or Connect
AVS › AVS Forum › Display Devices › Ultra Hi-End HT Gear ($20,000+) › Digital Active Speaker Thread.....
New Posts  All Forums:Forum Nav:

Digital Active Speaker Thread..... - Page 8  

post #211 of 481
Thread Starter 
Quote:
Originally Posted by John Kotches View Post

An apodizing filter has absolutely nothing to do with crossovers. It's used to remove artifacts of the initial process -- which causes pre and post echo in the impulse response. Removing the pre-echo results in a much more natural sound.

I was just wondering out loud if it could be applicable in the future.
Quote:

As far as digital crossovers, M uses a 48dB per octave crossover, and the only artifacts are 130+ dB below the main level signal. That's literally inaudible.

Didnt realize they'd switched. I remember the older ones to be at 24dB/octave. I wonder if that is why the DSP7200s sounded notably cleaner and more precise to me.
post #212 of 481
Thread Starter 
Quote:
Originally Posted by Chuck V View Post

I would say that is not nearly as simple as that statement.

It's true however, IF one listen to ONE speaker in anechoic chamber on axis (no stereo effect involved). If listening is done in a real room, one can not just look at a FR-graph and say A is more tone accurate then B. At least that's not my belief.

Well, the speaker's job is to be accurate. You do not want a perfectly flat response at the listening position or it will sound thin and aggressive.
post #213 of 481
Quote:
Originally Posted by Alimentall View Post

Power compression in the low midrange won't give you fatigue.

Probably so. After all, if the volume at a certain frequency range is 5dB of more below where it should be, that part of the frequency range isn't likely to be fatiguing. You call that an advantage?

You never cease to amaze me. You sing the praises of the Xd's FR, but then practically ignore its power compression. Anyone with half a brain will realize that significant power compression means that the FR is far from flat when the system is played loud. And it is far from flat on the louder passages / transients when the system is played at more moderate levels. I understand that with music, short-term peaks of 15-17dB above the average level of a recording are pretty common. So even if one is listening at an average level of 90dB, the music is calling for peaks of 105dB. At 90dB, the Xd already exhibits compression. At 95dB, it is much worse. One wonders how bad it gets at 100dB and 105dB.

Difference @ 95dB, 50Hz - 20kHz (measured @ 2m)

Curve: difference from 70dB at 95dB

So, if the Xd were perfectly flat at 70dB, the curve shows the FR at 95dB. Again, one can only imagine how bad it gets at 100dB or 105dB. Pretty sad above 100Hz, I'd say, and unusually bad in that frequency range for a speaker in the price range of the Xd (even the new, inventory-liquidation price). Oh, and if the listening distance is more than 2m? Even worse power compression.

This post isn't really about the merits of an active crossover approach. But since many of your posts are really about how great the Xd is, please, John, stop making the Xd out to be a giant-slayer. People on AVS are mostly smarter than that.
post #214 of 481
Thread Starter 
Quote:
Originally Posted by syswei View Post

Probably so. After all, if the volume at a certain frequency range is 5dB of more below where it should be, that part of the frequency range isn't likely to be fatiguing. You call that an advantage?

No need to be sarcastic. Subjectively, it is an error of omission, rather than the additive errors on most speakers so it is extremely benign. Yes, you can hear it, but it's relatively subtle. It is also improved with the updated crossover. If it were up to me, I would have used a tower format with multiple 8" woofers and a 300Hz or so crossover. Problem solved. Still, when you look at the overall picture, it still does many things better than very expensive speakers.

I don't know why people insist on putting the words "giant killer" or "giant slayer" in my mouth. It shows poor reasoning and logic skills to do so, aside from the fact that this was not my claim.
post #215 of 481
Quote:
Originally Posted by Alimentall View Post

I don't know why people insist on putting the words "giant killer" or "giant slayer" in my mouth

hmmm...

