Disturbing time alignment results

User avatar
Dahlberg
Posts: 90
Joined: 19 Oct 2019 23:46

Re: Disturbing time alignment results

Post by Dahlberg » 17 Nov 2020 14:23

UliBru wrote:
17 Nov 2020 13:26
If you have prepared the XO filter with EQ you can send me them too.
And then the measurements as described above with these XOs applied.
PM sent
Best regards. Bengt
User avatar
Dahlberg
Posts: 90
Joined: 19 Oct 2019 23:46

Re: Disturbing time alignment results

Post by Dahlberg » 17 Nov 2020 16:02

This is a .pdf describing the filter construction.

http://www.dahlbergaudiodesign.se/filter+eq.pdf

The description is for a NT filter, I'm using the UB jpol11 1:ord...
Best regards. Bengt
User avatar
Dahlberg
Posts: 90
Joined: 19 Oct 2019 23:46

Re: Disturbing time alignment results

Post by Dahlberg » 20 Nov 2020 10:56

I got some excellent support from "Uli" regarding time alignment using sine wave convolution and I added it all to my "to remember document".
Goes like this:
Time alignment with sine wave convolution.
Measure pulse responses, all with the same full range sweep signal but with applied XO/EQ , zero rotation, for:
- tweeter
- tweeter + midrange
- tweeter + bass
- tweeter + midrange + bass
To exclude drivers, simply run the sweeps and switch off/on amplifiers for bass + midrange.
Make sure the measurements are reliable by inspecting the pulse response for each channel, you may need to do a couple of sweeps with different sweep length. Pulse amplitude is supposed to be equal on both sides. If it looks strange, measure again. There can also be some faulty equipment involved.

After measurements:
Left side:
load tweeter pulse in curve1.
load tweeter + midrange pulse in curve2.
subtraction curve2 - curve1 into curve3.
generate sinewave 350 Hz in curve4.
convolve curve1 + 4 into curve5.
convolve curve3 + 4 into curve6.
zoom into centre of convolution result sinewave, use markers to find neighbour peaks of curve5 + 6. Get the peak position (right side of rime chart).

Result: 50 samples difference (example).
delay curve6 (midrange) by +50 samples -> sinewaves are aligned.
IMPORTANT! now zoom into beginning of sinewaves. Curves 5 + 6 are first in opposite but aligned = ok!
result: midrange xo needs rotation by + 50 samples.

Do the same for the right side. You may not get the exact same results even if the room is symmetrical and speakers are identical.

Proof :
rotate curve3 (midrange) by + 50 samples.
add curve1 + curve3 into curve 4.
deselect all curves except 2 and 4 in frequency chart.
result curve4 has higher amplitude in 350 Hz range -> tweeter + midrange are in phase.

Part 2:
Rotate XO2L48 by +50 and XO2R48 by +50 samples in the original XO filter and measure tweeter + midrange with these crossovers .The same procedure is repeated for the next XO position starting with tweeter + midrange in position 1 sub + tweeter in position 2. Bass + tweeter is already recorded. Position 6 is used for changes with this approach.
Thank you so very much Uli. :) This was very helpful and I'm pretty sure i would have given up before finding this out myself :oops:
I belive I got most of it even if there can be bit of a struggle to get both the transition area and steady state to do what's expected.
Best regards. Bengt
Post Reply