which point you still don't know the answer to the question, does it matter in speaker crossovers? Is the difference audible, and if so, in what way?

Timbers doesn't say "Trust me." He says, this is what people report hearing. And in forums he says, try it.

That's honesty, not snake oil. You're welcome to do the homework on published articles and report back, and I may do the same, out of curiosity. But no amount of refereed articles describing the electrical phenomena will answer the question, is it audible?

CLARIFICATION: there are plenty of studies showing that humans can identify frequency response differences of as little as +/- 1db (and that's trivial compared to the pitch/tone feedback of a world-class violinist), so there is NO DOUBT that the FR graphs posted by Zilch and others correlate to subjective experience. I just wanted to make clear that I'm completely on board with using measurements to improve sound. BUT I'm very skeptical that we currently have an exhaustive list of all the measurements that matter.

 

[Bauhausler]

IIRC, one (wish I had bookmarked it) of the technical analyses of distortion in CC crossovers stated that adding bias slightly INCREASED the IMD. They also said that the IMD was not causing the audible improvement from CC.
This is a red flag for me when someone says that the only measured effect is detrimental yet the net change is an audible improvement. What were you measuring? Why? What did you expect to find?
I'll get behind this and push if I have a little evidence that it's going somewhere worthwhile.

 

[tubino]

Surely that is testable. I sold my scope many moons ago, when I stopped designing (digital) circuits, but I may just go on the 'Bay and pick one up to settle this. It shouldn't be that tough.

A month in the laboratory can often save an hour in the library. (F.H. Westheimer)

 

[Robh3606]

Originally Posted by Paratima
Y'see, if we can measure it, it's science and we can understand it. If we can't, it's voodoo, like Monster Cables at $100 per foot. Trust me; it works! Just hand over your cash and sign here.

Tubino: I've read it several times. Greg Timbers saying, "Trust me".

They are not trying to sell you anything JBL doesn't sell these networks. They just use them in their TOTL systems. They are telling you why they use CC crossovers in their $60K speakers.

So don't try it. Your loss.

Rob

 

[jackgiff]

I really don't know if charge coupled crossovers sound better or not. They probably do or why would JBL spend the extra bucks to build them? On the other hand, they are not used in any low end speakers that I know of, just the TOTL guys that sell for some massive bucks. Could be just snake oil.

There are a couple problems with the explanations offered by Greg Timbers, assuming that he actually wrote the descriptions offered in Tubino's post about CC networks.

On page 12, the third paragraph, where biasing above zero crossing is touted, the paper states:

"By biasing the capacitors so that the signal does not reverse current flow through the capacitor, zero crossing distortion is eliminated."

How do they get any signal through the capacitor if the current doesn't reverse? With no signal present. the caps are charged to the battery potential, and no current flows through them. For the caps to pass an AC signal, the current through them must change directions. Current into the caps would present a change in output voltage in one direction, and current out of the caps would present an opposite change. The voltage across the caps may not pass through zero, but to conduct an AC signal, the current better reverse it's flow. Did Timbers rewrite the basic laws of physics?

In the first paragraph on the right hand side of page 11 it states:

Capacitors do not exhibit hysteresis, per se, but they have a behavior with similar results, called dielectric absorption. When an audio signal passing through a capacitor changes polarity, that is, it passes through zero volts, the current flow changes direction. But it does not do so immediately: due to dielectric absorption, the capacitor "remembers" it's previous state and resists change. This leads to blurring of transients in the drive signal, a highly audible distortion.

Dielectric absorption doesn't just happen as a capacitors voltage passes through zero. It happens any time the voltage across the capacitor is changed. The dielectric properties that cause dielectric absorption are always there, and are more prevalent in different dielectrics. Mylar is a terrible dielectric to use in an analog sample and hold, because of it's high dielectric absorption. Polystyrene was among the best dielectrics for building S & H's when I worked in the data conversion field many years ago. That is probably still so. If dielectric absorption was the culprit causing distortion in crossovers, changing to the proper dielectric would have much more effect in limiting the distortion than would charge coupling. But 10 uf capacitors made with polystyrene would be huge. Dielectric absorption is always taking effect as the audio signal changes the voltage on the capacitors, not just at zero crossover.

And the papers claim that biasing the caps is analagous to the difference between Class A and Class B amplifier performance leaves me shaking my head.

Perhaps Greg Timbers really did a great thing when he developed the Charge Coupled concept, but it is not clear from these white papers that he really understands why.

 

[Bauhausler]

They are not trying to sell you anything JBL doesn't sell these networks. They just use them in their TOTL systems. They are telling you why they use CC crossovers in their $60K speakers.

So don't try it. Your loss.

