Redoing an old build

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
One of my earlier builds has come back to the roost. It's a 4th order BP design build around a driver from Madisound called the INF-10. It was a surplus driver, looks like from an Infinity subwoofer that had decent published specs for vented alignments, but when I tried using one in a vented alignment, it behaved VERY badly, exhibiting a loud cracking noise if even slightly over-driven. I ended up using it in a 4th order BP design, which maintained quite a bit more control over excursion at low frequencies. The design is document here - The Subwoofer DIY Page v1.1 - Projects : An INF10 Bandpass Subwoofer

Years ago I'd lent it to Dad for his HT system, but recently he upgraded it with a new amp that does not provide an amplified subwoofer output, so I retrieved it to do some tests on it.

The interesting thing here is that the measured performance of the build doesn't match the sim, and when I tried re-simming it in Hornresp instead, it STILL doesn't match. Fb is lower than predicted, the passband is wider than predicted, and the measured impedance is much more damped than the Hornresp sim (in Loudspeaker Wizard) suggests.

The thing is, Hornresp does not appear to allow the user to export the resulting impedance curve when filling in used. The export routine is only available after using the "Calculate" option, and that option apparently does not take filling into consideration. So, I'm going to take a "Mountain to Mohammed" approach and remove all of the filling from the box, and then compare the impedance curves to see how they differ, and see what else I can do to bring the build more in alignment with the sim (or vice-versa).
 

Attachments

  • 20180128-INF10-BP.png
    20180128-INF10-BP.png
    88.3 KB · Views: 516
Well, here's the unstuffed INF10-BP's impedance curve compared to the Hornresp sim (which I've tweaked a little to reflect the same Fb). That upper impedance peak was way more damped than expected. I'm not quite sure what could be causing this. it kind of suggests that Fb is too high for this build, when in fact it's a bit too low. I'm modifying this build to increase it by at least 6~8 Hz.
 

Attachments

  • 20180131-INF10-BP (no stuffing).png
    20180131-INF10-BP (no stuffing).png
    48.1 KB · Views: 466
Well, here's the unstuffed INF10-BP's impedance curve compared to the Hornresp sim (which I've tweaked a little to reflect the same Fb). That upper impedance peak was way more damped than expected. I'm not quite sure what could be causing this. it kind of suggests that Fb is too high for this build, when in fact it's a bit too low. I'm modifying this build to increase it by at least 6~8 Hz.

I get this with my simulator (note scale) using Ql=15, Qb=10 and Qp=20 (just random selections). F3l ~33, F3h~92Hz.

Could be a leaky box?
 

Attachments

  • inf10 impedance.PNG
    inf10 impedance.PNG
    29.9 KB · Views: 445
I get this with my simulator (note scale) using Ql=15, Qb=10 and Qp=20 (just random selections). F3l ~33, F3h~92Hz.

Could be a leaky box?

I thought that might be a cause, but if the box was leaky, I wouldn't expect that the minimum point between the two impedance peaks wouldn't be so low. And then there's the extra extension of the passband above 100 Hz that's not predicted by the sim.
 
Administrator
Joined 2004
Paid Member
If in doubt, test. Don't guess.

Can you measure the T/S parameters of that raw driver. I suspect you will find your answers there.

Eons ago when I was designing speaker boxes and systems I learned I had to measure the drivers first. Then plug those values into a program I wrote. It didn't take long before the results matched the predictions very closely.

-Chris
 
Can you measure the T/S parameters of that raw driver. I suspect you will find your answers there.

I did that and there was a slight parameter shift. I adjusted the Hornresp sim accordingly. What I've presented here is the adjusted sim.

The only observable difference between the sim and the build that I can see is that the sim assumes straight vents that are mounted to the outside of the front volume, while the vents are actually flared vents that are mounted inside of the volume and terminate on the front. This box would built back when I was still using a lumped-mass model to simulate the results. If I was using Hornresp back then, I'd have a spent a bit more effort to ensure the build was as close to the sim as possible and likely used a shelf vent instead. I've adjusted the front volume in the sim to reflect the volume displaced by the vents, and also adjusted the vent length in the sim to achieve the measured Fb (that's another big difference - the vents are much shorter than expected for the measured Fb). And the upper impedance peak is still quite a bit off as is the expected cutoff at the upper end of the passband.
 
Administrator
Joined 2004
Paid Member
Hi Brian,
Something is off in the program then. I know for sure the math works. I wrote the program into a TI-58C when they were new. I worked off the raw equations and got my T/S parameters using the added weight method and the changed resonance in box method to make sure the numbers made sense. I even corrected for elevation for the speed of sound.

So either the math for the T/S parameters is off, or the box calculations are off. Have you tried another program? A freebie one might be worth more than an expensive one. When I was doing this, there weren't any programs you could buy except for one written in basic. The compiler in my real IBM PC wouldn't run it. Later I wrote the same program in basic for the PC and got the same numbers (but it did make graphs of expected response, that was cool). The PC was also new (5 grand back then, a "luggable").

-Chris
 
Try sealing the vents and measuring impedance. There is a way to back out tuning frequency with the vented and sealed peak frequencies. Trying to measure tuning frequency with the impedance minima doesn't always work, especially in lossy boxes or high inductance drivers.

When putting a driver in a box vs free air, there can be parameter shifts, slight changes in mass, also Qms can change appreciably with frequency. You might learn something from the sealed measurement, extract the parameters and compare them to theory with sealed 20L boxes on either side, measure Ql, etc... I'm thinking you should get a peak near 78Hz.
 
So, I basically went back to basics, and carefully measured the box again, measured the impedance of the driver in the sealed section (with the vented section open) and compared with the Hornresp sim, measured the driver's t/s parameters and made further adjustments until I got a pretty good correlation between the Fb predicted by Hornresp and the measured Fb. I also shortened the vents a bit to bring Fb of the vented section up a bit (it was too low). Then I put the vents back in, closed up the vented section, and measured the FR again, and compared it to the adjusted Hornresp sim.

This time I'm quite a bit closer, in that the location of the slopes on either side of the passband appear to be correct. The slope at the lower end of the passband is a bit steeper, but I put that down to box losses. But what happened in the passband? The measurements suggest up to 8.5dB difference between what Hornresp predicts and what I measured. I expected 2~3dB due to box losses, not 8.5dB! And the tilt of the passband is different too. And of course the big differences in the predicted vs. measured impedance curves still exist.

Something of interest has taken place since I shortened the vents a little. The big peak that was apparent around 400 Hz in the original build has diminished significantly in magnitude and there's a pretty large dip just a bit higher up in frequency. This leads me to suspect that the resonance at just above 400 Hz can be negated by selecting the appropriate combination of box dimensions and vent dimensions. Who knows - it might be possible to null it out entirely.
 

Attachments

  • 20180202-INF10-BP (no stuffing).png
    20180202-INF10-BP (no stuffing).png
    100.5 KB · Views: 391
  • 20180202-INF10-BP (no stuffing)-IMP.png
    20180202-INF10-BP (no stuffing)-IMP.png
    52.6 KB · Views: 383
This time I'm quite a bit closer, in that the location of the slopes on either side of the passband appear to be correct. The slope at the lower end of the passband is a bit steeper, but I put that down to box losses. But what happened in the passband? The measurements suggest up to 8.5dB difference between what Hornresp predicts and what I measured. I expected 2~3dB due to box losses, not 8.5dB!

Ok, I just realized that I hadn't adjusted the relative level of the Hornresp sim properly. Looks like the difference is more like 6dB at its peak. Still significant...
 

Attachments

  • 20180202-INF10-BP (no stuffing).png
    20180202-INF10-BP (no stuffing).png
    100.5 KB · Views: 94
Administrator
Joined 2004
Paid Member
Hi Brian,
I wonder if the 400 Hz peak isn't coming from the back of the woofer and angled just right to come out the port.

I like your measured response better than the sim. It may not reach the predicted SPL, but trading that for some section of flat-ish response is fine.

-Chris
 
Here's a comparison between measured FR and a "Lossy Le" version of the sim. There is much better correlation between the sim and the measurements at the low frequencies, but it goes out of whack above 60 Hz and the upper cutoff slope does not match.

Hi Brian,

Would it be possible to post a screenprint of the input parameters window or alternatively a copy of the exported record file, for the simulation? Thanks.

Kind regards,

David
 
Hi Brian,

Would it be possible to post a screenprint of the input parameters window or alternatively a copy of the exported record file, for the simulation? Thanks.

