Can you smell what Brando's cooking!?

[augerpro]

This is a somewhat budget design and $250 driver doesn't fit. This winter I will probably do a version with top of the line drivers, woofer will be a TD12M but the CD is undecided. Possibly the CP385ND, 4552ND, or the rumored new Beyma. Until then I have no need for all these drivers so they have to go to recoup some of the money I've blown.

[looneybomber]

TD12m I like seeing any build with the lambda woofers.

Speaking of which, if you decided to do a 3-way with TD15h's, let me know. I need to put these woofers to use

[CllessuR]

This is a somewhat budget design and $250 driver doesn't fit. This winter I will probably do a version with top of the line drivers, woofer will be a TD12M but the CD is undecided. Possibly the CP385ND, 4552ND, or the rumored new Beyma. Until then I have no need for all these drivers so they have to go to recoup some of the money I've blown.

I would also be very interested in this high end version as well!

CllessuR

[augerpro]

Yeah well, as behind schedule as I already am on these two projects I wouldn't hold your breath

[CllessuR]

Yeah well, as behind schedule as I already am on these two projects I wouldn't hold your breath

I've only acquired parts, so I'm glad there will be an opportunity for project money to collect before then. Those are two drivers that really have my interest, am glad to see there is thought on using them.

CllessuR

[augerpro]

Well I'm working the crossover now and hit a snag on the No Quarter. The DE160 acoustic center is back much farther than I was hoping compared to the woofer, even with the LP in place. At any given distance the DE160 phase is wrapping very fast so it and the woofer share a small area where they are in phase - with textbook slopes. Since I can now see far off axis at a 10 degree resolution I may be able to get creative with assymetric slopes. There are some other tricks I know but they usually center around lessening the woofer's phase wrap for the same rolloff (damping the shunt cap) or ladder delays for the tweeter. Trying to find some similar tricks for the DE160 but the difference is pretty large.

Outside of that I'm investigating mounting the driver to the back of the baffle, but that will require a new test box as I have no removable panel to do it and it won't fit through the horn cutout. I will be using a DCX2496 for prototyping so I can adjust the delay, but this project will have a passive crossover in the final version so I have to figure something out.

For those that are already building this I would make provisions for the possible rear mounting of the woofer. Or wait a little bit until I figure out what I'm going to do.

Oh yeah, I scratched the front of my pretty baffle moving a saw stand around too...

The BBV2 project is promising though. Not real far on the crossovers, but everything looks like it will work out well.

[Amphiprion]

You could always add a DSP with a delay element on the woofer. It's just a PCB, some soldering, and a little programming

[CllessuR]

Well I'm working the crossover now and hit a snag on the No Quarter. The DE160 acoustic center is back much farther than I was hoping compared to the woofer, even with the LP in place. At any given distance the DE160 phase is wrapping very fast so it and the woofer share a small area where they are in phase - with textbook slopes. Since I can now see far off axis at a 10 degree resolution I may be able to get creative with assymetric slopes. There are some other tricks I know but they usually center around lessening the woofer's phase wrap for the same rolloff (damping the shunt cap) or ladder delays for the tweeter. Trying to find some similar tricks for the DE160 but the difference is pretty large.

Outside of that I'm investigating mounting the driver to the back of the baffle, but that will require a new test box as I have no removable panel to do it and it won't fit through the horn cutout. I will be using a DCX2496 for prototyping so I can adjust the delay, but this project will have a passive crossover in the final version so I have to figure something out.

For those that are already building this I would make provisions for the possible rear mounting of the woofer. Or wait a little bit until I figure out what I'm going to do.

Oh yeah, I scratched the front of my pretty baffle moving a saw stand around too...

The BBV2 project is promising though. Not real far on the crossovers, but everything looks like it will work out well.

What method was used in calculating the accoustic center of the DE 160 driver? I've often wondered how this was done. ;x(

CllessuR

[Saurav]

The DE160 acoustic center is back much farther than I was hoping compared to the woofer

Déjà vu I'm sure you know far more tricks to deal with this than me, so I hope you figure it out, because I'm looking forward to learning how to do this.

[Dennis H]

Well that sucks. How far do you figure the AC of the tweeter is behind the woofer?

As I recall, a 3rd order with reversed polarity has the null above the listening axis and phase wrap like a 2nd order. Your AC offset might be a good thing with that kind of alignment moving the null up well above ear level. Just thinking out loud....

