I said it depends on the precise units chosen.
There are treble units where 2.5kHz xover is appropriate but I don't like them as much as those where 5kHz is proper. Also depends on the mid/bass unit too. YMMV.
I've measured a LOT of off-axis responses. More than 2 decades of my previous life was spent designing speakers for the largest maker in Europe.
Perhaps you should read the paper to find out.
I'm not denigrating Linkwitz. I said you need to know a bit more about speakers & xovers to move the main lobe where you want it. Good speaker designers have been doing it for decades before he appeared cos they measure a lot of off-axis responses.
BTW, for those of you designing supa dupa active & DSP LR4 xovers, a small reminder that it's the acoustic response that has to be LR4 bla bla
Last edited:
Steven, I'll answer this one first and the rest later (maybe much later as I'll be on a boat till Tue 7jul so pls remind me by email when I get back)
You use xovers to get the best sound. If you bi-amp, the xover which gives you the best increase in sound quality is 1kHz.
BiAmp (Bi-Amplification - Not Quite Magic, But Close) - Part 1
has the right idea but he's basing it on wonky data. (I worked with the chap at Fane who provided his table in 1.3)
That table is average power when what we are interested in is peak power for which 1kHz is the dividing point.
Rod's sections 1.3 - 1.7 are mostly OK though often not for the reasons he states. There is wonky advice in much of the rest. eg there's no problem achieving a seamless xover at 1kHz if you know a bit more about speakers & xovers than .. oops! Sorry vac!
You must have at least a capacitor between amp & treble unit. IGNORE THIS AT YOUR PERIL.
Once you do this, you might as well do a full Arthur-Smythe 3rd order. With most treble units, you don't gain anything if you have a separate amp so why not do the mid/bass to treble passive too. It will be greatly simplified cos you only need to get the xover smooth .. not flat. (Flat is done with your DSP EQ and of course you need to get the main lobe bla bla).
The inductors are likely smaller than 0.5mH and the caps less than 10u so you can do a really good one without too much $$$.
In my previous life, I had the luxury of designing my own drive units from scratch so they could match these requirements.
____________________
For the speaker you show early in the thread, we could get away with eg a MAC laptop (which are some of the very few these days with good h/p audio out. Suggestions for Windoze machines which have these please. Da old IBMs used to but alas I think, not any more ), a big, maybe 2x200W channel stereo amp and your speaker.
The laptop (if Windoze) would run JRiver Media software, or Windows MP with Sourceforge Convolver for Windows ... something which can load & use a lo...ong FIR (which we dream up with DRC or Accourate).
PS Robert Arthur-Smythe of Celestion formalised the AS3 .. what most of us horny handed speaker designers have been crudely doing for a long time. A treble unit is a 2nd order filter so the simplest ACOUSTIC response that is practical is 3rd order. There's good reason why this is so popular among real speaker designers. RAS designed the Celestion SL6, SL600 & SL700 among loadsa other good stuff.
Last edited:
Windoze or other Media Players which can do IIRs?
Barley has suggested JRiver Media software, or Windows MP with Sourceforge Convolver for Windows as good Media Players that can load & use a long FIR.
But are there any players that can load & use an arbitrary IIR? By arbitrary I mean where the Feedforward & Feedback coeffs can be anything. So those EQs which are just a bunch of parametrics are out. A bunch of parametrics would be Ramos & Lopez referred to in my IIR paper.
________________________
The player should ideally have the nice indexing & cataloguing facilities for music found in Microsoft's Windoze Media Player & Apples iPlayer. This is a BIG disadvantage for some nice players like VLC.
The user objects to losing his 'favourites' list etc and is likely to stay with lesser players for this reason
_________________
For those who intend to program this into a player (VLC please .. and don't forget the indexing & cataloguing stuff too), the requirement is very similar to implementing a FIR which just involves a Multiply & Accumulate (MAC) loop.
The IIR just needs another MAC loop and then we add or subtract that to the FIR sum. 256 feedforward + 256 feedback would be easy and nice.
