Dan, what's apropos to a crossover's radiation pattern is the behavior of the crossover's electrical outputs relative to each other. Converting one to linear phase and leaving the other warped phase won't preserve phase tracking. To preserve a consistent radiation pattern the FIR patch up you describe will end up applying the same phase correction to the tweeter as was applied to the woofer. So you might as well save yourself the hassle and just light up the active button on the Arbitrator channel. 
Managing output phase shifts due to Fs, ports, or acoustic center alignment is a separate set of corrections. You're correct they're important for getting the overall response flat phase, but as a design consideration that's distinct from the radiation pattern effects of switching from one crossover to another. Typically Fs for midranges and tweeters is low enough it doesn't cause signficant phase shifts within their pass or transition bands. Hence Arbitrator's UI assumes the applied port/free air/bandpass phase correction only affects the woofer. That doesn't hold for pro drivers, particularly in four ways, and I'm trying to convince Jan some improvements would be helpful.
SoundEasy has an FFT based crossover emulator. It doesn't do FIR and you have to design a passive crossover and plug in driver models to use it. Good for planning passive crossover builds, very cumbersome as a digital crossover.
Managing output phase shifts due to Fs, ports, or acoustic center alignment is a separate set of corrections. You're correct they're important for getting the overall response flat phase, but as a design consideration that's distinct from the radiation pattern effects of switching from one crossover to another. Typically Fs for midranges and tweeters is low enough it doesn't cause signficant phase shifts within their pass or transition bands. Hence Arbitrator's UI assumes the applied port/free air/bandpass phase correction only affects the woofer. That doesn't hold for pro drivers, particularly in four ways, and I'm trying to convince Jan some improvements would be helpful.
SoundEasy has an FFT based crossover emulator. It doesn't do FIR and you have to design a passive crossover and plug in driver models to use it. Good for planning passive crossover builds, very cumbersome as a digital crossover.
Right, my point was not so much that the radiation pattern would be different, but that the frequency response in general would be different. Even if you change both both the electrical high and lowpass sections to linear phase, the phase tracking will be different than with min. phase filters because the electrical portions were probably not symmetric. Their contributions to the final acoustic response was not equal so the new response with linear phase filters could be different.
Yeah, point well taken. I haven't used Arbitrator for a while, it's time to do some more A/B comparisons with it again. I was interested in cuibono's filtering setup because he had not been that impressed with Arbitrator but seemed pretty happy with the linear phase filters he's using now.
Dan
cuibono,
I think you mentioned the Tang Band 1320's were just about the right size to cross to your tweeters to balance radiation pattern while not being too small that there are air flow issues at the back of the driver. Did you look at any even smaller drivers than the 1320? If so, which ones and what did you find?
Dan
I think you mentioned the Tang Band 1320's were just about the right size to cross to your tweeters to balance radiation pattern while not being too small that there are air flow issues at the back of the driver. Did you look at any even smaller drivers than the 1320? If so, which ones and what did you find?
Dan
In the general case, yes, hence my remark about the phase tracking needing to remain the same. Personally most of the crossovers I've encountered are symmetric, with LR2, B3 (Butterworth), LR4, and LR8 being most common.
I took a look at active and passive crossover design and concluded I might as well go direct to digital, but along the way I got the impression the main reason for using asymmetric slopes in analog crossovers is time aligning drivers. That's of limited utility in dipoles as time aligning the front wave degrades back wave alignment; better to get the physical alignment correct. And, with a digital crossover, it's easier to change the channel's delay setting than to vary slope. The other reasons I've seen for asymmetric slopes are reducing component cost and development time and maybe trying to dodge a problematic driver behavior. IMO these don't really apply in digital as the incremental cost of using and tuning higher order crossovers is low.
Sorry for not responding - I moved last week...
Dan - I've looked quite a bit at 3" drivers, and there are a few I'm interested in. But the thing is, I really doubt that you could use only a 3" driver as a mid in a 3-way. The other thing is that there are no guarantees on the radiation pattern for the 3", so the only way to find out is buy one and measure.
Dan - I've looked quite a bit at 3" drivers, and there are a few I'm interested in. But the thing is, I really doubt that you could use only a 3" driver as a mid in a 3-way. The other thing is that there are no guarantees on the radiation pattern for the 3", so the only way to find out is buy one and measure.
Whoa!
I just realized, 3.5m in the air is over 11 ft. high. How did you suspend your speakers and mic that high?