Quote:
Originally Posted by Alimentall View Post

the Xds pretty well trounced anything I've compared it to in the $15K-$25K range

Quote:
Originally Posted by Alimentall View Post

If you ever wanted the sound that $20K+ speakers aspire to have for next to nothing, a used set of Xds is the best speaker bargain on the planet, bar none.

Quote:
Originally Posted by Alimentall View Post

NHT's Xd is still my favorite all around speaker. I've heard nothing at any price I'd rather have.

Quote:
Originally Posted by Alimentall View Post

I don't sell Xd anymore, but they're still one of the best speakers you can get at any price.

Quote:
Originally Posted by Alimentall View Post

I can tell you that Xds are, IMO and measurably, better than Thiel CS6s, Genesis 5.1s, B&W 800D/801D/802D, Meridian DSP5500s/6000s, Vandersteen 5As, Focal Electra Bes and Linn Espek actively triamped speakers off the top of my head. Maybe that isn't big leagues to you, but I would say it is.
post #216 of 481
Thread Starter 
I don't consider B&Ws or Thiels or any of the speakers that they handily beat (measurably, subjectively) to be 'giants'. The industry may. I consider them to be flawed, overpriced, even 'mass-market' speakers. But it is what people think of as 'the big leagues', at least in reputation. The Rockport Arrakis probably qualifies as a 'giant'. Xd won't kill them, though it will still do a few things better. In order to be a 'giant' killer, first you have to find a true giant, not a Napoleon, then you have to match it or beat it in every important way. Xd can't do that to the best speakers. HOWEVER, it does do many things better and the areas where they get beat aren't that important to me. It still *IS* my personal favorite speaker and I *do* think it is one of the best speakers available at any price, though I do think they could easily be improved upon with another DEQXed design and I think I would likely replace them with the Tikandi at some point if nothing cooler comes out.
post #217 of 481
Quote:
Originally Posted by Alimentall View Post

The Rockport Arrakis probably qualifies as a 'giant'. Xd won't kill them, though it will still do a few things better.

If you haven't measured the Rockport Arrakis, how can you say the Xds will do a few things better? Perhaps you mean the Xds theoretically might measure better in a few respects. Who cares what a discontinued speaker theoretically might do.
post #218 of 481
Thread Starter 
Quote:
Originally Posted by faberryman View Post

If you haven't measured the Rockport Arrakis, how can you say the Xds will do a few things better? Perhaps you mean the Xds theoretically might measure better in a few respects. Who cares what a discontinued speaker theoretically might do.

A lot of people seem to care and even seem aggravated about it. As in "how DARE they measure better in *any* way!". One look at the design and you know they won't do vertical dispersion well. Does it matter? Maybe not, especially for their intended 'giant room' audience. But *subjectively* haven't heard a D'Appolito design that I like, so I will always seek a single midrange or 4-way design with a separate midbass and and an upper midrange, especially for midfield listening.
post #219 of 481
Quote:
Originally Posted by Alimentall View Post

A lot of people seem to care and even seem aggravated about it. As in "how DARE they measure better in *any* way!".

There isn't a single member of AVS that cares about it or is aggravated. You're tilting at windmills.
post #220 of 481
Thread Starter 
Then how come I always seem to be the windmill? I refuse to change my opinions based on years of experience, just to 'play nice' with people who are intolerant of the opinions of others. Besides, it has little to do with the Xd itself. It is a proof of the DSP concept, what it can do, even with a low budget solution, and a window on the future. "A harbinger of things to come", I think is what Stereophile repeatedly called it.
post #221 of 481
Quote:
Originally Posted by Alimentall View Post

Then how come I always seem to be the windmill?

You'd have to ask your therapist.
post #222 of 481
Thread Starter 
Quote:
Originally Posted by faberryman View Post

You'd have to ask your therapist.

At least I have knowledge to contribute.
post #223 of 481
Thread Starter 
Speaking of which, as Syswei pointed out with the compression issue, even with steep crossovers and impulse response correction, you really have to stretch things to get the best performance. Every driver has limits. The best ones have an octave, maybe two where they behave in a nearly ideal fashion, with wide, smooth dispersion, low distortion and low cone resonance.