More detail in my IIR paper.
__________________
I'm not going into the IIR vs FIR. What I know for a fact is that 100+100pt IIR is better all round than 200pt FIR or 1024 pt FFT FIR. All in my IIR paper and the references.
What I have no experience of and which has only been available this Millenium is the 64K long FIRs that Accourate & DRC are capable of dreaming up.
My experience with simulating Room Responses since the 80's leads me to take some of their claims with a large pinch of salt but I have no practical experience. Eric Benjamin, ex-Dolby, and one of my BLaH collaborators has recent experience with some of the big $$ room EQ boxes and of course Lake, one of the first to do this, is now a Dolby Co.
But using IIRs except for very specialised applications is moot until there's a common free Media Player that will load & use one.
Plogue Bidule will do it but its hardly a consumer Media Player. see the BLaH Ambisonic paper for details and also how to do LR2 & LR4
(God! I hate it when someone comes late and hijacks stuff that we've been doing for years without a fancy name).
Barley has suggested JRiver Media software, or Windows MP with Sourceforge Convolver for Windows as good Media Players that can load & use a long FIR.
But are there any players that can load & use an arbitrary IIR? By arbitrary I mean where the Feedforward & Feedback coeffs can be anything. So those EQs which are just a bunch of parametrics are out. A bunch of parametrics would be Ramos & Lopez referred to in my IIR paper.
________________________
The player should ideally have the nice indexing & cataloguing facilities for music found in Microsoft's Windoze Media Player & Apples iPlayer. This is a BIG disadvantage for some nice players like VLC.
The user objects to losing his 'favourites' list etc and is likely to stay with lesser players for this reason
_________________
For those who intend to program this into a player (VLC please .. and don't forget the indexing & cataloguing stuff too), the requirement is very similar to implementing a FIR which just involves a Multiply & Accumulate (MAC) loop.
The IIR just needs another MAC loop and then we add or subtract that to the FIR sum. 256 feedforward + 256 feedback would be easy and nice.
More detail in my IIR paper.
__________________
I'm not going into the IIR vs FIR. What I know for a fact is that 100+100pt IIR is better all round than 200pt FIR or 1024 pt FFT FIR. All in my IIR paper and the references.
What I have no experience of and which has only been available this Millenium is the 64K long FIRs that Accourate & DRC are capable of dreaming up.
My experience with simulating Room Responses since the 80's leads me to take some of their claims with a large pinch of salt but I have no practical experience. Eric Benjamin, ex-Dolby, and one of my BLaH collaborators has recent experience with some of the big $$ room EQ boxes and of course Lake, one of the first to do this, is now a Dolby Co.
But using IIRs except for very specialised applications is moot until there's a common free Media Player that will load & use one.
Plogue Bidule will do it but its hardly a consumer Media Player. see the BLaH Ambisonic paper for details and also how to do LR2 & LR4
(God! I hate it when someone comes late and hijacks stuff that we've been doing for years without a fancy name
).
Last edited:
(God! I hate it when someone comes late and hijacks stuff that we've been doing for years without a fancy name
Yes, I think most filter combinations will have been made at some stage, based on theory and experimentation, without the developer knowing that it cloud be classified as LR, Bessel, what have you. In that context, I am very interested to learn more about the Arthur-Smythe 3rd order you mentioned a couple of posts before, never heard about it, no reference in any book I have, nor on the internet.
The only reason I ever saw to use 3rd order filters, is to correct for path way differences between drivers.
Vacuphile,
I did a search myself and couldn't find the Smythe filter either. What I think kgrlee is counting on with using that 3rd order filter is that the driver used would have a natural roll-off of 6db per octave at the crossover point supplemented with a 3rd order electric filter giving a fourth order acoustic slope. I very rarely see a speaker that has that smooth a roll-off or happens to fall just where you would want that to happen. As I said earlier I find I would much rather use a 4th order filter in the pass-band where the device is flat and have all of the filter slope being electrical. I guess it is just a different approach to get to the same final result of 4th order slope on both sides of the xo point giving a flat total acoustic output.