[mazurek]

I once did an active crossover where the tweeter required perhaps 160microseconds of delay, but rather than delay the tweeter, I delayed the woofer so that its delay time corresponded to 180degrees out of phase from the ideal target at the crossover point and then flipped the polarity. I know that is heresy to some people, but its easy enough to implement on the DCX 2496 that you can try it for yourself rather quickly and A/B test to see whether you can tell for your implementation.

Another trick I've used in the active world is to take more of the natural acoustic filter into account. For example, say your tweeter has an Fs of 500Hz, rather than just highpassing at 2000Hz, and ignoring the natural second order rolloff at 500Hz, you can use a shelving filter that incorporates the existing rolloff and moves it to a new location(same idea as Linkwitz Transform). The shelving filter will have marginally less group delay than the filter that ignores the natural rolloff. Your software process may already be taking this into account, but it can be surprising how relatively minor looking differences in the frequency domain can add up to meaningful phase differences.

[CllessuR]

With Altec, the manuals showed where the AC (allegedly) was physically, on the driver. They gave a measurement as to where on the top of the box equated to the woofers AC. this was with the A7 box.

I suppose there's something a little more scientific, and if there is it would be good to know.

For now I'm dealing with it the same way Zilch did:

[Dennis H]

Okay here's one of my fake LspCAD polar maps. Design axis is the center of the tweeter. Tweeter AC is 2" behind the woofer. BW3 XO at 1300 with the tweeter inverted. The AC offset centers the lobe pretty nicely.

[augerpro]

Here is the phase for the woofer (green) and tweeter (red) with LR4 @ 1300hz filters in place:



In most cases I could juggle slopes and polarities to get the end textbook behavior I want acoustically, even if the means is not exactly textbook (as mazurek alluded to), but this is just a lot of seperation causing very fast phase wrap for the tweeter relative to the woofer. Yeah I can get LR4 agreement for about half of the crossover region but it goes out phase very quickly at the top end of the band.

Dennis I added some delay via Hilbert Bode Transform to the woofer and 12cm looks pretty good

I'm going to recheck the speaker stand setup, it had a slightest bit of tilt backwards. I thought I had it pretty straightened out, but maybe it was still there and added to the tweeter time of flight.

Mark any progress on those filters? I might need them more than I thought so I can use the DCX to do some troubleshooting of this problem :roll:

[---k---]

Brandon,
I like your idea of mounting the woofer on back side of the baffle. That was my first thought. I bet you could salvage the box. The horn hole looks plenty big to fit the woofer through. Fill the recess in with bondo. do a cool roundover. Mount the woofer behind, and your set. Will look even more fit, polished, and sexier. Though, I don't know if moving it back 1" is enough.

[Dennis H]

Yikes, 12 cm !!! That's like (doing math in head) .3 msec. Are you seeing that much difference in the impulse responses?

[augerpro]

You could always add a DSP with a delay element on the woofer. It's just a PCB, some soldering, and a little programming

This may just end up fully active as the only solution :evil: Man if you could help me at all with those Mathcad sheets I would VERY MUCH appreciate it.

Yikes, 12 cm !!! That's like (doing math in head) .3 msec. Are you seeing that much difference in the impulse responses?

Yep...as soon as I saw it when I was doing the measurements I thought "that's gonna be tough to deal with."

Ryan> it doesn't fit through the horn cutout That was the first thing I checked.

[augerpro]

Wonder if ladder delays are advisable on a woofer? More so than tweeter I would think from an audibility POV even though it's tweeters that they are usually used.

[CraigJ]

Sorry Brandon, would two parallel notches in the woofer cutout allow the woofer to be installed inside the cabinet (assuming the magnet fits through)? Is it worth removing the baffle with four passes through your table saw? 8O

Again, thank you for the excellent work.

Craig

[Dennis H]

Wonder if ladder delays are advisable on a woofer? More so than tweeter I would think from an audibility POV even though it's tweeters that they are usually used.

Daryl over at the PE board did a series of sims showing that ladder delays aren't worth it. You end up with the same phase response going steeper with the XO.

[Amphiprion]

Brandon,

I have been looking for a project to do in my spare time. If you want to design an active XO I'd be totally happy doing the schematic/layout in Eagle and getting boards ordered. Just let me know what sort of solution you have in mind (active analog, digital, whatever). FWIW this is sort of what I do for a living, just sent out a 6-layer 12" x 4" board to be made this morning in fact. I use Altium at work but do all my hobby stuff at home with little old Eagle.

