If it were as easy as mentioned in the following patent
Espacenet - Original document
- we would all build transient perfect loudspeakers with higher order Linkwitz crossovers.
Unfortunately it is not that easy (some engineering joke with Polish tourists aboard a plane popped up in my mind).
No wonder the patent is in the status "Awaiting Applicant's Response" !
Regards
Charles
Espacenet - Original document
- we would all build transient perfect loudspeakers with higher order Linkwitz crossovers.
Unfortunately it is not that easy (some engineering joke with Polish tourists aboard a plane popped up in my mind).
No wonder the patent is in the status "Awaiting Applicant's Response" !
Regards
Charles
It's easy enough to do these sum and difference type filters electronically. Isn't the problem that the drivers are unlikely to have a perfect and flat response in the crossover region?
Most of the pioneers in this sort of "ability to reproduce a square wave correctly" field like Bud Fried ended up looking for suitable drivers. That was the hard part.
I really must try a series crossover on my regular build. They have some strengths with realistic dynamics according to Bud. Fried Audio | Cutting-edge Audio Speakers
Most of the pioneers in this sort of "ability to reproduce a square wave correctly" field like Bud Fried ended up looking for suitable drivers. That was the hard part.
I really must try a series crossover on my regular build. They have some strengths with realistic dynamics according to Bud. Fried Audio | Cutting-edge Audio Speakers
The driver's response isn't really a problem. One has just to develop a filter circuit that gives the desired response combined with the drivers.
But it is indeed tricky to find a crossover principle to start with that has steep slopes and the least overlap. If it would work then the filters in the patent shown would be a gift from heaven.
Unfortunately we don't live in the world that we'd like to have - instead we live in the real one. In this world the shown filters are not possible.
Now let's wait and see until someone tries to simulate one. I am looking forward to disappoint him/her ......
Regards
Charles
But it is indeed tricky to find a crossover principle to start with that has steep slopes and the least overlap. If it would work then the filters in the patent shown would be a gift from heaven.
Unfortunately we don't live in the world that we'd like to have - instead we live in the real one. In this world the shown filters are not possible.
Now let's wait and see until someone tries to simulate one. I am looking forward to disappoint him/her ......
Regards
Charles
I think the only way to solve this problem in practise is to listen on headphones!
Filter slopes introduce group delay. It's how it works. Maybe you could equalise that with a very complex bit of digital filtering, But even then, it would only work at one listening position unless you had a concentric driver.
There is a lot of good in time aligning drivers IMO:
18W-8434G00
Filter slopes introduce group delay. It's how it works. Maybe you could equalise that with a very complex bit of digital filtering, But even then, it would only work at one listening position unless you had a concentric driver.
There is a lot of good in time aligning drivers IMO:
18W-8434G00
I've had very good luck with properly adjusted third order elliptical filters that approach first order amplitude and phase response in their passbands, but that have a response notch an octave to three octaves into their stop bands which can often be used to suppress out of band driver peaks better than much higher order conventional filters can. By the time the notch dip decreases farther beyond the xover point, the driver response is usually down at least 30 db. Overall, this gives them the equivalent selectivity of second or third order conventional filters and a close approach (differentially within about 25 degrees) to the xover phase response of a classic first order network.
I don't feel having one optimal vertical listening plane to be much of a restriction. If someone is expecting perfect driver integration at all points, he is probably distracted from fully appreciating it by other variables, anyway.
I don't feel having one optimal vertical listening plane to be much of a restriction. If someone is expecting perfect driver integration at all points, he is probably distracted from fully appreciating it by other variables, anyway.
Last edited:
I built 4th order 3kHz allpass filters out of NE5534 opamps at college. We couldn't hear them either!
On the other hand, Bud Fried could always pick out steep filters just by the sound. Maybe you need a good trained ear. I mean, what exactly are we listening for? A sort of ringing I suppose. I think he listened to the hammer thingie striking the strings on Classical piano.
Troels Gravesen uses "Siri's Killer Note"!
BTW, I don't find reflex bass hard to pick out at all. You can just feel the lack of timing and slam compared to faster closed box. Especially on tight drums.
