Announcement

Collapse
1 of 2 < >

Midwest Audio Fest

It’s that time audio enthusiasts!

Registration for the 2019 Speaker Design Competition is now open! Visit midwestaudiofest.com for details and to list your speaker project.
We are excited to see all returning participants, and look forward to meeting some new designers this year, as well!

Be sure your plans include a visit to the Parts Express Tent Sale for the lowest prices of the year, and the Audio Swap Meet where you can buy and trade with other audio fans.

We hope to see you this summer!

Vivian and Jill
2 of 2 < >

Midwest Audio Fest

It’s that time audio enthusiasts! Registration for the 2019 Speaker Design Competition is now open! Visit midwestaudiofest.com for details and to list your speaker project. We are excited to see all returning participants, and look forward to meeting some new designers this year, as well! Be sure your plans include a visit to the Parts Express Tent Sale for the lowest prices of the year, and the Audio Swap Meet where you can buy and trade with other audio fans. We hope to see you this summer! Vivian and Jill
See more
See less

Trouble on both ends of the FR spectrum

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

  • Trouble on both ends of the FR spectrum

    I'm back to a CC design after a couple weeks. Thought I had it, bought the parts, built the XO. DIdn't measure anything like the mock-up with parts out of the parts bin... Fought with it for a little while and decided to take a fresh set of measurements. Somehow, both the woofer nearfield and tweeter (in its entirety) measured (and sounded) quite different. I gave up on the first design and started from scratch - these measurements were entirely repeatable.

    Seas W18E001 - OW2/MDM55 - Seas W18E001.

    Box is 22.5x9x13 with a 1.5" front baffle. Should be 27.7 gross L (assuming I did the math correctly), call it 25L net when everything is in. Sealed.

    It needs a final tweak, but I'm having trouble getting past the extremes.

    Regarding the low end, did I miscalculate something when modeling the box? I did a nearfield measurement of 1 woofer and the 70Hz bump seems very real. My room is 16 feet wide in 1 direction, could that somehow be the room? That isn't the whole issue in the low end - its still higher than I expected in 100-200Hz region.

    Regarding the high end, the OW2 has a bit of a bump above 10K, and it's definitely showing up. I'm sure I could notch it out, but is there an easier way? I tried throwing a little .05mH coil across the padding resistor in front of the crossover, but that seemed to lift the whole response somehow, not what was simulated. I've used a cap to lift dropping top octaves, stood to reason that the opposite should work with an inductor.

    W-M is 4th order acoustic, M-T is 2nd order acoustic (yes, tweeter is wired backwards).

    I'd welcome any thoughts. Thanks!


    .
    System Response Box Model LP

  • #2
    Is the frequency response a sim, or measurement?

    I see some odd things in your sim. It may be PCD acting up.

    If your woofer filter is 3rd order, where is the second coil?

    On the tweeter, there's a blank box below C9.

    On the tweeter response. Try padding it more with a larger resistor, and then increasing the cap value to bring up the lower end of the tweeter range.

    I've found that the low end of a center channel will change when you place them near a screen, or wall.

    Comment


    • #3
      The PCD is GUI acting up (although it remains accurate, when I rebuilt the XO, both in phase and out of phase measurements were essentially exactly as simulated). The labels get out of sync with the components that are enabled. If anyone happens to have ideas on that, I'm all ears by the way. Woofer filter only has the components that are valued, its second order. Similarly, the coil in the tweeter circuit is real.

      I'll give the resistor/cap a try. The phase alignment is touchy which is why I was chasing the other end. Couldn't get it to sim right before I started building but some more focused effort is certainly warranted.

      The response is measured with Omnimic. The OOP measurement was saved as an overlay and I turned the amp up a tiny bit to offset the measurements.

      Comment


      • #4
        I'm surprised that it took a 4th order high-pass for the Morel mid. I used one years ago with crude measurements. I seem to recall that it drops rapidly below about 800hz. Does you sim show much difference with a 3rd order electrical? (I'm not suggesting it's wrong. I'm just curious.)

        Your measurement looks like it has a bit more baffle step than needed. Does it sound like you want, or could it be improved by making it sound slightly more forward? In other words, would it be an improvement to raise the mids to about 87dB? That would in effect lower the 100hz to 200hz range 2dB.

        Comment


        • #5
          I'm not quite happy with the sound yet, definitely could use a bit more midrange - I was fighting the LP-MID XO level, and that was what I was setting the level from. A little less inductor and a little more cap lets me lift that, so a new route to explore.

          The MDM55 in this box didn't drop until about 550Hz. It starts falling above 1K, then rises back up. The 2nd inductor is doing a lot to get the slope steep enough. Looking at the raw response, a notch may work, similar to what I used on the other side of the MDM55. Back to the sim...

          Comment


          • #6
            Originally posted by Adam_M View Post
            I'm not quite happy with the sound yet, definitely could use a bit more midrange - I was fighting the LP-MID XO level, and that was what I was setting the level from. A little less inductor and a little more cap lets me lift that, so a new route to explore.

