Keith_W
Master Contributor
That looks nice. Incidentally, that weird tweeter roll-off has disappeared. What happened?
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Can anyone help with Alignment?
- Thread starter Robbie010
- Start date
- Thread Starter
- #162
That is a mystery......
One thing I'm a little confused about...... when aligning the drivers, if using the tweeter as the reference would you not first align the midrange to the tweeter and then align the woofer to the newly combined tweeter/midrange?
That is what I did here:
Not sure what the blip in the Phase at 4khz is?
One thing I'm a little confused about...... when aligning the drivers, if using the tweeter as the reference would you not first align the midrange to the tweeter and then align the woofer to the newly combined tweeter/midrange?
That is what I did here:
Not sure what the blip in the Phase at 4khz is?
3ll3d00d
Senior Member
- Joined
- Aug 31, 2019
- Messages
- 416
- Likes
- 303
The description reads as if you have added the mid twice (ie it is the sum of TM + MW = WMMT), is it what you did?
Keith_W
Master Contributor
Vance Dickason (that book again) explains it very well. I have created a simulated driver for you to illustrate this point:
This is a 4th order Linkwitz-Riley bandpass with corner frequencies of 100Hz - 2kHz, simulating a woofer in a 2-way system. The upper graph shows the amplitude response, and the lower graph shows the group delay. Notice how the GD is lowest at the upper end of the bandpass, and highest at the lower end. This is typical minimum-phase driver behaviour. All your drivers will behave something like this, with different amounts of GD obviously.
Think of the Group Delay as the "Zero Delay Plane" (ZDP) - i.e. the acoustic centre of the driver. This becomes more obvious if we flip the GD trace on its side, and include a driver in the illustration:
As you can see, the ZDP varies with frequency. So you have to think about what you are aligning when you time align drivers. What you really want to do is hitch the ZDP's together - the lower part of the tweeter ZDP gets hitched to the upper part of the midrange ZDP. If you do this, there will be no phase cancellation.
Remember that the GD trace is another view of phase. In fact, the GD is the negative derivative of a system's phase response with respect to angular frequency. Don't worry about the exact definition, all you need to know is that REW takes the phase, applies some maths to it, and shows you a GD plot
Or put another way, what you want to do is align the phase of two drivers so that they do not cancel when they sum. Drivers can be several cycles misaligned in time. This is why you align the impulses first before you adjust to remove phase cancellation.
Now to your question. First, we align the impulse of the mids to the tweeter, then we align the phase. You seem to understand the rationale for this. For the woofer, it does not matter if we align the woofer to the mid or to the tweeter. Either way, the "extra" delay cycles, if they exist, are removed. After all, this is an initial alignment, not the final alignment! Then we align the phase of the woofer to the phase of the mid. This is the final alignment. When it is properly done, you should see no phase cancellation at any of the XO points, and you will see a smoothly rising GD plot.
I hope this makes sense.
Discontinuities in the phase graph are usually regions of excess phase. If you want to confirm, do a phase extraction ("Excess Phase Copy") and view the Group Delay graph. Minimum-phase regions are flat areas of the graph, and excess phase look like spikes. Don't worry too much about EP for now, that's an advanced DSP topic, and not something you can correct with the minimum-phase filters you are using. Something you can revisit in the future.
- Thread Starter
- #165
So here is what I did:
Open Alignment Tool and chose Tweeter & Midrange
Apply delay to Midrange ( -7.00ms)
Click "Aligned Copy" for both
Open Arithmetic, chose Aligned Copies and Sum A+B:
However, when I repeat process with the Woofer and Midrange it doesn't quite work........ the best results is arrived at by speeding up the woofer by 0.50ms, not delaying it....... the problem is, I don't have the option to do this in my DSP, I can only apply delays......
If I view the overlayed individual impulse responses with a timing ref, that seems to suggest the the woofer is indeed the earliest impulse and the start of the rise between the woofer and midrange is indeed 0.50ms. So, does that mean that I need to increase the woofer delay by 0.50ms to match the newly aligned tweeter and midrange?
