ZDT3.5 (sort of): Duelund crossover design (I hope)

[fjhuerta]

Hi all,

Maybe some of you remember I built the ZDT3.5 some time ago, but I used the Vifa DQ25 tweeter instead of the small Neo dome. There was a reason for it - I had the Vifas, and I thought I could manage to fix the crossover as needed to use them.

Indeed, the little Vifas worked fine, and I was happy for a while. Then, a comment by John (Zaph) made me curious. He said that the Vifa would handle a 2nd order crossover.

So, I was curious too, and started modding the crossover. Months later, the Duelund calculator software was posted in this forum, and I noticed how similar the curve of my crossover was to a Duelund.

So I decided to give it a try. It's not a "pure" Duelund implementation (that'd require aligning the drivers to 1mm from the listening position!), but it follows the curves pretty closely enough.

So, I'm posting the design. In a way, it ended up being somewhat similar (at least in concept) to Jon's center channel.

[Jed]

Months later, the Duelund calculator software was posted in this forum, and I noticed how similar the curve of my crossover was to a Duelund.

Are your slopes LR2ish? Dennis H posted how using LR2 is close to a "true" Duelund crossover transfer function if using the correct FCs.

By the way Jon's Center uses slopes that are closer to 4th order on the tweeter and woofer sections IIRC. There's a lengthy explanation about why he chose those slopes to approximate the Duelund curve. So in your design, if it uses LR2 slopes is not the same as Jon's even if they share similar Duelund characteristics. Not saying there is anything wrong with that.

[fjhuerta]

OK, so the RS-52 can't go much lower (if at all) than 850 Hz, as used on the ZDT. So I began from there. Considering the size of the frame of the RS-52 and the Vifa, a 4 KHz x-over point seemed to fit perfectly with the requirements of a Duelund crossover, and were pretty close to where I was at that point.

So I loaded the curves on the Duelund web software calculator.



Apparently, everything will be fine. Other than a problem with some extra output from the tweeter (I didn't do anything about its Fs), it looked doable.

Little did I know how big and heavy the end result would be!

[fjhuerta]

The crossover schematic is pretty simple. The woofer has a second order filter and an RC shunt to shape the response a bit in the upper end. The midrange is a bit more complex, but still it keeps the elegant simplicity of Zaph's original design; except for a huge (105 uF!) series filter, which helps tame a midrange peak which, I suspect, is diffraction from the enclosure (and since I was going for perfection, and had a couple of spare caps, I went for it). The values were tweaked a bit because I felt the speakers were a bit too forward (considering my hardwood floors), so I added a bit more BSC. Finally, the tweeter is a standard 2nd order with a parallel filter, which eliminates the Fs resonance impedance hump

It's big, heavy, and it was expensive. Or so I think, since I managed to get it built with recycled components. Recession busting, indeed.

So, how did it all work out? Pretty good, I believe.

The FR: As flat as I could get it to be.



Two RS-180s in parallel are never going to win any Best Friends award from an 8 Ohm rated amplifier, but other than that, the curves are benign enough.



Simulated off-axis is smooth, as a 3-way should be.



And finally - the impulse response. According to Duelund's papers, I'd have expected a sharp rise followed by a slow and steady decline. I don't know whether this simulation is correct.

[fjhuerta]

Are your slopes LR2ish? Dennis H posted how using LR2 is close to a "true" Duelund crossover transfer function if using the correct FCs.

By the way Jon's Center uses slopes that are closer to 4th order on the tweeter and woofer sections IIRC. There's a lengthy explanation about why he chose those slopes to approximate the Duelund curve. So in your design, if it uses LR2 slopes is not the same as Jon's even if they share similar Duelund characteristics. Not saying there is anything wrong with that.

Yep, you are right. I believe my choice of the a variable made my design a 2nd order one. I didn't see Dennis H's post, but my experience mirrored his post - LR2 was very, very close to the Duelund version of the crossover with a=2.6. I had to swap a couple of components only.

[fjhuerta]

But, is it really a Duelund?

Only one way to find out for sure.

I downloaded the transfer functions from the Duelund Synkron web calculator site, and loaded them into LSPCad.

These are my results. I tried to get them as close as possible to the transfer functions, but due to the speakers' own response and the fact that the crossover was simply huge at that point made me make some compromises.

Still, pretty close if you ask me.





Note: I added the upper and lower woofer and eliminated tweeter and mid so you could get a better idea of how, when they sum up, they follow the Duelund topology.

[fjhuerta]

Phew.

