Announcement

Collapse
No announcement yet.

The Defiants - InDIYana 2019 Build

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

  • KEtheredge87
    replied
    OK... I think I did a better job of measuring my drivers and preparing for XO design. Basic steps were:

    1) Measure nearfield woofer, and farfield tweeter, farfield woofer, and farfield woofer + tweeter
    2) load woofer nearfield and farfield into blender, apply baffle diffraction, blend appropriately, extract minimum phase and export FRD
    - I picked 0.75" radius... but I don't know how that translates to a variable length facet cut.
    3) load tweeter farfield into blender, apply baffle diffraction (again with 0.75" radius), extract minimum phase and export FRD
    4) load minimum phase FRDs into PCD7, apply overlay of woofer + tweeter farfield curve
    5) set measurement distances and offsets in PCD, determine Z-offset.

    As it happens... PCD said my z-offset was literally 0.000mm

    Even after doing all that over, something still wasn't quite right, but I kinda lucked out and arrived at a solution. A pretty darn good looking solution in fact! Both Xsim and PCD are telling me my tweeter needs more padding, and that my reverse null isn't quite as sharp as the OmniMic is telling me... but this sounds really nice so far I tend to give speakers the "smile on my face" test... which is to say I tweak until I can't help but smile and think "Man that sounds good!"... and I'm smiling right now

    Assuming I make no changes to this, I think the only component robustness issue might be R2 in my tweeter... if I really wallop these guys with 100W of 4 kHz energy, XSim says that resistor will be dissipating over 50W of power. That dissipation result sounds a bit excessive to me, but I don't have any other way to estimate that currently.

    In the photos below, the Omnimic traces without reverse null have 1/6 octave smoothing applied. I did this because I read Jeff Bagby's whitepaper that states the human brain interprets frequency differences on a 1/6 octave smoothing level anyway, so that's a decent approximation for how flat I think it is by listening.

    Time for some sleep!
    Click image for larger version

Name:	Measured Design 2 PCD full system.PNG
Views:	45
Size:	233.1 KB
ID:	1409622Click image for larger version

Name:	Measured Design 2 PCD rev null.PNG
Views:	42
Size:	349.2 KB
ID:	1409624Click image for larger version

Name:	Measured Design 2 Xsim Power Dissipation.PNG
Views:	42
Size:	110.5 KB
ID:	1409623Click image for larger version

Name:	OmniMic frequency response screengrab MD2.png
Views:	38
Size:	114.9 KB
ID:	1409626Click image for larger version

Name:	OmniMic frequency 1 sixth smoothed response screengrab MD2.png
Views:	40
Size:	100.5 KB
ID:	1409625

    Leave a comment:


  • rpb
    replied
    In theory, you would want a deep reverse null at 1.5 to 2 meters on the tweeter axis. That might not be the same null you get measuring at less than 1m. It's just the geometry changing the distances from the mic, to each driver.

    Leave a comment:


  • KEtheredge87
    replied
    Thanks jhollander and rpb . Neither of you have said anything that came off negatively, so I'm just being goofy with this meme below... I remember John's lessons now. I'll go back and take proper measurements, run them through the blender and extract minimum phase for both woofer and tweeter, and give this another whirl! Thanks for jogging my memory!

    Click image for larger version

Name:	study yo crossovers.jpg
Views:	35
Size:	71.7 KB
ID:	1409591

    Leave a comment:


  • rpb
    replied
    Can you show us a measurement of the woofer and tweeter (with filters) individually on the same graph? The shape of the roll-offs can tell you a lot.
    I like the shape of your tweeter roll-off. Nice and gradual. If the woofer has more of a knee, that may not tend to have as deep of a null, when matched with the smooth tweeter roll-off.

    Leave a comment:


  • rpb
    replied
    Here's what you can do. Move the mic up about ten or fifteen degrees, and see if the null gets deeper. Then, if not, move it downward about ten or fifteen degrees, and check again. This will tell you if the tweeter is ahead or behind. Tweak the filters for a deeper null on axis, then switch back the polarity, and audition. If neither is deeper, then you may be off a lot. In that case, you may need an additional x-over part, or one less.

    Leave a comment:


  • jhollander
    replied
    Whoa, if you are using a blended woofer file the phase is not accurate and you are not using "as measured" phase. You are back to extracting minimum phase for the woofer blended file, then the tweeter, and finding the offsets. Once you extract minimum phase for the woofer you need to do that for the tweeter as well. You can't mix minimum phase with as measured phase files.

    Sounds like you may have not extracted min phase for the tweeter file.

    Also note that if you changed the measurement distance between your simulation measurements and your "with crossover" measurements the null will not be the same.

    Leave a comment:


  • KEtheredge87
    replied
    Originally posted by jhollander View Post
    If you are using as measured phase you do not need offsets or the T+W frd
    Thanks John,

    What I think you just told me was that I am unwittingly screwing myself up by applying a z-offset, measurement distance, and X/Y woofer offset in PCD. I should just take my Blended near/far woofer response and my farfield tweeter response and design from there.

    I still need my in-box ZMA measurements, right?

    Do I understand you correctly? (You'd think I'd have this down pat by now... especially for the number of times you've helped me on it!)