I could never use a 1Khz crossover point wth my two way system, the dome tweeter would sound terrible trying to go that low if it was capable at all and the cone driver can easily reach up to the 2.5 khz xo point. Actually the cone goes much higher but gets ragged up higher as most drivers do if they can reach so low, so that concept without a third device just wouldn't work in my eyes for a simple two way system with a dome tweeter and not a horn loaded compression driver to try and cover that range and then I don't know of a compression driver that would go that low and still have flat upper octave output if it could even get over 16khz in reality.
I did a search myself and couldn't find the Smythe filter either. What I think kgrlee is counting on with using that 3rd order filter is that the driver used would have a natural roll-off of 6db per octave at the crossover point supplemented with a 3rd order electric filter giving a fourth order acoustic slope. I very rarely see a speaker that has that smooth a roll-off or happens to fall just where you would want that to happen. As I said earlier I find I would much rather use a 4th order filter in the pass-band where the device is flat and have all of the filter slope being electrical. I guess it is just a different approach to get to the same final result of 4th order slope on both sides of the xo point giving a flat total acoustic output.
I could never use a 1Khz crossover point wth my two way system, the dome tweeter would sound terrible trying to go that low if it was capable at all and the cone driver can easily reach up to the 2.5 khz xo point. Actually the cone goes much higher but gets ragged up higher as most drivers do if they can reach so low, so that concept without a third device just wouldn't work in my eyes for a simple two way system with a dome tweeter and not a horn loaded compression driver to try and cover that range and then I don't know of a compression driver that would go that low and still have flat upper octave output if it could even get over 16khz in reality.
Don't bother looking. Not everyone is eager to put his name to a filter. It exists as an internal Celestion Engineering Memo. I did the MathCAD 'proof' but it didn't survive several HD crashes when I became a beach bum.
A cuboid box speaker usually has the acoustic centre of the bass/mid behind the treble. A LR4 would then point the main lobe down if the treble is on top as would all the even order xovers.
An odd order xover however, can be designed to point the main lobe up, so it is 'horizontal'. There are also versions for treble below the mid and for various assumed speaker & listener heights.
There is one specific target function called AS3 at Celestion & KEF, but its really a philosophy which encourages you to look at what happens off-axis .. something that good designers have done for a loo...ong time.
Julian Wright at Celestion dreamt up the '3D' off-axis display that every speaker design/measurement package has to have as a bell or whistle today.
My 1981 pontificating on phase has quite a bit of the thinking behind it.
AS3 is 3rd order acoustic.
It starts of as 3rd order electrical until the resonance of the treble unit .. below which it becomes 1st order electrical which combines with the natural 2nd order LF roll-off of the unit. You have to match the damping of the treble 'bass resonance' but if you get it right, the acoustic response is perfectly 3rd order all the way.
Mike Berman of KEF was the first to do this in the late 70. He gave us the circuit analysis tools to match passive xovers to drive units but he never made the mental leap to speakers as a 3D source. IIRC its described in a KEFtopic of the late 70's.
You can't ignore the natural roll-off of the unit.
It has a PHASE response that you have to take into account ... more than a decade from the 'bass resonance'. You need to do all this if you want to 'design' the off-axis response.
For a box speaker, you don't want an even order xover cos the main lobe will not be horizontal.
____________________
Just another gentle reminder that its the acoustic response that has to be LR4, AS3, Butterworth, Bessel or whatever you want. In most (all?) cases, having a supa accurate LR4 active xover has no bearing on this. It's nearly ths same as buying a passive xover from RadioShack.
Last edited:
Kgrlee,
I understand that you are really then looking at the acoustic response only then when you are doing the xo tuning, the acoustic response taking whatever electrical response is needed to create the dispersion pattern that you target. What about those who are using much higher brick wall type filters, those going up to say a 48db per octave type of filter? Is there ever a point where you wouldn't care about the resonance frequency of the tweeter, wouldn't that area of the output be so low down in the acoustic sense that you could disregard that or would you still do some type of correction even with the very high order filters?