So, how do they sound? This is pretty difficult to explain. The best way I can do it is - they sound like a true full range crossoverless speaker at a time. The first few times I listened to them I felt that maybe the tweeter level was set too low. Then I noticed that, quite simply, I was missing the usual emphasis on the tweeter - every little detail was there, but it wasn't calling attention to itself. Same thing with the midbass - every driver blends perfectly well.

I enjoyed the experience a lot. In fact, my next design will probably be a Duelund from the start!

[syncroniq]

Hi,

Great to see the calculator is being used
Nice filterjob there.

Best Henrik Høegh

[ch83575]

Javier,

That is a really impressive project! I bet they sound great. Looking at the off-axis plots make me think that with a little bit of waveguide action on the mid-dome this could be perfectly CD. Even without it, the power response looks phenomenal. I have no idea how big the waveguide would need to be, and obviously the tweeter would need to be installed mid-waveguide, but its an idea anyways.

Again, great work. It makes me want to build one!

-Chad

[JonMarsh]

Phew.

So, how do they sound? This is pretty difficult to explain. The best way I can do it is - they sound like a true full range crossoverless speaker at a time. The first few times I listened to them I felt that maybe the tweeter level was set too low. Then I noticed that, quite simply, I was missing the usual emphasis on the tweeter - every little detail was there, but it wasn't calling attention to itself. Same thing with the midbass - every driver blends perfectly well.

I enjoyed the experience a lot. In fact, my next design will probably be a Duelund from the start!

Ah Hah! The proof is in the pudding, and you've got the right flavor pudding!! :T No frequency range should draw attention to itself, unless the recording was made that way deliberately!

The actual shapes for the filters will vary as you vary the aleph coefficient and the peak level and spread of the midrange. You've used a bit higher center frequency, at 2 kHz; it wasn't easy to pull off 1600 Hz, but this worked better for the dispersion requirements of a speaker to be used both as a mains and center channel.

I think you will find that the more you listen to these, and the more you tweak or improve other aspects of your system, you will be increasingly rewarded in a musical sense.

I think you're only the second person I've seen on HT Guide or elsewhere to do a Duelund derived project or update in the last two years. If there are any others, of course I'd like to hear about them.

On other way to check your LspCAD simulations is to plot the phase response of the three drivers, ideally they should overlay each other (as the drivers should be in the same relative phase at all frequencies (of interest- my addition); in practice, I think this means keeping good amplitude and phase tracking down to about the -20 dB level- more if possible. Your amplitude tracks are pretty close.

The only suggestions I might make would be to add a trap filter to stomp on the RS180 top end breakups above 5 kHz, the tope end should sound even cleaner then. and you could relax the roll off rate a skosh on the RS180 to 4 kHz so that it tracks better the desired amplitude and phase.

As Steen says, you should plan to spend more on the crossover than the drivers; he usually used the whole back panel of the speaker for laying out and building the crossover. Not hard to see why, depending on the driver choices and crossover frequencies.

[benchtester]

Great project!

I am a little confused, though. I thought that with the Duelund crossover all the drivers were in phase. The crossover posted shows the mid installed with reverse phase?

[syncroniq]

Hi,

Heres a picture of a real duelund filter

Image not available

Heres another one

Images not available

Btw, when you meassure, do you use minimum phase? Do you delay the response of the bass driver accordingly? By using minimum phase, its as if the drivers where time aligned. I think you need to delay the bas and mid by substracting the stepresponses. Or am i wrong?

But super cool project.

Heres my 3way scanspeak project in and out of phase (the mid)

[fjhuerta]

Syncroniq, I took measurements on tweeter axis, so I assume the woofer delay is already there.. am I right?

One thing I learned about the project is that time-aligning the drivers with a 2nd order filter is a must. If I wanted it to be better, I should have used a time ladder delay circuit, or used a stepped baffle. As it is, I have a nice null between mid and tweeter (since their acoustical centers aren't that far off, I think that's to be expected) but the null isn't as deep between the woofers and midrange.

No wonder the Duelund circuits are complex! But they are definitely worth the time, money and effort.

[syncroniq]

Hi,

I made a small tutorial on lspcad and clio.

Clio

http://forum.speakerbuilder.dk/forum_posts.asp?TID=24

lspcad

http://forum.speakerbuilder.dk/forum_posts.asp?TID=12

And here how i put the microfone, and why

http://forum.speakerbuilder.dk/forum_posts.asp?TID=5

I hope you can get eg google to translate it as its in Danish.

Minimum phase is frequency dependent. If you move the mic, the freq does nok change, så the minimum phase does not ether. Only when close (bafle step) or off axsis.