[augerpro]

Thanks for the offer Mark. At first I would prefer to use an off the shelf active crossover. So at this point I'm really looking to have the ability to convert the transfer functions into a set of equivalent filters you can input into a digital XO like the DCX2496 or Hypex amps, like you had first mentioned. Once you have the transfer functions you can of course just build your own active XO, but I don't think that is a realistic option for most diy'ers. Have you looked at the transfer functions I sent and determined how to build the worksheets? This may be an opportunity to bring something new to the diy community, since there is an interest in active crossovers, but not really any published designs.

Dennis is there anything other info you can pass along about ladder delay's? I've modeled them twice previously in projects to see if they could help, but never actually used them. So I'm decent at tuning them, but don't really know much about them, other then they seem to eat alot of sensitivity once they are EQ'ed to whatever slopes you are doing.

[Saurav]

Have you looked at pre-existing XO boards? Linkwitz sells a few that contain lots of building blocks, you control the slopes/frequencies by picking the R/C values. You'll get a PCB that can do delays, shelving, notches, and straight LR4 HP/LP. Bob Ellis had an active filter PCB group buy that was similar, I used a pair of boards designed by Michael Price that were also similar.

Not to negate what Mark is offering, but if I understand you right, you're trying to keep the cost/complexity down for someone who might want to build this. Using already-available boards might be one option (you'd still need Mark's help to design the circuits and spec the component values, he just wouldn't be designing a new board). Also, I believe there are PCB prototyping shops that will build small batches of PCBs for not too much money, ao mini-group buys may also work. Mark would know much more about that than me, I've only looked into that very briefly.

[Dennis H]

Brandon, I don't remember much about the thread. I remember doing a sim in LspCAD to check Daryl's claims and they checked out but that was with an allpass on the tweeter and a relatively small AC difference. Nothing like what you're trying to do here.

[augerpro]

Not to negate what Mark is offering, but if I understand you right, you're trying to keep the cost/complexity down for someone who might want to build this.

Several people already are building it, so I'm trying to come up with solution for them too. I think building an actual active crossover from the board up is a cool idea, and I'm not saying I don't want to do that. I think we should if Mark is interested. I'm just saying that few people will want to build an active crossover, and since there was a lot of push to have an active option (using DCX or similar) for this project I'd like to turn this problem into an advantage and focus on how to incorporate an off-the-shelf active crossover into this design as the first priority.

Second priority is solving it passively, and I think rear-mounting and/or a baffle rework will get us there.

That said active is something I'd like to experiment with in the future, but electronics not being my strong suit I'd like to find a way to convert transfer functions into an equivalent set of filters that can be used with these off the shelf crossovers. Mark had mentioned being able to build a worksheet to do this, and I'd certainly be happy to compensate Mark for his time in helping me develop such a worksheet.

[Saurav]

I guess if the people building it and interested in an active solution already have a DCX. Otherwise, for me, soldering opamps and resistors to a PCB is a lot easier than soldering / assembling a passive crossover. With good instructions, it's pretty much connect-the-dots. And would be cheaper than a DCX I think. Designing PCBs is hard, but I view a pre-designed PCB as pretty much an off-the-shelf crossover. But that's just me, I'm more comfortable with the electronics than the woodworking (i.e. I suck less at the electronics than the woodworking).

So, yeah, I see what you're saying, and if the people who want to go active already have a DCX, or are more comfortable with that approach, then that's probably where you should start.

[Paul W]

Brandon,
Very interested in how you solve this problem. I'll be facing the same issue in a month or so, but with even more offset . I hadn't run sims on it but, in my mind, I had it solved using ladder delay on the LF drivers...oops!

I planned on starting out active, but wanted to switch to passive to be rid of an extra ADA conversion.
Paul

[augerpro]

Well Paul I think as it's such a large physical problem, a physical solution is going to have to play a large part. I knew this would be an issue, but I was surprised at the magnitude. So, beware.

Another reason I'd like incorporate the DCX "properly" is it would help me troubleshoot and prototype this sort of thing on the fly. I've had one for 1 1/2 years and it's still sitting in the box. I'm not sure how many filters it can cascade per channel though. Like I said I think this problem is just presenting an opportunity to do something different and new so I'm half frustrated and half excited

[Dennis H]

Do I remember correctly that SE can model a DCX and then upload the filters to the unit? If so, you don't really need any math from Mark.

[augerpro]

I believe that function has been broke since a firmware upgrade some time ago.

