That would be great! Maybe for starters, you could post some pictures of your current configuration so we can get an impression of how you arranged the radial alignment of the drivers. And of course it would be interesting to see your current XO and EQ settings - but I don't want to seem greedy... ;-)
In a nutshell, yes. Ultimately, it's the *system* response that matters, not really the response of an individual driver. Because of the pre-ringing that is inherent in linear phase filters, designing a multi-way system that uses linear-phase xovers is tricky; it relies on the pre-ringing artifacts cancelling between drivers, which basically can't happen both on-axis and off-axis. It's genenerally easier to get benign off-axis behavior using conventional IIR filters; if you are careful with your design and xover topology, phase correcting this in a single global FIR filter is viable.
This is why I put the Dueland out there as an example. It's a unified 3-way xover topology that has a system phase response similar to an LR2 2-way. Correcting this is reasonably straightforward either with an FIR stage or even with an Arbitrator style approach.
Yes, a Dueland is hard enough to pull off with conventional box loading; I'd imagine that a dipole would be really tricky. I really like the basic concept, though.
No, Danley is unique. IMHO the most interesting speakers being built. It's a multi-entry horn topology. Danley Sound Labs - Home of Tom Danley - Innovative Designer of Pro Loudspeakers & Subwoofers
I'm speaking from the perspective of a conventional 'dedicated listening' high-end system. I don't want a computer and a GUI etc in this type of environment. MPD has a nice web interface I can control from my Nokia web pad; the computer is basically just an embedded appliance, and Linux does this far better than Windows. I'm not aware of a Windows based media player that's really easy to control from this type of Web interface, but I also probably haven't looked that hard.
For my desktop system, I'm fine with the Foobar+Reaper setup.
What is pre-ringing?? I mean I have heard of the phenomenon, but what does it sound like??
If you have it, you will hear it type of thing?? For the record, I use a deqx (which is prob why I have heard of pre ringing) but what I hear is magnificent. If that it true, is it also true that my settings are not pre ringing??
hope the question made sense.. ie 'it can happen, if it does then you can hear it, if you cannot hear '''it'''' then you don't have it???'
can you describe the ''it''
If you have it, you will hear it type of thing?? For the record, I use a deqx (which is prob why I have heard of pre ringing) but what I hear is magnificent. If that it true, is it also true that my settings are not pre ringing??
hope the question made sense.. ie 'it can happen, if it does then you can hear it, if you cannot hear '''it'''' then you don't have it???'
can you describe the ''it''
Yep, I'd hit the website. Didn't see anything which discussed the crossover implementations, though the product lineup suggests they're DSP based.
My laptop doesn't have enough drive space to rip my collection to FLAC, so I just stream off my CD player's S/PDIF out. Accomplishes essentially the same thing. A little interaction with the laptop is required to wake it up at the start of a listening session but for me that's a negligible overhead.
Agree FIR, time reversed IIR, and FFT methods all involve time smearing. I'm unsure how much preringing is a problem if you're not running high order FIR, though. With Allocator I'm measuring around 100us and with Allocator+PLParEQX3 it's around 150us. In both cases it's 30-40dB down. Don't really see a good way to A/B test to isolate any audible effects. But at least on this scale it seems any subjective preringing deteriment is small compared to the improvement from having linear phase.
Curious to see the responses to Terry's questions since I don't exactly have answers.
Actually, Duelund's method is general to an arbitrary number of ways and produces what, for all practical purposes, are linear phase crossovers when used within appropriate parameters. Subsequent correction of the crossover should be unnecessary. The difficulty is using minus signs and driver inversions to knock out algebraic terms in the crossover transfer function only admits certain solutions at certain orders for certain ways. Digital IIR based Duelund is theoretically attractive as very high orders and correspondly increased flexibility become feasible but I couldn't see a way to crunch the polynomials into biquads. Theoretical math's not my strong point, though, but I've a hunch Lake's linear phase brick walls (purchased by Dolby) may be based on something like Duelund polynomials. See Duelund's limits of possibility paper for more.
I wonder too what pre-ringing sounds like.
I too use a DEQX. It produces a smooth, yet detailed and dynamic sound.
Obviously there ISN'T any pre-ringing on the listening axis if it is the measuring axis.