This what the impulse looks like with 0.200ms delay applied to the Midrange and 0.650ms delay applied to the woofer:
This is the summed response in all variations: WM, MT, WMMT, W/MT & WM/T
And this is the Phase Graph of the better responses (W/MT & WM/T)
Phew!
Note that if you add/change filters, it will change the timing once again, because of the group delay of the filters.
My experience is that the REW alignment tool does not work for LF / Woofers or sub woofers (works fine for Mid to Tweeter). Like you I get nonsensical results like delaying the woofer or inverting drivers when I know that is not right. I believe the issue is that in room LF timing measurements are worse than nothing due to reflections and gating and FFT limitations at LF which basically throws out 95% of the information leaving you with nothing of value. I am not sure what the real experts do but I either "go outside" to get some better LF measurements or use near field woofer and mid range measurements and look at the timing and use a tape measure between the drivers and the LP and calculate the delays that way. Of course that leaves you guessing on where exactly is the "acoustic center" but it will get you close. The only good news with all of this is that woofer timing is not as critical as mid/ tweeter timing.
- Thread Starter
- #168
Yep, I've really struggled with this today, even with all the great advice.
The main problem I've come up against is that, as you say, the timings don't seem to marry up. Anyway, after lots more measuring and tweaking I stuck a 7.5ms delay on the midrange and ended up here:
Now, I appreciate this is very messy, the measurement was taken at 2m in my living room but its the best I can do at the minute, I've no chance of hauling this lot outside. The Impulse looks baaaaaad, don't faint when you see it......
I think I'm going to have just keep plodding away at this bit by bit, next thing may be to see what it looks like at my seating position. However, I've lots of reply's on here to catch up on......
Thanks again for everyone's input thus far.......
The main problem I've come up against is that, as you say, the timings don't seem to marry up. Anyway, after lots more measuring and tweaking I stuck a 7.5ms delay on the midrange and ended up here:
Now, I appreciate this is very messy, the measurement was taken at 2m in my living room but its the best I can do at the minute, I've no chance of hauling this lot outside. The Impulse looks baaaaaad, don't faint when you see it......
I think I'm going to have just keep plodding away at this bit by bit, next thing may be to see what it looks like at my seating position. However, I've lots of reply's on here to catch up on......
Thanks again for everyone's input thus far.......
I would leave everything the same except remove the Mid/ Woofer delay and see what the impulse looks like.... 7.5 ms is a big delay and I don't see why such a large delay would be needed. The only time I have seen delays that large are when you are using active subwoofers that have a separate DSP built in which can add a lot of latency. In your case I would surprised if you needed anything near that much.Yep, I've really struggled with this today, even with all the great advice.
The main problem I've come up against is that, as you say, the timings don't seem to marry up. Anyway, after lots more measuring and tweaking I stuck a 7.5ms delay on the midrange and ended up here:
View attachment 481209
Now, I appreciate this is very messy, the measurement was taken at 2m in my living room but its the best I can do at the minute, I've no chance of hauling this lot outside. The Impulse looks baaaaaad, don't faint when you see it......
View attachment 481210
I think I'm going to have just keep plodding away at this bit by bit, next thing may be to see what it looks like at my seating position. However, I've lots of reply's on here to catch up on......
Thanks again for everyone's input thus far.......
- Thread Starter
- #170
The 7.5ms is delay is midrange to tweeter, no delay to the woofer.
I am using a digital dsp crossover.
Since you are using the same DSP device for all the drivers it should not change their relative timing except for whatever the individual driver filters do.
I see a little over 8 ms timing offset between the first and second impulse response which is close to your 7.5 ms offset (which also seems way too much for a Mid to Tweeter in my experience). I would try getting rid of all offsets and see what you get. I am thinking the correct offsets are going to be less than 1 ms woofer to mid and much less than that for mid to tweeter.
