I think correcting phase and frequency is one thing and minimising distortion is another thing. Distortion in speakers comes from an unlinear BL curve, unlinear suspension and variation of inductance over throw plus rubb and buzz. Prof. Klippel spend halve his life in researchning posibilities to model dynamic spaekers and find a way to digitally correct distortion in speakers. Theoretically improvements are posible but peak throw and rubb and buzz put an ultimate limit on it. I refer you to the Home website to study some of the incredible body of work he did. I belong to the Klippel Advanced Users Group and attended many seminars and had the chance to talk to Wolfgang many times. To put it simple, what he does goes over my head.
Soon, you can study the "fold down effect" when you maesure a hard dome tweeter that has a peak slightly over 20kHz. When you maesure only to 20kHz everything looks fine.
If you measue up to 40Khz for example and make a CSD it can happen that the peak broadens in the time domain and spreads out to lower ( maybe audible ) and higher ( ultrasonic ) frequencies based on the Q of that resonance.
Soon, you can study the "fold down effect" when you maesure a hard dome tweeter that has a peak slightly over 20kHz. When you maesure only to 20kHz everything looks fine.
If you measue up to 40Khz for example and make a CSD it can happen that the peak broadens in the time domain and spreads out to lower ( maybe audible ) and higher ( ultrasonic ) frequencies based on the Q of that resonance.
Here a paper from Prof.Hawksford with off axis plots. I think i already posted his paper about IIR and FIR filters.
http://www.essex.ac.uk/csee/researc...docs/C50 Digital crossover impaired polar.pdf
http://www.essex.ac.uk/csee/researc...docs/C50 Digital crossover impaired polar.pdf
In a tweeter I think that some of this peaking may not be actual dome breakup, but rather is the transition region where, due to dome geometry, the time delays between rim energy and dome tip energy is a significant percentage of a wavelength. Move off-axis slightly and this response may change significantly. It is a distortion that peaks on-axis, but may change significantly off-axis. As such it's distortion contribution is highly axis dependent.
Seems to me that it is not easy to determine how much is due to geometry and how much is dome breakup. I suspect, but don't know with assurance, that the actual dome breakup (resonances in the dome) is in the region above the first peak due to geometry. This is usually partially limited by "phase shields", of course.
I would think that one of the Klippel systems could separate these. The modeling of the geometry should easily show where the time delta issue comes into play.
Dave
Joachim,
Yes i would like a pair of 3liter spheres. Using WinISD beta i calculate a needed volume of 2.6L for the planed OEM, Vifa PL11 style, mid-range drivers. 3L’s calculates out to a Qtc ~ 0.69
No worries, there is no big rush, i do this for fun. The information i posted said 5 inch. The diameter measured inside the rolled rubber surround is ~ 75 mm.
Just for interest, have you calculated the Qtc for your mid-range in the spheres you plan to use?
What thoughts do you have regarding stuffing in your mid-range? There was stuffing in the water bottle enclosure photo that you posted. All that energy storage stuff you know.
DT
All just for fun!
Yes i would like a pair of 3liter spheres. Using WinISD beta i calculate a needed volume of 2.6L for the planed OEM, Vifa PL11 style, mid-range drivers. 3L’s calculates out to a Qtc ~ 0.69
No worries, there is no big rush, i do this for fun. The information i posted said 5 inch. The diameter measured inside the rolled rubber surround is ~ 75 mm.
Just for interest, have you calculated the Qtc for your mid-range in the spheres you plan to use?
What thoughts do you have regarding stuffing in your mid-range? There was stuffing in the water bottle enclosure photo that you posted. All that energy storage stuff you know.
DT
All just for fun!
DT i bought a 6 liter sphere because i am planning to try a second driver in the back that comes in at lower frequencies to restore some output in the lower reaches. This bigger sphere also gives the tweeter some baffling so it is all about loosing not to much effiency.
The woofer is quite efficient and i may not be able to get enough efficiency out of the mid-treble array. In that case and when the experiment with the second midrange fails i plan to do the woofer section active. You see there is a lot more work to do and i have other tricks in stock so this project is just in the experimental phase but i think the brainstorming is done mostly. For a single midrange 3 liters is better concerning usefull Q and it will work with your Viva. For stuffing i use a material called Twaron. It is sold under the name "angel hair" but it is also used as a bymix for car tires and in Pampers for babys plus other uses. I stuff the cabinet and watch the phase of the impedance curve. I put in just as much until the small wrincles in the phasecurve (and to a smaller visible degree in the impedance magnitude ) disappear. I do not like overstuffing but no stuffing material makes the sound quite hollow.
The woofer is quite efficient and i may not be able to get enough efficiency out of the mid-treble array. In that case and when the experiment with the second midrange fails i plan to do the woofer section active. You see there is a lot more work to do and i have other tricks in stock so this project is just in the experimental phase but i think the brainstorming is done mostly. For a single midrange 3 liters is better concerning usefull Q and it will work with your Viva. For stuffing i use a material called Twaron. It is sold under the name "angel hair" but it is also used as a bymix for car tires and in Pampers for babys plus other uses. I stuff the cabinet and watch the phase of the impedance curve. I put in just as much until the small wrincles in the phasecurve (and to a smaller visible degree in the impedance magnitude ) disappear. I do not like overstuffing but no stuffing material makes the sound quite hollow.
I'm very much interested in your findings with this Scan Speak as I'm pondering different loading methods at the moment for use of the same driver in a new design.
This is the link Joachim provided that was the subject of Michael's comment.
Issues in speaker design - 2
If you examine the section towards the end, it discusses the tweeter's "diffusor cap".
Not sure how you got onto the midrange test link.
This is the comment;
"No SL's "stored energy" term isn't interesting at all - its mere B*S*T - and anybody is jumping at it because it applies to each and everything that makes up for the slightest wiggle in FR and because it was coined by an audio icon - sorry Siegfried ..."
I went back to the orginial discussion of "stored energy" Yes I did read it all!
DT, Siegfrieds gated toneburst test is a valuable one. I found it has better resolution then an "over all" CSD or Wavelet for that matter. Many people have not tried this test because the lack of equipment and then they have the guts to critisise it based on theoretical asumptiions. War is much easier then piece.
The CSD etc. methods are good for overall visualization. Siegfried used a different window that takes out what he feels are artifacts. I like to look at CSD using rectangular window.
The tone burst method is good for detailed understanding of behavior at specific frequecies and verification of solution to a problem. I'm sure I've mentioned this before, it's always best to look at various data and cross reference.
Whenever we use a specific method, we must know why we use it.
The tone burst method is good for detailed understanding of behavior at specific frequecies and verification of solution to a problem. I'm sure I've mentioned this before, it's always best to look at various data and cross reference.
Whenever we use a specific method, we must know why we use it.
That is exactly what the UE is capable of doing, with or without phase correction. This, of course only applies on the design axis, but if the design is CD then it will follow that the corrections will also apply off axis. But if there is a 3dB dip on axis that becomes a 3dB rise off axis, then eqing flat on axis will result in the obvious 6db rise off axis. There is still a 6 dB difference at that frequency between on ans off axis in that case and which sound better, er'ed flat of left as a 3dB dip on axis, with depend on other factors, like listening positition, etc.
That is why it is better to strive for CD type design (over some reasonable off axis window and frequency range) where the off axis behavior follows the on axis, like this:
It really isn't anything hard to do, or a particularly earth breakingly new idea.
I have only tried equalizing as a system. But since it was a small full range system, the result was much better than I could do passively. I do realize that sound card quality becomes also very critical. It is true that the off axis profile is going to be also critical. I did realize that the equalization range can for each driver can be set differently from the crossover points, but had not explored it yet. The fun for me was getting proper low end out of a well behaved 3", and saturating the sound card outputs before the drivers bottomed. Some of my discs have to be re-ripped because I encoded them to use up full 24bit range. Now I realize how dumb it was, and only resample to 48KHz so that I can have everything the same rate during playback.
To me, the UE is the next best thing since sliced bread as I've never cared to work with DSP until it became available. Prior to the UE I used SoundEasy (the origin of the UE) to simulate and audition classic filters, but still did not care to venture into things like dipoles. It is so easy to create crossovers with things like dipole EQ or Linkwitz Transforms that I now do it routinely. It's a tool, an excellent one.
This is another example of, in this case, totally misunderstanding the debate on minimum-phase. That was in regard to a raw driver's response, the fact that software can model the response, that this response must therefore be minimum-phase as it's a requirement for accurate modelling and the fact that it is a requirement in order for crossover networks to be able to apply EQ appropriately. It had nothing to do with the final response of a system, and second, what the final response is with software such as the UE, if one chooses that option. You are again misinterpreting what was said and misunderstanding it as well.
One can either EQ the magnitude only, as has been described, whereby the system response will be the classic LR8 with it's classic phase or one can choose to EQ the phase as well. In that case, the phase response can be made linear, not just minimum-phase. If you paid any attention to the actual measurements I posted, you'd have seen the linear phase response in each section. The transient response of an LR8 with EQ of magnitude and phase through good DSP such as the UE is that of a linear phase system and unquestionably superior to the passive or active minimum-phase versions. As I have said once already, whether or not it's audible is another debate. You are missing or forgetting details of the discussion.
Now if you care to find out the basis on which an LR8 acoustic magnitude response worked best for my case, fine. This was detailed in a thread on the PE board, I've seen a post by you there so I know you're familiar with that board. I did not say much here because this thread was not about my design. I provided input as example of the capabilities of the software as one alternative in the DSP avenue.