[CraigJ]

I've had one for 1 1/2 years and it's still sitting in the box. I'm not sure how many filters it can cascade per channel though. Like I said I think this problem is just presenting an opportunity to do something different and new so I'm half frustrated and half excited

Well, don't be afraid of it, open the box up and get that baby hooked up. If needed, I'd be more than happy to walk you through it over the phone. I'm by no means an expert, but I can get sound out of it. :W

Craig

[Dennis H]

Firmware in the DCX? If so, you could still model it in SE and set the filters manually. That's how you have to do it in LspCAD.

One caveat, the shelving filters are wrong in LspCAD's DCX model. Don't know about SE. LspCAD is doing the 'standard' shelving filter where you set the center frequency but in the DCX the frequency is the 3dB point.

[Paul W]

You'll find the DCX to be a great little box. I've been using them on and off for several years and, other than reliability issues with a couple of early units, no issues. I've probably never used more than 3 PEQ filters for any driver, but never run short on DSP horsepower.

Dennis...yes, SE will upload to V 1.14 DCX, but that old DCX FW is buggy. I haven't actually tried uploading SE transfer functions to newer DCX FW, but even the latest V16 SE still references only 1.14 DCX...of course V16 still requires XP

[Dennis H]

Paul, I don't think firmware or SE versions will matter if you're setting the filters manually, right? Does SE get the DCX shelving filters right -- frequency is the 3dB point?

[Amphiprion]

Brandon,

I totally missed that I had a PM in my inbox, my bad! I just sent you an email. How does the DCX2496 take the input parameters? Is it a series of biquad filters, one big polynomial equation, ? I don't own one so I don't know how they do things.

Another option, that removes the finicky nature of the Matlab library, is to design an active crossover (analog) using op-amps and R's and L's and C's (yes I said L's. Even though you would never design a filter this way that you intended to build, it makes the math easier later). Then I can just take the Laplace transform of your circuit (laplace of an L is a lot easier than the laplace of the simulated L that uses an opamp and RC) and with a bilinear transform on pencil and paper and get a darn near identical match in the discrete time domain. That probably didn't come across all that clear, let me see if I can find a link. I did a lot of work on this stuff in college but nothing came of it.

[Amphiprion]

Here, this should better explain the active filter straight to DSP coefficients method. Sorry about the quality of the scans, I did these a long time ago in college and the scanner in the lab was junk (my student dollars at work).

Basic idea is to find the continuous time transfer function for a circuit block. This works best if you limit yourself to 1st and 2nd order blocks, and build a system out of the cascaded transfer functions. So I take the Laplace transform of the circuit, and then do a substitution for s with an expression in z to go from continuous time H(s) to discrete H(z). Then a bunch of simplifying to get it into an expression in terms of (crap) + (crap2)z^-1 + (crap3)z^-2 for the numerator and denominator, where (crapX) is the coefficient for a discrete biquad that matches the analog active filter. It does deviate a little bit at the filter extremes, but it's a pretty good way to go from continuous to discrete. Oh, and the R's and L's and C's along with the sampling period (1/Fs) are what determine the actual polynomial coefficients.

Hope this makes sense. I have more like these images, I actually did this for every active filter type that was present in LspCAD 5. So if you want those I can post them on my web page.

[Paul W]

Dennis,
Agree, manual settings shouldn't care about SW/FW versions. It has been, literally, years since I looked at the SE interface for the DCX and I don't remember much about it. (Last month, I used SE for the first time since 2006 for surround driver measurements...which I promptly dumped into LspCAD :W )
Paul

[augerpro]

Mark I decided to RTFM again and in chapter 9 I *think* I've found a solution - and I *may* not need you to do a bunch of work needlessly. One feature of the digital filters in SE uses "F-components" when modeling for the DCX. I've only ever used the active and passive modeling and kind of forgot about this. But it appears this is how you generate the transfer function for digital filters for upload into a DCX. These can be used by the optimizer too, they are not just generic BW3, LR4, etc. filters. So I think I can just model using these F-components just as I would a normal passive, and then input the resulting filters into the DCX manually. Pretty much what you had proposed doing with the Mathlab worksheet.

Paul since you've used this before, is my understanding correct?

Give me through this weekend to figure out if this is going to be the solution or not. BTW I can still send those transfer functions if you want to play around anyway?

[Paul W]

I honestly don't remember.

[Robh3606]

but this is just a lot of seperation causing very fast phase wrap for the tweeter relative to the woofer.