If you look at the SEAS W15, it is flat and smooth as can be, textbook so, from about 125Hz to 500Hz. Two octaves. But you have to cross it over higher if you want to remove distortion and/or compression. 250Hz, really (NHT pushes it to 110Hz-150Hz depending on filter, I believe the Tikandi crosses it a bit higher, like 180Hz if I remember correctly). Fortunately, distortion and compression are somewhat benign, so engineers often choose that over other factors like cone resonance. So, looking up in the range, from about 500Hz to 4000Hz, it is a bit inaccurate with an ~5dB bell curve thing happening around 2kHz. This, fortunately, is easily fixed by the impulse response correction in DEQX, so it's not a notable issue. However, once you get to 2kHz, dispersion narrows. This is common in most speakers near the tweeter crossover because a low pass crossover doesn't stop high frequencies, it just progressively lowers them. So a 12dB/octave 2000Hz crossover has tons of treble (and associated resonances/narrowing of dispersion) coming from the midrange . But if you cut off the signal steeply enough, you can push the crossover to 2kHz and retain very wide dispersion.

So, that leaves us with really 2 octaves of excellence from this driver. NHT pushes it for another octave and a half, for better or worse, out of necessity for this design and for the limits of a 3 band speaker. Legend moves both crossovers higher because it is a tower configuration and they know where the woofers are. The treble crossover is 2.6kHz. So, again, about 3.5, almost 4 octaves. Either way, because of some pretty horrific driver behavior starting at about 5kHz and peaking an 8kHz, you want to cut this off sharply so that *no* energy gets to the speaker at these frequencies. Both Legend and NHT do this, the secondary result being virtually no acoustic interference between the drivers. In both cases, the tweeter has to handle 3-3.5 octaves, which means they will both begin to narrow in dispersion by 10kHz. Of course, this is just accepted with most any design, even exotic ones. A smaller tweeter would do better, but would have to be crossed even higher. Legend is says that it gets better overall performance at a higher crossover point, allowing some narrowing of the dispersion, assumedly for better tweeter performance.

In addition, the woofers are flat above 50Hz, but need to be EQed below that to remain flat. Most speakers just peak and then rolloff, but unless you have a big cabinet and big drivers, you can't do this that easily without some EQ.

So Xd has some small compromises in the first octave, the third/fourth octaves and the very top octave, but is nearly ideal everywhere else. Imagine how hard this is to do with a 3-way speaker without steep crossovers. This is why most companies give up a little and use larger drivers and/or multiple drivers for low distortion, but poorer off axis quality and greater cone resonance. IOW, most speakers have compromises throughout nearly every octave, not just in a few places. 4-way and 5-way design can help eliminate these, BUT, passive crossovers aren't terribly transparent and speaker designers often try to minimize the number of crossover points.
post #224 of 481
Quote:
Originally Posted by Alimentall View Post

Well, the speaker's job is to be accurate. You do not want a perfectly flat response at the listening position or it will sound thin and aggressive.

Yes, but "accurate" is a bit more complicated then a measured FR-curve on axis in a anechoic chamber. If you use the speakers for stereo reproduction (aka fantom projected sound) the response is not going to be accurate, even if the speakers FR is dead flat.

I know that a flat response at the listening position is not going to be psychoacoustic flat.