Im still new to minimum phase, and i will do a small experiment tomorrow if i get the time (have a 4month old daugther). It should be easy to test. I will post the result here then.

Have you meassured the final speaker? How does it meassure with the mid out of phase?

[fjhuerta]

Yes, I've taken measurements. The FR is identical for all practical purposes to the simulated one. With the mid out of phase I get severe cancellation in the upper midrange, IIRC. Since I used a very short gate, I didn't trust the lower frequency response.

[TacoD]

Good Job! Nice to see more Duelund projects.

I think you're only the second person I've seen on HT Guide or elsewhere to do a Duelund derived project or update in the last two years. If there are any others, of course I'd like to hear about them.

I did two prototypes as well, very promising approach.

Unfortunately, I'm still looking for a proper mid driver to do a full pistonic duelund speaker. Tweeter will be a Scanspeak 9800, for the bass I will use 2x excel W22EX. At the moment for the mid I'm sampling a Monacor SPH-130AL. All drivers are bought and stored, now to find some time.

[JonMarsh]

Great project!

I am a little confused, though. I thought that with the Duelund crossover all the drivers were in phase. The crossover posted shows the mid installed with reverse phase?

A Duelund filter is an all pass topology; there is one complete phase rotation to 360 degrees; the drivers are all in the same relative phase at any given frequency; the mid must be wired out of phase with the woofer and tweeter.






Depending on the aleph coefficient used, varying levels and bandwidth of the midrange result, and a varying slope of phase change in the all pass transition.



The ultimate slope is 4th order, but for some values of aleph an LR2 is a reasonable approximation target.

[fjhuerta]

Syncroniq, I've been reading the documents you posted, and I have a question - if your listening axis is at tweeters' height, shouldn't you place the measurement microphone exactly at that point? I'd see no problem with that, unless your listening axis is in the mid point between tweeter and woofer. Also, should I assume that your listening axis in a 3+ way is between the mid and tweeter?

[Face]

Hi,

Heres a picture of a real duelund filter

Image not available

I see some Duelund components. 8)

[syncroniq]

Hey,

fjhuerta > If you meassure at tweeter hight, the distance will change, when you get to the listening possition. A 3 way is more difficult yes

If i should give a "duelund answer" i would say "well, then you cant listen at tweeter height" :P


The problem is that you have to meassure the units with the samme distance difference, as they would have at the listening possition. Else, you will be simulating on a wrong phase. Duelund said that if you hadnt timealigned your speaker, it was worthless. I dont think its that critical, as long as you meassure the right phase, you can compensate for the time difference in the filter. Though it would not be a real duelund filter. My 3way is not time aligned, but still meassure close to a duelund.

I can recommend this seas as a large mid.

http://www.seas.no/index.php?option=com_content&task=view&id=63&Itemid=87

[cjd]

Measuring from a fixed position should net the correct timing in the measurements automatically - whether time aligned or not. Tweeter height is a design decision and suggests that the tweeter should be at ear-height when listening.

[fjhuerta]

That's what I thought, C. As long as your design axis equals your listening axis, you shouldn't have any problems (or so I think).

[Dennis H]

I agree with syncroniq. Measuring from a single position works but only if you allow for the geometry of, for example, measuring at 1m and listening at 3m. Here's a simple spreadsheet to figure out where to put the mic for a 2-way. Of course a 3-way is more complicated but the W-M alignment isn't usually as critical as the M-T alignment.

[cjd]

I think you're only the second person I've seen on HT Guide or elsewhere to do a Duelund derived project or update in the last two years. If there are any others, of course I'd like to hear about them.

I'd have to run the numbers through to see how close it got, but the in-wall InKhanNeatos are pretty damn close... ish. maybe...

And the depth of the soundstage with them in the wall was eerie. Very captivating.

C

[syncroniq]

Hi,

Sorry about my bad English. Its not my first language :lol:

I mean, if you measure at tweeter height, the difference of the distances between the drivers, will change, when you change the distance to the speaker.

See here.

Images not available

I know the units have to be time aligned for the above to be 100% true, but its alot closer. The ideal would be to measure along the ZDP axis, and that it hit the listening poss too.


But... when using minimum phase. I thing you remove the distance, as if you where measuring inside the voicecoil. So LspCAD would think they are time aligned. I will try a small simple experiment.

[syncroniq]

Hi,

I just did a fast test on minimums phase.

I measured a tweeter and woofer raw without crossover. Then i made a 1. order filter and measured it. Then i simulated the filter on the raw units with and without the delay between the two peaks of impulse responses.

