What speaks against ring tweeters?
http://www.speakercity.com/Merchant2/merchant.mvc?Screen=CTGY&Store_Code=SC&Category_Code=VifaXT
http://www.speakercity.com/Merchant2/merchant.mvc?Screen=CTGY&Store_Code=SC&Category_Code=VifaXT
Hi,
exactly. Perhaps you didn't noticed but you put two (possible) reasons in your assumption.
First:
in this kind of CSD resonances at higher frequencies look not so bad as at lower frequencies, because of the linear time scale. Linkwitz explains this (Issues in speaker design, Section M - I think). So if one tweeter has a resonance freq. @1kHz and the other @2kHz, the latter one seems to have a better behaviour.
A better CSD is proposed by Linkwitz, but there I only know three measurement systems that can display that, and they're all from good ol' Europe: ATB (does a measurement with shaped tone bursts, slow), ARTA (next version, convolves the measured impulse response with a shaped tone burst, fast) and esweep (does it like ARTA, f'ing fast if you know how to handle this fine piece of software). SoundEasy (used by Zaph) can also convolve the impulse response with a shaped tone burst but can only display it with a linear time axis instead of normalizing it to periods (as ARTA and esweep do), AFAIK.
Second, and I suppose that's the point here:
a resonance frequency with a Q>0.5 is able to oscillate and this can be seen in every CSD. The frequency response of both Peerless and Seas looks like a low Q (<=0.5) so there is a clean CSD.
Of course, your conclusion is right: with a XO this will be completely different, or, as I said before: same (complex) frequency response, same dynamic behaviour, thus same CSD.
---
Ring radiators like the Vifa's or Scan Speak are a fine thing, for voodoo and marketing. Their HD (2nd order, 3rd order is very low) is high, and they tend to "scream" at relative low levels. Last year I did some series measurements on them (sorry, I can't publish it, was a job) and I think I will never use them in any project.
Although I'm not a friend of ribbons (or magneto-/elektrostatic speakers) because of their usually high distortion a big ribbon may be a solution for you if you want to have a linear energy response, because they will match better to your 6" Usher.
The mentioned Audaphon is a really good one, with low distortion >2kHz. Actually, the HD is low above 1kHz, but there seems to be an internal resonance corresponding two some flaws in the CSD and the HD at 1.5kHz, with a F3@1%@90dB. And its quite expensive
Another candidate would be the B&G Neo8, low distortion >1kHz, flat response up to 18kHz.
But I think your selection is quite good, but you should really have a look at the Audax tweeter, as it is all you want. And there is nothing bad about this, I checked it up to abnormal levels (sorry, no measurement, just with my ears, so no guarantee, see my post above) and there was no audible distortion.
Bye
Baseballbat
Jay_WJ said:
exactly. Perhaps you didn't noticed but you put two (possible) reasons in your assumption.
First:
in this kind of CSD resonances at higher frequencies look not so bad as at lower frequencies, because of the linear time scale. Linkwitz explains this (Issues in speaker design, Section M - I think). So if one tweeter has a resonance freq. @1kHz and the other @2kHz, the latter one seems to have a better behaviour.
A better CSD is proposed by Linkwitz, but there I only know three measurement systems that can display that, and they're all from good ol' Europe: ATB (does a measurement with shaped tone bursts, slow), ARTA (next version, convolves the measured impulse response with a shaped tone burst, fast) and esweep (does it like ARTA, f'ing fast if you know how to handle this fine piece of software). SoundEasy (used by Zaph) can also convolve the impulse response with a shaped tone burst but can only display it with a linear time axis instead of normalizing it to periods (as ARTA and esweep do), AFAIK.
Second, and I suppose that's the point here:
a resonance frequency with a Q>0.5 is able to oscillate and this can be seen in every CSD. The frequency response of both Peerless and Seas looks like a low Q (<=0.5) so there is a clean CSD.
Of course, your conclusion is right: with a XO this will be completely different, or, as I said before: same (complex) frequency response, same dynamic behaviour, thus same CSD.
