[FlorianO]

Well,  the subject says it all 

First, let me say outright that this feels that should be part of "speaker builder 101" i.e. already addressed and discussed somewhere. If so, pointers and resources highly appreciated.

Second, this is just my novice understanding, but I would rather be wrong then keep guessing 

To the subject: In different shapes  or forms ppl want to augment FRs with (super)tweeters and/or bass. An issue that showed up in (this thread is cancellation due to phase shifts when integrating tweeters with FRs. In very simple terms the FR has useful output well above the XO frequency, and due to the phase mismatch btw the tweeter and the FR  cancellation at the summed output occurs.

(Btw: Another very interesting discussion, albeit in the context of comb filtering -- which in my mind is a related,  special type of phase  cancellation -- is discussed at length by Lynn Olsson in this post at DIY audio in the "Beyond the Ariel" thread. About multidriver OBs, nonetheless  )

Now, in my understanding, the kind of cancellation presented in the graphs at the thread above is down right suspect: In my mind, for "cancellation" to occur, two things have to happen:

1) Both drivers have  to have useful output accross that range. That is (unfortunately?) no problem with (most) FRs.

2) The phase difference btw. the output of the two drivers output -- represented as phasors -- should be more than (resp. less than) 90 degress (resp. 270 degrees). "Worst" cancellation occurs when the phase shift is always 180 degrees (but in that case we simply reverse the polarity of the tweeter and live happily ever after).

Now, since the two drivers are not identical, the phase difference varies with frequency. Otherwise we could simply offset the tweeter accordingly and be done with it.

Moreover, phase mismatch has another detrimental effect (other than cancellation): Even if the output is "in phase" -- i.e. within +/- 90 degrees -- having the two outputs differ should be detrimental to the "time coherency" of the summed output (two waves arriving at slightly different times)

Where my understanding stops is how is this effect mitigated with proper  XO design (ideally a minimalistic passinve XO). IOW, how does one go about and

a) Choose a proper XO frequency.
b) Choose the right type of XO
c) How to phase-optimize the XO such that those effects are mitigated ? Where does one start and how does one go about this ?

Again, I might be wrong, but this is my current understanding. Pointers and resources highly appreciated.

Martin ? John K? Care to weigh in?  Thanks kindly...

Florian

P.S. One thing I haven't thought about is why exactly is this mostly relevant in the HF, but not in the LF ? It should be related to the wavelength, but not sure how... 

[scorpion]

Florian,

Just take your Mahler 5th with you and turn up at my place to listen. You will go no further. Sorry, do not mean to dampen the thread. 

/Erling

[MJK]

Florian,

CAUTION : OPINION

Unfortunately on the Internet forums, and in particular that DIYaudio thread, there is a lot of talking and not much engineering/designing. People tend to make rocket science out of fairly simple concepts to prove how smart they are to others in the discussion. One thing that sticks out in that thread is some of the OB people not actively participating in the discussion. Lots of talk, lots of opinions, very little design engineering or analysis.

END OF OPINION

I don't know any way of addressing your questions short of doing the actual work, the analysis required to design a speaker. I am designing another OB system at this time that does use a passive crossover and is smaller so that it is more appealing to the average DIYer. Hopefully, I will build it later this year. The steps I am taking, as I have on all previous projects, are as follows.

1. Feed the manufacturer's specs into my MathCad OB worksheet (highly modified for this application) to narrow the driver choices.

2. Look at the interaction of each driver with the baffle (size and shape concerns). This sets the basic baffle geometry and size.

3. Analyze the drivers on the baffle to nail down positions and text book crossovers (frequency and slopes) to produce a smooth combined response.

4. Using the driver impedance curves (from manufacturer's data sheets) design the passive crossover using real values of components purchased from PE.

5. If design looks promising (it does at this point) buy the drivers and measure the SPL response and actual impedance.

6. Substitute measured driver results into the MathCad OB worksheet and make any final tweeks of geometry and crossover.

7. Build the system.

8. Measure the actual response to check correlation with MathCad prediction (this has always worked out well in the past).

9. Document and post results.

I don't know any other way of designing a speaker system. Talking, debating, and agonizing over decisions without doing any engineering is a waste of effort. You learn more by doing the work then you will ever learn sifting through the information and misinformation on the Internet trying to detemine what is factual and what is speculation.

Martin

[FlorianO]

Martin,

I understand and have outmost respect your (and John_in_CR) learning-by-doing approach. Seriously, I really do.

Unfortunately I come from a theoretical background. As such I have a hard time letting go of the urge to understand the theory first...  That's just me. Bend the reality to conform with the theory . If you took a look in the A126 build thread there you might have seen that I spliced-and-diced and put Ron's kind patience at test in helping me understanding the ins and outs before even start building...

The link to the DIYaudio thread was for reference purposes only. FWIW, I also agree that it shouldn't take a thread of over 700 posts just to nail that OB down, and the gems of designs and analysis quite few and far between. It's then understandable that ppl in the know (like yourself) have better things to do than simply to contribute to the noise.

