Here are raw measurements of Faitail HF108 in the 32cm free standing Sandhorn from MaBat. It sounds "normal" for me, but I have it just for home/indoor listening. There is a large thread about ATH waveguides https://www.diyaudio.com/community/threads/acoustic-horn-design-the-easy-way-ath4.338806/
You're mistaken. Its from this thread, post #3 in 2016 -
In the case of the NSD1095N + XT1086, it drops below 100 dB/1m past 12k @ 1W input.
Using a passive shelf network, the combo would need to be linearized to under 95 dB as the output would be under that level at 20 khz.
If you propose a wide band notch to flatten it, the result would be little difference at that wide of BW to a shelf EQ. The relative phase in the upper octave would be negatively affected with a wide notch, moreso than a shelf, which is simpler to shape IMO and would facilitate a smooth, constant phase response past upper cutoff. This is important on higher end systems which prioritize timing linearity across the entire FR. Temporal uniformity is important for several reasons in a reference grade system. In pro audio its usually not that important or observed.
Last edited:
@miero The 17k peak in that horn with the HF108 is interesting. I've never observed that on other horns using that driver. There is a small peak at 15k in the factory impedance curve, but nothing like that has come through in my tests. Maybe its because my horns weren't that effective up there.
I will add that the HF108 does need a little tweaking from its out of box performance. Its an excellent driver for use with a 12" coax because it can cross very low.
An important note about this driver - I've noticed the diaphragm sometimes isn't centered very accurately from the factory. It benefits from a little TLC, especially if you want to use it very low past the recommended cutoff. I've run it down to 800hz a few times and it dealt with this quite well given SPLs weren't pushed too hard. The driver itself has a primary resonance at 700 hz which should be notched if you want to push the HP down under the Faital recommended 1300 hz.
I will add that the HF108 does need a little tweaking from its out of box performance. Its an excellent driver for use with a 12" coax because it can cross very low.
An important note about this driver - I've noticed the diaphragm sometimes isn't centered very accurately from the factory. It benefits from a little TLC, especially if you want to use it very low past the recommended cutoff. I've run it down to 800hz a few times and it dealt with this quite well given SPLs weren't pushed too hard. The driver itself has a primary resonance at 700 hz which should be notched if you want to push the HP down under the Faital recommended 1300 hz.
Attachments
Last edited:
I'm not mistaken - that graph is from my measurements. https://lscon.tripod.com/xt1086/You're mistaken. Its from this thread, post #3 in 2016 -
I am confused here? Are you saying the notch effects the acoustic phase? Phase and Group Delay follow the frequency response. In a well designed 2 way there won't be any audible temporal issues. The place to look is the the step response at crossover and group delay of the system.
Well yes of course and it would be in an active as well, You are always are a slave to the driver with the lowest sensitivity which is typically the woofer. In an active or passive you would be attenuating the mid band SPL response to match the woofer through crossover.
Rob 🙂
I have the dbx PA2 but had to abandon using it with my horns after comparing the sound quality with a discrete analogue external crossover (with same DSP/EQ curves applied in EQ APO).
I also know from communicating with Joseph Crowe that he eventually abandoned using miniDSP 2x4 due to sound quality not being good enough.
What I'm doing is measuring the room and figuring out the dsp/eq points for my horns/room using dbx, then hooking up my analogue crossover and applying those corrections on my server using equalizer APO, which is a pain in the **** but the results are worth it..
@Robh3606 I'm talking about the area above the (parametric) notch where it continues. That phase trend is important. I've found distinct differences in behavior between a wide notch vs a shelf at the limit of the top octave. Some RIAA EQ can have a similar issue, whether the filter slope continues on or eventually attempts to swing back to the original phase and timing.
Unfortunately you can't always look at just the step response to see this unless the measurement isn't cut off at 20khz. Some DSP systems treat EQ on the upper octave as a shelf by default, but thats a different issue with unavoidable consequences. When you just crank up the notch center frequency and have the upper end sit past the cutoff limit, the system may still treat it as a shelf past that point. Depending on the digital filtering, it may be a linear phase trend, but you can't tell much higher after the digital brick wall cuts it. At that point all the upper stuff is all gone, which may be beneficial in some ways.
In the analog active domain, with a shelf, the filter slope will continue on infinitely up to the RF LP on the output stage. The parametric notch will swing back to the starting phase point, just as with any analog filter. This is an issue which affects several things negatively in an audible way. The delay can be significant enough to ruin imaging and affect group delay above hearing cutoff.
In the passive xover domain, the same issue holds true as with the active analog domain. The amplifier and the tweeter inductance curve will be the limiting factor here. If you've worked hard to preserve timing in your crossover, extended FR well past hearing cutoff and use very wide band amplification, your efforts will be undone after applying a parametric notch up top. The shelf distinctly sounds better here and preserves spacial details, imaging and timing cues. Its not what you don't hear up top. Its the phase already swinging around past where it should in theory be zero.