3ll3d00d
Senior Member
- Joined
- Aug 31, 2019
- Messages
- 416
- Likes
- 303
if you simply adjust delays using REW estimated IR delay (which, given the high frequency content in each driver, looks a reasonable choice) then you get
T = 0.797
M = 0.5839
vs the W
i.e. they are later
these values look pretty sensible to me, perfect alignment is not likely to be found without better measurements so this is good enough to work with for now
if you apply adjustments to get those to zero and vector sum them you get the following
W and M are moderately well aligned but the filter slopes are not right hence the bump
M and T have phase slopes roughly in alignment at the crossover point but they are rolling off at different speeds hence their slopes are not well aligned throughout the stop band hence why they sum varies so much
this one is easier to see if you unwrap it
the slopes align basically at the xo point itself but they're dramatically different before hand which you can see in the FR
I suspect you'll find it helpful to overlay these measurements with the target slopes so you can actually see what you're aiming for, i.e. use software that is actually intended for speaker design.
so basically
1) the filter slopes aren't good acoustically
2) the measurements for the M and T are really messy (I don't think anyone could really say "this is the alignment" because of the amount of mess)
3) making a crossover without directivity data is going to be sub optimal unless you're incredibly lucky (which you won't be) but of course it's fine to ignore it for the purposes of getting your measurements working as surely no one wants to spend hours on useless measurements
did you check the sim matches the measurement btw?
T = 0.797
M = 0.5839
vs the W
i.e. they are later
these values look pretty sensible to me, perfect alignment is not likely to be found without better measurements so this is good enough to work with for now
if you apply adjustments to get those to zero and vector sum them you get the following
W and M are moderately well aligned but the filter slopes are not right hence the bump
M and T have phase slopes roughly in alignment at the crossover point but they are rolling off at different speeds hence their slopes are not well aligned throughout the stop band hence why they sum varies so much
this one is easier to see if you unwrap it
the slopes align basically at the xo point itself but they're dramatically different before hand which you can see in the FR
I suspect you'll find it helpful to overlay these measurements with the target slopes so you can actually see what you're aiming for, i.e. use software that is actually intended for speaker design.
so basically
1) the filter slopes aren't good acoustically
2) the measurements for the M and T are really messy (I don't think anyone could really say "this is the alignment" because of the amount of mess)
3) making a crossover without directivity data is going to be sub optimal unless you're incredibly lucky (which you won't be) but of course it's fine to ignore it for the purposes of getting your measurements working as surely no one wants to spend hours on useless measurements
did you check the sim matches the measurement btw?
OP @Robbie010,
Do you like to stick to REW all the way?
REW is REW; it is a really wonderful tool, but you need to know that REW is not always almighty...
As @levimax kindly pointed in his post #167, he wrote;
"My experience is that the REW alignment tool does not work for LF / Woofers or sub woofers (works fine for Mid to Tweeter). Like you I get nonsensical results like delaying the woofer or inverting drivers when I know that is not right."
I generally and essentially agree with him.
And, therefore, in some cases and situations, the comparison of REW-methods/results with other independent methods would give you further insights and understandings, as well as hopefully better results, especially for time alignment (at your listening position!) among the SP drivers (subwoofer, woofer, midrange, tweeter, super-tweeter) in DSP-based multichannel system.
This is the main reason for my suggestions in my post #53.
Do you like to stick to REW all the way?
REW is REW; it is a really wonderful tool, but you need to know that REW is not always almighty...
As @levimax kindly pointed in his post #167, he wrote;
"My experience is that the REW alignment tool does not work for LF / Woofers or sub woofers (works fine for Mid to Tweeter). Like you I get nonsensical results like delaying the woofer or inverting drivers when I know that is not right."
I generally and essentially agree with him.
And, therefore, in some cases and situations, the comparison of REW-methods/results with other independent methods would give you further insights and understandings, as well as hopefully better results, especially for time alignment (at your listening position!) among the SP drivers (subwoofer, woofer, midrange, tweeter, super-tweeter) in DSP-based multichannel system.
This is the main reason for my suggestions in my post #53.
Last edited:
Furthermore,,,
We are in ASR Forum, and objective measurements and tunings are always essential and critical.
