Another quick video. Enjoy :)
https://youtu.be/ZoZyDsDSaGY
Printable View
Another quick video. Enjoy :)
https://youtu.be/ZoZyDsDSaGY
Very helpful... Thanks for making this
Stupid question.... Lets say that you have a phase-induced dip (whether at the crossover or elsewhere). Is there a difference between fixing the issue with an allpass filter vs fixing the issue with EQ (assuming the end result is the same in terms of freq response)?
I'm still trying to understand and learn how to utilize allpass filters properly.
If a processor didn't have the ability to use allpass filters, is addressing the issue via EQ "appropriate"?
Also - how to decide when to use a first order or secord order allpass filter?
Thank you!
If it is a phase induced dip, then using EQ to boost it can be bad for the speaker and won't work very well anyway. The speakers are still playing those notes loud, they just get cancelled out by the other speaker. So boosting it, is just going to cause it to continue to get cancelled out and now you are pushing that speaker even harder for those notes.
This is all assuming that the individual speaker EQ is actually done well, and knowing what your speakers measure in REW from your posts, that shouldn't be your issue. If your individual speaker EQ is off then yes, that could cause phase issues and EQ could fix those.
In my case, if I EQ the individual speakers to my custom curve perfectly (and they match each other very well), then I've found that I get a few dips when playing them in pairs (really not sure if it is a phase issue, a reflection issue or both). I've just been EQ"ing those few dips (they respond well to EQ) in order to make the L+R speaker pair response match the curve correctly (so I worry more about the speaker pairs matching my curve than the individual speakers)..
I'll have to try using an allpass filter on the dips and see if that resolves them or not - would be nice if an allpass filter solves the dips without needing any EQ boost at all. Haven't really messed with allpass filters yet, so at the very least, it will be a learning experience. :-)
https://www.youtube.com/watch?v=sCEXh0Sv5TA
Great video Skiz, Peter plugs you a couple of times in his newest video.
Two of the best, giving us free knowledge. Thank you!
I've been wanting to make a video like this for ages. This is exactly what I've been jumping up and down about for years. Bravo man, very well done. Thank you for fighting the good fight.
I didn't think about using the overall target curve and then using the EQ module in REW to set the crossover points. That's pretty clever and avoids loading and reloading the target curves for each driver. I might have to adjust the instructions on the tuning companion spreadsheet.
I moved the tool over here to continue supporting it.
Now you all can have the "half-whitledge" target curve too!
https://www.caraudiojunkies.com/show...0209#post60209
As i changed my radio a few days ago (byebye SYNC3, hello Kenwood via Maestro RR) i redid my tune.
Did as you showed in your video (LR24 slopes and all) and then when playing left speakers together (and afterwards exactly the same for the right side speakers) i realized i had perfect dips in the xover. Switched phase on midbass by 180°, bam, perfect summation.
So either my speakers are wired incorrectly or my measurements for distance are way off, but all in such a way off, that it perfectly fits a 180° phase switch.
Question now is: should i just let the midbass flipped by 180° or switch them back to 0° and add some more delay to both tweeters first, followed by mids+tweets together to the sub?
When you say 24db slopes and all, is that electrical (dsp settings) or acoustical (measured via rta)?
Electrical and measured...
And verified against the EQ function in REW with the appropriate settings (24db et. al.).
Attachment 10967
You can always check your distance measurements, but if they are close, then you should be good. If you have a ford with easy entry/exit (where your seat moves backwards when key is removed and forwards when it is placed in the ignition) make sure your keys are in the ignition prior to measuring. I messed that up a couple times and had bad measurements.
If the distance measurements are good, get a 9v battery (or any battery but 9v makes it easy to see) and test the speakers for polarity. Hook the positive side that goes to the speaker to positive on battery, and negative to negative. This should make the speaker move out (reverse the connection and it moves in), all speakers should either move out, or all should move in (doesn't matter as long as they all move the same way). If a set moves opposite, you have them wired wrong.
If you can't do this, there are some apps for polarity testers, but they aren't quite as good, but work for speakers you can't see easily. I still prefer doing the 9v and using a flashlight to look through the grills to make sure, but that does require a second set of hands to connect the battery for most people.
Not that any of this really matters if you do have them summing correctly.
I have a Mustang, so no easy entry in that car. Measurements were done with the new headunit. And this was the only thing that changed.
Did some tunes in the past with SYNC3, but never had to flip polarity on any driver with 24db slopes. This was a first for me and i still think i made something wrong.
Thinking about doing it again scares me a bit as my wife really dislikes me sitting in the car and waving the mic around my head.
And as we all know... happy wife, happy life, but the opposite of a happy wife... well...