/Chuck
post #225 of 481
The strange thing about a flat FR is that in the Frequency domain it is the ideal situation... It adds or subtracts nothing.. So why does too often a flat FR at the listening position translates into too bright, tinny, fizzy , thin and a lot more negative epithet? Why does a tilted curve that attenuates smoothly the treble often appears to "sound" better? It has begun to dawn on me that we must consider the temporal response of a speaker as well. I don't know if there are any paper or study on the subject but it seems that ONLY trying for Flat FR at the listening position may screw some parameters that our perceptive system deems important... there again the capacity to manipulate the signal that Digital Crossovers afford the designer can be put to good use. It should also be noted that expediency is not in itself an advantage of Digital Crossovers.. They do not allow you to do things necessarily quicker from the Designer point of view actually as Morbius has remarked if indirectly they may create too many choices and that in itself require more knowledge to deal with properly.. So the first iteration of a product will have the designer latching on one particular parameters, in the case of the XD it is clear that Flat.. exceptionally flat FR was the goal and it was achieved but it might have been at the expense of some other characteristics that for the listeners approximate more the reality of an acoustic event... I sincerely believe we are far from measuring all that matters... yet it looks like DSP with all that it can provide the speaker designer is a very powerful weapon to have in their arsenal. It can only further the quality of Music reproduction in our home ... I am off talking about the Xd the thread is about Digital Crossovers...
post #226 of 481
Thread Starter 
Quote:
Originally Posted by Chuck V View Post

Yes, but "accurate" is a bit more complicated then a measured FR-curve on axis in a anechoic chamber. If you use the speakers for stereo reproduction (aka fantom projected sound) the response is not going to be accurate, even if the speakers FR is dead flat.

Well, yes, there is an accuracy disadvantage to stereo that makes the sound a bit thinner than what is normal. Multi-channel helps this considerable, which is why Meridian and Lexicon are proponents of converting stereo to multi-channel.
Quote:



I know that a flat response at the listening position is not going to be psychoacoustic flat.

We are accustomed to distance from the sound source, which means a little more rolled off upper mids/treble and room gain in the bass that makes the sound a bit richer and warmer than perfectly flat. Perfectly flat is like sitting outside with the speaker 3' from you.
post #227 of 481
Thread Starter 
Quote:
Originally Posted by FrantzM View Post

The strange thing about a flat FR is that in the Frequency domain it is the ideal situation... It adds or subtracts nothing.. So why does too often a flat FR at the listening position translates into too bright, tinny, fizzy , thin and a lot more negative epithet? Why does a tilted curve that attenuates smoothly the treble often appears to "sound" better? It has begun to dawn on me that we must consider the temporal response of a speaker as well. I don't know if there are any paper or study on the subject but it seems that ONLY trying for Flat FR at the listening position may screw some parameters that our perceptive system deems important... there again the capacity to manipulate the signal that Digital Crossovers afford the designer can be put to good use.

You seem to be talking, not about 'temporal' problems and more about room EQ/listening position target curves.
Quote:



It should also be noted that expediency is not in itself an advantage of Digital Crossovers.. They do not allow you to do things necessarily quicker from the Designer point of view actually as Morbius has remarked if indirectly they may create too many choices and that in itself require more knowledge to deal with properly..

Depends. You can always take an unlimited time massaging anything. OTOH, I could take a speaker into an anechoic chamber, get a 90% of maximum result within a day and get the product to market. Then offer a download 6 months later with further tweaking. Of course, when you're building your *first* DSP speaker, you have a lot more options going towards your ideal acoustic design. You have more options, but most all of them are better than could be achieved with passive design and could be brought to market more quickly. Now that NHT has done their first digital speaker, they could much more easily and quickly build a second. Well, you know, if they were to stay in business.
post #228 of 481
Quote:
Originally Posted by Alimentall View Post

If you have fewer constraints, you can have more than one right answer to the remaining problems. That is a good thing. One unique solution is not the desire goal. The goal is to remove as many design constraints as possible so you can find the *best* of multiple solutions.

John,

Unfortunately, when you have too few constraints you oft get grossly INFERIOR
results because you get spurious solutions.

For example, we do a lot of hydrodynamics simulation. One of the problems you can get
in hydrodynamics is "hour-glassing". Consider three zones - one atop the other. The
equations may constrain the sum of the volumes of the three - but not the distribution.

