DiyMobileAudio.com Car Stereo Forum banner

1 - 20 of 58 Posts

·
Registered
Joined
·
845 Posts
Discussion Starter #1
I was hoping that someone can help me understand allpass filters a littler better - basically, how to identify when/where an allpass filter may be beneficial and how to determine the allpass parameters to use in when/where needed (1st order, 2nd order, Q, etc).

Most of the info I find when I search for allpass filters is in regards to audio mixing, when is a different use-case, I think.

I use a Helix DSP.3, which does aupport allpass filters. I just don't know when/where to use them or even if I should use them.

Thank you!
 

·
Registered
Joined
·
158 Posts
I was hoping that someone can help me understand allpass filters a littler better - basically, how to identify when/where an allpass filter may be beneficial and how to determine the allpass parameters to use in when/where needed (1st order, 2nd order, Q, etc).

Most of the info I find when I search for allpass filters is in regards to audio mixing, when is a different use-case, I think.

I use a Helix DSP.3, which does aupport allpass filters. I just don't know when/where to use them or even if I should use them.

Thank you!
Yes! I am also wondering about this. This would be a great topic for a video or really anything else :)


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
845 Posts
Discussion Starter #4
Just to clarify, the phase lines shown in the Helix are not the actual "measured" phase in the car. That is just showing you how the phase changes with the EQ changes that you've applied. So just because the phase lines in the Helix PC-Tool app "line up", that has nothing to do with the actual phase you're getting in your car.

I don't think there is an easy way to measure phase with inexpensive gear. Maybe via sweeps? I've never really tried. I just know you can't really measure phase with pink noise. I do believe that you can see phase issues when measuring multiple speakers together (if you see peaks or dips in a multi-speaker measurement that don't exist with the single-speaker measurements), but that isn't actually measuring phase itself.
 

·
Registered
Joined
·
506 Posts
Just to clarify, the phase lines shown in the Helix are not the actual "measured" phase in the car. That is just showing you how the phase changes with the EQ changes that you've applied. So just because the phase lines in the Helix PC-Tool app "line up", that has nothing to do with the actual phase you're getting in your car.

I don't think there is an easy way to measure phase with inexpensive gear. Maybe via sweeps? I've never really tried. I just know you can't really measure phase with pink noise. I do believe that you can see phase issues when measuring multiple speakers together (if you see peaks or dips in a multi-speaker measurement that don't exist with the single-speaker measurements), but that isn't actually measuring phase itself.
I expand on that quite a bit in the link.

Sent from my SM-G965U1 using Tapatalk
 

·
Registered
Joined
·
918 Posts
Just to clarify, the phase lines shown in the Helix are not the actual "measured" phase in the car. That is just showing you how the phase changes with the EQ changes that you've applied. So just because the phase lines in the Helix PC-Tool app "line up", that has nothing to do with the actual phase you're getting in your car.

I don't think there is an easy way to measure phase with inexpensive gear. Maybe via sweeps? I've never really tried. I just know you can't really measure phase with pink noise. I do believe that you can see phase issues when measuring multiple speakers together (if you see peaks or dips in a multi-speaker measurement that don't exist with the single-speaker measurements), but that isn't actually measuring phase itself.
You can indeed use pink noise to measure phase but only with certain software and hardware... I use a Dayton xlr mic and a second hand interface I bought 👍🏼 smaart and the like use pink noise to measure live phase, basically it can see the phase of the noise looped back into the sound card and also the phase that’s sent back by the mic, so it can do big amounts of maths on the fly and compare the two to give phase information

as for all pass filters, you can’t use an all pass to correct a single speaker where a reflection is causing a phase issue (anything you do the direct sound will also be applied to the reflected sound which is causing the cancellation)

what you can apply them to is when one sides phase issue causes a big dip at the listening position when playing with another speaker (or as in nicks example when playing through the crossover region with another driver... however id rather avoid phase anomalies in the crossover region if I can by adjusting crossovers up or down), in that situation you can sometimes apply an all pass to the troubled side/driver and adjust the phase so it’s back in phase with the opposite side or neighbouring driver when played together

it’s a little bit hit or miss if you don’t have phase information to see that you know is also good and accurate

I have several pictures of phase traces which show issues of varying sizes... just not on my phone right now...

All pass filters are very powerful, however I will also add that they are very easy to mess up with 👍🏼
 

·
Registered
Joined
·
158 Posts
This thread was fairly useful on APFs.


After some reading, as I understand them they are a way to let frequencies “pass” through with their gain untouched but change the phase for those frequencies. Pretty cool!

It seems they are very useful when correcting a specific range of frequencies in a car that are out of phase with another speaker. (As dumdum said)

I found it helpful when checking crossover region when the phase issue can’t be corrected with a simple flip of speaker polarity for example.

I actually just ran into a use case last night.

I was tuning my car and checking phase alignment as in Nicks video above. It wasn’t summing properly (see green line below):



You can see that part of the crossover region summed well but the lower frequencies did not.

I tried adjusting the gains on the individual drivers which did not help. A simple flip of polarity did not help (it made it worse).

I decided to try an all pass filter (APF) and found that a second order APF at 160 Hz and Q of 2 worked nicely. Second order means it changes the phase by 180 degrees which seems most useful when it appears you have a focused area of cancellation going on as above. First order means 90 degrees and I think would be more helpful if you’re noticing a regular repeating pattern of cancellation or constructive interference (regular peaks and valleys) such as in a comb filter (but please correct me if this is not correct).

I assume Q sets the “width” of frequencies for which the APF is applied as in other filters.

As dumdum said, this was done with trial and error as I’m only using REW with UMIK-1.

Here is the plot with the filter in place:


You can see it is summing nicely now (the orange line).

Pretty powerful stuff.

Interestingly, I didn’t need an all pass filter with the GS25 installed but needed one with the HAT Unity U2. Not sure why it’s happening in a speaker specific way. Maybe an issue with the individual driver tunes is causing it? Maybe it’s due to the size difference in the drivers? (Between the GS25 and U2)?


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
845 Posts
Discussion Starter #9
Can't you measure phase with REW using sweeps (instead of pink noise)? I could have sworn that I read that somewhere. Haven't spent the time to look into it any further yet though.

I would just love to be able to actually measure phase without needing $1k software (Smaart), if possible. Would be nice to know for sure what are phase issues (vs reflection issues or some other issue). Maybe it's just not possible though.

Every time I''ve tried to use an allpass filter to correct a dip that exists when playing multiple speakers together (that doesn't exist when playing each of the speakers individually), the allpass filter didn't change anything at all, so based on that, I'm assuming that the issue(s) I'm talking about aren't actually phase issues, but probably reflection-related issues?
 

·
Registered
Joined
·
1,664 Posts
The "Measure" function (sweeps) in REW will produce phase plots. I have not tried it for myself. I've been told by people with way more experience than me that it doesn't produce consistent results. :unsure:
 

·
Registered
Joined
·
304 Posts
I decided to try an all pass filter (APF) and found that a second order APF at 160 Hz and Q of 2 worked nicely.
I'm curious - you made a nice improvement in a line on the graph. Do you think you can detect this with your ear ? If you defeat the all pass filter does it even sound different ? does it sound better or worse ?

I too am interested in how all phase filters are used - personally when i see all the lines on REW I'm happy to just get a decent response, I don't think my results are consistent enough or my understanding coherent enough to properly apply such. And then I start wondering if it makes any difference in the music ? Just curious what listening revealed vs REW plots.
 

·
Registered
Joined
·
158 Posts
I'm curious - you made a nice improvement in a line on the graph. Do you think you can detect this with your ear ? If you defeat the all pass filter does it even sound different ? does it sound better or worse ?

I too am interested in how all phase filters are used - personally when i see all the lines on REW I'm happy to just get a decent response, I don't think my results are consistent enough or my understanding coherent enough to properly apply such. And then I start wondering if it makes any difference in the music ? Just curious what listening revealed vs REW plots.
I'm curious - you made a nice improvement in a line on the graph. Do you think you can detect this with your ear ? If you defeat the all pass filter does it even sound different ? does it sound better or worse ?

I too am interested in how all phase filters are used - personally when i see all the lines on REW I'm happy to just get a decent response, I don't think my results are consistent enough or my understanding coherent enough to properly apply such. And then I start wondering if it makes any difference in the music ? Just curious what listening revealed vs REW plots.
Yes, but it was subtle. It’s a difference of about 2-3 dB at those frequencies so just barely perceptible. Making sure the crossover region sums nicely is definitely noticeable with getting the bass upfront, however (for between sub and mid bass). It’s something that you may not notice until you get it right and then you’re like “oh, THAT’S what upfront bass can sound like”


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
1,587 Posts
I was hoping that someone can help me understand allpass filters a littler better - basically, how to identify when/where an allpass filter may be beneficial and how to determine the allpass parameters to use in when/where needed (1st order, 2nd order, Q, etc).

Most of the info I find when I search for allpass filters is in regards to audio mixing, when is a different use-case, I think.

I use a Helix DSP.3, which does aupport allpass filters. I just don't know when/where to use them or even if I should use them.

Thank you!
View "predicted phase" tab in REW's "overlays" window, select only the speakers you want to match, unwrap the phase (gear icon, top right), open any speakers eq filters window, add an all pass filter, adjust it and watch what happens. The all pass filters will increase the slope of the phase (as the freq increases), so you will want to leave the steepest slope as-is and try to match the rest.
 

·
Registered
Joined
·
845 Posts
Discussion Starter #15
Yes!
I should have mentioned this in my previous post. Hanatsu's REW guide shows how to get useful measurements from multiple sweep measurements.
Thank you! I had forgotten all about that guide (it's probably where I see that it could be done!). :) I did find a .pdf version of it on my Google Drive, so here is a link to it in case anyone is interested in reviewing it. I'll have to revisit it myself and try the phase-related stuff. I haven't done any sweep-based measurements yet.

Anyway, here is a link to the Hanatsu REW Guide in PDF format from my Google Drive:
Hanatsu's REW Guide.pdf
 

·
Registered
Joined
·
845 Posts
Discussion Starter #17
Ok, I think I'm finally making some progress with Allpass filters! :) Hopefully, I'm using them like I'm supposed to. Here are some pictures...

This first picture is of my door and dash speakers - individual left and right measurements (Red and Blue) and the summed L+R response for each speaker pair. Notice the significant dips in the L+R combined measurements at approx 200hz, 400hz, 600hz (door speakers) and 700hz (dash speakers). There are a few other areas that could be fixed with the dash speakers, but let's stick with the 4 areas for now.





So then I tried using allpass filters at those 4 locations - and guess what? A major improvement! Completely removed those significant dips without using any EQ whatsoever!


Is this a proper use of allpass filters? Does the fact that I have the phase-induced dips to begin with tell me that something is wrong with time alignment? In this case, I put the allpass filters on the LEFT speaker in both cases.

When I originally tried using allpass filters for this, I was putting the allpass filters on both the left AND right speakers, so I think they were basically cancelling each other out - which is probably why they did absolutely nothing. :)

However, I still have an issue that I don't quite understand.... Even though the L+R door and dash speakers are summing good now without any big dips, look what happens at the crossover (500hz) when playing door, dash and sub all at the same time:


The crossover itself seems to be summing well, but not sure why I have those deep dips on each side of the 500hz crossover?! Any ideas on that?

All images at 1/12 smoothing in order to show the dips well. Smoothing too much tends to "hide" the dips.

Thank you!
 

·
Registered
Joined
·
158 Posts
At 400 and 600 it looks like your left and right are a bit off. I’m wondering if you were able to match those up better would those dips improve?

I noticed that in some cases if I could match the L and R a bit better sometimes these phase issues would go away. As the gain is adjusted with the EQ filters the phase adjusts as well so sometimes these small adjustments can make a difference. But sometimes they don’t! Just worth a try...

Also what all pass filter settings are you using?


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
158 Posts
Just to add to this, I found that around the crossover region, the closer I’m able to get the response to match the smooth target curve the better. If I can get it to be less jagged in this region the signals seem to sum much better in my experience.

This doesn’t always do it, but I’ve found it helps in some cases.


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
158 Posts
Is this with the GS690? :) just wondering if you were able to install them yet


Sent from my iPhone using Tapatalk
 
1 - 20 of 58 Posts
Top