Yes, there is a difference between using an all pass filter vs attempting to fix the issue with EQ. A dip caused by a phase cancellation causes a loss in amplitude (which is what you're measuring), but it also effects our localization of those cancelled frequencies. Let's say you have a wide phase related dip at 500Hz when you measure two speakers. If you listen to content centered around 500hz, you will likely hear it as two distinct copies coming from each speaker, rather than a single, coherent center image. Or, the center image will be awfully diffused and unfocused. Our brains treat "out of phase" frequencies this way because of their mismatch in time (since phase is directly related to time). Fixing the phase issue with an all pass filter may fix both the lost magnitude and our localization of the sound. You also need to ask why there is a phase mismatch in the first place (something I'm trying to figure out). Attempting to fix a phase induced dip with EQ will cause more distortion from wasted power and still fail to fix the localization problem. You have to be careful with using all pass filters, though. They aren't a band-aid for phase issues that should be used whenever you see a dip. Just my experience.
That's what i thought, too, but the speaker curves from your spreadsheet don't match up with what REW is doing if using the overall curve and then setting the acoustical xover points in REW for the corresponding speaker.
Don't know which one is off now, but there's definitely a difference.
See for yourself.
Attachment 10981
Orange-red is the unaltered curve for a middbass crossed 80Hz highpass, 2500 Hz lowpass both LR24db from your spreadsheet and blue is what REW is doing when using the overall curve from your spreadsheet, setting it to subwoofer and adding a 2500 Hz lowpass and 80Hz highpass.
Maybe that's one of the reasons why i had to flip polarity on my midbass to have the correct summing? Never had to do that when i was using the single curves from your spreadsheet, 'cause the slopes do not match.
When using the "Bass Limited" or "Subwoofer" "Target Type" settings in REW, it seems to use a crossover type other than Linkwitz RIley (not sure which one it uses off-hand). However, what you can do is select "Speaker Driver" as the "Target Type" in REW - that way it lets you select the crossover type manually. If you want REW to use Linkwitz Riley / 24dB slopes, you can select "Speaker Driver", then "LR4" for the "Crossover HP/LP type" fields.
I think if you do that then the REW target curves will match the curves from the spreadsheet.
Hope that helps.
Thanks for that tip. Just verified that the curves match, and they do.
Guess my wife will hate me next week. :P
On the other hand i don't really know if i need/want to do it over again as the system measures and sounds good right now.
So i don't really know if i want to do it again just for the sake of getting rid of a phase flip.
Don't kid yourself - you know you're going to do it over again. :-) I have a feeling that you are similar to me, in that the phase flip will continue to "bother" you until you "fix" it. Even though the end result would most likely be the same... It's just the way we're "wired" (pun intended). ;-)
What's funny is that I recently changed all of my front stage speakers and have been re-tuning - and I couldn't figure out why my crossovers weren't lining up correctly. Here it's because I forgot about REW using some other crossover type for "Subwoofer" and "Bass Limited"! So I just made the exact same mistake even though I already knew about it! You post made a light-bulb go on and I realized what was wrong. :-) So thanks for bringing it up!
Thing is... the overall response is better than it ever was when using the driver specific curves from the spreadsheet (sorry Jazzi). It‘s most probably user error in regards to tuning after the curves, but this time after flipping the phase it matched perfectly without the need to change the eq in the xover region, unlike before.
But yes, someday next week I’ll do everything again on a different preset. :nana:
I learned this from Andy Wehmeyer who goes into this exact same thing in his blogs and tutorials but Nick does an excellent job explaining it with REW and using examples. Nice job man.
Holy crap dude, thanks for the vid. This made a unbelieveable difference for me retuning my system, took me awhile to figure the all pass stuff out but wow what a difference it made.
Redid it today on another preset with Jazzis spreadsheet.
For whatever reason i cannot get the midbass and tweeters sum up to 6db where when i use "subwoofer" in REW for the slopes it sums perfectly.
See for yourself.
Attachment 11055
Attachment 11056
First is with choosing "subwoofer" in REW for the slopes and phase flip and second is with Jazzis spreadsheet without the phase flip.
Distances (from tape measure) are the same, electrical crossovers are the same.
First one is with sub, while i left out the sub in the second shot.
And yes, first one is only 3-4 DB summation, but sums flat, while the LR24db slopes also only add up to 3-4 db, but i cannot reach the target curve.
With Jazzis curves i fiddled with the delay of the tweeter on both sides, i.e. in 0.02 ms steps and remeasured after each step, but that was the best i could achieve. When i flipped the phase on the tweeter i got a huge dip with Jazzis curves as expected.