What will happen is the middle zone can shrink in volume while the top and bottom
zones grow in volume. The three zones take on the shape of an "hourglass" or a
woman's figure - wide at top and bottom and slender in the middle.

That's what one often gets when one "under-constrains" the solution - you get the
opportunity for the system to behave VERY BADLY

You want to have the system properly constrained - not over-constrained and most
certainly not under-constrained.
post #229 of 481
Quote:
Originally Posted by AndreYew View Post

... and hardly a figure-8 since they do not have enough horizontal dispersion for the front and back wavefronts to meet appreciably. Again, look at the wavelengths involved, and the dimensions of the ribbon (horizontally).

Andre,

The ribbon dipole HAS to have a "figure-8 like" dispersion due to the physical symmetry.
The dipole looks the same front and back except that the two side radiate with opposite
phase. If the direct radiation doesn't have a null at the side - then which side wins?
If the radiator is a true dipole - you CAN'T have one side winning - because then the
radiator wouldn't be a dipole because it wouldn't be symmetryic.
Quote:


And ribbons would only be infinitely tall if they go from floor to ceiling, and both floor and ceiling are perfect reflectors.

They approach looking infinitely long at much lesser lengths. You need to a get a good
book on wave theory which includes expansions in spherical harmonics.

http://www.kettering.edu/~drussell/Demos/rad2/mdq.html
Quote:


Sorry, but I'm not buying this. There are many, many problems with this line of argument. For example, the room where the cymbal is played greatly affects its sound, as does the miking technique, mic response, mastering, and mixing. If it's also played by a drummer whom you've never heard live, then the musician's technique is another uncontrolled variable. How can you tell any of these effects apart from what you think the speaker's doing to the signal, and whether the speaker's doing the right thing or not?

The room response is separated in time. I'm talking about how the cymbal sounds when
it is first struck. The room response doesn't enter into that - because the sound waves
haven't had a chance to hit the room boundaries.

The drummer isn't going to be able to appreciably affect the dynamics of the response
of the cymbal. The drummer hits the cymbal - and the rest is dictated by the physic
of the cymbal. The drummer can't alter the speed of sound in the cymbal metal no
matter what he/she does; nor the geometry - and THAT'S what gives rise to the effects
that I am noting.
Quote:


And you keep bringing up the Keith Howard article without addressing the fact that experienced audio listeners had a very hard time hearing the effects of his filters. Please explain how that jives with what you've said about ringing.

I don't know about the others - but I sure can hear the difference. I think it has to do
with knowing what to listen for. If you understand the mathematics, and know what the
effects of the artifacts are - you can train yourself to listen for them.
post #230 of 481
Quote:
Originally Posted by terry j View Post

just a quick question on the ringing of steep filters (I won't go into audibilty etc).

re the deqx, it can go up to 300 db etc, but afaik even deqx does not recommend (usually) that steep a slope.

How steep is too steep? (with digital x-over slope I mean) I guess morbius would say 300 is too steep, but is 48 db too steep? 100?

This I guess relates back to the 'uncertainty principle' in audio? the more we tighten the FR the more we screw the time, and reverse. Is it all down to the individual case, where for a given set of drivers what works best is not the same as another set?

Terry,

Verry good!!! There IS an "uncertainty principle" here. It is closely related to the
Heisenberg Uncertainty Principle in Physics - and the mathematics that dictates the
existence of these principles is the same. You can't know the position and momentum
of a particle simultaneously to arbitrary accuracy. The position of the particle is analogous
to the temporal response. The momentum of the particle is analogous to the frequency
response. If you attempt to determine the momentum to "perfect" accuracy - you
destroy the information on the position of the particle.

Likewise, just as terry states; if you attempt to be "perfect" in the frequency domain;
you are being inaccurate in the temporal domain.

Quote:


Do we reckon Dave Wilson got his desired results in an afternoon?

If all Dave Wilson and company were doing is tweaking their crossovers in SPICE or
some other circuit modelling program; then they could get a result in an afternoon.

