I recently ran an experiment with a small box and a midrange driver. The goal was to test the effects of different stuffing materials on the response. I shot these sweeps on REW and kept all the variables the same besides the materials used for stuffing. I'm not really sure how I should be interpreting the phase response graph and what is ideal.
Test 1: No stuffing, empty box
Both the frequency response and phase response are jagged. The pattern is broken up around 13khz, I'm assuming that is some kind of driver breakup causing a spike in FR.
Test 2: lightly stuffed with polyfill
In this test, the phase response seems to have flattened out significantly. My assumption is that this is a good thing since that means the different frequencies are more or less aligned closely?
Test 3: Tightly stuffed with polyfill
With the tightly stuffed polyfill, the phase response has gotten very smooth and linear. The FR has also smoothed out quite nicely.
So my question is, what is an ideal phase response? Subjectively speaking, what should a good phase response sound like?
Test 1: No stuffing, empty box
Both the frequency response and phase response are jagged. The pattern is broken up around 13khz, I'm assuming that is some kind of driver breakup causing a spike in FR.
Test 2: lightly stuffed with polyfill
In this test, the phase response seems to have flattened out significantly. My assumption is that this is a good thing since that means the different frequencies are more or less aligned closely?
Test 3: Tightly stuffed with polyfill
With the tightly stuffed polyfill, the phase response has gotten very smooth and linear. The FR has also smoothed out quite nicely.
So my question is, what is an ideal phase response? Subjectively speaking, what should a good phase response sound like?
Linear phase. Not flat, but a straight line tilted down, where the negative slope is the time delay.
The phase traces look to have a different timing reference for each of them.
Do you timings relative to loopback, or acoustic timing reference?
Sure, a straight line tilted down IF looking at phase vs frequency when frequency is on a linear scale (which is not so common i think).
If the excess constant delay, time-of-flight, etc) is removed from such a measurement, then the tilt will go away, and the phase trace will then lay straight flat at zero degrees..
And lay flat at zero on either the linear scale, or the more common log scale.
So once the excess constant delay is removed...which it should be to make any sense of phase traces ...linear phase is a flat line at zero degrees.
I've come to think of linear phase as simply a special case of minimum phase....the special case being it's flat at zero (again, when constant delay is removed as a given)
Cool. What were the relative to Loopback times on each of them? Shown on the measurement tabs.
Sorry I'm a little out of my depth here, but the measurements pane shows a timing offset of 0. Not sure if that helps.
Cool again....glad to try to help.
The experiment you've done might prove very interesting and illuminate how our measurement programs work, if my hunch as to what's going on is correct.
The delay i'm asking about is highlighted in yellow on this snip..
I'm guessing/hoping it is different on your 3 measurements.
The experiment you've done might prove very interesting and illuminate how our measurement programs work, if my hunch as to what's going on is correct.
The delay i'm asking about is highlighted in yellow on this snip..
I'm guessing/hoping it is different on your 3 measurements.
All of them? With zero timing offset in the measurement pane like shown in your post #8?
The graph you show here in #10 with 0.0 delay vs loopback....what is the timing offset in the measurement pane?
Timing offset and delay vs loopback can't both be zero......that would mean there is zero distance between mic and speaker.
If both 0 timing offset and 0 delay vs loopback are what you have, something is messed up.
The graph you show here in #10 with 0.0 delay vs loopback....what is the timing offset in the measurement pane?
Timing offset and delay vs loopback can't both be zero......that would mean there is zero distance between mic and speaker.
If both 0 timing offset and 0 delay vs loopback are what you have, something is messed up.
Yeah, they all note 0ms delay in both the measurement pane and the notes. It does make sense that there is a problem though because a delay of 0 would mean the sound reaches the microphone instantaneously when the distance was actually around 50cm. Unfortunately, I don't have the parts with me right now to reshoot the test and see what might be the problem, but I would still like to hear the theory that you mentioned before.
Have you connected one of the soundcards output to the right input for proper loopback?
I usually make one measurement, then estime IR Delay, and then click "shift and update IR Delay". As long as i don't move the mic or driver, all the following measurements should have correct timing.
I usually make one measurement, then estime IR Delay, and then click "shift and update IR Delay". As long as i don't move the mic or driver, all the following measurements should have correct timing.
It probably makes sense to assure timing is right before i go off into what i think might be contributing to the different phase traces. Don't want to be barking up a tree Lol.
Here's another method to keep timings consistent, in addition to the one skogs offered.
I like to make a first measurement with 'Use loopback as timing reference' and Timing offset: set to 0.00
Then, whatever Delay is reported in the Measurements Tab, insert that exact value into the Timing offset:
Run measurement again, and verify that Delay in the Measurement Tab is 0.00 or damn close.
(I like this method a little more than 'estimate IR shift, etc.')
Like skogs says, as long as nothing moves, it's all good.
I'm not an expert at phase, but as I understand it, the goal is to minimize the phase rotation to less than 90 degrees per octave.
This basically means that:
1) you might consider FIR filters and something like rephase
2) you might consider first order filters (they have 90 degrees of phase shift per octave)
I personally prefer hybrid solutions. I like the LeCleach solution; basically he used third order filters but he shifted the drivers forward to offset the delay. For instance, a 3rd order filter introduces 270 degrees of delay. So LeCleach would move the midrange forward of the tweeter, offseting the electrical delay introduced by the crossover.
Also, I tend to get mixed up on the degrees, so you'll want to mess around with this in vituixcad. And all of this is complicated by the impedance of the drivers and their frequency response.
This basically means that:
1) you might consider FIR filters and something like rephase
2) you might consider first order filters (they have 90 degrees of phase shift per octave)
I personally prefer hybrid solutions. I like the LeCleach solution; basically he used third order filters but he shifted the drivers forward to offset the delay. For instance, a 3rd order filter introduces 270 degrees of delay. So LeCleach would move the midrange forward of the tweeter, offseting the electrical delay introduced by the crossover.
Also, I tend to get mixed up on the degrees, so you'll want to mess around with this in vituixcad. And all of this is complicated by the impedance of the drivers and their frequency response.
In that case, I'll redo the measurements tomorrow to see if I can fix the problem, and hopefully, we can make some interesting insights.
Not having access to this equipment, I found the graphs at the beginning very interesting ( many thanks Michael 1 103 ) I presume that the zigzag in the phase response ( 1st graph ) is caused by the internal reflection superimposed on the cone output, the shift around 170 ( 3rd graph ) is the drive fc, the dip at 1,700 possibly due to internal standing waves and peek at 11,000 from edge diffraction or cone beaming?
It's interesting how the fc seems to rise in the tightly stuffed box, possibly overstuffed, reducing the volume.
It's interesting how the fc seems to rise in the tightly stuffed box, possibly overstuffed, reducing the volume.
I definitely agree with you about the internal reflections superimposing on the driver's output. I probably should have mentioned that this is a planar driver (GRS PT6825-8) which I would assume makes it more sensitive to that kind of interference due to the lightweight nature of the driver.
In an application scenario, the dip at 1700 is probably the only issue thing holding this back from being a really good midrange. A lot of speaker-building literature (mainly guides on YouTube) seems to think that adding foam/stuffing materials is enough to eliminate standing waves, but maybe that isn't the case. Maybe a differently shaped box would be better for eliminating standing waves? something trapezoid-shaped or triangular.