Announcement

Collapse
No announcement yet.

Budget Duelund Three-way

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #46
    Originally posted by gregrueff View Post
    Luckily I hadn't destroyed the tweeter in the other cabinet so crossover prototyping began today!

    First the woofer.
    A nearfield mic was able verify the target slope.

    Click image for larger version  Name:	3,100,3,.165.jpg Views:	0 Size:	109.4 KB ID:	1463753

    I have real issues with the crossover values in my models versus what ends up measuring appropriately. This isn't the first time either. It always manages to humble me.
    My model called for a 4.0 mH coil with a "tank" bottomless notch using a 0.17 uF cap. On the shunt circuit my model called for a 80 uF cap with a 3 ohm resistor to relax the rolloff a little and adjust the phase for good integration with the midrange.

    The slope above ends up using a 3.0 mH coil with a 0.17 uF "tank" value, and a 100 uF cap with a 3 ohm resistor.

    All in all though, the Duelund slopes are achievable with simple second order electrical filters. A Duelund slope really only deviates from a LR2 pretty far down the rolloff where it starts to approach LR4.

    I did play around with the "tank" cap value to see how critical it is to use 0.17 uF, because I couldn't find a single cap with a value close enough.
    Unfortunately it does make a difference, because to achieve 0.17 uF (0.165 uF), two (2) 0.33 uF caps need to be wired in series.
    The comparison below shows that both 0.1 uF and 0.2 uF attenuate the woofer breakup around 5 dB less than nailing it with the right value.

    Click image for larger version  Name:	comparison of bottomless notch.jpg Views:	0 Size:	131.6 KB ID:	1463754

    I have not conducted any sort of listening test to determine if it matters or not. It's entirely possible it's inaudible. Using a 0.1 uF value or 0.2 uF value also affects the rest of the crossover negligibly- it does not affect the phase at all so integration with the rest of the system is unaffected.
    Three-way x-overs can be a little tricky. I measure close mic on the woofer, but also from 3' or so. I don't think I could describe precisely how I arrive at my filter. That said, the 100uf cap and 3 ohm resistor might cause the impedance to be very low, and the resistor might tend to get hot. Will this speaker have any BSC?

    Comment


    • #47
      Originally posted by dlr View Post
      That grill is positioned proud of the front baffle? There's a gap under it, that is?

      dlr
      I would say the oak "frame" of the cabinet is 1-2 mm proud, with the grille recessed just a touch.
      It is possible the grille does not sit completely flush on the baffle and there also exists a small gap between them. The knobby attachment things I'm using to affix the grille snap all the way shut, but the receiver on the baffle probably sits on top of the baffle by a millimeter or two.

      I've read over your experiments with felt many times over the years.
      I might be able to stuff some between the baffle and the grille. Do you think that would help?

      Cheers,

      Comment


      • #48
        Originally posted by rpb View Post

        Three-way x-overs can be a little tricky. I measure close mic on the woofer, but also from 3' or so. I don't think I could describe precisely how I arrive at my filter. That said, the 100uf cap and 3 ohm resistor might cause the impedance to be very low, and the resistor might tend to get hot. Will this speaker have any BSC?
        My model is predicting that the impedance stays nice and healthy in the bass until the 4 ohm midrange impedance starts to take over. Refer to the first post for the prediction. I had really wanted to keep this an 8 ohm nominal design, but that doesn't look possible with this midrange selection.

        This basically has full BSC, but it's more complicated than that. The lowpass filter applies around a 1-2 dB bump just because the filter is so close to the driver resonance.
        I did a lot of modeling with the Jeff B. boundary simulator, and even with "bookshelf" placement, boundary reinforcement doesn't really kick in appreciably until under 100 Hz. No matter where I placed them in the model, I still lost around 6 dB from 1 kHz to 100 Hz. This lead me to believe that "full BSC" was still called for given my potential room placements.

        I also have a rather unfounded or unverified belief that BSC is more applicable to a 2-way design where intermodulation distortion is more of a factor. One of the main reasons I wanted to pursue a 3-way was to get the bass out of the midrange driver. I haven't brought it up, but a large part of this whole effort is my belief that intermodulation distortion is a bigger factor than people realize.

        So I've also been meaning to start a whole other thread to get to the bottom of this whole power rating / design voltage requirements for crossovers as it relates to the passband.

        Can an E.E. please explain to my lowly M.E. self how the power through the crossover varies with frequency? It's been a while since I've brushed up on my Kirchoff...

        Comment


        • #49
          Originally posted by gregrueff View Post
          I would say the oak "frame" of the cabinet is 1-2 mm proud, with the grille recessed just a touch.
          It is possible the grille does not sit completely flush on the baffle and there also exists a small gap between them. The knobby attachment things I'm using to affix the grille snap all the way shut, but the receiver on the baffle probably sits on top of the baffle by a millimeter or two.
          Poor wording in my question, but yes, I was wondering if the grille is flush or not. Your test with the cardboard cutouts looks to be flush. A gap like that might by part of the problem. Easy way to tell, fill the opening with anything. I'm thinking that if it's part of the problem, it will be the part above the tweeter because the that open area and the box frame could create a resonant chamber underneath the grille. Something to test fairly easily. Cotton will do, too, cotton balls stretched some to fluff it a bit first.