Here's the result.
Red is the measured response with filter.
Blue is the simulated response of the same filter.

With delay

Image not available

Without delay

Image not available

So, minimum phase IS as measuring time aligned drivers. You have to add the delay to the bass, as the sound starts about where the voicecoil is attached to the cone.

I measured between the drivers.

You can do the same if you measure at tweeter height, but then it would not be correct at the listening position.

[---k---]

This is why everyone should just build big WMTMW's.

Edit: Actually, I don't think that solves the mic distance problem that syncroniq is describing. It just makes it so you can't move the mic location to correct for listening distance.

[Jed]

How to set delay and use on-axis minimum phase measurements is CLEARLY outlined in the LSPcad manual. Of course time of flight is accounted for by adjusting the driver locations in the coordinates for each driver. These coordinates can be found accurately by taking a measurement of all the drivers at the same time and from a fixed mic location... usually on the tweeter axis. Then after taking all the individual measurements of the drivers and they are imported into a parallel circuit, the driver offsets are adjusted until the SPL curve of the parrallel circuit's response (no passive components) matches the measured curve of all the drivers summed on the listening axis.

Then the simulations will accurately show the response of the drivers. Someone can also see how the response will vary given different listening distances by specifying that listening distance in the settings.

[cjd]

This is why everyone should just build big WMTMW's.

Edit: Actually, I don't think that solves the mic distance problem that syncroniq is describing. It just makes it so you can't move the mic location to correct for listening distance.

You should measure at listening height, always. You'll always be mucking with time alignment as you move off this axis, but this is the one that you can get closest to fixed, always. If you're doing something where you're moving up and down while listening... I suspect the listening isn't your top priority at that moment... :B

WMTMW does balance the lobing issues, and it means that the closest to "balanced" distance is in fact at tweeter height, whether that's listening height or not...

[Jed]

You should measure at listening height, always. .

For final system measurements, yes; but not for minimum phase measurements of mids and woofers (in the same system). There's more than one way to get good results.

[Jed]

By the way is anyone working on a single point source system and why do line arrays sound so good? I'd imagine something like the long planar systems would have a lot of phase related cancellations.

[cjd]

For final system measurements, yes; but not for minimum phase measurements of mids and woofers (in the same system). There's more than one way to get good results.

Yes, of course - if you're using a fixed mic position for all measurements, that fixed position should be listening height. I wasn't trying to suggest that was the only way to get good measurements. I was simply disagreeing with syncroniq - in fact, in his illustrations I'd say both were being measured at the wrong spot. I think we often assume the tweeter will be at ear height when we talk about taking measurements aligned with the tweeter.

[JonMarsh]

By the way is anyone working on a single point source system and why do line arrays sound so good? I'd imagine something like the long planar systems would have a lot of phase related cancellations.

Nearfield they definitely do- the situation becomes better the farther you're away from them, which is why they don't follow the 6 db law for drop off with distance- it's also why mating them with a sub or point source of any kind (such as a super tweeter) is so problematical- you can only get it right for one listening distance.

I think folks are attracted to the dynamics and sound stage height that are possible. Having heard some of Jim Griffin's designs built by others, I'm not completely enamored of them myself, but I expect it's one of those things that under certain conditions they can be a lot of fun. Like my old X1-SLAMM clones- they had dynamics and SPL out the wazoo, but lacked a bit in refinement. The danger in the dynamics and SPL capability is that you turn up the system and because you don't get the usual system distortion cues, you wind up listening at higher SPL's than are really good for your ears long term. But it can be fun for a while- like when ThomasW I finished the first one and listened just to Aerial Boundaries over a Benchmark DAC1 straight into an Aragon 8008 into the speakers. Good party speakers, in an audiophile kind of way.



As the Russians say about military matters, quantity has a certain quality of it's own.

[Dennis H]

You should measure at listening height, always.

Only if you measure with the mic farther away than the typical 1m. It's pretty easy to introduce phase errors as big as 40 degrees at the XO frequency by measuring at 1m and listening at 4m.

[cjd]

Only if you measure with the mic farther away than the typical 1m. It's pretty easy to introduce phase errors as big as 40 degrees at the XO frequency by measuring at 1m and listening at 4m.

True.

It's so difficult to sum something up relative to a specific discussion where folks can't quote you and have you look like a dolt. (in this case, me being the dolt). ANY time you measure from a single point you have this issue.