But that is NOT what they do. Dave Wilson and his engineers LISTEN to their product
and they make tradeoffs. They may allow a degree of non-flatness in the frequency
response curve in order to get a better temporal response. Dave and his staff are being
good engineers. They know FR is not some "Holy Grail" of audio quality that has to be
optimized at all costs. They are doing what all good engineers do in designing a product;
they are making tradeoffs.

Good audio engineers do that. They don't just "go by the numbers". They know what
the metrics like FR mean - and know the limitations of those metrics. In his review of
the dCS Scarlatti in the July / August 2008 issue of The Absolute Sound, issue #183;
review Jonathan Valin makes the same comment about the engineers at dCS. He
states they know the mathematics behind the workings of their processors; but that
they also LISTEN to them.

If an electrical engineer comes to me with an amp design that has vanishingly small
THD; it propably means he / she loaded the amp with tons of negative feedback as
was done back in the '70s. The engineer has designed the amp to measure well on a
THD test - but that's about all it will do well. The negative feedback doesn't allow the
amp to respond quickly to changes in the signal; and that's what the music demands.

We have an engineer that optimized on a SINGLE metric - and created a LOUSY audio
amp in the process.

Engineers can't be slavish to the metrics. There is no metric that measures absolute
audio quality. There a number that serve as guides to audio quality. However, if one
picks a single metric or a couple and optimizes on those; then that is NOT good
engineering.
post #231 of 481
Thread Starter 
Quote:
Originally Posted by Morbius View Post

Unfortunately, when you have too few constraints you oft get grossly INFERIOR
results because you get spurious solutions.

You want to have the system properly constrained - not over-constrained and most
certainly not under-constrained.

C'mon! That's utterly ridiculous and completely without merit. Paging Dr Grant! Weird science alert! Dr Grant, please come to the thread!


It's this simple - a speaker design attempts to solve a dozen (if you group them) or more different and difficult variables simultaneously. If some of the variables are 'pre-solved' by DSP and others are addressed more easily, then that leaves only a few difficult ones to deal with. It's ridiculous to assert that solving a dozen difficult problems with one solution is "INFERIOR" to solving maybe 4 of those same problems with one solution. This isn't a simple math equation. Yes, you have more possible solutions, but all of them are equal to or, more likely, *better* than the ones that work when attempting to solve more and more difficult problems.
post #232 of 481
Thread Starter 
Quote:
Originally Posted by Morbius View Post

The drummer isn't going to be able to appreciably affect the dynamics of the response
of the cymbal. The drummer hits the cymbal - and the rest is dictated by the physic
of the cymbal. The drummer can't alter the speed of sound in the cymbal metal no
matter what he/she does; nor the geometry - and THAT'S what gives rise to the effects
that I am noting.

I think you should stop trying to discuss drumming technique or the sound cymbals make while you're only several miles away from the truth.
post #233 of 481
Thread Starter 
Quote:
Originally Posted by Morbius View Post

Verry good!!! There IS an "uncertainty principle" here. It is closely related to the
Heisenberg Uncertainty Principle in Physics

Not really.
post #234 of 481
Thread Starter 
Quote:
Originally Posted by terry j View Post

How steep is too steep? (with digital x-over slope I mean) I guess morbius would say 300 is too steep, but is 48 db too steep? 100?

This I guess relates back to the 'uncertainty principle' in audio? the more we tighten the FR the more we screw the time, and reverse. Is it all down to the individual case, where for a given set of drivers what works best is not the same as another set?

It just comes down to experimentation. Nice thing about DEQX is that you can load three different crossover slopes and then switch between them and see what sounds best. You can expect that, as the slope steepens, cone resonances are lowered, lobing/acoustic interference is lowered, dispersion is increased in both vertical and horizontal domains (unless you already have the horizontal optimized), motor distortion goes down, dynamic range goes up. So you have to listen to what the improvements are in these areas versus added ringing in the crossovers and YMMV.