Yeah, at some point soon I'll have to do that. From skimming over this and other threads I'm thinking the Neo3 matched to either the Neo8 or TB1320 is a good place to start with open baffle. On using a 3" driver in a 3 way, I was thinking more along the lines of monopole subs with 3-way dipoles running from somewhere around 100-150Hz. That might allow the use of one or two 8" or 10" drivers to run to a higher crossover point that a 3" driver could handle. Still, that's probably better considered a 4way since the subs might need to run off a different signal that the dipole woofers.
Dan
I just realized, 3.5m in the air is over 11 ft. high. How did you suspend your speakers and mic that high?
Yeah, at some point soon I'll have to do that. From skimming over this and other threads I'm thinking the Neo3 matched to either the Neo8 or TB1320 is a good place to start with open baffle. On using a 3" driver in a 3 way, I was thinking more along the lines of monopole subs with 3-way dipoles running from somewhere around 100-150Hz. That might allow the use of one or two 8" or 10" drivers to run to a higher crossover point that a 3" driver could handle. Still, that's probably better considered a 4way since the subs might need to run off a different signal that the dipole woofers.
Dan
Hi Dan - to measure speakers, I basically have a small table with 8' beams for legs. For the mic, I have a camera tripod, the legs of which I taped long poles to. Pretty home made, but works fine. I've taken pictures, just not posted them.
If you only want to go down to 150Hz, there are lots of 8" and 10" possibilities. That is one of the major benefits wrt the mid section of OB speakers - if you split the 150-2500Hz range between two drivers, your selection of appropriate drivers goes from almost nill to almost any.
If you only want to go down to 150Hz, there are lots of 8" and 10" possibilities. That is one of the major benefits wrt the mid section of OB speakers - if you split the 150-2500Hz range between two drivers, your selection of appropriate drivers goes from almost nill to almost any.
cuibono - since my last post to this thread I started my own experimental OB project, and in the course of driver testing I seem to reach every point you already got to. I hope you take my gradually copying your design as a compliment.
At the moment I'm discovering the low end limits of the 1320s, and I'd be very glad to hear about your experiences when driving them at higher SPLs. My own findings are that those little gems turn into nasty little noisemakers when driven below 400 Hz and above 80-85 dB SPL @ 0.5m - can you confirm that? At what levels are you driving your system?
And apart from that: Did you change anything in the setup, or are you still using the 3-way configuration?
Thanks for an update on that!
At the moment I'm discovering the low end limits of the 1320s, and I'd be very glad to hear about your experiences when driving them at higher SPLs. My own findings are that those little gems turn into nasty little noisemakers when driven below 400 Hz and above 80-85 dB SPL @ 0.5m - can you confirm that? At what levels are you driving your system?
And apart from that: Did you change anything in the setup, or are you still using the 3-way configuration?
Thanks for an update on that!
Hi Wowo,
The model I use is the w4-1320sj - did you get the same? I haven't had any issue with their low end performance. I drive them really loud.
It looks like you are only judging by measurements, and you're not too sure of your setup. I would check by listening too, before making judgement. I do non-linear distortion tests of the drivers, and had no issue. I didn't save the data, so I can't show it though...
For the last month, I've been away from home at a new job. I'm not currently doing any audio stuff, but a break is okay...
Good luck with your driver. I had serious issues with the TB driver I was using before the 1320sj - it had manufacturing issues - I had 4 of them, and they all had the same issue, some sort of buzz. You might be seeing something similar...
The model I use is the w4-1320sj - did you get the same? I haven't had any issue with their low end performance. I drive them really loud.
It looks like you are only judging by measurements, and you're not too sure of your setup. I would check by listening too, before making judgement. I do non-linear distortion tests of the drivers, and had no issue. I didn't save the data, so I can't show it though...
For the last month, I've been away from home at a new job. I'm not currently doing any audio stuff, but a break is okay...
Good luck with your driver. I had serious issues with the TB driver I was using before the 1320sj - it had manufacturing issues - I had 4 of them, and they all had the same issue, some sort of buzz. You might be seeing something similar...
Thanks for your reply, cuibono!
I'm using the W4-1320SI, which is basically the model you're using without the flange cut-offs. Concerning the measurements: Yes, after getting these data I was pretty unsure and rechecked everything, but after going through that and finding no other culprit, I listened closely to driver (playing test tones), and at levels > 85 dB there was a clear (and beyond 90 dB SPL downright nasty) rasp at frequencies around 300 Hz and some ringing around 700 Hz, although less pronounced.