What happens if, theoretically, the off-axis response is identical to the on-axis response?
In the real world, what happens if the response has very smooth constant directivity off-axis (e.g. a good example of a dipole with a waveguide)?
I believe that DEQX makes the linear-phase FIR correction globally after the crossover filters have been generated.
What happens if the correction is not made for one point in space, but if measurements are made optimally for each driver individually (which involves close measuring, correcting group delay as well as freq response), with the linear-phase correction made afterwards (which can be done)?
Really a set of questions about FIR filtering.
David
I too use a DEQX. It produces a smooth, yet detailed and dynamic sound.
Obviously there ISN'T any pre-ringing on the listening axis if it is the measuring axis.
What happens if, theoretically, the off-axis response is identical to the on-axis response?
In the real world, what happens if the response has very smooth constant directivity off-axis (e.g. a good example of a dipole with a waveguide)?
I believe that DEQX makes the linear-phase FIR correction globally after the crossover filters have been generated.
What happens if the correction is not made for one point in space, but if measurements are made optimally for each driver individually (which involves close measuring, correcting group delay as well as freq response), with the linear-phase correction made afterwards (which can be done)?
Really a set of questions about FIR filtering.
David
I think I have heard preringing - I've been playing with Waves LinEQ, and I'll post about it soon. But till then, I can say that I measured it, and I could hear it, particularly on close mic'd vocals - it made them sound a little raspy - like there was an unpleasant texture to the voice. It took me a little while to identify, and then I changed the treble's 'method' to 'low ripple', and the raspiness went away. I wasn't using steep slopes.
I've been using the Waves LinEQ as well. Ringing is very obvious with steep slopes, if the filter Q is above 2. "Low ripple" may affect it as you describe as well.
Would be interesting if you could test the trial version of PLParEQ. I did, and found that it sounded smoother and more "analog" than Waves. The penalty is CPU util, and its certainly not for free either....
Refined Audiometrics Laboratory
Would be interesting if you could test the trial version of PLParEQ. I did, and found that it sounded smoother and more "analog" than Waves. The penalty is CPU util, and its certainly not for free either....
Refined Audiometrics Laboratory
hey david, another deqx user, cool.
got a link to your build?? (if you're interested, here's mine.http://www.diyaudio.com/forums/multi-way/127653-frankenstein-becomes-cindarella-i-hope.html..or not too)
I don't use particularly steep slopes, around 72db from memory. Having said that, I cannot say I don't have preringing, but there is nothing to me that has ever suggested I do have it. A case of ignorance is bliss????
What I CAN say is that there is no doubting the benefits it brings.
Perhaps the secret revolves around the steepness of the slopes.(?)
got a link to your build?? (if you're interested, here's mine.http://www.diyaudio.com/forums/multi-way/127653-frankenstein-becomes-cindarella-i-hope.html..or not too)
I don't use particularly steep slopes, around 72db from memory. Having said that, I cannot say I don't have preringing, but there is nothing to me that has ever suggested I do have it. A case of ignorance is bliss????
What I CAN say is that there is no doubting the benefits it brings.
Perhaps the secret revolves around the steepness of the slopes.(?)
VERY interesting build terry j.
For those of you using Waves LinEQ, when you detect pre-ringing, does the pre-ringing sound the same as you move around the room? I presume, theoretically, that it should change? Where are your measurements taken? In particular, are they far field, at the listening position, or nearfield?
With DEQX it IS possible to configure the system so that it shows a filter with predicted pre-ringing. However, there is an optimum slope for each combination of drivers, which is not intuitive, and I can't tell yet what the correlating conditions are, but that produces a predicted "minimum" case of pre-ringing.
When I use Praxis to double check the step response (DEQX always gets it right), if I really zoom in on the leading edge under these conditions there is no apparent pre-ringing.
In some cases this might mean slopes equivalent to 48 dB per octave. especially for the low crossover, but, surprisingly, high slopes sometimes produce the best results at the high crossover (with the current configuration I am using an equivalent to 150 dB per octave !).
David
For those of you using Waves LinEQ, when you detect pre-ringing, does the pre-ringing sound the same as you move around the room? I presume, theoretically, that it should change? Where are your measurements taken? In particular, are they far field, at the listening position, or nearfield?
With DEQX it IS possible to configure the system so that it shows a filter with predicted pre-ringing. However, there is an optimum slope for each combination of drivers, which is not intuitive, and I can't tell yet what the correlating conditions are, but that produces a predicted "minimum" case of pre-ringing.
When I use Praxis to double check the step response (DEQX always gets it right), if I really zoom in on the leading edge under these conditions there is no apparent pre-ringing.
In some cases this might mean slopes equivalent to 48 dB per octave. especially for the low crossover, but, surprisingly, high slopes sometimes produce the best results at the high crossover (with the current configuration I am using an equivalent to 150 dB per octave !).
David
StigErik - I'm definitely going to look into other software options. Waves does have a number of issues, but it seems to suffice, at least for the time being.
David - the measurements I did weren't acoustic - I took them straight off the output of the computer, as I was rebuilding the transfer functions. I can't tell you if the sound changed depending on my location, but I wouldn't expect it too much.
So, I wanted to post my impressions of using 'phase linear' crossover/EQ software, but I want to first point out that I haven't done before/after phase measurements, so I can't say precisely what has changed. But the change in sound is obvious, and I won't be going back. When I can do more rigorous measurements, thing may continue to improve, or at least change, so consider this part of the evolution. Previously I had been using Ultrafunk Sonitus:fx EQ (which is regular, non-linear phase software), and then moved to Waves LinEQ.
I listened to various types of music, over several days before sitting down, doing A/B comparisons, and taking the following notes. There was no need to do the comparison blind - the difference was quite easy to hear. The following is straight out of my note book, in the order I observed things:
- increase in depth (of soundstage)/particularly in bass
- increase in clarity
- bass noticeably better defined
- increase in realism
- improved timing/instrument's (transient) attack
- soundstage seems larger
- substantial increase in realism
- much more coherent
- regular filters sound muddy and weak in comparion
- poorly (multi-mic'd, heavy EQ) recordings don't show as much difference
So 'linear phase' filters have made a big change, and the notes describe it pretty well. I would surmise that phase has a lot to do with instrument placement in the soundstage, coherency of the soundstage, and transient response/coherency, all very welcome changes. OB speakers already have such an open spacious soundstage, but this really took things to the next level.
'Linear phase' wasn't as apparent on some music - one recording I listened to was of a rock band I really like, and I attended the recording. It was done with about 15 cheap mics in a small room, and then probably really tweaked. The music is awesome, but I couldn't detect any change between the two filter types. I assume that the phase information of each instrument was seriously degraded due to the multi-miking and EQ, so there was no difference to hear.
But other music was another story, including some better recorded rock. But when it came to classical music, WOW! It really separated the men from the boys, when it came to recording quality. I really felt I was looking at things right from the perspective of the microphones - I've never heard that level of realism. 'Linear phase' filters are much more enjoyable to listen to if the recording is done well.
Moving around the room, I assumed things wouldn't be as obvious, but they were. The bottom line is if you're a serious listener, you want to give this the effort it takes to get it right. Its big.
So that means I won't be moving the filter 'out of the box'. Thats okay, less work for me I guess. I'll just have to make the whole thing a little less obtrusive.
David - the measurements I did weren't acoustic - I took them straight off the output of the computer, as I was rebuilding the transfer functions. I can't tell you if the sound changed depending on my location, but I wouldn't expect it too much.
So, I wanted to post my impressions of using 'phase linear' crossover/EQ software, but I want to first point out that I haven't done before/after phase measurements, so I can't say precisely what has changed. But the change in sound is obvious, and I won't be going back. When I can do more rigorous measurements, thing may continue to improve, or at least change, so consider this part of the evolution. Previously I had been using Ultrafunk Sonitus:fx EQ (which is regular, non-linear phase software), and then moved to Waves LinEQ.
I listened to various types of music, over several days before sitting down, doing A/B comparisons, and taking the following notes. There was no need to do the comparison blind - the difference was quite easy to hear. The following is straight out of my note book, in the order I observed things:
- increase in depth (of soundstage)/particularly in bass
- increase in clarity
- bass noticeably better defined
- increase in realism
- improved timing/instrument's (transient) attack
- soundstage seems larger
- substantial increase in realism
- much more coherent
- regular filters sound muddy and weak in comparion
- poorly (multi-mic'd, heavy EQ) recordings don't show as much difference
So 'linear phase' filters have made a big change, and the notes describe it pretty well. I would surmise that phase has a lot to do with instrument placement in the soundstage, coherency of the soundstage, and transient response/coherency, all very welcome changes. OB speakers already have such an open spacious soundstage, but this really took things to the next level.