KEEP IN MIND that the ringing is essentially equal and opposite! If you have a 'positive' ring coming from the tweeter, you will have a 'negative' ring coming from the midrange and they will largely cancel at the listening position and any place to the sides in that horizontal axis. I have not heard any ringing as an obvious artifact. Atkinson thought he was getting an echo once. I have heard this effect once or twice on power up and it was a software issue where the algorithm didn't load properly and the problem went away as soon as the box was restarted. This is an obvious "what happened to my sound?" thing, not a subtle artifact that may or may not be audible. Any ringing is well below the signal and the self canceling attribute only lowers that further. What I do know for sure is that the improvement in off axis response is SO tremendous that I've hear NO passive speaker, except maybe the MBLs that can compete when you're outside the typical sweetspot. The ability to move anywhere in the room, including directly between the speakers or even above them just points to the magnificent job DEQX does and how solving 4 or 5 major audible problems by allowing a small amount of another problem is still a gigantic leap forward in performance, both objectively and subjectively.

As for the Heisenberg Uncertainty Principle, no, Greg's got it wrong there. That's an entirely different concept and obviously not what you meant. Digital crossovers are basic algorithms with predictable results.
post #235 of 481
The ability to play with alternative crossover slopes and parameters is a nice feature on the DEQX

While I fully understand the math of shallower vs. steeper slopes, I was surprised by how little the sonic differences were between various settings that get argued to death. In my case, the sonic differences between 24db and 96db/octave were very minor for the tweeter to midrange, 24db vs. 200+db was modestly detrimental, moving the crossover point from 1850hz to 2250hz was highly audible, and minor changes to PEQ settings were even more audible.

I've easily spent as many hours tweaking my curves as I would have spent designing a passive crossover (and that was true in the past with my Tact gear.)

It's similar to dialing in a great fully analog two channel setup: it takes many many hours to get positioning, toe ins, stands, etc. "just right". Similarly, it takes many many hours to dial in a digital crossover.

For inveterate tweakers that's a blessing and a curse.
post #236 of 481
Quote:
While I fully understand the math of shallower vs. steeper slopes, I was surprised by how little the sonic differences were between various settings that get argued to death.

Agree, some most minor topics are really discussed to death from the theoretical side. Obviously we are in desperate need of more blind and level matched A/B comparsions. I would dare to say some people currently fighting about real triffles to theoretical death would become best friends and fall into each others arms. No wonder I love the pro world
post #237 of 481
Quote:
Originally Posted by Alimentall View Post

Not really.

John,

There sure is an Uncertainty Principle just as Terrry states. It is also very much related
to the Heisenberg Uncertainty Principle - because the SAME mathematics gives you both.

In the Physics world, position and momentum are conjugate variables. Their operators
don't commute. Likewise, with energy and time; those operators don't commute either.

Many times the Heisenberg Uncertaintly Principle is taught saying it has to do with our
abilities to measure. Actually, Mother Nature herself doesn't know the answer any better.

In quantum mechanics, the momentum is related to the wavelength. If you have a
particle whose momentum you know to arbitrary precision; then the wavefunction will
be a single frequency / single wavelength wave for that particle. However, that wave
is unbounded in space - it goes to infinity in both both directions - so you have absolutely
no idea where the particle is.

The only way you know where the particle is would be to build up a wave "packet" - you
add waves of many frequencies together so that the wave function "peaks" in one place.
Now, you have a pretty good idea where that particle is - it is somewhere in the range of
that peak. However, to get that peak - you had to add together waves of many different
frequencies, hence wavelengths, and therefore momenta. So now you, nor Mother
Nature has a precise knowledge of what the momentum is.

The mathematics of digital filters does the same thing. I know you "think" that you can
specify the flatness of the frequency response to arbitrary precsion, while simultaneously
dictating a perfect temporal response. However, you just plain can't do that.

The frequency response and temporal response are related. Just as Keith Howard shows
in his article; Terry is correct that when you constrain the frequency response to be
flat to arbitrary order - you mess up the temporal response and visa versa.