            The MDM55 in this box didn't drop until about 550Hz. It starts falling above 1K, then rises back up. The 2nd inductor is doing a lot to get the slope steep enough. Looking at the raw response, a notch may work, similar to what I used on the other side of the MDM55. Back to the sim...
            Try increasing the resistor value on the existing mid notch to about 4 ohms or so. It looks like you will better match your target if you do. Might pick up a little spl at the tweeter hand-off.

            The mids dip at 4k is probably caused by the woofer cavities. I've seen that on an mtm I built a while back. It probably goes away off axis. What looks like a peak up higher, is just the normal mdm55 response. It's not a resonance peak that needs to be hammered.

            Factory graph. https://www.parts-express.com/pedocs...ifications.pdf

            There's a hundred ways to get the response you want. The harder decision is deciding what you want.

            On the mid notch. If you double the coil value, and half the cap value, you will get a narrower notch, and it may match the target a little better.

            Comment


            • #7
              You were right about the resistor in the notch, I arrived at the same place through experimentation in PCD. Makes sense about the dip at 4k - will check off axis when I take the next set of measurements.

              There was definitely more efficiency to be had - the sim has found +2dB across the board except the extremes, which is exactly what I was hoping to do. I'm also trying to keep the impedance high enough that the amp doesn't need to double as a welder - there were configurations that were ugly with fairly broad dips below 2 ohms. Its obviously going to be a 4 ohm speaker, just trying to keep it ...polite. Dropping the pad resistor on the mid to lift the response caused problems, although I'm working to bring them up. The dips are now about 2.5 ohms at 950 (very narrow) and 3500 (a bit broader)

              I'm targeting XO's below 1K and 4K - although just below is ok. The LP-BP in particular shows a lot of improvement when going from even 1050-1000- much less audible lobing.

              Thanks for all of the help so far.

              Comment


              • #8
                Try adding about 2 ohms to the 68uf cap to ground. Or, on the woofer tank filter, you could add a 1.5 ohm resistor in series with the cap. I don't think PCD will sim that though. Maybe xsim will. The impedance should come up at the frequency of the notch by 1.5 ohms. (Any resistor value is OK. 1.5 was for an example.) You can measure the response change even though you can't sim it on the tank filter.)

                You could add an ohm or two on the coils in the mid high-pass, and see if it helps. That dip in mid response near 1k is also diffraction. If you can match a lower frequency target without trying to fill the 1k dip, that might bring up the impedance. Take a look at the filter transfer functions. If there is a sharp knee in the transfer, that's going to cause a lower impedance. (I think.) I suspect that's what is happening at 950hz.

                Comment


                • #9
                  I didn't get a chance to measure last night, just sims.

                  Regarding the resistor in series with the 68uf cap - that seems to alter the shape of the knee quite a bit - I'd actually like to make the knee a bit sharper and have been toying with a smaller inductor and larger cap. It seems part of the too much BSC is too big of an inductor. I'm trying to find a balance between a dip in the low midbass and the sharpness and exact rolloff point of the knee. Basically, if I try to roll the woofer off earlier, I induce a dip around 200Hz. Next step is to try 3rd order electric, but I didn't get there yet - this is all in an effort to push the LP-BP XO a bit lower, it seems to gravitate towards about 1050-1100Hz.

                  Can you help me understand what the resistor would do in the tank circuit? I thought that an (ideal). LC wired in parallel produced infinite resistance at the resonance frequency? I of course could have that backwards and then the resistor makes perfect sense.

                  Adding resistance to the first coil in the mid HP circuit directly affects the impedance dip around 950Hz - doing exactly as you suspect, softening a very sharp knee in the filter transfer function. It necessarily messes with the FR and phase, but seems to be able to be dealt with.

                  By adjusting the BP circuit, I was able to soften both knees a bit which gave me more output on either side of the peak where the BP filters meet and allowed for a larger padding resistor on the mid. As an aside, the filter actually had positive gain at this peak where they met and a corresponding impedance dip. Making the filter gain-neutral at this point directly lifted the impedance at the peak. That seemed to pull the impedance up over 3 ohms in all parts of the BP circuit - and the phase seems benign. The tradeoff is a slight - maybe 1.5dB droop at both Fc's. Given that that droop is likely diffraction, when I build it I'll take some off axis measurements and see if it fills in. If it does, sounds like a viable circuit!

                  As a side note, I did experiment with a topology that matched both the LF and HF targets of the mid, ignoring the diffraction ripple entirely. Surprisingly, I was able to do that with a single cap and a single coil. It meant wiring the mid out of phase (and tweeter in phase) for some reason, but I couldn't get the phase in particular to line up, because (presumably) with the phase being directly derivable from the FR, the anomalies in the FR also cause corresponding phase anomalies. I'd love the idea of a BP this simple and think it could merit further exploration. Would it be best to get some slightly off axis measurements for the mid to soften the diffraction and change the design axis by just a few degrees?

                  This speaker has been quite the learning experience and a lot of fun so far. All the advice given that 3 ways are far more complex than 2 ways is absolutely true. Its not as simple as 2 2-ways developed independently in the same box...