          Also, all of the layers of cardboard are rather porous, stepped and not very reflective The cardboard may be channeling the wave. It might be performing better than solid wood of the baffle, despite it being shaped. It is primarily reflective. Just more random thoughts.

          I've read over your experiments with felt many times over the years.
          I might be able to stuff some between the baffle and the grille. Do you think that would help?
          Test it by stuffing it as I suggested. If it is acting as a chamber, just a small amount of felt under it close to the cutout edge would probably suffice.

          dlr
          WinPCD - Windows .NET Passive Crossover Designer

          Dave's Speaker Pages

          Comment


          • #50
            Originally posted by gregrueff View Post

            My model is predicting that the impedance stays nice and healthy in the bass until the 4 ohm midrange impedance starts to take over. Refer to the first post for the prediction. I had really wanted to keep this an 8 ohm nominal design, but that doesn't look possible with this midrange selection.

            This basically has full BSC, but it's more complicated than that. The lowpass filter applies around a 1-2 dB bump just because the filter is so close to the driver resonance.
            I did a lot of modeling with the Jeff B. boundary simulator, and even with "bookshelf" placement, boundary reinforcement doesn't really kick in appreciably until under 100 Hz. No matter where I placed them in the model, I still lost around 6 dB from 1 kHz to 100 Hz. This lead me to believe that "full BSC" was still called for given my potential room placements.

            I also have a rather unfounded or unverified belief that BSC is more applicable to a 2-way design where intermodulation distortion is more of a factor. One of the main reasons I wanted to pursue a 3-way was to get the bass out of the midrange driver. I haven't brought it up, but a large part of this whole effort is my belief that intermodulation distortion is a bigger factor than people realize.

            So I've also been meaning to start a whole other thread to get to the bottom of this whole power rating / design voltage requirements for crossovers as it relates to the passband.

            Can an E.E. please explain to my lowly M.E. self how the power through the crossover varies with frequency? It's been a while since I've brushed up on my Kirchoff...
            I'm not an EE, but see if this sounds logical. Imagine the voltage to the speaker is 16 volts. Look at your impedance. (Pick a frequency.) If the impedance is 4 ohms, the current will be 4 amps. If the impedance is 8 ohms, the current is only 2 amps. Power is equal to current squared times resistance. 64w vs 32w.

            Comment


            • #51
              Okay, I think I get what you're saying.
              If I look at the woofer's impedance network below, it does in fact dip to around 2.8 ohms after 2 kHz.

              Click image for larger version  Name:	Woofer Impedance.png Views:	0 Size:	24.7 KB ID:	1463825

              So that should mean over 2 kHz the network is seeing 1.78 Amps given a 5V input signal, right?

              So even though the speaker's total impedance never dips that low, it will in the woofer only circuit?

              Click image for larger version  Name:	Total Impedance.png Views:	0 Size:	21.3 KB ID:	1463826

              So is it correct to say this is why the capacitor, inductor, and resistor in the woofer network needs to be able to handle the current /voltage resulting from the network's overall impedance curve from 20 Hz to 20 kHz (or whatever frequency content is in the source signal)?

              Doesn't this mean that a tweeter network's components could actually have a more robust design criteria than the woofer if it has an impedance that drops even lower than the woofer's?


              Comment


              • #52
                I'm getting lost in the details, so not sure. However, the coil in the woofer circuit increases the impedance as the frequency increases.

                I thought the 100uf would be too large for the situation. I guess it all depends on the transfer function needed. To more or less flatten the woofers rising impedance, try something like 8uf, and 8 ohms. Then for the filter, use a coil and cap suitable for an 8 ohm load.

                There's a lot of ways to get the response you need. The low frequency hump can be annoying to deal with. My attempts a 3-ways usually end up with a large xo part count.

                Comment


                • #53
                  Well I tried your suggestion out, but an 8 uF cap just doesn't have the oompf to hit the slopes I need.

                  I did some additional research on some well known and respected designs. These high values are not unprecedented.
                  -One of Jon's first concept crossovers for the Wavecor Ardent features a 5.6 mH coil and a 78 uF cap on the woofer.
                  -Curt's Statements design uses a 7 mH coil with a 78 uF cap.
                  -Zaph's SB12.3 uses a 4 mH coil and 30 uF cap, but is not crossing over as steep.
                  -Zaph's ZDT3.5 uses a 1 mH coil for the first stage with a 62 uF cap and a separate 80 uF one in the zobel.

                  I guess on that point- yes this resembles a typical zobel, but it's more acting as a response shaping leg than an impedance flattening circuit.
                  But also to your point, Zaph doubles up both the capacitor and the resistor on the ZDT3.5, and Jon uses five (5) resistors in parallel in his design.

                  Finally, I must point out that I considered it a great achievement to hit the woofer's target transfer function using only four (4) components. Yes, they are relatively large but the network is simple. That was one of the goals of this design.