I understand that you are really then looking at the acoustic response only then when you are doing the xo tuning, the acoustic response taking whatever electrical response is needed to create the dispersion pattern that you target. What about those who are using much higher brick wall type filters, those going up to say a 48db per octave type of filter? Is there ever a point where you wouldn't care about the resonance frequency of the tweeter, wouldn't that area of the output be so low down in the acoustic sense that you could disregard that or would you still do some type of correction even with the very high order filters?
Absolute Listening Tests-Further Progress tells us that the most important characteristic of a good speaker is it's Room Interface Profile.
Brickwall xovers give an abrupt change in directivity which adversely affects this and don't sound good. One of the earliest was a student of Malc. Hawkesford at Essex U. who converted a Celestion SL6 IIR to this. I was external examiner for his degree but we also did listening tests for this digitally EQ'd & xovered, bi-amped bla bla item.
There's also a lot of BS about 'pistonic' diaphragms.
In fact a good sounding cone speaker MUST break up in a controlled manner and I've done a lot of work on getting cones to behave like I want.
It's this breakup which sorta happens for ALL cone speakers that make Rice & Kellog's invention pre-eminent from cheapo iPod speaker dock to the $$$ Golden Pinnae monolith.
The breakup 'shrinks' the size & mass of the effective radiating diaphragm coupled with the horn effect of the cone and other factors.
Last edited:
I've never personally played with those higher order filters nor have I heard any so I will accept your knowledge as truth. I guess in a horn application if all the horns were conic with the same angular dispersion then perhaps it wouldn't have the same effect, but that is pure speculation on my part.
I thought you were supposed to be sailing right now?
If you read my email to you I would think that you would understand that I got over very quickly the idea that you truly wanted an extremely rigid cone in a speaker, it causes nothing but trouble in my eyes. A false premise that is often repeated until most people believe this. The same with diamond type dome tweeters, that just seems like it would cause some serious issues at very high frequencies.
I don't know the sound of the B&W kevlar cones, and they use the kevlar as the surround also but I imagine that what I have done is more in that realm of things.
I'll wait for you to return and then we can get back to some of those questions that you seem to have so many answers to. A fine mind to pick indeed.
I thought you were supposed to be sailing right now?
If you read my email to you I would think that you would understand that I got over very quickly the idea that you truly wanted an extremely rigid cone in a speaker, it causes nothing but trouble in my eyes. A false premise that is often repeated until most people believe this. The same with diamond type dome tweeters, that just seems like it would cause some serious issues at very high frequencies.
I don't know the sound of the B&W kevlar cones, and they use the kevlar as the surround also but I imagine that what I have done is more in that realm of things.
I'll wait for you to return and then we can get back to some of those questions that you seem to have so many answers to. A fine mind to pick indeed.
but has a MAJOR influence on a speaker system's directivity.
Steven, this is probably my last post for a week or so. I'm sailing today.
There's a number of concepts that need retiring eg the importance of response 1m on treble axis bla bla. Even vac's talk of 'delay' in units is misleading cos its a 1D view. A speaker has 3D output so its much better to think "is the response in ALL important directions good?" "Is what the speaker spewing into the room good?"
For the rest of yus guys, I apologise for wanking but this is a subject that was a big part of my previous life.
Returning (??) to the civilised world, I find much has changed (eg huge computing power) but other stuff, like know how to use this power, isn't that more advanced.
But this beach bum finds it difficult to recover from nearly 2 decades out of the mainstream in stuff like programming.
Last edited:
Barleywater, this is not how I understood Richard. He departs from the notion, that both drivers have significantly different radiation patterns, which will be most certainly the case if you xover as high as he advocates.
This is old school, you can't get away with speakers like that anymore. Controlled directivity came up, to ensure that both drivers have similar radiation patterns around xover.