Last edited:
Blauert and Laws give average thresholds for group delay audibility by frequency.
BR group delay is a problem for me wrt critical listening, but found if the BR resonant point is sufficiently lower than 40 hz with amplitude down a few db at that point relative to the mid-bass that most of that issue was mitigated. The overall effect then could resemble more the initial rolloff of a closed box, although without as much of a sensation of extreme low frequency bass energy.
BR group delay is a problem for me wrt critical listening, but found if the BR resonant point is sufficiently lower than 40 hz with amplitude down a few db at that point relative to the mid-bass that most of that issue was mitigated. The overall effect then could resemble more the initial rolloff of a closed box, although without as much of a sensation of extreme low frequency bass energy.
Last edited:
I built 4th order 3kHz allpass filters out of NE5534 opamps at college. We couldn't hear them either!
After reading a some about how Bruno Putzeys describes the problem: If I've understood correctly the problem isn't the phase itself but that the phase difference on the lower / higher driver. This difference causes the off axis response to not sum gracefully on a non linear phase speaker. There is still the group delay though, but at 3 khz that's almost nothing. He discussed that linear phase crossovers at 1.5 khz was barely audiable, but in the bass region it was much more audiable. In the bass it might be group delay that is audiable though and not the phase itself.
Gilbert Briggs...rubbish!
Harold Leak...rubbish!
Now poor old Bud Fried gets it!
What about Peter Walker? Those old Quad electrostatics...
Hi,
Peter Walker really knew his stuff. But FWIW his intransigence
regarding "BS" about components cost the company dearly.
(Pointless penny pinching in good kit is not what people
want to see but that is what Quad did, unwisely IMO.)
Famously he arranged a listening session that proved IT
of Linn was full of "BS", but still lost the commercial battle.
rgds, sreten.
Last edited:
Peter Walker and Bud Fried both agreed that what made for "the closest approach to the original sound" was high dynamic range.
This might sound a bit bizarre, but I think that the moment you split a signal into two elements, woofer and tweeter, and recombine them at the ear, the overall sound wave is modified in some way such that you can detect this.
The time signature changes, the phase changes and the lobing changes. It'll sound different at different positions. So we may really just be looking at trade-offs as usual. It's the quantum theory of waves applied to sound. When you interfere with a system by filtering, you change it forever. It's how it works.
People claim great things for series filters, as opposed to the regular parallel filter. There IS a measurable difference, and they do seem to align phase better. I don't know really. But impedance curve is totally different, and the rolloff is gentler. Perhaps another trade-off.
See below for example filters that do similar things to frequency response and phase.
This might sound a bit bizarre, but I think that the moment you split a signal into two elements, woofer and tweeter, and recombine them at the ear, the overall sound wave is modified in some way such that you can detect this.
The time signature changes, the phase changes and the lobing changes. It'll sound different at different positions. So we may really just be looking at trade-offs as usual. It's the quantum theory of waves applied to sound. When you interfere with a system by filtering, you change it forever. It's how it works.
People claim great things for series filters, as opposed to the regular parallel filter. There IS a measurable difference, and they do seem to align phase better. I don't know really. But impedance curve is totally different, and the rolloff is gentler. Perhaps another trade-off.
See below for example filters that do similar things to frequency response and phase.
Attachments
Feel free to discuss audibility issues here, although my intention was to point at a ridiculous patent of which I don't have the slightest idea why it was filed in the first place.
If it was indeed as easy as shown in that patent at least all analog active speakers would be transient perfect. It would not stop at the crossover however: We could even make any type of woofer alignment without any group-delay at the lower end simply by adding a subsonic filter with compensating phase response. But it definitely shows that even large multinational companies are not immune against patenting stuff that doesn't work.
I actually wanted to do exactly the same thing 25 years ago. Then my prof just smiled and said: Keep an eye on the poles !
The solution pointed at by "Thoriated" is indeed one of the working ones. As soon as the weather permits to make outside measurements I will try something a little similar with an asymmetric subtractive crossover with a notch in the derived branch.