Our final goal, however, is to enjoy listening to music (not to rapid Fq sweep or tone bust test sound) in optimal sound quality (in all aspects and parameters) in your own acoustic environments. In this perspective, you always need to go back-and-forth between objective measurements and subjective "music listening" evaluation; you may understand that the optimal objective tuning would not always gives you the optimal/best subjective results.
You need to have, therefore, your own consistent "Audio Reference/Sampler Music Playlist" consists of various genres of music tracks in excellent recording quality. For your reference, in my case, I shared my such playlist here, and if you would like, I will be more than happy sharing the intact music tracks of my playlist.
Measurement microphone and tuning tool are not identical to your ears and brain, as repeatedly pointed by many audio-enthu people and big-name professional authorities, including Dr. Toole, in ASR Forum...
We are in ASR Forum, and objective measurements and tunings are always essential and critical.
Our final goal, however, is to enjoy listening to music (not to rapid Fq sweep or tone bust test sound) in optimal sound quality (in all aspects and parameters) in your own acoustic environments. In this perspective, you always need to go back-and-forth between objective measurements and subjective "music listening" evaluation; you may understand that the optimal objective tuning would not always gives you the optimal/best subjective results.
You need to have, therefore, your own consistent "Audio Reference/Sampler Music Playlist" consists of various genres of music tracks in excellent recording quality. For your reference, in my case, I shared my such playlist here, and if you would like, I will be more than happy sharing the intact music tracks of my playlist.
Measurement microphone and tuning tool are not identical to your ears and brain, as repeatedly pointed by many audio-enthu people and big-name professional authorities, including Dr. Toole, in ASR Forum...
Last edited:
- Thread Starter
- #177
Thanks.
Brilliant, thanks.
I have just managed to reproduce your results myself using the same or very similar delays. I will remeasure tomorrow and see if the results stack up.
I'm certainly open to using other software so I'll look in to that.
In terms of the target curve I'm initially aiming for, its fairly basic, +2db, flat across the mids and a -2db slope:
Brilliant, thanks.
I have just managed to reproduce your results myself using the same or very similar delays. I will remeasure tomorrow and see if the results stack up.
I'm certainly open to using other software so I'll look in to that.
In terms of the target curve I'm initially aiming for, its fairly basic, +2db, flat across the mids and a -2db slope:
I agree speaker design software is the way to go which is why I suggested VituixCAD but it is quite a learning curve. To accomplish target slope matching with REW you can create the crossovers you want in "filters" and then overlay your measurement so you can see what the acoustic slopes you are aiming for look like. That should be enough to get the speakers up and running. At that point you can study and practice to decide how you want to take better full measurements including off axis and load them into a CAD program which will help you take your design to the next level.
Nice to hear so!
You would not always need dedicated "software" for your measurements and tunings of interest.
In some cases, you only need suitable test tone signals and typical sound analysis tools like Audacity and/or ADOBE Audition, just like as I have done (ref. #53)!
Last edited:
Then, and,,,
Room acoustic mode environment and tuning thereof are closely related to your time alignment and Fq-SPL tunings; sometimes more important and critical/essential than the audio gear tunings; for example:
- Not only the precision (0.1 msec level) time alignment over all the SP drivers but also SP facing directions and sound-deadening space behind the SPs plus behind our listening position would be critically important for effective (perfect?) disappearance of speakers: #687
Consequently, towards your (final?) optimization of your audio setup, you need to go through further long spiral journey of audio gear tuning and room mode tuning, with spiral (back-and-forth) of objective measurements and subjective evaluation using your own "Reference/Sampler Music Playlist".
Room acoustic mode environment and tuning thereof are closely related to your time alignment and Fq-SPL tunings; sometimes more important and critical/essential than the audio gear tunings; for example:
- Not only the precision (0.1 msec level) time alignment over all the SP drivers but also SP facing directions and sound-deadening space behind the SPs plus behind our listening position would be critically important for effective (perfect?) disappearance of speakers: #687
Consequently, towards your (final?) optimization of your audio setup, you need to go through further long spiral journey of audio gear tuning and room mode tuning, with spiral (back-and-forth) of objective measurements and subjective evaluation using your own "Reference/Sampler Music Playlist".
Last edited:
