Announcement

Collapse
No announcement yet.

Speaker Design Question #2 - Measuring Driver Offset (z-axis) - WinPCP

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

  • Speaker Design Question #2 - Measuring Driver Offset (z-axis) - WinPCP

    First and foremost, thanks to all that responded to my BSC question.

    My next question is regarding measuring driver offset. The speaker I'm currently practicing with is Bamberg Audio 2TM (SS-Kit), ,which allows for bi-wiring.

    My setup = Room EQ Wizard - out to Focusrite 2i2 - out to Trends 10.1 T-amp - Left Channel connected to Tweeter-side crossover terminals, and Right Channel connected to Woofer-side crossover terminals
    Input = MiniDSP UMIK

    These are the procedures I followed:

    1) Took a full-range measurement of the speaker - On tweeter axis - Distance of 20.25" - 5.25ms Gating

    2) Took Woofer only measurements - Still on Tweeter Axis - Distance of 20.25" from speakers baffle - 5.25ms Gating

    3) Took Tweeter only measurements - On Tweeter Axis - Distance of 20.25" from speaker baffle - 5.25ms Gating

    4) Took all measurements and converted in "FRD" files (which contained frequency, amplitude, and phase data)

    5) Uploaded the files into "Acoustic Offset tab of WinPCD

    6) Set measurements distance to 0.514 meters (because 20.25" = 0.514 meters)

    7) Set "Driver 1" (aka the Tweeter) - X-offset = 0, and Y offset = 0 (because tweeter was used as the reference point)

    8) Set "Driver 2" (aka the Woofer) - X offset = 0, and Y offset = -0.1524 (because the woofer is mounted 6" = 0.1524 meters below tweeter)

    9) I believe the next step is to adjust Driver 2 (the Woofer) Z-offset until the mathematical summed response of Step 2 and Step 3 above match the measurement in Step 1 above.

    Based on these steps, I calculated the Woofer Z-offset to equal -.020m or negative 20mm?

    I'm not sure I'm confident in that result. I believe the measurements should have been closer to 40mm +/-.

    The tweeter is ScanSpeak D2604/8300 and the woofer is ScanSpeak 18W/8434

    Any thoughts on what I might be doing wrong? Some of my questions:

    a) Should gating not be used?

    b) Do the tweeter/woofer measurements represent minimum phase measurements? Or do I need to use one of Jeff Bagby's spreadsheet to derive minimum phase?

    Click image for larger version

Name:	Bamgerg 2TM  9-27.jpg
Views:	1
Size:	247.0 KB
ID:	1346776 Click image for larger version

Name:	Bamberg Woofer w phase.jpg
Views:	1
Size:	232.5 KB
ID:	1346777 Click image for larger version

Name:	Bamberg Tweeter.jpg
Views:	1
Size:	303.2 KB
ID:	1346778 Click image for larger version

Name:	winPCD measurements.jpg
Views:	1
Size:	385.3 KB
ID:	1346779 Click image for larger version

Name:	Bamberg Fullrange w phase.jpg
Views:	1
Size:	190.2 KB
ID:	1346780







  • xmax
    replied
    A good example under time alignment.

    http://www.sbacoustics.com/index.php/technical-notes/

    Leave a comment:


  • dkalsi
    replied
    Thanks Dlr - your above post helps tremendously.

    In essence, the acoustic offset isn't necessarily the measured (like with a ruler) the delta between the radiating point of the two drivers mounted on a vertical plane, it is instead the acoustic delay embedded in the frd files created. Is that correct?

    So long as I then use the same files are utilized in my crossover design (while specifying the corresponding z-offset related to those files), then the simulated response should be very close?

    __________________________________________________ ____________________
    EDIT:


    Dlr - I apologize. I just went back to your earlier post and noted that you have already covered this - I feel like such an idiot :-( Didn't mean to exhaust your effort explaining this to me over and over. For those that are curious, I am alluding to the following explanation previously provided by dlr in an earlier post:

    "Remember that offset is relative, that is, it's a value required so that the two minimum-phase files sum to the measured sum. If you change either file, the relative offset needs to be updated to account for the change.

    The absolute value of the offset isn't really important. All it amounts to is the difference between the two measurements that "aligns" the time delays between the measurements to correspond to the difference required for the generated phase of each driver file. There is no "correct" offset, only the offset needed for the measurement phase as generated for the SPL of the files."
    __________________________________________________ _____________________

    I think I'm now past asking the same question over and over about minimum phase :-)

    My next step is to figure out why my simulated response only closely matches the measured response, which was taken at .775meter, when I set the "Measurement Distance" in System tab to 2meters. I figured that if I have specified the baffle layout (i..e, the x,y,z coordinates of the driver) as well as provided the radiating diameter of the drivers, the program should be able to approximate how the speaker sound perform at 0.775 meters.

    It will be helpful to understand what may be causing that variance.

    Leave a comment:


  • dlr
    replied
    Originally posted by dkalsi View Post
    All that being said, "z-offset" couldn't possibly have multiple values could it?
    Yes. The offset is a function of delay to the "apparent" acoustic center. That "apparent" center changes if the minimum-phase changes, which it will if a tail is changed in the SPL. SPL and phase are mutually dependent.

    Theoretically, there should be only one exact value for z-offset right?
    Yes and no. For any given phase response of two files, there will be one. The no is because offset, as mentioned above, is dependent on the phase of both files. If one changes, the offset will change, a little or a lot. That's why I so often say to use the same files for design or at least use the same tail if you post-process the files additionally, such as adding a different box from software (such as Unibox) or if you were to average tweeter measurements for a windowed response. Few do the latter AFAIK.

    If that is the case, then I know I just need to learn more about setting "tails".
    The key is to be consistent. You can set tails in a random manner if you wish as long as you use the same tails in any later post-processing. Keep in mind that the phase change for a box spliced into a woofer or possibly even a midrange won't have a huge affect on the offset. It's the lowpass (driver top end tail) that has the biggest effect.

    I was also hoping you (or others) can chime in and possibly explain why my simulated response doesn't match the measured response. If I have measured the value of each individual crossover component, if I have specified the measurement axis to be the same exact point where the measured response (inclusive of crossover) was taken, if I have specified the correct crossover layout, shouldn't simulated response = measured response? I only seem to get my simulated response to be close to the measured reference response (@ 0.775meters) when I adjust the measurement distance to 2 meters.
    If the procedure for measurements and the offset are done correctly and the measurement system provides consistent results, the correlation should be very close.

    dlr

    Leave a comment:


  • dkalsi
    replied
    Originally posted by dlr View Post
    The goal is to measure at a known point (whatever point you want), make all three measurements, then determine offset at that point. The results should be a nearly identical calculated summed response overlay with the measured sum.
    Duly noted Dlr - Once set, I do NOT touch the mic for all three measurements. And I do get near identical summed response when adjusting for z-offset as you indicated. I'm glad you also found that selecting different tails did result in different indications of z-offset.

    All that being said, "z-offset" couldn't possibly have multiple values could it? Theoretically, there should be only one exact value for z-offset right? If that is the case, then I know I just need to learn more about setting "tails".

    I was also hoping you (or others) can chime in and possibly explain why my simulated response doesn't match the measured response. If I have measured the value of each individual crossover component, if I have specified the measurement axis to be the same exact point where the measured response (inclusive of crossover) was taken, if I have specified the correct crossover layout, shouldn't simulated response = measured response? I only seem to get my simulated response to be close to the measured reference response (@ 0.775meters) when I adjust the measurement distance to 2 meters.

    Leave a comment:


  • dlr
    replied
    Originally posted by dkalsi View Post
    I did exactly as you said - I did not add a LP tail to the tweeter response. I am assumting the way to do this is to not specify the slope for the LP tail in Frequency Response Blender.

    Now - the result was that I was getting an acoustic offset of +0.008m.

    Just so I understand your comment about the design axis (as intended by the manufacturer) - right now, I was only hoping match the simulated response to the measured response. If my measured response was at the tweeter level, then is it a safe assumption to think that the simulated response (one specified to be on the 0 vertical and 0 horizontal tweeter axis) should match the measured response?
    Design axis means the axis on which you intend to make your design. It has nothing to do with any measurements, those are a separate issue. Typical design point is somewhere on the tweeter axis, but doesn't have to be.

    I only now tried out Jeff's Blender. It appears that there is a default on the high end tail of -12db @20Khz.This is lower than I have found typical of tweeters. FYI, what the lowpass slope (top end tail) does is add delay, the steeper the slope, the more it delays. What this means is that for any increase in slope, there's more delay, consequently for the acoustic offset with another driver (e.g. a woofer), the offset (z-axis) will be increased correspondingly. A low enough slope could require a positive offset of a woofer to compensate for the (possibly unrealistic) amount of delay added by using a very steep slope. In the end, it doesn't matter as long as you use the same slope and Fc on any subsequent measurement file you may substitute with newly generated minimum phase.

    One more than I forgot to mention was that the one thing that also helps immensely is if change the "Listening / Measuring Distance" to = 2meters, as oppose to 0.775m (i.e., the distance at which the files were actually created - also with the distance where I currently measured the crossover included response as a reference to check my simulated response against). To be clear, this is in the "System" tab, and NOT the "Acoustic Offset" tab - I understand that is not something one is supposed to change from the actual distance when trying to determine the acoustic offset.
    When you are determining acoustic offset, it is best to set the distance to that which was used for the measurements to get the most accurate value. The goal is to measure at a known point (whatever point you want), make all three measurements, then determine offset at that point. The results should be a nearly identical calculated summed response overlay with the measured sum.

    dlr

    Leave a comment:


  • dkalsi
    replied
    Earl,

    I did exactly as you said - I did not add a LP tail to the tweeter response. I am assumting the way to do this is to not specify the slope for the LP tail in Frequency Response Blender.