Well, either way, you are only getting 3-4dB of summation at the crossover, which I believe is the actual problem. The difference is just that REW uses 24dB Butterworth crossovers when selecting "Subwoofer" - which I believe would normally (if you had full summation) have a "bump" at the crossover point. In your case, the butterworth crossover is coming out "flat", when it really should result in a "bump".
So you do not appear to be getting full summation at the crossover for whatever reason.
Have you ever measured L+R speaker "pairs"? For example, left+right tweeters and left+right midbass? In my case, when I play L+R speaker pairs, the resulting "shape" of the response is slightly different that the individual speaker measurements. I have a few dips (one wide and one narrow) in my midbass L+R pair response that don't exist when I measure the speakers individually.
I'll post some examples shortly. These dips don't seem to be phase-related - applying an allpass filter like Nick shows in the video has absolutely no effect whatsoever on these dips. Maybe the are reflection-related? They do respond normally to EQ, so I can easily apply an EQ filter to "fix" it - although, I'm not really sure if I even should fix it.... Again, will post some example pictures shortly (I'm not experiencing a nice flat summation at the crossovers either, so the post will also be on-topic).
Well, i'm still fiddling with EQ and time alignment, and yes, my wife is looking angry at me whenever i go back to the car.
Looks a bit better than before as the lowpass on both midbasses were not exactly where i wanted them to be. Same for the highpass of the tweeters.
Now i get 4-5 db summing for both left+right which looks a lot better.
Bass is spot on in the xover region though.
Still have to measure them in pairs, but now i got to work first.
It really sounds good now though, but it always could be better, right?
Well, i'm still fiddling with EQ and time alignment, and yes, my wife is looking angry at me whenever i go back to the car.
Looks a bit better than before as the lowpass on both midbasses were not exactly where i wanted them to be. Same for the highpass of the tweeters.
Now i get 4-5 db summing for both left+right which looks a lot better.
Bass is spot on in the xover region though.
Have to sill measure them in pairs, but now i got to work first.
It really sounds good now though, but it always could be better, right?
Attachment 11077
edit: nevermind... fucked up the slopes while dealing with the EQ. Am narrowing to a 4th order butterworth again.
How do i deal with something like this if my DSP doesn't have allpass filters?
F.e. the sub/mid integration if it's only summing up to 3 db with a LR24db alignment. Add time on midbass/tweeter until it matches? Maybe in 0.4ms increments (if 360° phase is 12.5 ms at 80Hz --> 11,25° would be ~0.4ms) and when it gets better do finer adjustments?
Any tips?
Can you share a screenshot of the names of the target curve files you are using from my spreadsheet? That might help.
Can you also share a screenshot of the equalizer settings within REW? Like the one below:
Attachment 11086
I overlaid them to see the difference. The one with the wider and more shallow gap between speakers has a poor summation.
Attachment 11085
It's the same custom overall curve for both. Made a custom curve as both whitledge and half of whitledge didn't fit my taste in my Mustang.
Whitledge had way too much bass and and not enough treble for my liking and half of whitledge not enough of bass and too much of treble. So i made up a curve that is somewhere in the middle between both, i.e. less bass than whitledge and less treble than half of whitledge.
Screenshots for Subwoofer attached. First is for the wider gap, second with the more narrow gap (I verified that it's a Butterworth alignment if you use any of the Subwoofer or bass limited presets). Please note that i only used the overall curve for both to not have to load the single driver curves everytime.
Attachment 11114Attachment 11113
Either way. Both are not summing correctly as the more narrow gap should have a bump in the summed response and the LR24db alignment should sum flat, right?
So i think i might have some phase issues in my system that should be solved regardless of the alignment i'm using.
The one with the more shallow gap uses LR24db alignment as shown below for the midbass driver against the midbass curve from your spreadsheet.
Attachment 11115
I think i put this in a seperate thread though as in my opinion it starts to get offtopic for this thread.
Hey Skiz,
First, Thanks a million for your constant contributions to the community.
I've got another install coming up in the next couple weeks and after seeing your comment the other day regarding the "punch in the chest" feeling coming from proper sub-midbass phase cohesion, I came to rewatch this video.
Though, it seems the video has been set to private :(
Is there any way I can view this video?
Thanks!
any chance to get an invite to watch the video? seems it went private and i lost out on knowledge. :(
Like they said ^^^ it got switched to private, if you can change it back this sounds like a great one.
Thanks for making this - this I'm sure hits on one of my big concerns.. both frequency and phase cohesion across a crossover.
I found two drivers I love together, and opens the possibility of a 1khz or 2khz Xover...
It's a big opportunity and just as big of a "yikes" from an imaging perspective.
Since that's smack where our hearing is most sensitive, I have a feeling I'll either be spending hours trying to align those - or abandoning it for a more traditional tweeter point.
Would love to see this vid.
Sent from my LM-G820 using Tapatalk