What's to explain? The added delay will be specifically the delay introduced by the circuit and it's interaction with the driver's complex impedance. A lowpass has little delay at high frequencies, so technically the impulse will for practical purposes still start to rise almost immediately. What we see as a change in the rise or change in the start if you see it that way is due to the reduction in amplitude of the higher frequencies, nothing more. That impulse start is essentially not changed. Woofers and even midranges look slower because they already have little output at the higher frequencies.
If you apply the same passive crossover to two distinctly different drivers it's a given that the resultant delays will be different because of two things. One, it's unlikely that the complex impedance of the two drivers is identical. Two, it's unlikely that the raw frequency response is identical to begin with as well, so the resultant acoustic response will have differing delays. You just aren't going to get the same acoustic response from two different drivers using the same crossover.
Small adjustments of phase are possible, but I wouldn't try to get much and this only via slight alterations in the stop-band. Passive delay circuits are of no interest to me. I almost always physically offset my drivers from the start, then the small adjustments in the crossover slope stop-band are sufficient.
Dave
Yes, but here we are over 1300 posts later and we still haven’t answered “why?”. Is it because of:
* Some still not fully explored peculiarity of human perception? Or
* Some inherent acoustic characteristic of the typical “small” (but not anechoic) room? Or
* A natural compensation for the almost always over-bright recordings that result from close microphone placement? Or . . .
And we also haven’t addressed the widely acknowledged fact of human auditory processing that we are not only able to, but regularly do, “hear through” all the artifacts of room, loudspeaker and recording to “hear” in our minds the overall “sound” of a familiar source (the recording is terrible, but the band sounds great).
And then there's the delicious irony that whenever you defend “design by measurement” (with which I mostly agree) it still always ends with “and they sound fine” . . .
I don't know if you were directly addressing me, but I never end with "and they sound fine" if the measurements are good then the statement is superfluous.
I did, at some point, address "why" there is a desire to slope the bass upwards at lower frequencies (small rooms) and the treble downwards (constant directivity). It's there in the posts somewhere.
Begging everyone's pardon, but I was checking out gedlee.com and found the statement, "The use of multiple subs in a small room is an absolute requirement if the best possible playback performance is desired." Could someone elaborate a bit on this for us? ie, Define "small room" and describe the one-sub limitations that are addressed, please? Even a few paragraphs would be helpful.
I've been puzzling over this technique because I don't understand how it would work. Wouldn't you end up with multiple notches or nulls because of comb filtering? How do you know which is the right one? Even in drivers that overlap by a couple of octaves, you'd still get major comb ripples if they were offset. Where the nulls occur should also change depending on the polarity of the drivers. Right?
Two identical drivers should cancel each other out if in reverse polarity (ideally) but with an offset there will be comb filtering. Drivers that are high and low passed will show the null at the crossover frequency if aligned. But without the stop band, I don't see how it would work. What am I missing?
I thought that was the point I have tried to repeatedly make, that we can hear through much of the room's acoustics to the direct sound. All the answers are in this thread, and in the published papers, if you can seperate the signal from the noise.
David S.
I just found this great article by John Broskie, which addresses the original question in this thread; Is flat what we want? Short answer is sure start with that and then make it sound nice with whatever. Long story (a fun read) here:
The Tube CAD Journal: Out of Control: The missing sonic controls
The Tube CAD Journal: Out of Control: The missing sonic controls
A quick search of the mult-way forum should turn up a good bit of info on this.
Well, one reason I did a crossover independent null to try to work out the acoustic offset between the two drivers more accurately than I had previously, is because I am using an odd order (3rd order butterworth) crossover, where (a) it's impossible to just reverse the polarity of one driver and adjust the tweeter position for an optimum null and (b) the time alignment and therefore phase between the drivers is much more critical than it is in an even order system - even relatively small phase errors result in a lot more change in the amplitude response than an even order filter.
The absolute amplitude response will change depending on the high pass filter, yes, but the relative amplitude and phase of the two drivers and the offset distance where the null occurs will not.
Another way to approach it is to use dual channel measurement - and use the reference channel to sample the signal at the drivers after the filter, which will compensate for the roll off of the filter to at least a couple of octaves below the cut-off - until noise starts to become a problem. This does work, but doesn't really provide any more information for the particular test because you're just looking for a null in a flat line instead of a null in a sloped line...
I agree, any individual tailoring of the response in the drivers other than the crossover slope would change the result, it's only in the case where both drivers are flat well across the crossover region and have very wide bandwidth would it be absolutely correct, I guess I'm lucky that both drivers have wide bandwidths and plenty of overlap, so the response is largely dictated by the crossovers.
Another way I haven't tried to address that concern in the case of an odd order filter, where you can't just reverse one driver to look for a notch, would be with the final crossover to move the tweeter in the direction of the nearest null at the crossover frequency, (in the case of 3rd order in phase, moving it backwards) then precisely move it forward the distance equal to 1/4 wavelength at the frequency of the notch.