It would probably be instructive for you to read the referenced AES paper as it includes a fair amount of discussion on the minimum-phase aspect of drivers and its basis in the software, any software that is designed to EQ speakers. I did not make up any of this.
The UE is only a tool, but it happens to be one of the easiest and best currently available IMO. You keep either inferring things not said or misunderstanding what has been said, both by John and by me. That is not our fault.
Dave
Last edited:
The LR8, like any LR crossover, sums flat with zero interdriver phase difference. That is the phase of the HP and LP sections track each other. As you note, a highe rorder LR crossover will generally have worse transient distortion due to greater phase rotation. However, with a dsp tool like the UE (not unique to the UE) the acoustic output of the combined filter and driver can be made to match the LR8 anplitude, and the acoustic phase can be made to be linear. Thus an acoustic crossover with LR8 HP and LP amplitude but linear phase can be achieved. Since the phase is linear there is no transient distortion introduced on the design axis. Off axis some transien error will be introduced but this is generally much less than that of a crossover based on MP filters.
At this point - may I also reference :
http://www.diyaudio.com/forums/multi-way/63078-how-pc-xo.html
for any in depth discussion regarding PC XO solutions using convolution techniques (UE not the only option to await / others being out for quite some time)?
Pretty sure Ultimate Equalizer will be discussed there at length as well (and "not only" as teaser / for advertising ) - *if* - it will become available one day.
Michael
http://www.diyaudio.com/forums/multi-way/63078-how-pc-xo.html
for any in depth discussion regarding PC XO solutions using convolution techniques (UE not the only option to await / others being out for quite some time)?
Pretty sure Ultimate Equalizer will be discussed there at length as well (and "not only" as teaser / for advertising
) - *if* - it will become available one day.Michael
SE v17 has an early version built in, just limited to 3-way systems. That's what got me started. It's under the DSP Tools menu, "Ultimate System Linearizer".At this point - may I also reference :
http://www.diyaudio.com/forums/multi-way/63078-how-pc-xo.html
for any in depth discussion regarding PC XO solutions using convolution techniques (UE not the only option to await / others being out for quite some time)?
Pretty sure Ultimate Equalizer will be discussed there at length as well (and "not only" as teaser / for advertising
Michael
Dave
Putting on my audiophile hat, I'd like to mention it was UE that prompted my final decision to load all my CDs onto a NAS. I'm not going to say it's the final word, but it does provide quite an improvement in the accurracy playback with very little effort. In the process, I also discovered that reading audio discs was not as error free as I originally thought, so this was a new experience for me as well.
Ever since SE had implemented the equalizer, I had continuously been bugging the author for a stand alone product that would do the playback based on SE designed project, as a result, we go something even better, a totally standalone product.
Putting on my engineering hat, I think Joachim would appreciate more ideas that evolve around the original design goal instead of argument about what tool is best. I personally am interested such discussion as well. Sorry that I brought UE into the discussion, but it just seemed to provide measureable performance in concurrence with what I first heard before I did EQed measurements, and also seemed to be in line with what Joachim is trying to accomplish, "speakers disapearing". I certainly hope that anyone that has tools with better measureable performance can stand out and show it.
Last edited:
He may need some time to digest - there's been brought quite something on the plate - LOL
I mean - at least he can not complain about not having gotten "food for thoughts" - given that he started out playing around with simple bowls
Michael
Last edited:
For anyone thinking of addressing the lion's share of response anomalies with EQ who might be buying into this continuous stream of rubbish - I will present once again the counterpoint, fact checking view. Passive crossovers as most of us again with a reasonable amount of experience know
can only absorb energy on a selective basis from that which is applied to the speaker terminals from the amplifier. EQ on the other hand has the potential to add energy to select bands. While the two methods can have essentially the same effect when subtracting energy, the same cannot be said for eliminating dips in a driver's response. And sometimes (some might argue often) when a dip occurs at one frequency, the energy "absorbed" at that frequency appears at a different frequency in the form of distortion. This behavior routinely occurs when drivers enter the breakup region. Hence when we add energy via EQ, we sometimes wind up with unintended consequences. Contrary to the above quoted ridiculous oversimplification, EQ is not equivalent to passive crossovers and the above claim - "any crossover is nothing more than EQ" - is patently false. An often cited advantage of actively controlled speaker systems is the maintenance of sensitivity. The natural tendency is to boost dips rather than suppress peaks when it comes to "evening out" response. So the distinctions and the warnings by people like Linkwitz and myself about the limited usefulness of EQ as a high performance loudspeaker management tool should be at least considered if not fully heeded. Active speaker control is still a potential advantage but the end result needs to be carefully listened to and measured for increases in distortion that do in fact occur - contrary to assertions of others within this thread.
And now, hopefully the person who assured us he was going "to bail" earlier will live up to his promise and allow this particular thread to cease becoming yet another advertisement for JohnK's Sound Easy instructional download or some other plug for Sound Easy v17.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.