2K said:
I didn't say a single component. What I meant was, for the exactly the same reason you mentioned above ("real world drivers" with natural rolloffs), to impelement a first order acoustic target, using a first order electrical filter PLUS notch filters is always better than using a higher order electrical filter (plus whatever response shaping).
In fact, there is hardly "enough" for a first-order transient-perfect design. Due to the natural rolloff of real world drivers, the result is almost always closer to "quasi transient-perfect 2nd order," as John Kreskovsky calls, than to true first-order, and the phase shift around Fc is greater than 90 degrees. See:
http://www.geocities.com/kreskovs/Quasi-transientP.html
http://www.geocities.com/kreskovs/CrossoverdocN.html
This is exactly what you described in your post---shallow slope around Fc but steeper slope far from Fc. For most real world drivers, if you use a higher order electrical filter, the result will be even worse---much steeper slope far from Fc and as a result more phase shift around Fc. This does not mean an on-axis flat response is not obtainable. But there is always a better way than this. I don't see any circumstances with real world drivers in which a greater than 1st order electrical filter is required to implement a 1st order T-P design (or quasi T-P 2nd order in most cases). As I said above, if you have a rising response, it is always better to use a notch filter than a higher order filter in this type of design.
Jay_WJ said:
Jay
I can't look at Johns site at work but I'll take a peek later.
There is no true TP 1st order design. It can't happen with real drivers since they are bandpass devices whose frequency response does not reach from light to DC.
The best you can hope for is the quasi TP or as I call it Transient OK. That is state of the art for 1st order arrangements commercial or DIY. It is capable of a good step response, something resembling a square wave and low phase rotation. It's far better in these respects to any other order filter if that's your goal.
As for the steeper slope far from FC this is primarily driven by the acoustic rollof of the driver itself. That will be generaly somewhere where from 2nd to 4th order. A little additional rolloff at that point from the filter isn't going to affect overall phase that much and a little extra attenuation can be usefull. Tweeters for instance like a little more rolloff down low.
You have the same issues with higher order crossovers as you know. Most driver responses don't stay on target two octaves past Fc.
I will look into your suggestion of using notch filter to flatten the response on top of the filter. I have used this technique before to tame an unruly response knee on a higher order filter but not in this application.
1st order filtering....
There are a few successful and respected manufacturers that use these. Thiel and Vandersteen are the ones that come to mind. Have you ever seen pictures of their crossovers? There are a lot more than one part in their crossovers to optimize the crossover for 1st order filtering. In fact, they my use more parts to make the 1st order filter work than most companies would use for a second, third or fourth order filter.
A few conventional two way designs I can think of are the Sonus Faber Extrema and the Audio Physic Step that took the minimalist approach to crossover design.
http://www.stereophile.com/standloudspeakers/915/
C
There are a few successful and respected manufacturers that use these. Thiel and Vandersteen are the ones that come to mind. Have you ever seen pictures of their crossovers? There are a lot more than one part in their crossovers to optimize the crossover for 1st order filtering. In fact, they my use more parts to make the 1st order filter work than most companies would use for a second, third or fourth order filter.
A few conventional two way designs I can think of are the Sonus Faber Extrema and the Audio Physic Step that took the minimalist approach to crossover design.
http://www.stereophile.com/standloudspeakers/915/
C
2K said:
I have used it in a 3way some years now ...Inductor-RC-notch or Incuctor-notch-RC...which one is better I dont know but I use it on both woofer and mid
Highpass are single series cap and a paralel Inductor with a series resistor...what to call it I dont know...IR ?...anyway I believe that downwards its affecting impedanse in the same manner as RC does upwards, controlling the series component
I can only say that it works, but having no measurement equipment nor sims its really nice to get an explanation as to why it sounds better with notch/RC rather than additional series inductor
That's what I said, exactly.
The question in actual application is, how large phase errors does that "additional rolloff" cause? In general, lower order xovers are much more sensitive to nearby phase errors than higher order xovers. For a real world quasi T-P design, there is no point making it "more quasi" by using a greater than 1st order electrical filter. That's what I mean. If you want some more protection for a tweeter, using a notch filter is a better option. An example is what Joe Rasmussen did for his Elsinore design, though this design is not 1st order but basically 2nd order LR:
http://www.customanalogue.com/elsinore/elsinore_17.htm
IMO, it's a little overprotection with a very deep notch, but should be okay since it's a 2nd order design.
I think the following says a lot about what Thiel does:
Figure 3:
http://www.thielaudio.com/THIEL_Site05/PDF_files/PDF_tech_papers/techpaper_general1.pdf
Figures A to D:
http://www.thielaudio.com/THIEL_Site05/Pages/Tech/amplituderesp.html
I.e. they use equalising (like notches for instance) combined with lead or lag filters. I.e. filters that don't continue to roll off beyond a certain frequency.
While this isn't much of a problem with normal woofers - you definitely won't do it with your average tweeter or midrange.
Regarding phase-shift: Although delay and phase-shift isn't the same thing it is still possible to phase-align the drivers to some degree by the use of time-delay (i.e. the slanted baffle).