Kind regards,

David

Here ya go. This is the current tweaked model that I'm working with. I've also included an image of the vented section with the access panel removed. This shows the only significant difference between the sim and the build, as the sim assumes that the vents are external to the vented section, when they're actually internal (and yes, the adjusted sim takes into consideration the volume displaced by the vents).
 

Attachments

  • 2018-02-03.png
    2018-02-03.png
    15.5 KB · Views: 230
  • INF10-BP.txt
    956 bytes · Views: 85
  • 20180203_070321.jpg
    20180203_070321.jpg
    179.4 KB · Views: 215
Here ya go.

An adjusted sim using (from your screenshot)
Fs 20.0 Hz
Qts 0.317
Qes 0.332
Vas 141.78 Liters
Re 6.89 Ohms
Dia 20.50 cm

Vr=21.3

Gives modified (caused by sealed side) parameters of:
Fs 55.3
Qts 0.824
Qes 0.919
Vas 18.518

Then using (guesses)
Vf=18.7
Fb=57
Ql=15
Qb=Qp=30

Gives this:
Impedance peaks at ~35 and ~90 just over 30ohms
F3l=33
F3h=102
slight droop in the passband
 

Attachments

  • inf10 impedance.PNG
    inf10 impedance.PNG
    31 KB · Views: 202
Here ya go. This is the current tweaked model that I'm working with. I've also included an image of the vented section with the access panel removed. This shows the only significant difference between the sim and the build, as the sim assumes that the vents are external to the vented section, when they're actually internal (and yes, the adjusted sim takes into consideration the volume displaced by the vents).

Sigh, I found a minor error in the model, see corrected information below and attached....
 

Attachments

  • 20180203-INF10-BP (no stuffing)-FR.png
    20180203-INF10-BP (no stuffing)-FR.png
    39.9 KB · Views: 187
  • 20180203-INF10-BP (no stuffing)-IMP.png
    20180203-INF10-BP (no stuffing)-IMP.png
    53.4 KB · Views: 91
  • 2018-02-03 (1).png
    2018-02-03 (1).png
    15.5 KB · Views: 103
  • INF10-BP.txt
    956 bytes · Views: 68
An adjusted sim using (from your screenshot)
Fs 20.0 Hz
Qts 0.317
Qes 0.332
Vas 141.78 Liters
Re 6.89 Ohms
Dia 20.50 cm

Vr=21.3

Gives modified (caused by sealed side) parameters of:
Fs 55.3
Qts 0.824
Qes 0.919
Vas 18.518

Fb is actually 54 Hz when the panel is left off the vented side, so that's pretty close. I basically took the panel off, took the vents out, laid the cabinet on its side (so the holes for the vents were blocked), then measured the impedance curve and compared it to a Hornresp model of a sealed cabinet loaded with a short "horn" with the same CSA and depth of the vented section.


Then using (guesses)
Vf=18.7
Fb=57
Ql=15
Qb=Qp=30

Gives this:
Impedance peaks at ~35 and ~90 just over 30ohms
F3l=33
F3h=102
slight droop in the passband

Vf looks a little small there (I worked it out to be 22 l), and Fb a bit lower than measured. Depending on whether you use the minimum impedance or the zero-phase crossing to determine Fb, it lies between 59 Hz and 62 Hz.
 