Most analog transient-perfect crossovers are indeed worse in terms of lobing than Linkwitz for instance. How much this is detrimental to reproduction is depending on the actual listening environment. And yes, the vertical window, on which the response is flat, is quite narrow. But that doesn't mean that the speaker becomes unlistenable below and above.
The reason why I stumbled over the patent was because I was searching for sources that are worth citing in an article on simple active crossovers with low group delay.
Regards
Charles
If it was indeed as easy as shown in that patent at least all analog active speakers would be transient perfect. It would not stop at the crossover however: We could even make any type of woofer alignment without any group-delay at the lower end simply by adding a subsonic filter with compensating phase response. But it definitely shows that even large multinational companies are not immune against patenting stuff that doesn't work.
I actually wanted to do exactly the same thing 25 years ago. Then my prof just smiled and said: Keep an eye on the poles !
The solution pointed at by "Thoriated" is indeed one of the working ones. As soon as the weather permits to make outside measurements I will try something a little similar with an asymmetric subtractive crossover with a notch in the derived branch.
Most analog transient-perfect crossovers are indeed worse in terms of lobing than Linkwitz for instance. How much this is detrimental to reproduction is depending on the actual listening environment. And yes, the vertical window, on which the response is flat, is quite narrow. But that doesn't mean that the speaker becomes unlistenable below and above.
The reason why I stumbled over the patent was because I was searching for sources that are worth citing in an article on simple active crossovers with low group delay.
Regards
Charles
Last edited:
Go to
Intellectual Property Office - Patent document and information service (Ipsum)
And then search for Publication Number GB2489440. You will then get the page with the legal status of the patent. At the upper right you can choose to see the download links for the documents belonging to that case. The one that shows the whole patent document is called the "Publication Document".
Regards
Charles
Intellectual Property Office - Patent document and information service (Ipsum)
And then search for Publication Number GB2489440. You will then get the page with the legal status of the patent. At the upper right you can choose to see the download links for the documents belonging to that case. The one that shows the whole patent document is called the "Publication Document".
Regards
Charles
Keep an eye on the poles indeed!
This is the guts of a third order allpass network. The zeroes can go either side, but the poles are always on the left in the real world. So group delay and group delay distortion is always positive and just gets worse every time you filter.
The ideal SINC or Dirac Delta function does not exist in the real world of filters except with some advanced digital techniques. The ringing must always follow the step, if you think about it. So you won't see anything like this one as a step or impulse response:
It's one of those mathematical niceties that a Linkwitz-Riley 4th order is two second order butterworths cascaded. I suppose you could filter a signal with BW2 and then play it backwards on a tape recorder and then filter it again with BW2. You'd end up with a LR4 with no group delay. Now Radar people might do something like that, but they'd do it digitally.
An externally hosted image should be here but it was not working when we last tested it.
This is the guts of a third order allpass network. The zeroes can go either side, but the poles are always on the left in the real world. So group delay and group delay distortion is always positive and just gets worse every time you filter.
The ideal SINC or Dirac Delta function does not exist in the real world of filters except with some advanced digital techniques. The ringing must always follow the step, if you think about it. So you won't see anything like this one as a step or impulse response:
An externally hosted image should be here but it was not working when we last tested it.
It's one of those mathematical niceties that a Linkwitz-Riley 4th order is two second order butterworths cascaded. I suppose you could filter a signal with BW2 and then play it backwards on a tape recorder and then filter it again with BW2. You'd end up with a LR4 with no group delay. Now Radar people might do something like that, but they'd do it digitally.
Last edited:
Running signals backwards through allpasses or allpass crossovers is indeed a method that is used with some active concepts. The method adds some overal delay in exchange for constant group delay. Another method is using a subtractive delay crossover which also introduces some delay in order to get flat group delay.
The really mean thing regarding circuits like those in the aforementioned patent is the fact that they simulate well with circuit simulators as long as only the frequency-response ist considered. As soon as a transient response simulation is run things get really "interesting".
Regards
Charles
The really mean thing regarding circuits like those in the aforementioned patent is the fact that they simulate well with circuit simulators as long as only the frequency-response ist considered. As soon as a transient response simulation is run things get really "interesting".
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.