Rob

Loss? maybe and maybe not. Not all applications are the same. The Ewave is not a K2.S5500. The crossovers are different in several details and the drivers are completely different. What is significant for one may not be for another.
To put it another way, if you have a Ferrari you'd better be putting premium gas in it because that's what it's designed for. My Honda derives no benefit from octane over the 87 it was designed for.
Technology that is essential for one application may have no effect in a different situation.

 

[tubino]

I just sent an email to a friend who teaches EE at our university. I may also contact the librarian at our engineering library to find published articles on capacitor behavior at zero crossing.

Folks, there are a lot of trained engineers in academia and industry with SOTA equipment who spend a LOT of time answering questions like this in detail for a very wide range of frequencies and signal amplitudes, and they only get their results published after review by other engineers. Talented though you may be, you really don't HAVE to play scientist to learn what's going on, and you don't have to limit yourself to short non-refereed 20-year-old papers either.

The hardest part of finding stuff like this in journals is that most academics -- even engineers -- consider these questions too uninteresting (basic). But industry journals might cover it. I'll try to get back with more data, less wank.

 

[Zilch]

Measurements posted on the Klipsch forum clearly show the benefit of biasing NPEs, AND it's a fun read:

http://forums.klipsch.com/forums/t/99466.aspx?PageIndex=1

[Well, for objectivist speakergeeks, it's fun.... :tongue: ]

 

[skywave-rider]

...

The hardest part of finding stuff like this in journals is that most academics -- even engineers -- consider these questions too uninteresting (basic). But industry journals might cover it. I'll try to get back with more data, less wank.

:thmbsp:
The thing about Econowave is that we can, or want to, or at least try to understand what's going on. We're not manufacturing companies, nor engineering institutions. To state the obvious.
We're speaker students. Not a trivial pursuit!

I feel lucky to have access to all your ideas.
Pour it on.

 

[dnewma04]

The problem is that a capacitor has a "memory". The dielectric material doesn't allow the signal to smoothly pass through 0 volts. When a cap is charged in a + direction the dielectric is stable. As the voltage goes down it discharges until you get to 0. At 0 there is a sticky point where it has to start recharging. Then you go in the - direction. The same will happen as you cross 0 going from - to +.

It's somewhat like static friction in a sliding joint. You have to overcome that threshhold before the joint will slide smoothly. Think of pushing a car. It's hard to get it going, but once it's moving it's not so hard to keep it going. It's the same type of effect. If you were going to vary the speed of a car you were pushing by 1mph peak to peak it would be easier and smoother to use 2mph as your reference and vary the speed between 1 and 3mph than it would to start from a stop, go up to 1mph, stop, pull backwards to 1mph, etc.

That's the same principal as in the CC crossover. You're keeping the capacitors from that 0 crossing where they stall.

Did that help any?

Ray

Perfect. You are the layman's term analogy master.

 

[jackgiff]

Measurements posted on the Klipsch forum clearly show the benefit of biasing NPEs, AND it's a fun read:

http://forums.klipsch.com/forums/t/99466.aspx?PageIndex=1

[Well, for objectivist speakergeeks, it's fun.... :tongue: ]

It was a fun read, but raises numerous questions. Since the circuits used weren't shown and the settings of the scope weren't detailed, who can tell if it is relevant. At least the polystyrene was shown to be an excellent dielectric.

Maybe more wank, less data?

 

[Steve O]

I really don't know if charge coupled crossovers sound better or not.

<snip>

Perhaps Greg Timbers really did a great thing when he developed the Charge Coupled concept, but it is not clear from these white papers that he really understands why.

I've read the linked Timbers white papers a couple of times and agree with your general assessment. As you point out, the paper contains many technical non-sequiturs and misapplied concepts. Probably the result of writing for the intended, non-technical audience. Ultimately the whole thing strikes me as a case of creating a product feature for the sake of market differentiation. The feature might have some basis in fact but in this app, it's marginal at best and benign at worst so why not?

Based on Tubino's recent post, it appears that we're going to get some academic input on the subject. Won't be surprised if there is indeed a well documented, detectable, dielectric crossover effect. I doubt that the audibility of the effect will be addressed so we'll be left where we are now WRT its significance in speaker xovers. Ultimately I’m left wondering why, if the effect is real and significant, the best solution isn’t development (or use) of a better cap.

 

[Kegger]

"the best solution isn’t development of a better cap"

No a better battery..

 

[Bauhausler]

The great and very readable Steve Bench wrote on the distortion and 'sound' of caps:

http://greygum.net/sbench/sbench102/caps.html

(EDIT)

OOPS. I was reading the thread on the Klipsch site and hadn't gotten to the post of Steve's capacitor test photos. My link includes Steve's accompanying text. The poster on the Klipsch forum really should not have taken those out of context. Read the original. It's worth it.