---
Ring radiators like the Vifa's or Scan Speak are a fine thing, for voodoo and marketing. Their HD (2nd order, 3rd order is very low) is high, and they tend to "scream" at relative low levels. Last year I did some series measurements on them (sorry, I can't publish it, was a job) and I think I will never use them in any project.
Although I'm not a friend of ribbons (or magneto-/elektrostatic speakers) because of their usually high distortion a big ribbon may be a solution for you if you want to have a linear energy response, because they will match better to your 6" Usher.
The mentioned Audaphon is a really good one, with low distortion >2kHz. Actually, the HD is low above 1kHz, but there seems to be an internal resonance corresponding two some flaws in the CSD and the HD at 1.5kHz, with a F3@1%@90dB. And its quite expensive
Another candidate would be the B&G Neo8, low distortion >1kHz, flat response up to 18kHz.
But I think your selection is quite good, but you should really have a look at the Audax tweeter, as it is all you want. And there is nothing bad about this, I checked it up to abnormal levels (sorry, no measurement, just with my ears, so no guarantee, see my post above) and there was no audible distortion.
Bye
Baseballbat
Baseballbat,
The exact Audax model you suggested is not available here. What I found are:
TW025A0 1" textile dome
TW025A2 1" textile dome (4 Ohm version of the above)
TW025A20 1" titanium dome (expensive $103)
All these tweeters do not have a rear chamber and have a very high impedance peak above 1 kHz. I don't think they are suitable for my project.
The exact Audax model you suggested is not available here. What I found are:
TW025A0 1" textile dome
TW025A2 1" textile dome (4 Ohm version of the above)
TW025A20 1" titanium dome (expensive $103)
All these tweeters do not have a rear chamber and have a very high impedance peak above 1 kHz. I don't think they are suitable for my project.
Hi Jay. I was wondering about your comment regarding teh notch filter that "filled in" the response dip. Since the dip is near the knee of where you intend to cross I assume you came up with notch filter that caused a bit of peak at that knee? If so I believe this can be source of ringing within the XO. The transfer function will tell you for sure. if it peaks at the knee it's high Q and will ring. How bad it is I couldn't say though...just thought I'd mention it.
Also this baby is flat as a board to below 800Hz: http://s139.photobucket.com/albums/q288/augerpro/ScanSpeak 9800/
Check the data sheet for excursion though, I think it was kinda small if I remember right.
Also this baby is flat as a board to below 800Hz: http://s139.photobucket.com/albums/q288/augerpro/ScanSpeak 9800/
Check the data sheet for excursion though, I think it was kinda small if I remember right.
augerpro said:
Yes, you're right. That's the method I used (actually the only way that can do this in a passive XO). I agree that it can "ring" if this happens in a range where there is no dip. That is, if this method causes a very noticeable bump, it will be clearly audible. But if it's used at a frequency range where there is an obvious dip, I think it's okay. This response rise by a notch filter, for example, was used to elevate the top end dip of the RS28A in Dave Brown's dB MTM, Roman B's RS180/RS28A MTM and Daedalus.
But I didn't go too wild on this: just a little bit elevation of the dip. Below are the net transfer function of my XO design for the Usher and RS28A, and its predicted FR and phase alignment:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
A little bump at the crossover frequency is intentional here. I found that the dip of the sample measured by Zaph is a bit shifted to the left, compared to another sample measurement I found. Also, note that a notch filter is also used in the tweeter net to address the top end dip of the RS28A.
For comparison, below is my simulation of an existing XO design for the Usher 8945P by A Feyz P. A tweeter used here is Seas 27TDFC:
An externally hosted image should be here but it was not working when we last tested it.