I haven't checked the behaviour at those levels when they are playing music, but I guess rasping with sinus tones doesn't promise any good for music...
So I might, in the end, have run into similar issues as you did with the drivers before the 1320. I guess I have to talk to the dealer (who luckily is located just a few bus stations from my home).
I'm using the W4-1320SI, which is basically the model you're using without the flange cut-offs. Concerning the measurements: Yes, after getting these data I was pretty unsure and rechecked everything, but after going through that and finding no other culprit, I listened closely to driver (playing test tones), and at levels > 85 dB there was a clear (and beyond 90 dB SPL downright nasty) rasp at frequencies around 300 Hz and some ringing around 700 Hz, although less pronounced.
I haven't checked the behaviour at those levels when they are playing music, but I guess rasping with sinus tones doesn't promise any good for music...
So I might, in the end, have run into similar issues as you did with the drivers before the 1320. I guess I have to talk to the dealer (who luckily is located just a few bus stations from my home).
Drivers arranged radially from the ear
Hi cuibono
Fantastic work, and thank you for sharing and explaining it so well.
(What would/ do SL or JK think?)
I was pointed here. This thread and a local audio guy changed the way i am approaching my OBs http://www.diyaudio.com/forums/multi-way/165051-3-way-open-baffle-shape.html I’ve now decided to use a small/ nude baffle; and just bought a different tweeter, to cross (instead of c 10 kHz) at about 2 - 2.5 kHz.
At this stage I'm planning my wood cuts - The principle of your drivers arranged "radially" is understandable, but how you did it in wood, physically - can you post a pic of the revised speaker please?
Hi cuibono
Fantastic work, and thank you for sharing and explaining it so well.
(What would/ do SL or JK think?)
I was pointed here. This thread and a local audio guy changed the way i am approaching my OBs http://www.diyaudio.com/forums/multi-way/165051-3-way-open-baffle-shape.html I’ve now decided to use a small/ nude baffle; and just bought a different tweeter, to cross (instead of c 10 kHz) at about 2 - 2.5 kHz.
At this stage I'm planning my wood cuts - The principle of your drivers arranged "radially" is understandable, but how you did it in wood, physically - can you post a pic of the revised speaker please?
Last edited:
Crossover - “Semi active” linear phase?
If you go active on the bass XO, but having a *significant investment in a very good (tube) amp, so prefer to use passive between the midrange and tweeter, is it still possible to do linear phase?
. . With any of
Thuneau's Allocator+ PLParEQX3+ Bidule (twest820)
PLParEQ (full version) (StigErik)
Reaper + Waves LinEQ (cuibono)
an Ultrafunk linear-phase EQ? (Stig, your Ultrafunk friend – a new product?)?
. . ie can linear phase still be implemented, without adding too much to the complexity (I’ll take Todd’s word, that excludes Octave or Matlab)?
Or might SoundEasy now do it??
Since that post (March 27th) SoundEasy v 17 has come out
From Home Page
"Digital Filter (FIR implemented as Fast Linear Convolution, and also as Partitioned Convolution ) function has been provided for auditioning your crossover design via sound card. Our implementation includes simulating sound propagation delays due to different spatial location of the acoustic centres of drivers."
i have a long learning curve ahead of me, but this and a look over SE's Reference Manual (ch 9) Filter And Crossover Design - Passive / active / digital FIR networks . . and (new ch 19) "DSP section, Includes Ultimate Equalizer" - my reading is that SE now does do FIR . . (but is it "fully"/ well?).
Whether it definitely does, and how it compares to Arbitrator etc . . . comments welcome
If you go active on the bass XO, but having a *significant investment in a very good (tube) amp, so prefer to use passive between the midrange and tweeter, is it still possible to do linear phase?
. . With any of
Thuneau's Allocator+ PLParEQX3+ Bidule (twest820)
PLParEQ (full version) (StigErik)
Reaper + Waves LinEQ (cuibono)
an Ultrafunk linear-phase EQ? (Stig, your Ultrafunk friend – a new product?)?
. . ie can linear phase still be implemented, without adding too much to the complexity (I’ll take Todd’s word, that excludes Octave or Matlab)?
Or might SoundEasy now do it??
Since that post (March 27th) SoundEasy v 17 has come out
From Home Page
"Digital Filter (FIR implemented as Fast Linear Convolution, and also as Partitioned Convolution ) function has been provided for auditioning your crossover design via sound card. Our implementation includes simulating sound propagation delays due to different spatial location of the acoustic centres of drivers."
i have a long learning curve ahead of me, but this and a look over SE's Reference Manual (ch 9) Filter And Crossover Design - Passive / active / digital FIR networks . . and (new ch 19) "DSP section, Includes Ultimate Equalizer" - my reading is that SE now does do FIR . . (but is it "fully"/ well?).