                  It's really just a simple second-order electrical with a tank across the inductor and a resistor in parallel with the capacitor to relax the slope. All the parts work perfectly to get the woofer's breakup to around 30 dB down and hit the target rolloff.

                  This is also close to about an LR2 lowpass at 450 Hz, so large values are just needed.

                  Cheers!

                  Comment


                  • #54
                    Here's what I came up with using PE data. These appear to have similar transfer function as your filter. I think these would have higher impedance, but perhaps not enough to matter.

                    Click image for larger version

Name:	xo 450 three.jpg
Views:	196
Size:	247.8 KB
ID:	1463913 Click image for larger version

Name:	xo 450 two.jpg
Views:	188
Size:	245.4 KB
ID:	1463914 Click image for larger version

Name:	xo 450.jpg
Views:	188
Size:	243.6 KB
ID:	1463915

                    Comment


                    • #55
                      I appreciate the time and effort you took to model this yourself.
                      I wasn't completely able to replicate your curves in my model which is curious. I wonder how much my as-built impedance is to blame for that.
                      But remember also, my PCD model showed a 4.0 mH coil and 80 uF cap following the curve perfectly, only for reality to be a 3.0 mH coil and 100 uF cap.

                      You have piqued my interest nonetheless and in doing so I have uncovered an error in the way I was modeling this crossover in PCD.
                      If you model it like this, the impedance drops to 2.8 ohms after 2 kHz:

                      Click image for larger version

Name:	Woofer Network Incorrect Impedance.png
Views:	149
Size:	300.0 KB
ID:	1463981

                      So while PCD is predicting the transfer function and FR curve correctly, this is an incorrect impedance.
                      You have to model it as follows in PCD for it to correctly predict the resultant impedance:

                      Click image for larger version

Name:	Woofer Network Correct Impedance.png
Views:	145
Size:	285.1 KB
ID:	1463982

                      I verified this with DATS, first measuring the woofer alone and then placing one more crossover component in the network each time I measured to ensure the impedance matched the PCD model.
                      You can see that with the whole network in place, the impedance drops to around 5 ohms, not 2.8 ohms.

                      Click image for larger version

Name:	DA215-8 impedance with various filters.png
Views:	139
Size:	45.9 KB
ID:	1463983

                      Comment


                      • #56
                        The 'comp' button in PCD is for the conjugate across the driver only, so it is not indicative of the resultant impedance- if that is what you are referring to.
                        Later,
                        Wolf
                        "Wolf, you shall now be known as "King of the Zip ties." -Pete00t
                        "Wolf and speakers equivalent to Picasso and 'Blue'" -dantheman
                        "He is a true ambassador for this forum and speaker DIY in general." -Ed Froste
                        "We're all in this together, so keep your stick on the ice!" - Red Green aka Steve Smith

                        *InDIYana event website*

                        Photobucket pages:
                        http://photobucket.com/Wolf-Speakers_and_more

                        My blog/writeups/thoughts here at PE:
                        http://techtalk.parts-express.com/blog.php?u=4102

                        Comment


                        • #57
                          Originally posted by Wolf View Post
                          The 'comp' button in PCD is for the conjugate across the driver only, so it is not indicative of the resultant impedance- if that is what you are referring to.
                          Later,
                          Wolf
                          Ha! Leave it up to my user error to completely misunderstand PCD.
                          Thanks for the tip Ben.

                          Comment


                          • #58
                            If you want to see the woofer alone, temporarily add a 999 ohm resistor in front of the mid, and tweeter, then look at the total speaker impedance.

                            Comment


                            • #59
                              I spent a lot of time today trying to diagnose the dip/peak combo around 7 kHz seen in my tweeter's response.
                              I tried applying felt wool under the grille to fill gaps, in between the tweeter and the midrange, on the grille waveguide portion, etc. All over the darn place wherever I could think.
                              No matter where I put the wool felt, I was unable to affect the anomaly.

                              So I took another set of measurements with the bare baffle, with grille attached but the cloth removed, and with the cloth in place.

                              Click image for larger version

Name:	all three3.jpg
Views:	123
Size:	126.0 KB
ID:	1464858

                              It almost looks to me like there exists an inherent 1 dB dip/peak at 6-7 kHz that is shifted in frequency and amplified with both the application of the grille and cloth.
                              With the application of the grille, it boosts to about a 2 dB anomaly, spread out a bit more and little bit higher in frequency.
                              Then the cloth furthers this trend making it a full 4 dB dip/peak.

                              I'm struggling to understand this. I'm thinking maybe this is just a product of using grille cloth.

                              I'm also wondering if this has to do with the total recessed depth of the tweeter relative to the apparent front of the speaker.
                              In OEM condition, the tweeter dome sits recessed to the faceplate with a round-over acting as a mild horn/waveguide. Applying the grille makes this more exaggerated.
                              Assuming I haven't made a silly math error, 7 kHz is close to 1.5" which is also pretty close to the chord length from the dome tip to the front facet of the speaker.

                              Comment


                              • #60
                                I get closer to 1.9"?

                                Comment

                                Working...
                                X