    Now - the result was that I was getting an acoustic offset of +0.008m.

    Just so I understand your comment about the design axis (as intended by the manufacturer) - right now, I was only hoping match the simulated response to the measured response. If my measured response was at the tweeter level, then is it a safe assumption to think that the simulated response (one specified to be on the 0 vertical and 0 horizontal tweeter axis) should match the measured response?


    Dlr,

    After using the tweeter axis file (i.e. 1a) - modified for minimum phase derived from not specifying a LP tail for the tweeter (as recommended by Earl) and file 1b - I now get the following results:
    Click image for larger version

Name:	WinPCD - Screenshot.jpg
Views:	1
Size:	318.8 KB
ID:	1348429

    Click image for larger version

Name:	New Response Tweeter and Woofer both at Tweeter Axis.jpg
Views:	1
Size:	264.8 KB
ID:	1348428



    One more than I forgot to mention was that the one thing that also helps immensely is if change the "Listening / Measuring Distance" to = 2meters, as oppose to 0.775m (i.e., the distance at which the files were actually created - also with the distance where I currently measured the crossover included response as a reference to check my simulated response against). To be clear, this is in the "System" tab, and NOT the "Acoustic Offset" tab - I understand that is not something one is supposed to change from the actual distance when trying to determine the acoustic offset.

    Click image for larger version

Name:	Simulated response utilizng a 2meter measurement distance.jpg
Views:	1
Size:	256.6 KB
ID:	1348427


    Leave a comment:


  • dlr
    replied
    Originally posted by dkalsi View Post
    Here is the response with the "Polar Repose" off with all driver measurements from tweeter axis (i.e., 1a and 1b from above - modified for minimum phase)
    You use one of the files for the tweeter with the tail added and use that one only afterwards. If you follow the directions shown in the Acoustic Offset tab of WinPCD, you should be able to get a near identical match with the summed measurement of the two raw driver minimum-phase measurements (1a & 1b). After that measured results should correspond very closely to predicted unless there is an issue with the measuement setup and/or software.

    Show the system tab of WinPCD to prevent any confusion and help others to help you. I'd like to see that vs. the line-by-line listing on page 1.

    dlr
    Last edited by dlr; 10-09-2017, 09:50 AM. Reason: The info I asked for is on page 1, I overlooked that.

    Leave a comment:


  • EarlK
    replied
    Hi Dhar,

    Point #3: What happens ( what offset do you get ) when you don't add any low-pass tail to the tweeter FR ? ( I believe that's the way most people operate )

    Also, I think your Swan 2.2a example was designed from a listening axis mid-way between the center of tweeter & center woofer cone ( that's were I get the best summation at crossover using your files from post #33 > this is true using both WinPCD & XSim ) .

    .

    Leave a comment:


  • dkalsi
    replied
    Here is the response with the "Polar Repose" off with all driver measurements from tweeter axis (i.e., 1a and 1b from above - modified for minimum phase): Click image for larger version

Name:	win polar response off.jpg
Views:	1
Size:	258.0 KB
ID:	1348299

    Leave a comment:


  • dkalsi
    replied
    Okay - back with some questions

    This is the experiment I am running:

    1) First, the following measurements were taken:
    a) Tweeter measurements without crossover at 30" on tweeter axis
    b) Woofer measurements without crossover at 30" on tweeter axis
    c) Tweeter + Woofer, wired in parallel, measurements without crossover at 30" on tweeter axis
    d) Woofer measurements without crossover at 30" on Woofer axis
    e) Tweeter measurements with crossover at 30" on tweeter axis
    f) Woofer measurements with crossover at 30" on tweeter axis
    g) Tweeter + Woofer, with crossover at 30" on tweeter axis

    2) All single driver measurements were exported into JB's Frequency Response Blender to extract minimum phase. New files with minimum phase were created

    3) Measurements "1c", along with "1a + 1b" - modified to include mimimum phase, were imported into WinPCD to determine acoustic offset
    ISSUE #1 - I think the selection of "tails" definitely impacts results - even if its way outside of the overlap region. What seemed to have the most impact on my result was whether I am using a 0db, 6db, 12db, 24db, 48db - "low-pass" tail for the tweeters upper-end response. (acoustic offset was calculated to be +0.0081, +0.0062, +0.0040, +0.0010, -0.0060, respectively)

    4) All crossover components were measured utilizing DATS, and the exact values were included in WinPCD crossover section.

    5) Zma files were created using DATS and included in WinPCD

    6) First, files 1a and 1b (both modified for minimum phase) were uploaded in WinPCD

    7) The resulting response from Step 6 was compared to the measured response in Step 1g (utilizing the "import overlay" functionality)
    ISSUE # 2 - My mind is telling me, that if I measured every crossover component, correctly measured my FRD and ZMA files, the simulated response would match identically to measured response. Maximum variance observed was at 1,720HZ - a delta of 2.3db between measured and simulated.

    8) Next, files 1a and 1d (both modified for minimum phase) were uploaded in WinPCD

    9) The resulting response from Step 8 was compared to the measured response in Step 1g (utilizing the "import overlay" functionality)
    ISSUE # 3 - Much closer than the results from Step 7 - BUT still not identical. When I compared 1f with simulated predictions, it matched perfectly. When I compared 1e with simulated predictions, the response did not match perfectly.Maximum variance observed was at 1,720HZ - a delta of 1.8db between measured and simulated.

    So - soliciting advice again.

    For Issue # 1 - How to you select tweeter LP tail? Positive woofer offset??!?!

    For Issue # 2 - Shouldn't simulated response exactly meet measured response if you know all unknowns (e.g., various crossover component values, etc) - as indicated in Step 9, my woofer simulated response exactly matched my measured response - however, this was not the case for the tweeter section

    For Issue # 3- I don't know if it helps answer my questions from an earlier post. My finding indicated that I get better results when I measure each driver on its respective axis when the idea is to minimum phase files and specifying the baffle layout and the drivers radiating diameter as oppose to utilizing the woofer's response that was captured on the tweeter axis for determining the acoustic offset. It is very subtle, but I thought I share my experience none the less.

    Ultimately - the best response I get is if I turn off "Relative Polar Locations" and simply use the files 1a + 1b ==> this is when measured response matches pretty close to simulated response. I think less than 1db variance

    Step 7 Results: ​ Click image for larger version

Name:	Step 7 Results.jpg
Views:	1
Size:	258.1 KB
ID:	1348295


    Step 9 Results ​ Click image for larger version

Name:	Step 9 Results.jpg
Views:	1
Size:	286.9 KB
ID:	1348296

    And with polar response off: Click image for larger version

Name:	win polar response off.jpg
Views:	0
Size:	0
ID:	1348297
    Attached Files

    Leave a comment:


  • dlr
    replied
    And the problem is......you have to invert the tweeter connection. It's a second order system. You must have inverted it in PCD as required for second order, but didn't in WinPCD. I've got a very sore forehead right now. This is with the Polar setting Inactive in WinPCD.

    This reminded me to have the polarity setting color change added to the System graph. I hope to have an updated version ready soon. The big change is it will now be "responsive UI" when running the polar plots. Currently that is in the same thread as the UI so the whole program is blocked during that time. It will be in its own thread in the new version so the UI will respond.

    dlr

    Leave a comment:


  • Pete Schumacher
    replied
    Originally posted by killa View Post

    I prefer his Frequency Response Blender. It lets you to add tails to the response to get more accuracy when you extract minimum phase. It also lets you easily and accurately combine near field and far field response including baffle affects.
    You can do the same thing with Response Modeler using the EQ settings and adding LP or HP filters to generate the tails.

    Leave a comment:


  • dlr
    replied
    I'd have to boot my old XP machine to run PCD (too late tonight), but I put the data into WinPCD. I get what you show only if I leave the Polar setting inactive, meaning the offsets are ignored. When I do set Polar active, I get something closer to what you show, but still not the same as what you show for PCD. There are some peaks/dips in the crossover area.

    Compare the phase of both woofer and tweeter with the crossover in place and see how close the two are. I haven't had any feedback of major differences between PCD and WinPCD when the settings and crossover values are the same.

    Edit:
    I'm finishing some updates for behind-the-scene changes primarily and was comparing the new one to two previous versions (1.510 and 1.511). The easy way was to save the project file to import into the other versions. The summed response was different. Not seeing any obvious reason, I imported the saved project into my new version. It now is the same as others. It appears that when files are manually loaded and data set manually, the summed response is not correct. I'll have to look into it further, but for now try saving/re-loading the project to see what the difference is. The results I see are that with polar inactive, the sum is close to the PCD sum that you show. With it active, as it should be if you are using minimum-phase files, the response has some peaks/dips in the crossover area, not the large peak shown.

    I have to investigate this.

    dlr

    Leave a comment:


  • killa
    replied

    You can generate the minimum phase versions of the files by using Jeff's Response Modeler spreadsheet.
    I prefer his Frequency Response Blender. It lets you to add tails to the response to get more accuracy when you extract minimum phase. It also lets you easily and accurately combine near field and far field response including baffle affects.

    Leave a comment:

Working...
X