[AJINFLA]

The issue of "time alignment" is 3, not 2 dimensional (as drawn above). The "listener position" cannot be a "point" like a microphone hole tip, as the head location occupies a larger spatial volume, as do the 2 pinnae and without a vice, is not stationary (fixed).

By the way is anyone working on a single point source system and why do line arrays sound so good? I'd imagine something like the long planar systems would have a lot of phase related cancellations.

Define "single" point source system . I usually wear those on my head. Perhaps Dr Griffin can comment, as he has heard both my quasi point sources and his line arrays, whereas the opposite is not true. The line arrays I have heard (at least a dozen) left me less than impressed, with the exception of the spatial distortion causing large scale symphonic works to sound more realistic. Smaller scale or single instruments fared worse.

cheers,

AJ

[Dennis H]

True.

It's so difficult to sum something up relative to a specific discussion where folks can't quote you and have you look like a dolt. (in this case, me being the dolt). ANY time you measure from a single point you have this issue.

Sorry Chris, I wasn't trying to make you look dumb. I know you know all that and a whole lot more. Just trying to fill in a bit of 'the rest of the story' that you might take for granted and others might not have thought of.

[cjd]

Sorry Chris, I wasn't trying to make you look dumb. I know you know all that and a whole lot more. Just trying to fill in a bit of 'the rest of the story' that you might take for granted and others might not have thought of.

I was just laughing at the piecemeal filling-in-of-the-story

[rsbonini]

I'd like to replicate this modifed version of the ZDT3.5. I have two questions though:

1. The ND20 and DQ25 are different sizes, which will result in different driver-driver, and driver-baffle spacing. Perhaps I have missed it, but what are the new driver locations on the baffle with the DQ25 tweeter?

2. I plan to build both a pair of towers and a matching centre channel. JonMarsh seems to indicate that the modified crossover may be acceptacle for both configurations (post #10). Is this the case? If not, what type of changes might be necessary for a centre channel version?

Any other input would be appreciated.

[JonMarsh]

Key issues for the baffle are the width, because this sets where the baffle step fall off in response as you go below 1kHz occurs. So, you must keep the baffle width the same, or art least within 5-8%.

I'll review the issues behind your questions about the part substitution and get back to you tomorrow- a ton of stuff going down at the day job today, it's almost 8 PM here in CA, and I'm texting colleagues on the East coast.... that's how important.

[rsbonini]

First of all, thanks for taking the time to reply, much appreciated! FYI, my initial plan was to build the Modula NeoD CC, but alas the tweeter for that is NLA...

I’m an EE (mixed signal / instrumentation gear mostly) and build amplifiers/electronics, but this is my first attempt at speaker building so please bear with me.

After reading Zaph’s ZDT page, and the three threads that fjhuerta has on this project, the baffle width being an issue is definitely something I picked up on. Since fjhuerta doesn’t appear to give (decipherable) dimensions, my idea is to use the baffle width (the entire cabinet design in fact) and round-overs specified in Zaph’s original design. The drivers would all be shifted downward the ½” or so to accommodate the DQ25 tweeter. In comparing the crossovers and design notes I *think* Zaph’s original woofer circuit/values could then be used. Is this correct?

The complication arises with the midrange section which fjhuerta seems to have modified partially in response to his baffle changes, and partially in response to tweeter changes. My plan there was to build the speakers with fjhuerta’s crossover, measure the responses, and then attempt to adjust the crossover using one of the freeware design/sim options.

Zaph’s changes between the ZDT3.5 and ZD3C amounts to reducing the midrange and tweeter pad resistor values, and reducing the woofer circuit to a single LR2 low-pass without a .5-way section and with a slight adjustment in Fc (from what I can tell). The plan for the centre channel crossover is basically to copy this approach with fjhuerta's ZDT3.5 crossover, and then try to measure and adjust as above.

The hope with the foregoing plan is to only require value changes in the crossover compoents to achieve same/similar results to those published by fjhuerta. Please feel free to point out my naiveté as appropriate.

Again, thanks for the input and for confirming the baffle step issue.

[Juhazi]

rsbonini, you seem to be well on the track! Minor changes is driver placement aren't a problem, but naturally all interference peaks and dips will shift a little. Perhaps you should also study the possibility to set all drivers along the midline and use more generous chamfering in Avalon style, which is Jon's favourite. The center version has different baffle and positioning, those issues were handled with said changes in xo.

Zaph's pages are very informative, have you read his design mantras too? http://www.zaphaudio.com/mantras.html

[rsbonini]

I've, made some progress on my version of this modified design here: http://www.htguide.com/forum/showthr...-Tweeter-Build