Attachments

  • 20180203-INF10-BP (open)-IMP.png
    20180203-INF10-BP (open)-IMP.png
    43.4 KB · Views: 94
f Port Zin
5.0 46.1 6.9
5.2 46.7 6.9
5.4 47.4 7.0
5.5 48.0 7.0
5.7 48.6 7.0
5.9 49.2 7.0
6.2 49.8 7.0
6.4 50.4 7.0
6.6 51.0 7.0
6.8 51.7 7.0
7.1 52.3 7.0
7.3 52.9 7.0
7.6 53.5 7.0
7.8 54.1 7.0
8.1 54.8 7.0
8.4 55.4 7.1
8.7 56.0 7.1
9.0 56.7 7.1
9.3 57.3 7.1
9.7 57.9 7.1
10.0 58.6 7.1
10.4 59.2 7.1
10.7 59.8 7.2
11.1 60.5 7.2
11.5 61.1 7.2
11.9 61.8 7.2
12.3 62.4 7.3
12.7 63.1 7.3
13.2 63.7 7.3
13.7 64.4 7.4
14.1 65.1 7.4
14.6 65.7 7.5
15.2 66.4 7.5
15.7 67.1 7.6
16.2 67.8 7.6
16.8 68.5 7.7
17.4 69.2 7.8
18.0 69.9 7.9
18.7 70.6 7.9
19.3 71.3 8.1
20.0 72.1 8.2
20.7 72.8 8.3
21.4 73.6 8.5
22.2 74.3 8.7
23.0 75.1 8.9
23.8 75.8 9.2
24.6 76.6 9.5
25.5 77.4 9.8
26.4 78.2 10.3
27.3 79.0 10.8
28.3 79.8 11.5
29.3 80.5 12.4
30.3 81.3 13.5
31.4 82.1 15.0
32.5 82.8 16.9
33.6 83.5 19.5
34.8 84.2 22.8
36.1 84.8 26.2
37.3 85.3 27.9
38.6 85.8 26.2
40.0 86.3 22.4
41.4 86.6 18.7
42.9 86.9 15.8
44.4 87.2 13.7
45.9 87.3 12.2
47.6 87.5 11.1
49.2 87.6 10.3
51.0 87.7 9.8
52.8 87.8 9.5
54.6 87.8 9.3
56.6 87.9 9.2
58.6 87.9 9.2
60.6 88.0 9.4
62.8 88.0 9.7
65.0 88.1 10.1
67.3 88.1 10.6
69.6 88.0 11.4
72.1 88.0 12.5
74.6 87.8 14.0
77.3 87.7 16.0
80.0 87.4 18.7
82.8 87.0 21.7
85.7 86.6 23.8
88.8 86.0 23.5
91.9 85.4 20.9
95.1 84.7 17.7
98.5 84.0 15.1
102.0 83.2 13.1
105.6 82.4 11.6
109.3 81.6 10.6
113.1 80.7 9.7
117.1 79.9 9.1
121.3 79.1 8.6
125.5 78.2 8.3
130.0 77.4 8.0
134.5 76.6 7.8
139.3 75.8 7.6
144.2 75.0 7.4
149.3 74.3 7.3
154.5 73.5 7.2
160.0 72.7 7.2
165.6 72.0 7.1
171.5 71.3 7.1
177.5 70.5 7.1
183.8 69.8 7.1
190.3 69.1 7.1
197.0 68.4 7.1
203.9 67.7 7.2
211.1 67.0 7.2
218.6 66.3 7.3
226.3 65.7 7.3
234.3 65.0 7.4
242.5 64.3 7.5
251.1 63.7 7.6
259.9 63.0 7.6
269.1 62.4 7.7
278.6 61.8 7.8
288.4 61.1 8.0
298.6 60.5 8.1
309.1 59.9 8.2
320.0 59.2 8.4
331.3 58.6 8.5
343.0 58.0 8.7
355.1 57.4 8.8
 
see corrected information below and attached....

Hi Brian,

Thanks for the information. Because the port tube is simulated using a horn segment rather than Ap and Lpt, no end correction is being added to the internal end. Attachment 1 may be a more accurate model, with segment 1 accounting for the internal end correction (0.1952 * Pi * Rp, where Rp is the port tube radius). The predicted system resonance frequency becomes 61.49 Hz.

Kind regards,

David
 

Attachments

  • Attach_1.png
    Attach_1.png
    53.8 KB · Views: 87
  • Attach_2.png
    Attach_2.png
    61.8 KB · Views: 101
Hi Brian,

Thanks for the information. Because the port tube is simulated using a horn segment rather than Ap and Lpt, no end correction is being added to the internal end. Attachment 1 may be a more accurate model, with segment 1 accounting for the internal end correction (0.1952 * Pi * Rp, where Rp is the port tube radius). The predicted system resonance frequency becomes 61.49 Hz.

Kind regards,

David

Hi David, I'd actually accounted for the end-correction in my sim. The vents are flared at both ends, and inside the box they terminate pretty close to the access panel and the sides (which are within 1 diameter of the vent opening), so I thought that the best way to work out the effective vent length was to measure Fb, and then work backwards to see what combination of S1-S2 and effective vent length (L34) would give a similar Fb. The actual "length" of the vents (minus part of the flare) is about 20cm each.

The build is actually in pieces at the moment, as I'm repainting the box (it is about 14 years old and the white paint had yellowed). but when I have it back in one piece again, I'm going to rerun the test with even shorter vents to see if I can at least flatten the passband. I would really like to know however where the up to 6dB of theoretical output in the passband has disappeared to.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.