However, my inquiry about the _theory_ behind FR and tweeter integration still stands. Physics is physics. I just was under the assumption that this should be a specific application of a basic speaker design knowledge (which I am lacking). As such I was simply trying to instill a discussion that I thought is relevant to FR-on-OB crowd here. Nothing more....

Best regards,

Florian

P.S. Looking forward for your reports on the 3-way OB...

[FlorianO]

Just take your Mahler 5th with you and turn up at my place to listen.

Erling,

That is a promise !! The reason I didn't post in your venetian blinds thread was because I'm was (still am...) blind of jealousy

While you still didn't accepted my invitation to a belgian trappist beer and some music in my cellar, I will _definitively_ accept yours. Sometime I will ship the family away, throw myself in the car and drive to Söderköping for a listen.

All the best

Florian
 
P.S. Attending the tonights performance at the Götebörgs Konserthuset was a mistake: It was _that_ good and I realized how far off my  speakers are to giving me _that_ kind of "performance"..

[markC]

I'm having great success with my tweeter + b200 design. I've simulated on spice software for hours and the results are getting so close to where I want them.
I take the approach of using the manufacturer's advertised specs. and plugging in a text book x-over. That is only a starting point, because text book x-overs are just that-a starting point.
Next, I play with the values of the components in the software, add damping resistors, add resistors to add to the dcr of the inductors, change the brand of inductor to lower the dcr, etc.
When I came up with something that I liked on the program, I ordered inexpensive components, hooked them up, and listened and listened. Changed a couple of values in the l-pads of the tweets and it's sounding real nice to my ears.
So nice that I am now waiting for my more expensive components to arrive with a couple of variables thrown in for good measure.
If it works out the way I think it will, I'll be building permanent x-overs and be glad I started experimenting.
I feel that the only way to get it right is to theorize, design, build, implement, listen, adjust, listen.
This is for me in my listening room, and when they're completed, I'll be taking them to a friend's house to try them there. If they sound good in his room, I'll consider them a huge success.

[MJK]

Florian,

Unfortunately I come from a theoretical background. As such I have a hard time letting go of the urge to understand the theory first... 

I am not sure what you are looking for in your posts. You want more than advice, you don't want to use somebody else's simulations or design, so I think you are saying you want to understand the fundamentals. If you have the theoretical background, the path to what you want seems clear to me. You need to do the work yourself and gain the level of understanding that will result. I can understand that, it was where I was many years ago and that is what I did. It just took some effort and time, I don't know any shortcuts.

Martin

[johnk...]

Your question is diffult to answer but I'll try. Start by considering two drivers with perfectly flat response radiation form the same point in space. Obviously there would be no problems here. The driver would have the same amplitude and phase and since they are at the same point in space there would be no delay problems to consider. Their output would just sum to + 6dB over a singe driver. Now consider that for some reason you want to have one driver cover the response below 2K Hz and the other the response above 2K Hz. There are any number of crossover which could be used. However, a correctly designed crossover has an all pass response. That is, what ever the phase relationship between the HP and LP section is, the amplitude sums to 1.0 exactly. For example consider a Linkwitz/Riley crossover. At the crossover point the amplitude is -6dB or 1/2. The phase of the HP and LP sections are identical. Thus, 1/2 + 1/2 = 1. If yo go to a frequency where the LP amplitude is 0.75 the HP amplitude will be 0.25 and the HP and LP phase still match, thus, 0.74 + 0.25 = 1.0. For an odd order Butter worth crossover the phase difference betwen the HP and LP sections is 90 degrees. Thus the HP and LP amplitude sum like a right triangle. Sum = sqrt (HP^2 + LP^2). At the crossover point the amplitude is 0.707 so we have Sum = sqrt (0.5 + 0.5). At some point away from the crossover point the HP amplitude would be given as Sqrt(1 - LP^2) and vise versa.

Now what happens when the driver are separated? When you are at some listening position where the distance to the LP and Hp drivers is the  same there is no differential delay issue and the things still sum correctly. However, when you move to a position where you are a different distance from each driver there is a differential path length difference, call it d. This gives rise to a time delay Td = d/c where c is the speed of sound. this delay of one driver relative to the other introduces a phase shift of 2Pi x F x Td. If you like we can define a "delay frequency" as Fd = 1/Td and write the phase shift as 2Pi x F/Fd. Now if you conside the behavior of the phase shift vs. frequency you will see that it varies for 0 at Dc to Pi/2 at F = Fd/4, to Pi at Fd/2 to 2Pi at Fd and continues to rotate as F goes higher. So if there is no crossover you can see that as the phase shift rotates with frequency at the listening point there Will be frequencies where the phase shift is Pi (Fd/2, 3Fd/2, 5Fd/2.....) and there Will be cancellation. When a crossover is in place  these same phase shifts will occur but the idea is that by the time a significant phase shift is generated the amplitude of one of the driver will be attenuated so that it doesn't contribute to the sum. The way to do this is the keep Fd large relative to the crossover frequency and to use steep slope crossovers. Keeping Fd high compared to F crossover means close driver spacing in general.