Whether it definitely does, and how it compares to Arbitrator etc . . . comments welcome
Yes; just use one of the Arbitrator channels in Allocator to correct the passive crossover. (So long as the passive crossover is a reasonable approximation of an odd order Butterworth or even order Linkwitz-Riley alignment.) To my knowledge this capability is unique to Arbitrator.
If you think about it you'll see why SoundEasy's crossover emulation would require you to get a second tube amp.
Thanks for the appreciation, I'm glad you got something out of this thread! Unfortunately, I'm 400 miles from my speakers, and will be for a while. But what I did was fairly simple, but maybe hard to explain:
The main thing is that the ACs of each section (woofer, mid, tweeter) are the same distance from your ear, and pointed at it. So I rigged up a string from the back of my chair, anchored at a point where one of my ears would be (or you could use the center of your head, if you wanted). I then marked a specific distance on the thread, and pulled it taught. I used the far end to find out where the AC of each section should be in space (the string keeps it a fixed distance), and what angle of tilt I needed for them to each point at my ears. The woofer section needed to tilt back about 25 degrees, and the mid and tweeter less, something like 17deg, IIRC. I then just cut supports for each section so they would stay where I wanted them (remember, I'm not using any baffle for the mid or tweeter - it was just a matter of cutting an angle into the bottom of the post they are attached to).
Hopefully, that is clear enough...
Its possible yes, but do I know how, no. The only reason I was able to adjust phase is because my active EQ compensated for it automatically - and it still has issues. Particularly, I know of no solution that allows you to specify a degree of phase rotation for a given frequency - which would be necessary to get true phase flatness (at a given location).
JohnK knows how to compute models that will do this, but I don't.
SoundEasy is definitely not going to 'play nice' if used regularly to listen to audio through. It isn't very user friendly.
Arbitrator has serious user issues too, and fairly limited options for phase correction. My simple experiments with it made me abandon it.
I'll be looking more into this in the future. I always discounted phase, but my recent experiments showed me that it has a big effect - if you are a serious/critical listener.
Interesting; can you be more specific? Other than dropping the signal stream when the correction order is changed (not a big deal as it's easy to bring the stream back online and changing orders is uncommon) it works just fine for me. I find UI quite simple and straightforward and light years easier to work with than SoundEasy.
The UI is fairly easy to deal with, but...
I haven't used Allocator in over a year, so I don't remember all the details - but the most infuriating thing was how difficult it was to get Allocator to work with other pieces of software. The total chain was a serious pain to get started - aka, connecting my music database (winamp) with allocator (via some other sketchy/free piece of software) and then from allocator to my soundcard (again with the same interconnection issues). I also found Allocator to be buggy - like if your driver file you imported had too many data points, the program would immediately close. At some point, the program wouldn't even open anymore, and I had to reinstall the whole thing.
And getting Soundeasy to interface with Allocator? Forget it not a CHANCE! That was what killed it for me.
Also, there is very little support for the product. I spent $150 on something that didn't work very well, when considering all the tweaking it needed to actually be functional. I wasted many frustrated hours trying to get it setup, and paid too much for it, IMO. Considering the cost, one would hope that it would actually be reasonably easy to implement.
Additionally, its ability to manipulate the XO and EQ is adequate, but nowhere near as powerful as other, free software (FREE!). Also, that free software has no setup issues! Hmmm....
At least SE isn't buggy. It is kind of primitive, in terms of UI, but at least it does its job. Actually, now that I know more about what I'm doing, I wouldn't buy SE either - I'g get ARTA. SE emphasizes more of the box/crossover modeling, which I don't use at all (it has no support for OB modeling). SE's meaurement suite, while powerful in some ways, is primitive in others (particularly data handling/manipulation). ARTA looks much more friendly in those regards. And MJKs worksheets are literally a godsend for OB modeling!
I haven't used Allocator in over a year, so I don't remember all the details - but the most infuriating thing was how difficult it was to get Allocator to work with other pieces of software. The total chain was a serious pain to get started - aka, connecting my music database (winamp) with allocator (via some other sketchy/free piece of software) and then from allocator to my soundcard (again with the same interconnection issues). I also found Allocator to be buggy - like if your driver file you imported had too many data points, the program would immediately close. At some point, the program wouldn't even open anymore, and I had to reinstall the whole thing.