See how noticeable the dip is. You can see a very similar but measured FR at http://members.fortunecity.com/pirimoglu/Design/km-u2/KM-u2.html
-Jay
The effect on the instantaneous response is not a problem (it's smoother afterall), but the effect in the time domain is. Honestly though I don't know how much of problem it is, since I'm still very new to this hobby. I'd be interested in hearing some more experienced advice on ringing in the XO. I wonder if you could get a compromise, like accept somewhat of a dip for better behavior in the time domain?
The high Q filter is exactly the same as a high Q box alignment for a woofer. The response will (mostly) track the impulse (whether a tone or actual music or whatever) but will not settle in time when the impulse stops. It will overshoot and keep outputting energy, ie, ringing. I suppose this nonlinear distortion.
augerpro said:
Hi Brandon,
Are you sure they are the same? I'm not sure, but here's what I think. High peak SPL due to high Qtc in a sealed box, for example, is caused by the interaction of both mechanical and electrical factors. Underdamping and less controlled cone movement (and resulting poor transient) in this case are caused not only by high SPL but also by changed mechanical Q in a sealed box. I think this is why the sensitivity at Fcb can be magnified beyond the driver's inherent free-air SPL. But high Q filter is purely electrical and transient response or group delay must be directly related to SPL only. As long as the summed response is flat, I don't think there is a transient problem. I think a hint is that the rise of SPL due to filter Q cannot exceed the driver's inherent SPL.
I don't understand this stuff very well. But I think this is different from high Q box alignment. What do you think?
-Jay
P.S. I just looked at Speaker Building 201 and LDC. I think the term "ringing" is used only for high Q of box alignment, but not for Q of XO filter. Hmm...
Mechanical, electrical, box, XO, they are all just filters, so Q means the same for all of them. Remember the equations used to determine a box alignment were originally used for...electrical networks. The big advancements in design came when I think Olson (or Small?) equated the loudspeaker box with known electrical filters.
Check Lintwitz's site. I think there is some article there that mentions network ringing.
Check Lintwitz's site. I think there is some article there that mentions network ringing.
I looked at the Likwitz website and also LDC. All I can understand is that a group delay or a transient response of a filter Q is a direct function of the filter's resulting frequency response (e.g., an exact LR4 target response has a little amount of group delay peak (or ringing) but this is minor and thus not audible). So what we have to look at is not the transfer function but the resulting acoustic FR. The "knee" in the transfer function in my case is a minor response correction that is intended to smooth out the woofer's LPF rolloff. The resulting curve is smoother than without the notch filter. I still don't see a problem in this approach.
Actually using a notch filter to achieve a smooth target filter slope is quite common. For example, see Zaph's ZD5. He used a notch filter centered at 1 kHz for the SS15W midwoofer. Here's the transfer function:
See something similar to mine? Although a bit mild, there is also a "knee." Does this ruin the otherwise good transient property of an LR2 filter? No, it's used there to make the response smoother.
The same story applies to mine. My notch filter is at 975 Hz. Its main purpose is to suppress the woofer's FR bump there, which is aggravated by a baffle diffraction effect. In this case, a notch filter always results in not only a notch but also a bump on its higher frequency side, whose magnitude depends on its notch depth. Although mild, this bump is also present in Zaph's notch filter use (I confirmed this using a simulation of his XO). In my case, this bump happens to be at the Usher's dip. By using an electrical third order and choosing a proper XO point, I was able to obtain a smooth high frequency roll off of the Usher.
See something similar to mine? Although a bit mild, there is also a "knee." Does this ruin the otherwise good transient property of an LR2 filter? No, it's used there to make the response smoother.
The same story applies to mine. My notch filter is at 975 Hz. Its main purpose is to suppress the woofer's FR bump there, which is aggravated by a baffle diffraction effect. In this case, a notch filter always results in not only a notch but also a bump on its higher frequency side, whose magnitude depends on its notch depth. Although mild, this bump is also present in Zaph's notch filter use (I confirmed this using a simulation of his XO). In my case, this bump happens to be at the Usher's dip. By using an electrical third order and choosing a proper XO point, I was able to obtain a smooth high frequency roll off of the Usher.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.