(EDIT AGAIN)

Very interesting pix from Steve showing biased electrolytics:

http://greygum.net/sbench/sbench102/caps1.html

 

[tubino]

IWon't be surprised if there is indeed a well documented, detectable, dielectric crossover effect. I doubt that the audibility of the effect will be addressed so we'll be left where we are now WRT its significance in speaker xovers. Ultimately I’m left wondering why, if the effect is real and significant, the best solution isn’t development (or use) of a better cap.

Maybe there is a formula to express time/phase delay at zero crossover as a function of dielectric absorption. Then we can get down to business: arguing about whether you can hear it or not.

 

[Steve O]

It was a fun read, but raises numerous questions. Since the circuits used weren't shown and the settings of the scope weren't detailed, who can tell if it is relevant. At least the polystyrene was shown to be an excellent dielectric.

Maybe more wank, less data?

As noted at some point in the linked discussion, the scope images and basis for the discussion are from Steve Bench's evaluation of caps done quite some time ago. Unfortunately, (IMO of course), the test results are usually taken out of context. He originally conducted the tests as a result of being asked to evaluate some paper-in-oil caps, apparently for a client. He found the PIOs sounded so good he had to find out why...thus the tests.

Although not clearly stated in his test set-up description, I believe what he's really measuring is the voltage coef of cap. I have no idea of the significance of this parameter to good sound.

Curiously, while as you note, polystyrene comes up good, the PIO caps he's being asked to evaluate come up best. Not an intuitively obvious result IMO. And potentially misleading since there are many other factors besides voltage coef of cap that may be factors in cap goodness.

 

[Paratima]

A month in the laboratory can often save an hour in the library. (F.H. Westheimer)

Tubino: I assure you I know that quote very well! :thmbsp:

...and I've often been on both ends of that particular stick.

 

[Paratima]

So don't try it. Your loss. Rob

You didn't read the bit where I said I almost undoubtedly will try building it?

Do you think I really want to buy a scope?

 

[Bauhausler]

As noted at some point in the linked discussion, the scope images and basis for the discussion are from Steve Bench's evaluation of caps done quite some time ago. Unfortunately, (IMO of course), the test results are usually taken out of context. He originally conducted the tests as a result of being asked to evaluate some paper-in-oil caps, apparently for a client. He found the PIOs sounded so good he had to find out why...thus the tests.

Although not clearly stated in his test set-up description, I believe what he's really measuring is the voltage coef of cap. I have no idea of the significance of this parameter to good sound.

Curiously, while as you note, polystyrene comes up good, the PIO caps he's being asked to evaluate come up best. Not an intuitively obvious result IMO. And potentially misleading since there are many other factors besides voltage coef of cap that may be factors in cap goodness.

My last post on teh previous page now has links to Steve's full articles, with descriptions of the test setup.

 

[jackgiff]

The problem is that a capacitor has a "memory". The dielectric material doesn't allow the signal to smoothly pass through 0 volts. When a cap is charged in a + direction the dielectric is stable. As the voltage goes down it discharges until you get to 0. At 0 there is a sticky point where it has to start recharging. Then you go in the - direction. The same will happen as you cross 0 going from - to +.

It's somewhat like static friction in a sliding joint. You have to overcome that threshhold before the joint will slide smoothly. Think of pushing a car. It's hard to get it going, but once it's moving it's not so hard to keep it going. It's the same type of effect. If you were going to vary the speed of a car you were pushing by 1mph peak to peak it would be easier and smoother to use 2mph as your reference and vary the speed between 1 and 3mph than it would to start from a stop, go up to 1mph, stop, pull backwards to 1mph, etc.

That's the same principal as in the CC crossover. You're keeping the capacitors from that 0 crossing where they stall.

Did that help any?

Ray

Ray, it is much more complex than you describe it. If your description was true, all an analog designer would need to do to build an accurate Sample and Hold would be to bias his holding capacitor, and all would be well. Tain't that simple. Dielectric absorption is not voltage level sensitive, it is always present, regardless of voltage level. Its effect at positive or negative voltage is the same as it is at 0 volts. It doesn't "free itself up" once through the "threshold." Biasing the caps is not the answer if the problem is dielectric absorption.

Maybe there is a formula to express time/phase delay at zero crossover as a function of dielectric absorption. Then we can get down to business: arguing about whether you can hear it or not.

tubino, passing an AC signal through a capacitor requires the current to reverse it's flow. Therefore, we need to differentiate between zero voltage crossing, and zero current crossing. Biasing the capacitors can only affect the zero voltage crossing, not the zero current crossing.

I think the entire charge coupled concept reeks of being something that sounds too good to be true. Like $2,000 speaker cables. Maybe, Marketing 101?