This would assume that the crossover slopes and any response tailoring of the drivers correctly summed to 90 degree phase shift. I may end up doing that in the final design, although I'm not sure how accurate it would be.
I agree that the individual centres can't be found to a high degree of accuracy, but as you correctly state, it really doesn't matter as long as the relative offset is known.
I know the relative offset is 50mm between the drivers, (give or take phase errors in the drivers as discussed above) and although the cancellation test can't determine the absolute acoustic centre, the impulse response can find it roughly.
I used dual channel mode in ARTA (which eliminates random delay variations in the sound card channel) and measured the exact delay to the nearest sample of the start of the impulse rise of the tweeter, which I think is more accurate than measuring it for the midbass driver due to the much steeper rise.
From that the sources of error are the time resolution of ARTA at 96Khz, and the ability to accurately measure the physical distance from the microphone capsule to the faceplate of the tweeter. The two together worked out to around +/-3mm of uncertainty so the figures I settled on for acoustic depth were 15mm for the tweeter (wave-guide ribbon) and 65mm for the full range driver, (8") but more importantly 50mm relative difference.
That's about as accurate as I could measure it with what's available to me.
But surely when you initially chose drivers and think about the physical design you take rough acoustic centre depths into account ? If not, what type of compensation do you apply - an all pass delay, or just a phase shift, which is only optimal at one frequency ?
I'm perfectly happy to use a computer for making accurate measurements, and I do use some prediction software as well, like WinISD Pro for T/S alignments, edge.exe for baffle diffraction, (part of SIRTA) a spice simulator to help with rapid iteration of passive crossovers, and so on, but I stop short of using "all in one" speaker design packages like speaker workshop etc, which try to do everything from measurements right through to designing the crossover for you...
While I see the usefulness of them, I'm not keen on design by numbers taken to quite that extreme, and I'd rather do the main design process manually and iteratively to make sure I really learn and fully understand what's going on at the fundamental levels, rather than relying too much on following a design that a computer program has come up with based on a few measurements and a simplistic design criteria.
It's just the way I prefer to work, especially when the learning process and any mistakes along the way are just as important as the end result. Sometimes "mistakes" can lead to new discoveries and insights that might not have been made had I followed "the path" strictly. 🙂
Last edited:
I appreciate it. My first search brought me to this thread. I really hadn't considered the multi-way forum.
No, not with the drivers connected in reverse phase. The reason is that connecting the drivers in reverse phase changes their relative phase shift by 180 degrees at all frequencies, while moving one driver back or forwards by "180 degrees" at the crossover frequency (about 43mm at 4Khz) only shifts it by 180 degrees at that one frequency (well a few specific spot frequencies actually) and by a different angle at all other frequencies.
So leaving the drivers in phase and moving one forwards or backwards until there is a notch at the desired crossover frequency will always result in a comb filtering effect. Reversing the phase will give a broad notch when the acoustic centres are aligned but if you additionally move the driver a full 360 degrees forwards or backwards at the crossover frequency you'll get a narrow notch at the crossover frequency, but comb filtering to either side of it. (Assuming there is enough overlap in the drivers for it to form)
Two ways - if there is a lot of overlap in the drivers, you'll see comb filtering if you're at the "wrong" notch point, but a more accurate way is to work out a rough acoustic centre using an impulse measurement, which will get you close enough to the right notch that you know which is the correct one to home into.
Another way is to use common sense - you know fairly closely where the acoustic centre of a tweeter will be, and even at 4Khz a 360 degree phase shift is 86mm - which means the next nearest false notches would estimate the acoustic centre of your low frequency driver to be 86mm too far forward (in front of the face plate on a typical 8" driver) or 86mm too far back. (behind the magnet) From this you can eliminate the notch on either side as being "not possible" as a candidate, and thus determine the correct one even with relatively little overlap.
Yep.
Just because two drivers have less than total overlap doesn't mean they can't still cancel each other out. If they're both flat over a significant overlap (as one would hope, if you're about to cross them over) then they can constructively and destructively interfere in this range.
You forgot to include tweeter directivity patterns in your list of possible suspects 😀
Could it be a result of bloom in the power response in the low treble region above the crossover typical of many speakers with a midrange driver crossed to a dome tweeter ?
It's only anecdotal but I've noticed on designs I've played with that dome tweeters that sound ok in a heavily damped room do indeed tend to sound over bright and unpleasant in reverberant rooms without shelving them down slightly, but I think the blame lies in the lower/mid treble region where their polar pattern is very wide, rather than the very top end.
With my ribbon tweeters I really don't notice the same effect - what sounds balanced in a heavily damped room and/or up close still sounds fine in a more reverberant room, even at the far end, and apart from them having more directivity in general than a dome (especially vertically) the wave guide will help to reduce the power bloom at the lower treble region near the crossover compared to a dome where the size of the dome is the only directivity controller, and thus only controls directivity at the top end.