'Linear phase' wasn't as apparent on some music - one recording I listened to was of a rock band I really like, and I attended the recording. It was done with about 15 cheap mics in a small room, and then probably really tweaked. The music is awesome, but I couldn't detect any change between the two filter types. I assume that the phase information of each instrument was seriously degraded due to the multi-miking and EQ, so there was no difference to hear.
But other music was another story, including some better recorded rock. But when it came to classical music, WOW! It really separated the men from the boys, when it came to recording quality. I really felt I was looking at things right from the perspective of the microphones - I've never heard that level of realism. 'Linear phase' filters are much more enjoyable to listen to if the recording is done well.
Moving around the room, I assumed things wouldn't be as obvious, but they were. The bottom line is if you're a serious listener, you want to give this the effort it takes to get it right. Its big.
So that means I won't be moving the filter 'out of the box'. Thats okay, less work for me I guess. I'll just have to make the whole thing a little less obtrusive.
Your subjective impressions of linear-phase vs minimum-phase EQ is very much like my own. I did use Waves Paragraphic, and then I changed to Waves LinEq. Later on I changed to PLParEQ, which I think sounds smoother than Waves. No by much, but noticeable.
Cool that you've been using Ultrafunk! One of the guys that was involved in Ultrafunk is actually a friend of mine. The did make a lin-phase EQ, but I dont remember what it was called. Ultrafunk has been out of business for a while, but they have been revived and new exiting products will come.
Cool that you've been using Ultrafunk! One of the guys that was involved in Ultrafunk is actually a friend of mine. The did make a lin-phase EQ, but I dont remember what it was called. Ultrafunk has been out of business for a while, but they have been revived and new exiting products will come.
Last edited:
Another update - a couple of days ago, I spent about 2 hours listening to music at 'loud' levels - I would guess the neighbors could hear it, at least if their windows were open. The news is that nothing blew up. No magic smoke escaped. So, as a general statement, these speaker can play 'very loud'. At 'insanely loud', I started noticing the sound becoming unpleasant, so I would guess the system was running into nonlinear distortion problems. Trying to put a number to this SPL level is difficult, but I'll try at some point. The bottom line is that nothing melted, so we know the system isn't thermally limited when it comes to music. I was listening to mostly classical and rock.
Another thing I've been noticing too, that I don't fully understand yet. I've been listening to much of my music collection for a number of years, and probably heard each piece a thousand times - but since using the 'phase linear' software, I've begun hearing clipping - both digital and analog. I'm still trying to track it down, but it seem to be in the recording, and I've just never noticed it before. It shows up at the same point on certain recordings every time, no matter of level. It isn't my computer clipping, it isn't my amps, and the drivers seem fine, nothing blown as far as I can tell. So it must be in the recording. It shows up at the same places in the recordings regardless of level - and it only happens in a couple of places. I'm just surprised I've never noticed it before, I've listened to those recordings so many times.
Similar story with low level noise on certain recordings - I'm hearing stuff I've never noticed before. Surprising.
Another thing I've been noticing too, that I don't fully understand yet. I've been listening to much of my music collection for a number of years, and probably heard each piece a thousand times - but since using the 'phase linear' software, I've begun hearing clipping - both digital and analog. I'm still trying to track it down, but it seem to be in the recording, and I've just never noticed it before. It shows up at the same point on certain recordings every time, no matter of level. It isn't my computer clipping, it isn't my amps, and the drivers seem fine, nothing blown as far as I can tell. So it must be in the recording. It shows up at the same places in the recordings regardless of level - and it only happens in a couple of places. I'm just surprised I've never noticed it before, I've listened to those recordings so many times.
Similar story with low level noise on certain recordings - I'm hearing stuff I've never noticed before. Surprising.
Can you make up a (short) list of recordings where you hear clipping? Preferably well-known stuff. I'd like to check it out....
I'm not suprised if its the recording. A lot of albums are compressed or peak limited a little bit too much.... to say the least! Even classical records are compressed and limited to some degree.