It's basically the same mathematics as the Heisenberg. If you constrain the precision of
momentum to arbitrary order - i.e. a single wave - then by necessity you've messed up
any precision you may have had as to the position.

If you say "Not really"; then I would really like to see your "proof" of "not really". Please
be mathematically rigorous in your proof and include all the requisite Fourier Transforms.
post #238 of 481
Quote:
Originally Posted by Alimentall View Post

As for the Heisenberg Uncertainty Principle, no, Greg's got it wrong there. That's an entirely different concept and obviously not what you meant. Digital crossovers are basic algorithms with predictable results.

John,

I didn't mean that there was any "uncertainty" in the response of the digital filters.

Frequency response and temporal response are related just like momentum and position
are. When you tighten down on the specification of momentum; you LOSE precision in
the knowledge of the position. That is what the Uncertainty Principle is about - you
can't specify the two non-commuting variables to arbitrary accuracy simultaneously.
Tightening down on one; screws up the other.

That's what happens with frequency response and temporal response; and for the same
reason - the same mathematics.

I never said that there was anything "uncertain" about the behavior of the filter.
post #239 of 481
Quote:
Originally Posted by Alimentall View Post

C'mon! That's utterly ridiculous and completely without merit. Paging Dr Grant! Weird science alert! Dr Grant, please come to the thread!


It's this simple - a speaker design attempts to solve a dozen (if you group them) or more different and difficult variables simultaneously. If some of the variables are 'pre-solved' by DSP and others are addressed more easily, then that leaves only a few difficult ones to deal with. It's ridiculous to assert that solving a dozen difficult problems with one solution is "INFERIOR" to solving maybe 4 of those same problems with one solution. This isn't a simple math equation. Yes, you have more possible solutions, but all of them are equal to or, more likely, *better* than the ones that work when attempting to solve more and more difficult problems.

John,

Actually this IS a math equation - albeit not a "simple" one depending on your definition
of "simple". Compared to the simulations that I work on - the audio speaker design
problem really is simple.

At the end of the day; in your design you had better have constrained all the degrees
of freedom that the problem has. If you don't constrain a degree of freedom; then
Mother Nature is free to do with that degree of freedom as she wills; and you probably
won't like the result.

Matching contraints and degrees of freedom is well known in the computational physics
and numerical modelling world. We know we get spurious solutions if we don't tack
down all the degrees of freedom that the system has.
post #240 of 481
Thread Starter 
Quote:
Originally Posted by Morbius View Post

The mathematics of digital filters does the same thing. I know you "think" that you can
specify the flatness of the frequency response to arbitrary precsion, while simultaneously
dictating a perfect temporal response. However, you just plain can't do that.

I never said that you could. The goal isn't to have a 'perfect' FR or temporal response, the goal is to have an *improved* FR and temporal response (at least at this point in the development). There is a verification procedure so that you can test the result versus the theoretical result. In the future, I am sure that systems will allow multiple testing procedures that build increasingly precise algorithms, even ones that are adaptive to the content or volume. The result, however, will still be 'certain', as in predictable, plus or minus a reasonable factor of error and beyond what is achievable with passive crossover design. If crossover design was perfect, in the digital or passive domain, the former wouldn't need a verification procedure and speaker designers would have a computer spit out the perfect crossover for any given speaker within seconds.

I mean, I see your point and apologize for misreading it, but this does not take away from DEQX's capability as you can set the parameters for correcting time and FR and focus on one, the other or both, not to mention with a lot of flexibility. NHT spent 5 years developing Xd and another year or two improving the existing filter set in response to consumer and reviewer feedback, for instance. Tis a shame few outside of actual owners have ever heard the improved firmware/filter set.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Ultra Hi-End HT Gear ($20,000+)
This thread is locked  
AVS › AVS Forum › Display Devices › Ultra Hi-End HT Gear ($20,000+) › Digital Active Speaker Thread.....