Unfortunately you can't always look at just the step response to see this unless the measurement isn't cut off at 20khz. Some DSP systems treat EQ on the upper octave as a shelf by default, but thats a different issue with unavoidable consequences. When you just crank up the notch center frequency and have the upper end sit past the cutoff limit, the system may still treat it as a shelf past that point. Depending on the digital filtering, it may be a linear phase trend, but you can't tell much higher after the digital brick wall cuts it. At that point all the upper stuff is all gone, which may be beneficial in some ways.
In the analog active domain, with a shelf, the filter slope will continue on infinitely up to the RF LP on the output stage. The parametric notch will swing back to the starting phase point, just as with any analog filter. This is an issue which affects several things negatively in an audible way. The delay can be significant enough to ruin imaging and affect group delay above hearing cutoff.
In the passive xover domain, the same issue holds true as with the active analog domain. The amplifier and the tweeter inductance curve will be the limiting factor here. If you've worked hard to preserve timing in your crossover, extended FR well past hearing cutoff and use very wide band amplification, your efforts will be undone after applying a parametric notch up top. The shelf distinctly sounds better here and preserves spacial details, imaging and timing cues. Its not what you don't hear up top. Its the phase already swinging around past where it should in theory be zero.
The PA2 is good for PA, At home i use a MiniDSP Flex that is higher resolution then our ears (officially tested at ASR). The (older) 2X4 that Joseph Crowe uses does it also not for me, but the Flex does, and is the first dsp that does not sound like a DSP in hifi setting to me.
But the PA2 was chosen on a budget, and to be rugged (that system will be used outdoors mostly) and is good enough for it's purpose. The same with the Crown XLi amp i use there. It's not up to my standards for home use, but great for a little party soundsystem like the on i build. At home i prefer nice colouring tube or class A amps or very clean neutral amps, and have both
The system i'll build for my own will probally start also with a MiniDSP Flex and clean NCore amps, but i may also try to make a good passive crossover for it. I'm not against it, but it needs to fit the purpose. And if i do, i surely will also connect it at least once to my 40W tube amp... I like experimenting with systems, that's why i started diy (it's not cheaper, i can assure you) and for me nothing is carved in stone but a scientific based approach to designing and building stuff.
I almost bought the MiniDSP Flex but was concerned about reports that its second analogue outputs don't measure as well as the first outputs. Any experience with that? I am in need of another xo/dsp/eq for a second system and would LOVE a xo/dsp that is as transparent as my discrete active x-over.
Ok guess I am being dense but why a notch up in the last octave? You are using attenuation to tailor the curve and use the most in the lower range of the compression driver. Your level in the last octave doesn't typically get attenuated so no notch. If anything it is the limiting factor you overall system sensitivity. Virtually all passive networks are set up this way with a CD type horn where the horn roll off is similar or matches the mass roll off of the driver. It's a tailored shelf, parametric or a combination to flatten the on axis response.
Why the emphasis on phase? If you have a reasonably flat response phase and group delay will follow.
Rob 🙂
I think in most cases either the responses don't really match, whether polar or on-axis, or the worse-sounding CD is overtaxed at the bottom of its range, or both. There are definitely other audible differences, but they don't seem worth considering without measurements showing it's not response or excursion.
More often than not, these all in one wonder DSP boxes people trust to work miracles with sub optimal drivers on like horns create various other unwanted artifacts and noises which are a product of limited processing power in the software code. Its prioritizing specific audio in a way which tends to cut off or drop bits as errors occur, filling in missing little pieces with smoothed over noise. This is usually derived from the error correction inserting copies of duplicate sound. Its the strategy most DBX DSP uses with their cheaper processors, The higher end Venu series is much better, but it also chops off bits to cope with more sources and outputs. Driverack PA is decent, but not designed for higher end Hifi.
We could discuss DSP all day, but let's keep that for an other topic. I was asking for why some CD's sound so harsh altough they measure good on first sight...
I would try to identify/isolate a frequency band in which that harshness is present. For example, is it there also if a driver is plays only up to 4kHz?
I don't buy all those drivers to test. It's just that i want specs to look at that indicate that the driver may sound harsh before buying them. Compression drivers are expensive, and i'm not that rich that i can buy a lot. I try to test and listen to compression drivers when i have the opportunity, but just like the majority of the people I don't have the time, money or means to test them all.
Do you look at the waterfall plots? That will certainly show you how controlled any resonances are especially up top.
Rob 🙂
How about just looking for a driver & horn combination with the most smooth non-smoothed measurements?
For example, this one could be extraordinary smooth:
- https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-7804444
For example, this one could be extraordinary smooth:
- https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-7804444
You are better off looking at a CSD where you can see where the resonances are and how long they persists after the test pulse.
Start at 48 minutes, 55 minutes The whole presentation is worth watching.
Rob 🙂