I'm not suprised if its the recording. A lot of albums are compressed or peak limited a little bit too much.... to say the least! Even classical records are compressed and limited to some degree.
I don't think you will run into trouble with W4-1320 until hitting the power limit. X-max will be sufficient and a reasonably steep crossover will get the rest. Here is a rms simu for cone travel at 100 dB with 12dB/octave crossover at 250 Hz, to get resonable real value multiply with 1.414 (SQR(2)).
/Erling
An externally hosted image should be here but it was not working when we last tested it.
/Erling
You mean CPU resources? I so - then the audio should be choppy all the times, thats how that problem sounds (I've been there...)
I'll find two or three songs, and post them. Having others check is a good idea.
It is clearly a clipping sound - not a cpu choking type sound.
Thanks Skorpion for the graph - I have MJK's sims too, and use them all the time. Very very useful. I don't know if I remember well enough, but my graphs looked just slightly worse (aka, more cone excursion).
I think these are the relevant parameters I used (besides the W4's TS parameters input at the beginning of the worksheet): LR4 at 275Hz; the driver's spec is 3mm Xmax, so that is equal to 2.1mm rms deflection, as a calculated maximum; the baffle size is a little smaller than the frame (due to round driver vs square sim'd baffle); often I look at 2m rather than 1m distance; no floor reflections; I set the power in to the max the driver is spec'd for; then I look at if I exceed the rms deflection value, and what dB SPL I get at about 300Hz. IIRC, things didn't look to hot, but in the end, worked out.
EDIT: I ran the sims - 50W power in wrt 7ohms, 2m away, no floor, 4inch baffle. I got a rms deflection of .6 (out of 2.1 max calculated), but only 85dB SPL at 300Hz (pretty low efficiency!!). The good news is that its doing about 95dB SPL at 550Hz, and 100dB SPL at 1kHz. The low efficiency in the mid-hundred Hz is one of the reasons for going to a 4 way...
It is clearly a clipping sound - not a cpu choking type sound.
Thanks Skorpion for the graph - I have MJK's sims too, and use them all the time. Very very useful. I don't know if I remember well enough, but my graphs looked just slightly worse (aka, more cone excursion).
I think these are the relevant parameters I used (besides the W4's TS parameters input at the beginning of the worksheet): LR4 at 275Hz; the driver's spec is 3mm Xmax, so that is equal to 2.1mm rms deflection, as a calculated maximum; the baffle size is a little smaller than the frame (due to round driver vs square sim'd baffle); often I look at 2m rather than 1m distance; no floor reflections; I set the power in to the max the driver is spec'd for; then I look at if I exceed the rms deflection value, and what dB SPL I get at about 300Hz. IIRC, things didn't look to hot, but in the end, worked out.
EDIT: I ran the sims - 50W power in wrt 7ohms, 2m away, no floor, 4inch baffle. I got a rms deflection of .6 (out of 2.1 max calculated), but only 85dB SPL at 300Hz (pretty low efficiency!!). The good news is that its doing about 95dB SPL at 550Hz, and 100dB SPL at 1kHz. The low efficiency in the mid-hundred Hz is one of the reasons for going to a 4 way...
Last edited:
Yes, my intention was to stress the usefulness of this TB driver. I also have a pair to test. In a Swedish DIY show they really shined in small closed boxes on top of closed bassboxes with Peerless elements, giving very natural, neutral sound.
As I have been playing around recently with Neo3PDR (non dipole) and 18Sound's 6ND430, which is a formidable combination, I think the TB's will pair very well with the Neos.
Your bass solution really gives thoughts to a pair narrow U-baffles, which yours are but perhaps with a little wider sides and top, in dipole configuration with some air in between. Also in Europe there might be 10" elements for a similar setup for very little money only element quality beeing the question.
/Erling
As I have been playing around recently with Neo3PDR (non dipole) and 18Sound's 6ND430, which is a formidable combination, I think the TB's will pair very well with the Neos.
Your bass solution really gives thoughts to a pair narrow U-baffles, which yours are but perhaps with a little wider sides and top, in dipole configuration with some air in between. Also in Europe there might be 10" elements for a similar setup for very little money only element quality beeing the question.
/Erling
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.