    Leave a comment:


  • jhollander
    replied
    If you are using as measured phase you do not need offsets or the T+W frd

    Leave a comment:


  • KEtheredge87
    replied
    Alright guys - I'm having some issues with Z-offset, one of the very basics of XO simulation. I'm simulating designs that appear to have good reverse nulls in PCD and/or Xsim, but have a non-existent reverse null in Omnimic when I measure. That tells me that my Z-offset calculation within PCD alone is incorrect, so I've tried using the OmniMic GUI as well as some guestimations based on Fusion360 X-section views. As expected... I get a few different results.

    Here's my PCD setup, and some result files with Z-offsets being compared in PCD.

    Measurement setup:
    24.25" away on tweeter axis. (0.61585m)
    woofer located -6.75 inches down / 0 inches over from the tweeter axis (-0.17145m)

    Measurement FRDs Recorded:
    1) Woofer alone
    2) Tweeter alone
    3) Woofer and tweeter together.

    Measurements of FRDs loaded into PCD without ZMA files (OmniMic FRDs contain phase info already):
    1) PCD's woofer FRD slot gets the RS180P-4 alone file
    2) PCD's tweeter FRD slot gets the RST28F-4 alone file
    3) Import overlay on graph gets the RS180P + RST28F file
    4) Offsets for listening distance and driver locations set appropriately with tweeter axis being the origin


    My observation is that my PCD-based results aren't lining up as nicely as Jeff's how-to instructions would indicate. Calculating Z-offset within the OmniMic GUI results in a value that I believe more... but PCD seems to disagree with it. This uncertainty makes me suspicious of everything...

    Without getting this Z-offset thing nailed down, none of my simulated crossover results are translating to actual measurements.

    Surely someone else has run into these issues before. Does anyone have some suggestions?
    Thanks in advance!

    Click image for larger version

Name:	What PCD thinks.PNG
Views:	47
Size:	333.0 KB
ID:	1409534Click image for larger version

Name:	Omnimic's results loaded in PCD.PNG
Views:	55
Size:	344.7 KB
ID:	1409533Click image for larger version

Name:	Fusion360 in PCD.PNG
Views:	46
Size:	374.8 KB
ID:	1409535Click image for larger version

Name:	Z-offset results from Omnimic.PNG
Views:	51
Size:	10.9 KB
ID:	1409537Click image for larger version

Name:	Fusion360 Guesstimate.PNG
Views:	51
Size:	288.0 KB
ID:	1409536

    Leave a comment:


  • Wolf
    replied
    I just assumed that was tweeter and its filter. Good call, RPB!
    Wolf

    Leave a comment:


  • KEtheredge87
    replied
    Originally posted by rpb View Post

    Looks like a second driver might be causing a cancellation, or you wired the x-over wrong. When you measured the tweeter, where did you disconnect the woofer? Should be disconnected before the woofer filter, not after. L2 and C2 might be creating a notch filter?????

    If you measure again, and get the same result, move the mic location, and see if the dip frequency changes. If it does, disconnect the tweeter, and run a sweep ro see if you hear a second source..
    I think you win bud! Last night I just disconnected the woofer AFTER the filter, not before it. Today I went to a 2nd order XO before I read your suggestion and tried a frequency measurement with the woofer disconnected before the filters and the 2nd order tweeter network left as-is. No more odd notch. Must have been my extra parts in the network causing chaos.
    Attached Files

    Leave a comment:


  • rpb
    replied
    Originally posted by KEtheredge87 View Post

    Thanks John, I'll give that a go. Jeff B suggested I was having a room reflection over on facebook. I guess the bit I still don't understand is if that dip is truly a room reflection, I don't get why it's only in the filtered response, not the raw driver response... quite curious.
    Looks like a second driver might be causing a cancellation, or you wired the x-over wrong. When you measured the tweeter, where did you disconnect the woofer? Should be disconnected before the woofer filter, not after. L2 and C2 might be creating a notch filter?????

    If you measure again, and get the same result, move the mic location, and see if the dip frequency changes. If it does, disconnect the tweeter, and run a sweep ro see if you hear a second source..

    Leave a comment:


  • KEtheredge87
    replied
    Originally posted by Wolf View Post
    1k = 13" That's where I'd start.

    Next, since you have a 1st order, I'd drop the parallel resistor and go second order. It may be the Fs of the driver and the interaction of the 1st order rolloff.

    Later,
    Wolf
    Thanks Ben, going 2nd order is on my docket for tonight's measurement and experiment session. I was hoping I could get away with just a cap and a single resistor on the tweeter, as the woofer starts to give me a pronounced hump at the knee if i increase that parallel capacitor in value. Having a shaping resistor there would be quite helpful... oh well... maybe I'll save that idea for XO version 2 for MWAF!

    Leave a comment:


  • Wolf
    replied
    1k = 13" That's where I'd start.

    Next, since you have a 1st order, I'd drop the parallel resistor and go second order. It may be the Fs of the driver and the interaction of the 1st order rolloff.

    Later,
    Wolf

    Leave a comment:


  • KEtheredge87
    replied
    Originally posted by jhollander View Post
    Might try tightening up the gating. It looks like there's a blip at 4.2 ms and your gating is at 5.? ms.
    Thanks John, I'll give that a go. Jeff B suggested I was having a room reflection over on facebook. I guess the bit I still don't understand is if that dip is truly a room reflection, I don't get why it's only in the filtered response, not the raw driver response... quite curious.

    Leave a comment:

Working...
X