And getting Soundeasy to interface with Allocator? Forget it not a CHANCE! That was what killed it for me.
Also, there is very little support for the product. I spent $150 on something that didn't work very well, when considering all the tweaking it needed to actually be functional. I wasted many frustrated hours trying to get it setup, and paid too much for it, IMO. Considering the cost, one would hope that it would actually be reasonably easy to implement.
Additionally, its ability to manipulate the XO and EQ is adequate, but nowhere near as powerful as other, free software (FREE!). Also, that free software has no setup issues! Hmmm....
At least SE isn't buggy. It is kind of primitive, in terms of UI, but at least it does its job. Actually, now that I know more about what I'm doing, I wouldn't buy SE either - I'g get ARTA. SE emphasizes more of the box/crossover modeling, which I don't use at all (it has no support for OB modeling). SE's meaurement suite, while powerful in some ways, is primitive in others (particularly data handling/manipulation). ARTA looks much more friendly in those regards. And MJKs worksheets are literally a godsend for OB modeling!
Ah, my experience is nearly all the routing pain goes away if you have an audio interface with loopback---it's a pretty common way of getting around these ASIO limitations that really have nothing to do with Allocator. Jack's worked well for me too, but IMO it's a bit of a hack. I have hit the crash on too many datapoints in Allocator (not an Arbitrator issue, which is what we were originally discussing) but that's the only bug I can remember encountering. I've found Jan's decent with support---better than one can expect most places---though he's more focused on finding workarounds instead of on making Allocator better than I'd like.
Interestingly, my experience is SoundEasy is riddled with bugs and crashes regularly. All sorts of stuff that I consider basic---like being able to do a frequency sweep or model an enclosure---failed more often than it worked, usually with errors that were either wrong or so vague as to be useless. In some cases I was able to successfully second guess the code to figure out the what the bugs were and, in some case, figure out a workaround. Usually I and everyone else on the SoundEasy eGroup would get stuck and Bohdan wouldn't bother to reply. I've also developed a deep mistrust of SoundEasy's MLS implementation as the results I get from it are unstable and prone to noise. I'm much happier with a PC crossover and HOLMImpulse. In principle not having SoundEasy costs me the ability to IMD sweeps, but they never worked anyway so it doesn't really matter.
Same planet, different worlds. I did make quite sure I could everything I needed in the demo version of Allocator before committing to it. I do everything with HOLMImpulse in ASIO; I'm confident loopback would route SoundEasy stimuli through Allocator but I'm not entirely sure about SoundEasy's ability to pick up the results as it lacks ASIO support.
Interestingly, my experience is SoundEasy is riddled with bugs and crashes regularly. All sorts of stuff that I consider basic---like being able to do a frequency sweep or model an enclosure---failed more often than it worked, usually with errors that were either wrong or so vague as to be useless. In some cases I was able to successfully second guess the code to figure out the what the bugs were and, in some case, figure out a workaround. Usually I and everyone else on the SoundEasy eGroup would get stuck and Bohdan wouldn't bother to reply. I've also developed a deep mistrust of SoundEasy's MLS implementation as the results I get from it are unstable and prone to noise. I'm much happier with a PC crossover and HOLMImpulse. In principle not having SoundEasy costs me the ability to IMD sweeps, but they never worked anyway so it doesn't really matter.
Same planet, different worlds.
I did make quite sure I could everything I needed in the demo version of Allocator before committing to it. I do everything with HOLMImpulse in ASIO; I'm confident loopback would route SoundEasy stimuli through Allocator but I'm not entirely sure about SoundEasy's ability to pick up the results as it lacks ASIO support.Specifically? Last I checked the Waves bundles which included linear phase equalization were a couple grand.
Thanks Cuibono
I hadn’t thought of something as simple string . . but if it works
Your measuring method is clear. It’d be interesting to see the tilted supports, when you get a chance.
Sonoma is Wine Country, hope the close proximity doesn’t distract you from getting closer to audio nirvana.
TW,
What do you mean by an audio interface with loopback?
Are you referring to the soundcard? I have the M-Audio Delta 410
I hadn’t thought of something as simple string . . but if it works
Your measuring method is clear. It’d be interesting to see the tilted supports, when you get a chance.
Sonoma is Wine Country, hope the close proximity doesn’t distract you from getting closer to audio nirvana.
TW,
What do you mean by an audio interface with loopback?
Are you referring to the soundcard? I have the M-Audio Delta 410
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.