This is old school, you can't get away with speakers like that anymore. Controlled directivity came up, to ensure that both drivers have similar radiation patterns around xover.
This is the exact opposite of what I meant.
One of my criteria for xover frequencies & slopes is so the directivity changes doesn't change across the xover.
I was doing 'controlled directivity' since the late 70's. The papers I quote all bear on this.
Would you like to post what treble & bass/mid units you use with 2.5kHz xover and perhaps some off-axis responses of the units and the complete systems? I'm seriously interested.
That's exactly what I said, two ways to do it: 1) waveguide, extreme example the Geddes approach. Problem is that you still have a sharp transition between the listening window, where the FR is relatively straight, and beyond that. 2) Other way is to cross over the mids before they start to beam, so that they are effectively in 4 pi radiation at xover. This is my approach, we can perhaps come back to why this desirable at a later stage.
This necessitates small mid-drivers and low xover frequencies. Does it work? Bet you it does. Please have a look at the following:
This is not your normal listening window, but a 180 degree measurement. Even at 90 degrees off axis, it is almost straight right up to 7kHz. The subsequent fall off is the natural beaming pattern of a 3/4" tweeter. For those who like colorful images, here is a directivity plot of the same speaker, but in another graphical representation:
This measurement goes from 0-110 degrees off axis (indeed, from slightly behind the baffle in the most extreme measurement).
Richard, I'd love to tell you which drivers I use, but I won't because of commercial reasons. The sizes are 3/4" for the tweeter and 2 1/2" for the mids in a d'Appolito configuration. The trick is in the enclosure, based on an old patent, with a new patent pending.
Vertical dispersion is pretty decent, but not completely to my liking. Even a 4th order filter keeps the mids going on for too long, resulting in a bit of suck out. Yet, the vertical listening window is about 30-40 degrees, which is not bad, but I want to get it better by using steeper filter slopes. This will require FIR filtering. Last picture:
Please note that these measurements are relatively old ones, and improvements have been made in the meanwhile. Time to book the anechoic room again and get some fresh data.
This necessitates small mid-drivers and low xover frequencies. Does it work? Bet you it does. Please have a look at the following:
This is not your normal listening window, but a 180 degree measurement. Even at 90 degrees off axis, it is almost straight right up to 7kHz. The subsequent fall off is the natural beaming pattern of a 3/4" tweeter. For those who like colorful images, here is a directivity plot of the same speaker, but in another graphical representation:
This measurement goes from 0-110 degrees off axis (indeed, from slightly behind the baffle in the most extreme measurement).
Richard, I'd love to tell you which drivers I use, but I won't because of commercial reasons. The sizes are 3/4" for the tweeter and 2 1/2" for the mids in a d'Appolito configuration. The trick is in the enclosure, based on an old patent, with a new patent pending.
Vertical dispersion is pretty decent, but not completely to my liking. Even a 4th order filter keeps the mids going on for too long, resulting in a bit of suck out. Yet, the vertical listening window is about 30-40 degrees, which is not bad, but I want to get it better by using steeper filter slopes. This will require FIR filtering. Last picture:
Please note that these measurements are relatively old ones, and improvements have been made in the meanwhile. Time to book the anechoic room again and get some fresh data.
Last edited:
Vacuphile,
That waterfall is really impressive, nothing wrong there. Why are you not happy with the vertical dispersion, 30 to 40 degrees would seem to cover most any listening position without putting much more energy towards the floor or ceiling? Those are fairly small drivers so I imagine you are using something below that for the bass section and the upper limit of spl may be a bit less than some demand but I think that is mostly just bs as most don't actually want to listen at concert level in their homes.
That waterfall is really impressive, nothing wrong there. Why are you not happy with the vertical dispersion, 30 to 40 degrees would seem to cover most any listening position without putting much more energy towards the floor or ceiling? Those are fairly small drivers so I imagine you are using something below that for the bass section and the upper limit of spl may be a bit less than some demand but I think that is mostly just bs as most don't actually want to listen at concert level in their homes.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.