Now the problem with adding a super tweeter to a FR driver or any system is that the x-o frequency is typically high (if an x-o is used). As a result, even a relatively small driver separation can yield significant combfiltering at all but the design point. Adding a ST with only a HP filter (allowing the FR driver to run full range) make the situation worse. To avoid the problem crossover frequencies in the 1 to 3K Hz are usually implemented, but this defeats the intent of a FR driver. The botton line is that adding a super tweeter to any system (not just a FR) isn't really a good ideal if wide, uniform dispersion is desired.

[FlorianO]

Martin,

You want more than advice, you don't want to use somebody else's simulations or design

I did used Ron A126 horn design, incredibly good; I use John K ABC dipole material; I do have your U baffle design (smth very much like Erling's) on my short list  (as I mentioned in the other forum). Looking forward for Rons comming hybrid design and your 3-way OB.

I think you are saying you want to understand the fundamentals.  You need to do the work yourself and gain the level of understanding that will result. I can understand that, it was where I was many years ago and that is what I did. It just took some effort and time, I don't know any shortcuts.

Point well taken. Initially I thought I could delve into very specific issues and get by with that  . Not least to help sorting out the good gems of engineering from the BS. Anyway you're right (again...): if one is _really_ serious_, there's no shortcut. I should have known that...

[FlorianO]

John,

Many, many thanks for your analysis. Highly appreciated.

Adding a ST with only a HP filter (allowing the FR driver to run full range) make the situation worse...The botton line is that adding a super tweeter to any system (not just a FR) isn't really a good ideal if wide, uniform dispersion is desired.

If not too much to ask, one last detail: Care to develop on why  an ST with a HP  makes the situation worse ? I was thinking the case where FRs usually roll-off (10k or so), not very much useful output above that, and the HP maybe even an octave lower....

Anyway. Kind thanks for your analysis, yet again.

Regards,

Florian

[Zero One]

 Hi Florian I understand your desire for mastery of the theory behind the set-up, in my area of business (digital imaging ) it is paramount, but I have found after much frustration that with OBs and audio even when the numbers are right the sound may be wrong.  One of the problems and especially with OBs is that the room has an awful lot of influence on the sound and may well swamp totally any issues to do with phase etc unfortunately it is not really possible to take all this into account at the design phase.  To a degree I feel the best appraoch is to build your whole system with the room in mind, in other words as a symbiotic whole.  The only way to really do this is get out the soldering iron, grab some drivers and take the long road, so far my road has taken about two years, and the sound is now very nice indeed but still room for improvement of course.

As for the phase issue I think a lot of this gets blown out of proportion, what is more of a problem is messing up soundstaging because the sound is coming no longer from a single point, in my case my super tweets are centered in front of the main FRs and I moved them in and out till they sounded right, as for the cross over well I tried different caps, Lpads etc till it sounded right, which turned out to be just below 9Khz with a 6db slope.

Other potential issues include your seating position as the highs will always beam a bit and what is heard from just in front where many make the measurments might be very different from what happens at your ears, then of course there is your own hearing response and reflections from behind the seating position and lots more as well...............

Probably in the end if we really want to nail phase issues and point source advantages the optimum answer is a top grade coaxial driver, and it surprises me that there are not a lot more of them out there. Have a look at the Paudio drivers from CommonSense  audio, that might be an option, as I think from another forum you were after a system which was able to do a big bold no holds barred performance? I've never heard these but from the specs I think they would be a wild wild ride if properly implemented.

[johnk...]

It's not so much adding a ST to an FR, it's adding a ST to ANY system. At 10 k Hz the wave length is about 1.3". The physical size of the driver would typically mean that the positions of the ST and the other driver will be separated considerably more than a wave length. When you start to triangulate from the center of the main driver and the ST to the listening position you will find that you don't have to move vertically very far off the design axis to start seeing cancellations due to the the changes in path length. The point is simply that the higher the crossover point is the more sensitive the radiation pattern is to driver offset, separation, and off axis listening positions.

Do a simple test for your self. Just mount two tweeters on a baffle about 6" appart and play some pink noise though them. Sit about 1 or 2 M away and move your head around. You will clearly hear the variations in sound due to destructive interference at different head positions.

[jlo]

Do a simple test for your self. Just mount two tweeters on a baffle about 6" appart and play some pink noise though them. Sit about 1 or 2 M away and move your head around. You will clearly hear the variations in sound due to destructive interference at different head positions.

I did a crossover as software (it's free), see http://www.ohl.to/about-audio/audio-softwares/ls-filter/
you can simulate a filter and directly hear/see the result of moving the speakers by changing the "delay"
hope the image is not too big !


By the way, I also did another software to listen to the effect of loudspeaker position near walls, with the possibility to simulate differents speakers (OB, closed, reflex). It'a very interesting to listen to the walls effect when you change the type of loudspeaker (OB is not bad at all...)
see http://www.ohl.to/about-audio/audio-softwares/room-sound/