Is the whole shelving down the tweeter thing really that universal, or is it only common with drivers with "typical" directivity patterns ?
If flat on axis sounds right in a dead room or in the near field but it sounds too bright in the treble in typical reverberant rooms, surely the cure is to increase the directivity of the tweeter in the x and y planes (not necessarily the same in each) until the shift in perceived balance is minimal, rather than tweak the on axis response ? If the tweeter is too directional you'll actually lose apparent treble listening in the far field (eg unassisted large full range drivers) but it stands to reason that some intermediate amount of directivity will give a tweeter that has a pleasing balance all the way from dead to live rooms.
From what I've heard in my own listening a small wave-guide loaded ribbon like the G2 comes a lot closer to this ideal than any typical dome.
Last edited:
"It's only anecdotal but I've noticed on designs I've played with that dome tweeters that sound ok in a heavily damped room do indeed tend to sound over bright and unpleasant in reverberant rooms without shelving them down slightly, but I think the blame lies in the lower treble region where their polar pattern is very wide, rather than the very top end."
That is a typical problem with common 2-ways ...
They work for a certain room - placement due to
sidewalls plays a role too - at a certain distance.
If directivity in the crossover region is not matched,
tweeter level (at lower highs) has to be changed
dependent on room/position/listening distance , also the
midwoofer's rolloff may be adjusted (due to listening
distance and damping of the room).
Many of the old "classic english style" 2 way speakers
have that gently falling (lowpass) slope in upper midrange,
compensating the the directivity of the woofer, thereby making
the power response problem even worse unfortunately...
I did a lot of 'usual' 2-ways and listened to them in many rooms.
You cannot live with them, without compromising on axis and
power response IMO.
The last ones had switchable components and jumpers at the
terminal plate to adjust them to my friend's homes, to where
i sold most of them as a student.
Oh my ... some decades passed since then.
The usual non DI matched 2-way is just playing
"acoustical roulette".
That is a typical problem with common 2-ways ...
They work for a certain room - placement due to
sidewalls plays a role too - at a certain distance.
If directivity in the crossover region is not matched,
tweeter level (at lower highs) has to be changed
dependent on room/position/listening distance , also the
midwoofer's rolloff may be adjusted (due to listening
distance and damping of the room).
Many of the old "classic english style" 2 way speakers
have that gently falling (lowpass) slope in upper midrange,
compensating the the directivity of the woofer, thereby making
the power response problem even worse unfortunately...
I did a lot of 'usual' 2-ways and listened to them in many rooms.
You cannot live with them, without compromising on axis and
power response IMO.
The last ones had switchable components and jumpers at the
terminal plate to adjust them to my friend's homes, to where
i sold most of them as a student.
Oh my ... some decades passed since then.
The usual non DI matched 2-way is just playing
"acoustical roulette".
Last edited:
Choosing the drivers does include a rough idea of the acoustic centers. This was key in my latest system, a modified dipole, although the tweeter directivity probably was a higher priority. The extra offset in the tweeter associated with this was killing two birds with one stone, so-to-speak. When I have the choice, I prefer drivers with more closely positioned centers.
Even so, there's almost always a small amount of compensation needed, but that can be achieved using the stop-band in the lowpass section. I've never used all pass delays.
The only "all in one" package that I could recommend is SoundEasy. I still largely measure with my old LAUD system, but SoundEasy has options that make it a good recommendation. You just need to be ready to spend a little bit of time with it. It's especially useful for auditioning and/or tweaking crossovers prior to building them if you're going passive.
Dave
I wouldn't presume to speak for Earl 😉, I was affirming that I would be happy to throw together two drivers with different and unknown acoustic centre depths, measure their amplitude and phase, and design whatever crossover correction is necessary to account for the unknown difference in acoustic centers.
I can't remember, of the many designs I've done, ever doing a system with offset baffles (i.e. offset to get the acoustic centers to line up). First off, I've never had difficulty getting good blending between drivers with any reasonable offset by choosing the appropriate rolloff slopes and phasing.
Secondly, if I did get acoustic center allignment, how would I be better off? The crossover I would add would retard the lower unit and advance the upper unit to the point where they were well out of allignment at crossover and I would have the same issue as starting with random driver depths. Unless you have a crossover scheme that only works with delayless units, there is no advantage to raw driver time allignment.
Here is a thread that I started at the Classic Speaker pages.
Crossover mods for the AR4x - The Classic Speaker Pages Discussion Forums
It is about upgrading a classic AR speaker but illustrates the process of getting offset drivers to pull into phase by choosing the right crossover orders.
Perhaps you should break this topic off into its own thread.
David S.
This may be what is needed to be read in order to understand exactly where we can EQ from in room response graphs:
AES E-Library A Model of Loudness Applicable to Time-Varying Sounds
Please read the preview before refutation or acceptance,
Dan
AES E-Library A Model of Loudness Applicable to Time-Varying Sounds
Please read the preview before refutation or acceptance,
Dan
- Status
- Not open for further replies.