DiyMobileAudio.com Car Stereo Forum banner

41 - 58 of 58 Posts

·
Registered
Joined
·
506 Posts
OK, so here are some pictures. This first set of pictures just shows my Door and Dash speakers - individually and the combined L+R response of each set (L+R Door and L+R Dash).

Here is what the combined responses looked like before applying any allpass filters, with little red arrows showing the phase-related dips (3 in combined midbass response and 2 in combined dash speaker response) - I could probably use another one for the dash speaekers, but...:



Here is what the combined response looks like after I applied allpass filters to those dips - they magically just disappear that easily! Awesome! This is with 1/6th smoothing (with 1/3rd smoothing, things look perfect):


I applied the following allpass filters to fix those dips shows with the arrows (no EQ, just an allpass filter per dip):

L+R Door (all 2nd order):
  • 197hz, Q of 4
  • 414hz, Q of 9
  • 630hz, Q of 9
L+R Dash (all 2nd order):
  • 672hz, Q of 6
  • 985hz, Q of 5
No idea if this is "normal" (to need to apply so many allpass filters), but they definitely addressed the big dips, so I'm happy.

So I figured - cool! I'm done - everything is lining up just right phase-wise. NOT. :) I forgot that I haven't deal with the DashDoor crossovers yet! So far I just fixed the phase-related issues when playing speakers in L+R pairs....

I'll be back to show the response of each side (Left Door+Left Dash and Right Door+RIght Dash). I can't seem to get those right - at all, which obviously affects the overall response...

It's becausse first order allpass filters don't affect the frequency you are setting them at. They roll the phase of things 1+ octaves above it. The second order has a much more pronounced impact at the frequency you set it at. You can see clearly where the filters affect phase if you watch this video. Then you can use first order filters as well, the next time! :)

 

·
Registered
Joined
·
175 Posts
I applied the following allpass filters to fix those dips shows with the arrows (no EQ, just an allpass filter per dip):

L+R Door (all 2nd order):
  • 197hz, Q of 4
  • 414hz, Q of 9
  • 630hz, Q of 9
L+R Dash (all 2nd order):
  • 672hz, Q of 6
  • 985hz, Q of 5
No idea if this is "normal" (to need to apply so many allpass filters), but they definitely addressed the big dips, so I'm happy.

So I figured - cool! I'm done - everything is lining up just right phase-wise. NOT. :) I forgot that I haven't deal with the DashDoor crossovers yet! So far I just fixed the phase-related issues when playing speakers in L+R pairs....

I'll be back to show the response of each side (Left Door+Left Dash and Right Door+RIght Dash). I can't seem to get those right - at all, which obviously affects the overall response...
THANK YOU for this write up !! This type of explanation is so super helpful to a DIY guy like me. Thanks again for taking the time to put your post together! You Rock the House Jtrosky!
 

·
Registered
Joined
·
845 Posts
Discussion Starter #44
With those huge dips at the crossover, did you try flipping the polarity of your widebands? I’ve thought polarity was all in sync and then I ran across that exact same problem and flipped polarity and boom! Fixed. May not do it, but work checking.
Yeah, unfortunately, it's not that simple - I did try that. Flipping the polarity on the dash speakers just moves the dip a tiny bit - but very similar results. I just did some more testing and found that I can either get the L+R speaker pairs to sum correctly (no big dips/nulls) or the dash/door crossovers to sum correctly - but not both - fixing one "breaks" the other. :) Not sure which one is "better" at this point. But I can't get all combos of all speakers to sum correctly with the same settings. I'll change things (add or remove allpass filter(s)) to get the door/dash crossovers to sum correctly, but then that change causes the L+R pairs to not sum correctly anymore. Or I can get door/dash crossovers to sum correctly, but then it creates issues with the L+R speaker pairs. Just can't get them all correct at the same time, with the same settings. :) Honestly, not so sure I can actually hear the difference anyway though, so I'm not going to stress over it anymore. I'll probably fix the door/dash summing and just deal with a null in the L+R speaker pair summing...


It's becausse first order allpass filters don't affect the frequency you are setting them at. They roll the phase of things 1+ octaves above it. The second order has a much more pronounced impact at the frequency you set it at. You can see clearly where the filters affect phase if you watch this video. Then you can use first order filters as well, the next time! :)
Right, but it this case, we're using all 2nd-order allpass filters - and it's still affecting an area outside of the area we're trying to impact (that one issue that @Mauian fixed at the crossover). Strange - especially since the Q is so wide for such a relatively narrow dip. So hard to know what's really going on with phase without being able to actually see the phase as you adjust things. Unfortunately, it's too expensive to buy Smaart just for stuff this.


Damn dude, 1/3 octave smoothing. You're really going for it, huh.
Not following? If you're saying 1/3rd isn't detailed enough - it's kind of funny - originally, I used to use mainly 1/6th and 1/12th for everything - then everyone got on me saying that I didn't need to use that level of detail. Now I mainly use 1/3rd for everything (with brief looks at 1/6th and 1/12th) - and now I'm getting slack for using 1/3rd. :) Can't make everyone happy! Everything I posted above was actually 1/6th though - except for the on pic showing how the crossovers sum.


THANK YOU for this write up !! This type of explanation is so super helpful to a DIY guy like me. Thanks again for taking the time to put your post together! You Rock the House Jtrosky!
Awesome - glad you found it useful. I was actually looking for help myself though, so I'm certainly not the expert on the subject! I'm in the process of trying to understand it all myself. But again, ,glad you found it useful. I wish more people posted graphs and how they deal with things - you can learn a lot just seeing how other people do things - and the graphs really help clear things up sometimes. You rarely see people actually post graphs though. Just a few people, for the most part.
 

·
Registered
Joined
·
506 Posts
REW nerds tend to use less smoothing on lower frequencies (no smoothing or 1/24 on sub) and will limit their smoothing to 1/6 on mids/tweets. The argument this author makes for 1/3 I think really suggest 1/6 is the way to go. If we can discern 1/3 octave changes then we can be sure 1/3 octave smoothing will drop some of that data, while smoothing at 1/6 would preserve far more. Other people argue that you can hear 1/6 octave changes. So then... Anyway, you can't really go wrong.

 

·
Registered
Joined
·
506 Posts
Right, but it this case, we're using all 2nd-order allpass filters - and it's still affecting an area outside of the area we're trying to impact (that one issue that @Mauian fixed at the crossover). Strange - especially since the Q is so wide for such a relatively narrow dip. So hard to know what's really going on with phase without being able to actually see the phase as you adjust things. Unfortunately, it's too expensive to buy Smaart just for stuff this.
Allpass filters have a very significant impact over a wide range of frequencies, it's very much unlike a parametric filter. A lower Q would mean the filter applies to a wider frequency band, isn't that what you are describing?
 

·
Registered
Joined
·
158 Posts
REW nerds tend to use less smoothing on lower frequencies (no smoothing or 1/24 on sub) and will limit their smoothing to 1/6 on mids/tweets. The argument this author makes for 1/3 I think really suggest 1/6 is the way to go. If we can discern 1/3 octave changes then we can be sure 1/3 octave smoothing will drop some of that data, while smoothing at 1/6 would preserve far more. Other people argue that you can hear 1/6 octave changes. So then... Anyway, you can't really go wrong.

Yeah, I read recently Andy (Audiofrog) recommends 1/3 for frequencies above 1kHz and 1/6 for below.


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
918 Posts
In the above case where you can’t get the phase Matching at the crossover, it’s very likely you have a phase issue at or around the crossover point, which will never help the summation and get decent summation, I have a phase issue at 300hz on one midbass, but the pair play nicely upto 200hz so I cross there, it keeps any phase issues well out of the crossover regions and gets me excellent summation and a very good blend so the two speakers melt into the tweeters and my stage is all 12” off the dash... I have midbass under my seats, and the sub is on the dash 90% of the time, I think I need to work on its eq and rattles as the phase trace overlaps perfectly from 30hz to 300+ 👍🏼

I measure with smaart by the way

I should also add i will be testing with 12db per octave crossovers to see if that further improves the blend as the midbass will play a little lower and the sub will play a little higher... although I’ll do this for Emma sq only to help the really awkward subwoofer test track at lower levels 👍🏼
 

·
Registered
Joined
·
158 Posts
Jtrosky, I wonder if you could have success if you didn’t do any all pass filters until playing L+R together AND dash/doors together - as in see where the phase issues are with all drivers playing and address them then. Could be worse, but maybe better? Worth an experiment? (I mean, you’re clearly having success already, but we’re talking about the little things)

I agree with dumdum. Moving the crossover could get you out of a trouble spot.


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
845 Posts
Discussion Starter #50
Actually, I went back out shortly after my last post (can't give up that easily!) and was able to get everything "in sync". Just had to find that "just right" combination of allpass filters. Basically, I just had to add a single allpass filter to each of the dash speakers. Now, the left side and right side have no more dips at the door/dash crossover and the L+R speaker pairs are also "dip free". :)

I'll post the updated graphs shortly.

Previously, I was concentrating too much on the midbass speakers to address the dips (since the dips were mainly on the "midbass side" of the door/dash crossovers). However, in the end, all that I needed was a single allpass on each dash speaker and all is good now everywhere. Now I can concentrate on some tonality and other minor details.

I will say - I think my system is sounding the best it's ever had with those CDT Unity 8.0 speakers. It's hard to put in words, but they are just so "smooth" sounding. Not a hint of "harshness" - and I hear every little detail. VERY happy with those little speakers so far!

Thank you all for the help - I appreciate it!

EDIT: Oh - by the way, these allpass filtes are awesome! Before I started using allpass filters, I would fix all of these dips with straight-up EQ (boosts), which was 100% the incorrect way to handle these issues. Now, I was able to fix every "unexplained" dip without using a single bit of EQ. Pretty cool how that all works and everything comes together.
 

·
Registered
Joined
·
158 Posts
Actually, I went back out shortly after my last post (can't give up that easily!) and was able to get everything "in sync". Just had to find that "just right" combination of allpass filters. Basically, I just had to add a single allpass filter to each of the dash speakers. Now, the left side and right side have no more dips at the door/dash crossover and the L+R speaker pairs are also "dip free". :)

I'll post the updated graphs shortly.

Previously, I was concentrating too much on the midbass speakers to address the dips (since the dips were mainly on the "midbass side" of the door/dash crossovers). However, in the end, all that I needed was a single allpass on each dash speaker and all is good now everywhere. Now I can concentrate on some tonality and other minor details.

I will say - I think my system is sounding the best it's ever had with those CDT Unity 8.0 speakers. It's hard to put in words, but they are just so "smooth" sounding. Not a hint of "harshness" - and I hear every little detail. VERY happy with those little speakers so far!

Thank you all for the help - I appreciate it!
Sweet! :)


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
474 Posts
Just a heads up based on my experience with all-pass filters. When you see a phase cancellation between two speakers, you may be tempted to automatically use an all-pass filter to get summation on the RTA. However, you need to consider what caused the phase between the two speakers to diverge in the first place. The original sound wave left each speaker at the same time (assuming correct delay settings), and the direct sound definitely reaches your ears in phase at all frequencies. The reason why there is a phase cancellation is due to a reflection that arrives AFTER the direct sound that is out of phase with the opposite speaker. When you use an all-pass filter to correct this issue, you're compensating for the phase change due to the reflection, but also ruining the phase response of your direct sound. Which is more important to correct, your direct sound, or the reflection? Remember that the RTA is a representation of frequency response OVER TIME. Direct sound and reflections get mixed together and you can't tell them apart. That's a tradeoff you need to experiment with.

For example, my midbass to midrange crossover point was 300hz. The midrange behaved as it should (minimum phase) and the phase trace was ideal. The midbass, however, had a nasty reflection some milliseconds later which I could clearly see ruined the phase trace. Naturally, I applied a 2nd order all-pass filter to the midrange channel and got the phase to match perfectly. I had a perfect 6db summation on the RTA afterward. When it came time to listen, frequencies around 300hz did not produce a tight center image and were unintelligble. I was baffled because the phase traces matched so perfectly. I undid the all-pass filter and went back to normal settings. Even though the dip showed up on the RTA, I had my tight center image back around 300hz. My best hypothesis is that the direct sound of the midbass driver had a greater influence on phase/imaging than the later reflection did. The measurement software unfortunately gives the reflection too much weight in terms of perceived importance.
 

·
Registered
Joined
·
918 Posts
Just a heads up based on my experience with all-pass filters. When you see a phase cancellation between two speakers, you may be tempted to automatically use an all-pass filter to get summation on the RTA. However, you need to consider what caused the phase between the two speakers to diverge in the first place. The original sound wave left each speaker at the same time (assuming correct delay settings), and the direct sound definitely reaches your ears in phase at all frequencies. The reason why there is a phase cancellation is due to a reflection that arrives AFTER the direct sound that is out of phase with the opposite speaker. When you use an all-pass filter to correct this issue, you're compensating for the phase change due to the reflection, but also ruining the phase response of your direct sound. Which is more important to correct, your direct sound, or the reflection? Remember that the RTA is a representation of frequency response OVER TIME. Direct sound and reflections get mixed together and you can't tell them apart. That's a tradeoff you need to experiment with.

For example, my midbass to midrange crossover point was 300hz. The midrange behaved as it should (minimum phase) and the phase trace was ideal. The midbass, however, had a nasty reflection some milliseconds later which I could clearly see ruined the phase trace. Naturally, I applied a 2nd order all-pass filter to the midrange channel and got the phase to match perfectly. I had a perfect 6db summation on the RTA afterward. When it came time to listen, frequencies around 300hz did not produce a tight center image and were unintelligble. I was baffled because the phase traces matched so perfectly. I undid the all-pass filter and went back to normal settings. Even though the dip showed up on the RTA, I had my tight center image back around 300hz. My best hypothesis is that the direct sound of the midbass driver had a greater influence on phase/imaging than the later reflection did. The measurement software unfortunately gives the reflection too much weight in terms of perceived importance.
It’s not that it gives anything any weight... it just measures everything, it’s impossible to get it to measure only direct sound as gating won’t be possible due to how enclosed a car environment is... I do agree though that a dip can sometimes be better left alone 👍🏼
 

·
Registered
Joined
·
845 Posts
Discussion Starter #54
Just a heads up based on my experience with all-pass filters. When you see a phase cancellation between two speakers, you may be tempted to automatically use an all-pass filter to get summation on the RTA. However, you need to consider what caused the phase between the two speakers to diverge in the first place. The original sound wave left each speaker at the same time (assuming correct delay settings), and the direct sound definitely reaches your ears in phase at all frequencies. The reason why there is a phase cancellation is due to a reflection that arrives AFTER the direct sound that is out of phase with the opposite speaker. When you use an all-pass filter to correct this issue, you're compensating for the phase change due to the reflection, but also ruining the phase response of your direct sound. Which is more important to correct, your direct sound, or the reflection? Remember that the RTA is a representation of frequency response OVER TIME. Direct sound and reflections get mixed together and you can't tell them apart. That's a tradeoff you need to experiment with.

For example, my midbass to midrange crossover point was 300hz. The midrange behaved as it should (minimum phase) and the phase trace was ideal. The midbass, however, had a nasty reflection some milliseconds later which I could clearly see ruined the phase trace. Naturally, I applied a 2nd order all-pass filter to the midrange channel and got the phase to match perfectly. I had a perfect 6db summation on the RTA afterward. When it came time to listen, frequencies around 300hz did not produce a tight center image and were unintelligble. I was baffled because the phase traces matched so perfectly. I undid the all-pass filter and went back to normal settings. Even though the dip showed up on the RTA, I had my tight center image back around 300hz. My best hypothesis is that the direct sound of the midbass driver had a greater influence on phase/imaging than the later reflection did. The measurement software unfortunately gives the reflection too much weight in terms of perceived importance.
Great post - thank you for that! As I was applying all of these allpass filters, I was kind of wondering if there were any unintended consequences - and it sounds like there could be....

To be 100% honest, I'm not even sure I could tell the audible difference between the original RTA response with the dips throughout and the "corrected" response after I fixed all of the dips with the allpass filters. :) The "after" RTA response definitely looks a lot better, but not so sure it really translates to actually sounding any better.

RTA measurements are really cool - but I think they also have you chasing your tail a lot of the times for things that you can't actually hear anyway. I guess that is what makes an experienced tuner so much better than a novice like myself - being able to identify the stuff that really needs addressed vs what doesn't really matter (or items that can actually make things worse!).
 

·
Registered
Joined
·
845 Posts
Discussion Starter #55
Is it considered somewhat "normal" to have 5 or 6 allpass filters in a tune? Or would that be highly "irregular"? I'm just curious if what I'm doing adding these allpass filters is a "common" practice or not....
 

·
Registered
Joined
·
158 Posts
Just a heads up based on my experience with all-pass filters. When you see a phase cancellation between two speakers, you may be tempted to automatically use an all-pass filter to get summation on the RTA. However, you need to consider what caused the phase between the two speakers to diverge in the first place. The original sound wave left each speaker at the same time (assuming correct delay settings), and the direct sound definitely reaches your ears in phase at all frequencies. The reason why there is a phase cancellation is due to a reflection that arrives AFTER the direct sound that is out of phase with the opposite speaker. When you use an all-pass filter to correct this issue, you're compensating for the phase change due to the reflection, but also ruining the phase response of your direct sound. Which is more important to correct, your direct sound, or the reflection? Remember that the RTA is a representation of frequency response OVER TIME. Direct sound and reflections get mixed together and you can't tell them apart. That's a tradeoff you need to experiment with.

For example, my midbass to midrange crossover point was 300hz. The midrange behaved as it should (minimum phase) and the phase trace was ideal. The midbass, however, had a nasty reflection some milliseconds later which I could clearly see ruined the phase trace. Naturally, I applied a 2nd order all-pass filter to the midrange channel and got the phase to match perfectly. I had a perfect 6db summation on the RTA afterward. When it came time to listen, frequencies around 300hz did not produce a tight center image and were unintelligble. I was baffled because the phase traces matched so perfectly. I undid the all-pass filter and went back to normal settings. Even though the dip showed up on the RTA, I had my tight center image back around 300hz. My best hypothesis is that the direct sound of the midbass driver had a greater influence on phase/imaging than the later reflection did. The measurement software unfortunately gives the reflection too much weight in terms of perceived importance.
I love this. Great idea to use a little band passed pink noise on the area with the phase issue / APF and see if it centers up better before or after. Thanks!


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
845 Posts
Discussion Starter #57
So here is where my tune ended up for now. Keep in mind that I've never had a tune as "complete" as this. Previously, before I understood allpass filters, I would just use EQ to boost areas where I had phase-induced dips, which is obviously not the correct way to handle those issues. :) That caused my ending "overall" response to look "correct", but it really wasn't since I was just using EQ to account for the phase-induced dips, instead of actually fixing the phase-related issues). I no longer use ANY EQ to "correct" any phase-related dips (as you can see from the picture below since the combined L+R measurements match the "shape" of the individual speaker measurements almost exactly).

This is what my tune looks like right now with CDT ES-690CFS 6x9 midbass speakers, CDT Unity 8.0 wideband dash speakers and a JBL BassPro SL under-seat sub. I do also have some Kenwood Excelon KFC-X174 6.5 coaxials speakers in the rear-deck, but they are not included in any of these measurements. I've also decided to let my 6x9 door speakers give more deep bass (set their HP xover to 50hz instead of 65hz, which boosted the ~40hz output from them significantly). As long as it doesn't cause any significant vibration-related issues (which it hasn't so far), I figured "why not?".

I definitely don't have any issues with my bass not being "up front" - since all of my deep bass actually comes from the front (front 6x9 door speakers and under-seat sub under passenger seat). "Up front" bass is a complete non-issue in this build. :)

 

·
Registered
Joined
·
90 Posts
Damn... i wish Mosconi would implement more than 5 filters in their V1 GUI for the D2 series of DSP amps and include allpass filters for them.
 
41 - 58 of 58 Posts
Top