As you are using Equalizer APO in testing you might want to try some PEQ to tilt the response down a little more and see if you like this as it is relatively easy to A/B compare. The slope of the PIR in your crossover is quite shallow at -0.8dB/oct. Tilting the response down so the PIR gets to -1 to -1.2dB/oct is something that I would try. -1.2 is the slope target that is preset in Vituix and I have found -1 to -1.2dB to be a good target to shoot for overall.
I think then I need to play with each channel separately, what is a mess, and then I will not know how to "translate" it to resistors, inductors and coils. It defeats purpose: to build a speaker with reasonable passive x-over
Also A/B compare is not so trivial - I loose ~1-2seconds when switching between EQ APO files with different settings, and those 2 seconds is a lot. Yes, with EQ it is simpler and I did that.
Is it the blue line or orange line?
Yes, it may need a little down slope for my taste, but I played lots of time with Vituix, and this is the best numerical score I came with. If I tilt down tweeter, score drops from 6,8-6,9 to 6,0-6,2 very quick. Mid frequencies are a bigger problem: it is still a bit too harsh there and I don't see simple solution to it.
My apologies, I understood only "the phase". Fc is crossover point? This one?
All the crossovers I tried - it has the same one, this one is the "closest" I can get.
I did mistake in measuring? As it is coax, I DID NOT MOVE mic on nearfield measurements and I merged tweeter also, as it plays pretty low here. I tried to simulate without tweeter merge, but then 1-2kHz is not real - it has more SPL there than vituix shows, so the merge kinda makes it more realistic
This is above my knowledge
Last edited:
Probably you are talking about tweeter phase switch?
I don't know. I measured 3+ times in the same way, there is no error in wiring, I checked it, and if you pad down the tweeter - this happens. I cannot say that I can hear this. Besides - my x-over is still the same as manufacturer, only with variation. There is nothing radical except extra coil on woofer, which does not impact this phase change.
Perhaps if you share your x-over we can compare. When you really invert polarity of one driver - simulation changes a lot, it makes almost impossible to simulate something.
I don't know. I measured 3+ times in the same way, there is no error in wiring, I checked it, and if you pad down the tweeter - this happens. I cannot say that I can hear this. Besides - my x-over is still the same as manufacturer, only with variation. There is nothing radical except extra coil on woofer, which does not impact this phase change.
Perhaps if you share your x-over we can compare. When you really invert polarity of one driver - simulation changes a lot, it makes almost impossible to simulate something.
No, it's because crossover filters will never be 100% symmetrical + even with coax drivers the real acoustical center is never 100% perfect.
Manufactures have to make a compromise here between proper freq resp and phase alignment.
I would personally not consider this as a real mismatch, but just an ever so slight mismatch.
Because the phase plot will be slightly different depending on the axis anyway.
Green vs red, there is 60deg phase mismatch at 3kHz, slightly less below 3kHz, but it is getting worse above 3kHz, so the drivers acoustically subtract at 4kHz.
I always try to make phase matching good and the phase following each other in broad range of frequencies.
There are types of the crossovers thar work with 90deg phase offset and though they have some advantages they never sounded right to me.
This Sica is definitely not easy to work with. I write from my phone, I will post my crossover later today.
Last edited:
Green vs red, there is 60deg phase mismatch at 3kHz, slightly less below 3kHz, but it is getting worse above 3kHz, so the drivers acoustically subtract at 4kHz.
I always try to make phase matching good and the phase following each other in broad range of frequencies.
There are types of the crossovers thar work with 90deg phase offset and though they have some advantages they never sounded right to me.
[/QUOTE]
Yes. Probably that is why it sounds a bit (just a bit) harsh there, but if there is no mismatch and cancelation - this coax will not work at all, as it has bump there due to tweeter 3rd order xover bump. That 4kHz region would be unmanageable. That is only my opinion, I am rookie there. Did not found a way to make it work on 2nd order tweeter. Only 3rd or 4th tweeter and 2nd or 3rd woofer.
Here is on "R5" the one I already put there, "R3" is close to "Full range woofer assisted by coax tweeter" as there is very little music going through tweeter.
"R3" - 2nd order woofer, 4th tweeter
Pros:
-No harshness in sound, softer, much more suitable for late evening lo-fi listening
-Sounds more "expensive" and somehow bigger
-Better protection for tweeter
-Overall - more pleasant
Cons:
-Bit nasal and little boomy sound. Will need to figure out which freq does that. According to SPL line should be the opposite.
-No speed
-No sharpness
-Much more expensive xover: 1-2 more items, couple front resistors instead of 1 on tweeter, as lots of W dissipated there, more precision items, as tweeter filter is very sensitive.
-Probably more distortion form woofer. Will need to measure.
-Harder to "locate" instruments
I am tempted to build both crossovers, then make some mechanical switch to switch instantly, but that would be expensive. Your advice will save me from additional expenses and headache...
I always try to make phase matching good and the phase following each other in broad range of frequencies.
There are types of the crossovers thar work with 90deg phase offset and though they have some advantages they never sounded right to me.
[/QUOTE]
Yes. Probably that is why it sounds a bit (just a bit) harsh there, but if there is no mismatch and cancelation - this coax will not work at all, as it has bump there due to tweeter 3rd order xover bump. That 4kHz region would be unmanageable. That is only my opinion, I am rookie there. Did not found a way to make it work on 2nd order tweeter. Only 3rd or 4th tweeter and 2nd or 3rd woofer.
Here is on "R5" the one I already put there, "R3" is close to "Full range woofer assisted by coax tweeter" as there is very little music going through tweeter.
"R3" - 2nd order woofer, 4th tweeter
Pros:
-No harshness in sound, softer, much more suitable for late evening lo-fi listening
-Sounds more "expensive" and somehow bigger
-Better protection for tweeter
-Overall - more pleasant
Cons:
-Bit nasal and little boomy sound. Will need to figure out which freq does that. According to SPL line should be the opposite.
-No speed
-No sharpness
-Much more expensive xover: 1-2 more items, couple front resistors instead of 1 on tweeter, as lots of W dissipated there, more precision items, as tweeter filter is very sensitive.
-Probably more distortion form woofer. Will need to measure.
-Harder to "locate" instruments
I am tempted to build both crossovers, then make some mechanical switch to switch instantly, but that would be expensive. Your advice will save me from additional expenses and headache...
I don't see why this wil result in harshness.
I also don't find them difficult to filter to be honest.
But from the Klippel and distortion data (shared earlier in this thread), it looks like the distortion as well as inter-modulation distortion is not great.
Especially Le(x) isn't great (at all), since there is not demodulation ring/sleeve.
H3 is a little high as well.
As long as you make sure the woofer doesn't have much cone excursion, it should result in an okay sounding speaker.
Not much worse than a lot of other woofers of the same size.
If I look at your freq resp, I think the "harshness" is more because the tweeter has quite a lot of extra output.
I would optimize the crossover a bit better.
It's definitely possible to get the phase decent, like any other speaker combination.
I meant to apply some shelving filters on top of your existing filter as a different preset, then switch between the different presets. With a wide ranging frequency balance EQ a few seconds should not be any issue in detecting whether you prefer one version or the other. If you listen from a computer source you can always have overall EQ presets available to switch for tonality changes without having to bake it into the crossover.
PIR is the orange line ER is the blue. Although looking at your gif image the slope shown in the upper part of the window is changing between the versions and is around -1.2 already. My comment before was based on the bottom part of window where the targets are. The -0.8 is not changing there so it does not seem to represent anything about your actual crossover.
33uF cap... That thing is expensive in this side of the pond...
Gating 10ms - does that mean you had opportunity to measure it at ~2.5m distance, right?
Now I have problem. I assume that it is mine problem, as the same crossover should recreate very similar sound, except in baffle related region, and box related region (mine ~7.7L, already counted ~0.4L as taken by driver and with external BR tube, tuned to ~59Hz) the slope should be the same. But what I have got:
It may be measurement mismatch of the initial data.
I have 3 mics at hand: Dayton, Sonarworks and dbx. First two are calibrated. It is a bit of a story behind them, as... my trust in at least one of them is not absolute. There is a thread about them which is not over (!!!). dbx measures much brighter than it is, so if I model with it, all my simulations are WAY too dark. My modeling was done with Sonar mic, which is a bit, just a bit, ~1.5-2db rising slope from 1.5kHz to 15kHz. Also there are 2 bumps at 3.5 and 4.5kHz, but other than that they are identical.
Another thing: I tried to model manufacturer recommended crossover - it is WAY too bright. Even for hi-fi showroom levels of treble too bright.
So... What mic you use for measurements?
Last edited:
Yes, exactly. The speaker was placed around 2m above the floor, in the middle of very large room. Mic distance 1m if I recall well.
I use Earthworks M30 mic.
Why I am in doubt? There are reasons:
First of all - my ears are well above average. Not 999,9 gold, but 585 😀 So if I hear if something is not right - most likely it is not right. That is why I doubted that the first one (uncalibrated) is right. It is WAY of in the high freq.
Second - I can easily check measured response of the speaker and compare it to the manufacturer spec or the similar measurements of similar speakers, as PKAudio. The most obvious are 5+kHz range, up to 15kHz, and as we see - I have much more treble in graphs, so my mic probably is measuring rising high freq, when actually it is flat. So I am dropping it in simulations, and to my ears it is too dark. That uncalibrated dbx is measuring even more high freq. Dayton is measuring 1-2dB lower in high than Sonarworks, so probably it is the most correct one in terms of slope. I still haven't got to compare any of them to the expensive one, which I would trust.
Maybe it’s time to get proper correction data on one or two of your mikes. In Germany there are a few budgetwise attractive options, like akulap.de and hifi-selbstbau.de. Don’t trust your ears, they have a habit of accomodating…
I use this mic, it is ok for me. The supply calibration is good. I also have a b&k 4155 not much difference in amplitude/phase. Note on a microphone you have a correction at high frequency, rising response due to a diffraction at the extremity. You have to apply the correction in your measurement software ! To avoid this correction the mic can be positioning vertically.
What ever the mic you use, your Sims are approximative. You have to listen and adjust.
Here you have a topic about a test with a lot of mics, the only difference are on distortion measurements. An instrumentation microphone gives the best result.
I started to doubt this mic. Really. It is still pretty realistic, but....
Now I have 3 mics: and if I measure with this Sonarworks mic - sound is more or less ok, just a bit dark, I would prefer rising response from 200 to 15000Hz total of ~2-3dB. I am comparing this to Vituixcad default score. I understand that I can adjust it, but if it is based scientifically - let it be.
Dayton mic is showing drooping response comparing to Sonarworks - if I make according to it, then I get somewhat more realistic 1-2dB raise from my sims, but still 1-2dB deficit (to my ears)
dbx is showing a raise of 3-4dB, so if I model according to it - the sims are WAY too dark. Like unlistenable dark.
NTI M4260 (circa year ~2010) which probably never recalibrated - showing dropping response o 100-20000Hz, around 3-4dB, so it is closest.
"rising response due to a diffraction at the extremity. You have to apply the correction in your measurement software ! " - please, can you elaborate this? Yes, I am using cal file. Simulation is done WITH calibration file corrected measurements.
I am measuring 1m, gating ~4,5ms if I remember, diffraction should not be at the extremes, as the baffle is rounded, like sphere round and the 10+kHz is mostly influenced by the coax woofer cone itself. Anything extra is well above my knowledge.
I have never heard, that some other correction is needed. Is there a paper, instruction or thread about this?
Yes, Sonar mic has calibration file for this. Why I have to measure vertically? I don't need info from reflections.
What is "instrumentation mic"? I have 3 of the mics, another one I got, but returned was faulty, and the I have compared one of my mics to old but expensive NTI. So I have info on horizontal measurements of 4 mics. And ALL of them are so much different. Not in 1dB range.
Personally I think that I am past the "good enough" phase in audio, and want to build smth which is accurate and based on science. Ears can get used to any sound but I simply don't want to.
Comparing my simulation to PKAudio with the same crossover kinda confirms that I am missing ~2-3 dB on the high end.
Last edited: