Trust Your Ears: My XO Tweaking Adventures!

Discussion in 'Archived Threads 2001-2004' started by Patrick Sun, May 31, 2001.

  1. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
    More tweaking 5/24/01 on my SunOnes
    Graph 5 (1m): I did lower the lower woofer crossover point to around 215Hz. Since I was able to free up some capacitors, I adjusted the zobel on the upper woofer. Sounds okay, but a little thin on the bass, and still hear a bit of blare/glare on some tracks.
    Graph 6 (1m): So with the adjust zobel on the upper woofer, I go back to the 350Hz crossover point for the lower woofer. A bit more bass on the low end returns, but still need more tweaking.
    Crossover breakthrough 5/29/01 - Trust your ears.
    I spent the last few days listening to various music sources to see if I could nail down the "something that wasn't right" when I was listening to the speakers. I finally decided to take some individual measurements of the woofers (in tandem) and the tweeter. I am using a dual-input terminal cup that allows me to just listen to either section if I remove the connection bars. So I made some cursory measurements of the woofers' crossover behavior on the frequency range over 2000Hz.
    What I found out was shocking. I discovered that my upper woofer filter was not doing its job! Wherever it was crossing over, it was nowhere in the vicinity of even 8000Hz! Basically I had no woofer filtering and what I had been hearing as blare/glare was the upper woofer's cone breakup in the 3000-5000Hz range due it playing butt nekkid and fancy free!
    Upon reviewing my calculations, I discovered I made a mistake in determining the impedance at the woofers' crossover point. I brain-farted and just used the zobel target impedance of 8 ohms when calculating for a 2nd order Butterworth. This is what threw off the filter component values I used for the upper woofer. I should have parallelled the upper woofer's zobelized impedance of 8 ohms with the lower woofer's non-zobelized impedance of 12 ohms for a nominal 4.8 ohms. Since I rearrange my components around, I fine-tune the capacitor value for the upper woofer's zobel with some of the freed up capacitors (this just means reducing the cap value from 24.4uF to 21.3uF which is closer to the calculated value).
    I noticed that the tweeter filter needed some work as well (yet another brain-fart where I forgot to use the zobelized impedance when calculating the filter - instead of using 8 ohms, I used 6.5 ohms @2000Hz from the spec sheets from Morel's website). So I whipped up a different crossover network for both the tweeter and upper woofer, and see if I actually get some filtering action on both drivers.
    5/30/01 - This is the start on the road to sonic recovery...
    Graph 7 (3m): I decided to attack the XO design a little differently by plotting the response of the woofer (in tandem) and the tweeter, as well as the composite frequency response. This allows me to see what needs to be tweaked (because I know I didn't get it right on the first shot). As you can see, I now get the filtering I expect to get with the new upper woofer filter in place, its response is shelved down starting around 2000Hz, and the rest of the response is consistent with its frequency response above 2000Hz to 6500hz. Bear in mind the lower woofer is still contributing (albeit not as much as the upper woofer) to the over frequency response. Take note of the tweeter's response slope which starts too far past the desired XO point of 2000Hz, it appears to sloping off around 3200Hz. This gives me more food for thought.
    Graph 7a (3m): Also, I've seen folks use graphs with the tweeters connected out of phase with the woofers to verify the phase coherency of the crossover point (you should see a big dip in the response around the crossover point). From this graph, you can see the dip around 2500-3000Hz is just past the target crossover point of 2000Hz. So this means more work on the tweeter filter.
    I decide that the tweeter's zobelized impedance is too high in the pass band that I am concerned about. The nominal impedance of the Morel MDT30 is 8 ohms, but that's simply too high for the frequency range from 1500-20000Hz. I decide to junk the tweeter zobel from the tweeter filter, and rely on the natural impedance/frequency response curves of the tweeter for the crossover. Luckily for me, Phil Abbate was nice enough to measure my tweeter and midwoofer on my baffle, and produce the impedance/phase "zma" file, and the frequency response "frd" file for use for modelling in speaker software design packages. I looked at the "zma" file and looked for the value of the impedance in the 2000Hz range and above. I found that the impedance was around 7.2 ohms, and then it slopes down to around 6.5 ohms before rising upwards - the inductance of tweeters don't have as much of a detrimental effect that woofers do, so I decided to remove the zobel, and use 7.2 ohms for the impedance when calculating a new tweeter filter. Luckily for me, I had the right parts on hand to make this filter a reality for both speakers (I've been really lucky in this respect).
    Graph 8 (3m): So with the new tweeter filter in place (ditching the zobel on the tweeter, and keeping it a simple 2nd order Butterworth with a 3.8dB l-pad), here is the frequency response: showing both the woofers, the tweeter, and the composite response. The tweeter filter is behaving as well as can be expected, the sloping is a bit harder due to the tweeter natural roll off at 2000hz. I thought about using a 1st order (12uF capacitor), but was more concerned with tweeter protection, so I went with the 2nd order BW filter. But this is okay because of the contribution from both woofers helps out in the crucial 2000Hz range. I leave the lower woofer crossover point around 350hz for now, and will evaluate it if I think the bass is too much.
    Graph 8a (3m): Here's the tweeter connected out-of-phase graph, along with the in-phase response. As you can see, the dip is centered around 2000Hz now, so that's a good sign that the new revised crossover network is making a difference in the over speaker performance.
    This part of the tweaking process is where the lightbulbs are going off in my head. Kind of a neat feeling. Now it's time for me to listen to more music CDs and see if there's anything that needs to be addressed in the crossover. Remember: Trust your ears.
    ------------------
    PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; LDs
     
  2. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
    The end is near!
    Graph 9 (3m): Well, after listening some more, I decided that I still have the blare/glare problem. Looking at Graph 8 shows why, still too much contribution from the midwoofer for frequencies above 2KHz. So this time I plop in original 4th order LR filter for the upper midwoofer (I remove the zobel since I needed the capacitors to make the 4th order filter). I also adjust the tweeter filter just a little bit, trying to extend the SPLs around 2KHz be lowering the tweeter filter's break frequency down a bit to 1.8KHz. As you can see, the region around 2KHz is a mess, a nice depression of SPL is noted. Also, there's still a bit of blare/glare.
    Graph 9a (3m): Just a comparison of the tweeter connected in-phase and out-of-phase on this graph. It looks better than Graph 8a, but the dip is a little further up the road around 2.3KHz.
    Graph 10 (3m): So here I try placing the zobel back on the tweeter (the highs sound smoother with a zobel on the tweeter, plus the impedance is a little more stable, making the filter calculations/components easier to realize with the parts on hand). This yields minimal improvement in the 2KHz region, but get rid of a weird dip in the tweeters response from Graph 9 around 4KHz. This XO won't be the final one, that's for sure.
    Graph 10a (3m): Can't expect much to change in the comparison of the tweeter connected in-phase and out-of-phase on this graph. Onward to the next XO!
    Graph 11 (3m): So, I ask myself, what should I do to squeeze out more SPL around the XO frequency? I try and reduce the order on the tweeter filter to a 1st order Butterworth. This should add a little bit of SPL output for the frequency range of 1KHz to 2KHz. If you compare Graph 10 to Graph 11, you can see the 1st order (just a 10.5uF capacitor) slope vs. the 2nd order slope by looking at the tweeter response. This yields just a little bit of improvement, but again, not what's needed to combat the blare/glare and the dip at the XO point.
    Graph 11a (3m): From the comparison of the tweeter connected in-phase and out-of-phase on this graph, at least the dip is centered around 2KHz.
    I do some more thinking about the blare/glare problem. I stare at the graphs over and over again. I go to bed and sleep on it. The next morning, it comes to me, my lower woofer is contributing too much in the 2KHz-6KHz region. The light at the end of the tunnel is getting brighter and brighter.
    Graph 12 (3m): I run some test to see what happens to the blare/glare when I use a 2nd order filter on the lower filter. From the graphs, it's evident that midwoofer output gets reduced by 3dB-4db, which is enough of a difference to clean up most of the output from the midwoofer in the 2KHz-6KHz region. Compare Graph 11 to Graph 12, and you'll see this reduction of midwoofer output. I go back to the 2nd order filter for the tweeter to get rid of the extra SPL from 1KHz-2KHz since it's not needed, and lower the L-pad to around -2.8B (I will probably have to raise it to around -3.8dB to reduce the high end output). The best part is that the voices are much clearer now that much of the blare/glare is gone. This latest XO gets me very close to a final XO design!
    Graph 12a (3m): From the comparison of the tweeter connected in-phase and out-of-phase on this graph, the dip is now well centered around 2KHz, which is desired in this type of graph.
    I finally figured out how to use Adire Audio's ResGraph program to graph the LAud files I had for my tweeter and midwoofer.
    Here's the frequency response of the tweeter measured on a LAud setup on the front baffle:
    Tweeter Frequency Response
    Here's the impedance/phase response of the tweeter measured on a LAud setup o the front baffle:
    Tweeter Impedance/Phase Response
    Here's the frequency response of the midwoofer measured on a LAud setup on the front baffle:
    Midwoofer Frequency Response
    Here's the impedance/phase response of the midwoofer measured on a LAud setup of the front baffle:
    Midwoofer Impedance/Phase
    ------------------
    PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; LDs
     
  3. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
    6/5/01 - More minor tweaking (yes, it never ends!)
    Fun with different filter types on the tweeter (all 2nd order filters):
    In my attempt to squeeze some output out of the tweeter at the XO point, I test out a few different filter types and see which one will be the most beneficial in the final design.
    Butterworth filter (Q=0.707):
    Graph 13 (3m): I decided that the l-pad of -2.8dB wasn't padding the tweeter enough, the highs were a little too high, so I backed down the l-pad to -3.8dB. This makes the SunOnes sound a bit more mellow, but this does cause the tweeter response to dip around the XO point more than desired.
    Graph 13a (3m): Just a comparison of the tweeter connected in-phase and out-of-phase on this graph. It looks a little worse than Graph 12a, with the dip a little wider and not as distinct.
    Linkwitz-Riley (Q=0.50):
    Graph 14 (3m):: This is the response once I install a 2nd order Linkwitz-Riley filter on the tweeter. With a Q=0.50, its output at the corner frequency of 2000Hz is pretty tame. This is very evident by the dip in the tweeter and overall response. Obviously this is not a desirable response for this application.
    Graph 14a (3m): The dip on the comparison of the tweeter connected in-phase and out-of-phase graph gets a little wider, and not as pronounced.
    Chebyshev (Q=1.00):
    Graph 15 (3m): Looking at graphs 13 and 14, it obvious that I need to go with a higher Q filter, so I give the not-so-orthodox Chebyshev filter a try. This is a very high Q filter, so much so that for my particular situation, it actually helped. (Of course, I'm not quite sure how the phase is holding up with this mix of filters, thus I'll have to do some serious listening). The graph now shows a much steeper output near the XO frequency of 2000Hz, so it flattens out the response in the XO region, and gets rid of the dip that I was measuring with the Butterworth and Linkwitz-Riley filter. I would guess that using 2 midwoofers both with LR filters, the combination of the driver outputs probably produces a high Q corner frequency response which mates with the new tweeter's Chebyshev filter much better than the BW or LR filters. I will listen to some more music material and see if I hear anomalies that would make me revert back to the BW filter for the tweeter (graph 13).
    Graph 15a (3m): From the comparison of the tweeter connected in-phase and out-of-phase on this graph, it appears the dip is centered around 2KHz again, perhaps a good sign.
    Comparison of the filter Q's:
    Graph Q (3m): Here's a graph showing the differences in filter Q's for the tweeter filter. This may be instructive to some people who like verification of textbook filters and their effects on frequency response.
    ------------------
    PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; LDs
     
  4. Julian Data

    Julian Data Second Unit

    Joined:
    Oct 5, 1999
    Messages:
    408
    Likes Received:
    0
    Pat:
    You have truly out done yourself in testing and tweaking and I learned a great deal based from your experiences. I really appreciate it. Perhaps you could write a memoir on your XO design? [​IMG]
    ------------------
     
  5. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
    Whaddya mean? It's all on the SunOne webpage! That is my memoirs!
    Actually it helps to document where I've been, and compare it to where I want to go. In my scrap pieces of paper, I have all the XOs I've tested so far.
    Once I do some more listening tonight, I can tell if this latest XO is a step in the right direction. [​IMG]
    ------------------
    PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; LDs
     
  6. Julian Data

    Julian Data Second Unit

    Joined:
    Oct 5, 1999
    Messages:
    408
    Likes Received:
    0
    I meant publish a book! LOL! [​IMG]
    What are the next plans for the future?
    ------------------
     
  7. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
    Retrofitting the center channel speaker with the same tweeter, the Morel MDT30, fixing up the XO on that speaker.
    Then, it'll be the rear speakers (same drivers that were used in the SunOnes) in a TM configuration. The XO was "outsourced", and I may do a XO design of my own just to see which is preferable by me. I have the drivers, ordered the XO parts, and just need to build the enclosures and XOs.
    ------------------
    PatCave; HT Pix; Gear; DIY Mains; DIY CC; Sunosub I + II + III; DVDs; LDs
     
  8. Julian Data

    Julian Data Second Unit

    Joined:
    Oct 5, 1999
    Messages:
    408
    Likes Received:
    0
    That sounds good, I can't wait for the documentation. [​IMG]
    Which business did you get the XO done?
    ------------------
     
  9. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
  10. Greg Monfort

    Greg Monfort Supporting Actor

    Joined:
    May 30, 2000
    Messages:
    884
    Likes Received:
    0
    Good work PS, and why I use horns or fullrange drivers with a woofer, supertweeter, and only an inductor and cap. [​IMG]
    I'm surprised you don't mind the hump in the upper HF BW where voice sibilance and treble overtones are reproduced.
    GM
    ------------------
    Loud is beautiful, if it's clean
     
  11. Mark Hayenga

    Mark Hayenga Supporting Actor

    Joined:
    Jun 11, 1999
    Messages:
    607
    Likes Received:
    0
    Just wanted to drop in and say I was simulating Pat's tweeter's FR with his LAUD data and his xover design, and the tweeter's simulated FR is flat +/-1.5dB from 2k to 10k. I'm guessing Pat's ears are telling him right and the HF anomalies are due to the RS SPL meter.
    Mark
    ------------------
    Mark Hayenga
    www.hayenga.com
    ----------------------
    "Saru mo ki kara ochiru"
     
  12. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424
    Mark, thanks for the sim. What I think is happening is that I'm measuring about 3 meters away, and the tweeter response is shelved down a bit near 2KHz (vs. the 1m measurement) and that's probably why using the underdamped 2nd order HP filter helps fill in that XO region from that far away. I'll see what happens when I measure from 1 meter away tonight.
    I listened to the current XO (graph 15) last night, and I find male vocals a bit more fuller than with the Butterworth 2nd order HP filter (graph 13). Female vocals sound just about right to me (but at high volume, Diana Krall can zing your ears on a few songs from "When I Look In Your Eyes").
    One of my test CDs for this particular region is U2's Joshua Tree. If Bono sounds thin, it shows up with a dip around 2KHz. Sting is also another singer that will reveal deficiencies in this region.
    ------------------
    PatCave ; HT Pix ; Gear ; DIY Mains ; DIY CC ; Sunosub I + II + III ; DVDs ; LDs
     
  13. Patrick Sun

    Patrick Sun Moderator
    Moderator

    Joined:
    Jun 30, 1999
    Messages:
    38,676
    Likes Received:
    424

Share This Page