Regards
Charles
Figure 3:
http://www.thielaudio.com/THIEL_Site05/PDF_files/PDF_tech_papers/techpaper_general1.pdf
Figures A to D:
http://www.thielaudio.com/THIEL_Site05/Pages/Tech/amplituderesp.html
I.e. they use equalising (like notches for instance) combined with lead or lag filters. I.e. filters that don't continue to roll off beyond a certain frequency.
While this isn't much of a problem with normal woofers - you definitely won't do it with your average tweeter or midrange.
Regarding phase-shift: Although delay and phase-shift isn't the same thing it is still possible to phase-align the drivers to some degree by the use of time-delay (i.e. the slanted baffle).
Regards
Charles
phase_accurate said:
I looked closely at what Thiel did for a tweeter's HP filtering (Figure 3) and for a midrange's LP filtering (Figures A to D). First, for the tweeter, it seems that they didn't even use a series cap as a means of 6 dB/oct HP filtering (see the network's transfer function). The network should only consist of a bunch of series and parallel notch filters. Very impressive.
Next, for the midrange, the network seems to use one primary coil and a bunch of notch filters (again, closely look at its trasnfer function). The response is faithful to the target up to 10 kHz!
Time alignment by slanted/stepped baffle (or ladder delay network on a flat baffle) isn't optional for low (1st or 2nd) order xover implementation. It's a must. Without it, phase will be misaligned to the extent that intended summation is ruined. For higher order xovers, asymmetric slopes work pretty well.
tinitus said:
I have used it in a 3way some years now ...Inductor-RC-notch or Incuctor-notch-RC...which one is better I dont know but I use it on both woofer and mid
Highpass are single series cap and a paralel Inductor with a series resistor...what to call it I dont know...IR ?...anyway I believe that downwards its affecting impedanse in the same manner as RC does upwards, controlling the series component
I can only say that it works, but having no measurement equipment nor sims its really nice to get an explanation as to why it sounds better with notch/RC rather than additional series inductor
A sharp peak is exceptionally audible because it rings. On a waterfall plot, it will extend forward for quite a distance, indicating a long duration after the original stimulus is gone, so the total energy is quite high. A correctly matching notch, of the same Q, not only notches out the peak, but the extended ringing.
I certainly agree with the Thiel comments on diffraction. If heroic efforts are taken to reduce diffraction, the resulting smoothness/musicality certainly justifies the effort.
I dont use drivers with pronounced ringing, just elevated FR(SS 15W)...ahh, not true, my SS 8" woofers does peak
But I understand what you are saying about the ringing...and I hear it too...I try to deal with it the best way I can, allthough I am not as skilled as you guys...the method with controlling Q with resistors does seem to work, but a real pain to get right
my "master" says..."no matter what you do electrically, the ringing is still there"...its barely audible, but still there
Ahhh, I think I misunderstood you...You say that a correctly Q alligned notch DOES keep the extended ringing under control...question is whether it is enough
But if 6db filters means using especially peaked drivers, I have my doubts...on the other hand I sometimes think that we need just a slightly little ringing to get some life in the music
is Q controlling a valid way to better sound ?
But I understand what you are saying about the ringing...and I hear it too...I try to deal with it the best way I can, allthough I am not as skilled as you guys...the method with controlling Q with resistors does seem to work, but a real pain to get right
my "master" says..."no matter what you do electrically, the ringing is still there"...its barely audible, but still there
Ahhh, I think I misunderstood you...You say that a correctly Q alligned notch DOES keep the extended ringing under control...question is whether it is enough
But if 6db filters means using especially peaked drivers, I have my doubts...on the other hand I sometimes think that we need just a slightly little ringing to get some life in the music
is Q controlling a valid way to better sound ?tinitus said:I dont use drivers with pronounced ringing, just elevated FR(SS 15W)...ahh, not true, my SS 8" woofers does peak
But I understand what you are saying about the ringing...and I hear it too...I try to deal with it the best way I can, allthough I am not as skilled as you guys...the method with controlling Q with resistors does seem to work, but a real pain to get right
my "master" says..."no matter what you do electrically, the ringing is still there"...its barely audible, but still there
Ahhh, I think I misunderstood you...You say that a correctly Q alligned notch DOES keep the extended ringing under control...question is whether it is enough
But if 6db filters means using especially peaked drivers, I have my doubts...on the other hand I sometimes think that we need just a slightly little ringing to get some life in the music
Slight misunderstanding perhaps. The peaking is undesirable; an ideal driver for low order crossovers would be flat and extended well above and below the desired range, with smooth rolloffs. But many good drivers do have peaks; with new, stiffer, cone materials superseding polyprop a few years ago, but lacking polyprop's internal damping, there were some drivers that were very difficult to design with, and which were discontinued quickly. However there are still some very good drivers with more reasonable peaks, and notching (or very steep filters like Joseph) is needed with them. The Thiels are well behaved compared to some other metal coned drivers. In any event, modern drivers have tended towards a livelier sound, so assistance should not be needed.
As a random observation, the woofer/mid transition is less critical, and 2nd order seems to work very well; and using the sealed cabinet/mid rolloff does get around the midrange series capacitor problem.
So the first step if you want a low order xo system is to choose the drivers very carefully with that application in mind. This tends toward more expensive drivers... and hands on evaluation of course. We evaluated about 8 mids before settling on the final choice.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.