                  Comment


                  • #10
                    Originally posted by Adam_M View Post
                    I didn't get a chance to measure last night, just sims.

                    Regarding the resistor in series with the 68uf cap - that seems to alter the shape of the knee quite a bit - I'd actually like to make the knee a bit sharper and have been toying with a smaller inductor and larger cap. It seems part of the too much BSC is too big of an inductor. I'm trying to find a balance between a dip in the low midbass and the sharpness and exact rolloff point of the knee. Basically, if I try to roll the woofer off earlier, I induce a dip around 200Hz. Next step is to try 3rd order electric, but I didn't get there yet - this is all in an effort to push the LP-BP XO a bit lower, it seems to gravitate towards about 1050-1100Hz.

                    Can you help me understand what the resistor would do in the tank circuit? I thought that an (ideal). LC wired in parallel produced infinite resistance at the resonance frequency? I of course could have that backwards and then the resistor makes perfect sense.

                    Adding resistance to the first coil in the mid HP circuit directly affects the impedance dip around 950Hz - doing exactly as you suspect, softening a very sharp knee in the filter transfer function. It necessarily messes with the FR and phase, but seems to be able to be dealt with.

                    By adjusting the BP circuit, I was able to soften both knees a bit which gave me more output on either side of the peak where the BP filters meet and allowed for a larger padding resistor on the mid. As an aside, the filter actually had positive gain at this peak where they met and a corresponding impedance dip. Making the filter gain-neutral at this point directly lifted the impedance at the peak. That seemed to pull the impedance up over 3 ohms in all parts of the BP circuit - and the phase seems benign. The tradeoff is a slight - maybe 1.5dB droop at both Fc's. Given that that droop is likely diffraction, when I build it I'll take some off axis measurements and see if it fills in. If it does, sounds like a viable circuit!

                    As a side note, I did experiment with a topology that matched both the LF and HF targets of the mid, ignoring the diffraction ripple entirely. Surprisingly, I was able to do that with a single cap and a single coil. It meant wiring the mid out of phase (and tweeter in phase) for some reason, but I couldn't get the phase in particular to line up, because (presumably) with the phase being directly derivable from the FR, the anomalies in the FR also cause corresponding phase anomalies. I'd love the idea of a BP this simple and think it could merit further exploration. Would it be best to get some slightly off axis measurements for the mid to soften the diffraction and change the design axis by just a few degrees?

                    This speaker has been quite the learning experience and a lot of fun so far. All the advice given that 3 ways are far more complex than 2 ways is absolutely true. Its not as simple as 2 2-ways developed independently in the same box...

                    I was off a little on my advice to add 1.5 ohms. That would help slightly at the tank notch frequency, but that's not where your low point is.

                    I ran a sim with my 8 ohm drivers, and your woofer filter. Then I tried to improve on the impedance, but really could not. It's just tough with the woofers in parallel.

                    Comment


                    • #11
                      Yep, I think that's the conclusion I'm coming to as well. The decision point seems to be a lower XO point or flatter response with a nicer impedance curve. I think I"m going to compromise on the XO point and let it float higher than I wanted and see how bad the lobing seems in this iteration. FR is quite flat and the impedance ...is what it is, but the lowest point is 2.9 ohms and I think is just a function of the W18's in parallel - not much can be done with the filter.

                      I have a model that pushes the woofer XO point to about 900, which would be great, but I lose about 1.5db of output from 600-1200, which I suspect will be audible. The impedance minimum also drops .25 or so ohms from 100-500 - not sure if that's academic or a problem.

                      I may build both and have a listen, and if the lower XO sounds better, just turn up the amp I'm using for testing and see if it gets mad. If the magic smoke comes out, so be it. Its a cheap amp and I know that's too aggressive of an impedance profile.

                      Comment


                      • #12
                        Well, I think we're on to something. I'm experimenting with the inductor size in the LP to see what sounds best in the 600-900 range, but this is getting close. Sounds much better than the first iteration. I'll throw the crossover up soon once I've listened a bit more. The impedance is even coming around.

                        Comment


                        • #13
                          Forgot to ask - do you think the peak in the low end is the room or could the box tuning be off that far?

                          Comment


                          • #14
                            Originally posted by Adam_M View Post
                            Forgot to ask - do you think the peak in the low end is the room or could the box tuning be off that far?
                            Room. The room can effect the bass drastically. I use gated measurements for x-over work, and don't pay any attention to bass below 300hz.

                            Comment


                            • #15
                              Spent quite a bit of time testing and listening, and I found that it was really was critical to keep the LP XO point below 1k. I wasn't able to get the FR from 600-1.2K flat with a lower XO point, but the ~1dB-1.5dB dip doesn't offend me in the slightest after listening - it actually makes the speakers really easy to listen to. I was able to fix the dip with a couple other compromises (notably higher XO point, but also slightly worse phase coherence and more complex XO, or a really nasty impedance dip) and it never sounded as good as this version. Crossover didn't end up being terribly complex either, down a bunch of parts from the early iterations.

                              Comment

                              Working...
                              X