Hello Augerpro

That phase measurement looks like the Excess Phase from the Measurement is still in it. What happens if you look at the Minimum phase with the Excess Phase from the flight distance removed. Can't you just measure the actual Imp and Phase of the compression driver horn combo on the baffle? You have to measure the woofer Impedance in the box so why not the horn as well.

Rob

[augerpro]

Hi Rob. Welcome to htguide! These are the actual measurements on the actual baffle. Both drivers are measured from the same mic position and with the same start for the FFT gate, so the relative phase is correct (at the mic position), which is what is important. You can't pick different gates or the relative phase will not be accurate. For example I could put one driver on one wall and one on the floor and if I started the gate at the beginning of the impulse for each (i.e. minimum phase) I would model a crossover where the drivers appear to be in alignment but in real life clearly are not. I don't if that was a good example or not, hope it was clear.

Now the design axis is whatever you want it to be, I thought I picked the point where the horn and woofer and meet. I have to put it back on the stand tomorrow and make sure the stand was not tilting and the mic wasn't too low (it likes to slide down on its own sometimes). Both of those could cause several extra cm of delay to be added to the CD. While you don't need to be in perfect time alignment to be in phase alignment, you can't be too different either or the phase is just wrapping too much on one driver to be in agreement with the other for long.

You dig?

EDIT: FYI the gate was started just before the woofer impulse since it was closest, 195cm or so. So the majority of the time of flight is removed (or else the screen would be a visual mess, although the model would still be fine) but as the impulse of the CD shows up later you still have some flight time included in its phase. But it has to be there or it will no longer be accurate relative to the woofer.

[augerpro]

BTW this digital filter modeler in SE is kinda cool. At least the portion meant to be used with a DCX. Already have something hammered out and ready to input into the DCX. The only thing I don't like is it doesn't appear to be able to change filter Q, it just uses generic filters. But I suppose this is because the DCX is the same.

BBV2 are all looking good, especially the fullrange which is surprising. The MTM has pretty good ripple in teh treble, but MTM's tend do this anyway, especially ones this small. The great thing is the response smooths right out off axis and is beautifully flat with a nice tilt out to 50 degrees! No dip/bump so typical of say a 6" woofer mated to a 1" tweeter. Having 0-50 degree measurements is proving very valuable for the MTM since I can see what is going on off axis when it is on its side as it would be as a center channel. First draft is using BW3, but you can wire the tweeter either way, but now that I see the difference on the off axis response (BIG difference) this helps me select which polarity I want to use.

Loving this lazy susan stand. Still working out the kinks, but I can't believe I went this long without it.

[Robh3606]

Hello Brandon

Thanks for the Welcome. I understand about the alignment and baffle measurements. I don't understand the concern about the wrapping of the phase of the compression driver vs the woofer. All that wrapping is a measurement artifact. It's not the true driver phase. You would have to remove the measurement flight time component, excess phase from the measurement to really see what you have. If you look at the impulse responses you can overlay them and measure what the actual delay is.

If I understand this correctly you are concerned the the wrapping doesn't line up long enough through the crossover region?? Is that correct?? All I am saying is the wrapping isn't real.

I use CLIO and here is a measurement I did of a tweeter. What you see is the wrapped Phase Response with the measurement time added Excess Phase Blue, the Minimum Phase Purple which is calculated and removes the Excess Phase/Measurement time and the on axis frequency response Green.

They look really nice! I bet the will sound great once you get them going.

Rob

[augerpro]

Phase is all relative. Relative between drivers, and relative to the listening position. You can't pick a point to measure at then use different starts for the gate. Is that what you are suggesting? If I did that you could play with around with gates and make the .3ms delay disappear. But it's real. Just like doing a design with LR2 slopes requires real effort in physically aligning the drivers. If we did as you suggest and took out any time of flight (including from the driver center to the baffle surface) everyone would be making wonderful looking LR2 designs. That would all measure horribly at any REAL listening position. REAL delay will cause a REAL problem. Otherwise nobody would ever have to bother with aligning drivers, and there would be no need for all those active pro monitors that use delay to align the driver responses.

[Dennis H]

BTW this digital filter modeler in SE is kinda cool. At least the portion meant to be used with a DCX. Already have something hammered out and ready to input into the DCX. The only thing I don't like is it doesn't appear to be able to change filter Q, it just uses generic filters. But I suppose this is because the DCX is the same.

Cool Brandon! You can use a parametric filter to change the shape (Q) of the highpass or lowpass shoulder but you already knew that. Do you have the Behringer software? That's the easiest way to input your settings and you can save a file that you can share with builders.