Put simply, allpass filters are EQ for phase. They don't impact the frequency response like a regular EQ filter, they only impact the phase at the frequency and for the Q specified by the allpass filter.

The effects of an allpass filter can be measured - they can be measured directly if you measure the phase of the individual speakers - or "indirectly" by measuring the frequency response when multiple speakers are playing. Measuring a single speaker will not show the effects of an allpass filter, since phase changes only impact the frequency response when multiple speakers are playing together.

For example... Here are some old pictures of my L+R midbass response before and after I applied allpass filters.

Here are the "Before allpass filters" measurements (Blue=Left, Red=Right, Green=L+R combined) - see the big dips in the green L+R combined measurement that don't exist in the L and R individual measurements?

Now here are the same measurements after I applied a few allpass filters to make the phase match (where the dips are) between L and R - see how the big dips are completely gone in the L+R combined response?

That is what allpass filters can do for you.

Sent from my ASUS Chromebook Flip C302 using Tapatalk

 

Nice. So, allpass filters can be configured to target certain frequencies . I did not know that.
Is there a maximum number of allpass filters that can be used at one time ?

 

I’m trying to do some research to get my head around them as well, two spots I’ve found some helpful info was JL audios training session on their tuning software, and Helix had a basic overview on their software guides

Equalizer

www.audiotec-fischer.de

 

just a heads up, you should not be throwing all pass filters at every dip you see on an RTA

 

Care to elaborate at all? When/why shouldn't you use allpass filters to correct multi-speaker summation issues (whether it be L+R combined or at crossovers)?

Personally, I mainly use them on my midbass and on the lower frequency areas of my widebands, if/when needed - and only if the multi-speaker response actually dips lower than the individual responses (which from my understanding, is clear evidence that there is a significant phase mismatch causing that dip). I have also used them to correct phase-related dips at crossovers as well.

I'm trying to understand why you wouldn't want to address these areas of poor summation? Why would you want to leave a significant dip there if it can easily be corrected with an allpass filter?

In my experience, where it can get tricky is when using allpass filters to address "L+R combined" phase-related dips near crossover areas. You can use an allpass filter to address the dip in the L+R combined response, only to find that it caused a new dip at the crossover. However, I can usually find the right combination of allpass filters to get everything summing correctly for both the L+R combined response and at crossover areas.

Personally, I haven't noticed any "cons" when fixing these kinds of dips with 2nd order allpass filters... The measured response clearly shows that I'm getting better summation - and I don't hear any "side-effects" when actually listening to the system, so I don't see why you wouldn't want to address the phase-related dips.

 

All pass filters work in the same place where boost doesn’t generally… , this often means a dip caused by a reflection or standing wave… and the eq doesn’t work so it means the reflection cancels any extra energy put in, this also effectively means there is a delayed energy that cancels the direct energy

all pass filters don’t correct this at all, what they do do is make the opposite side to the reflection phase error side have an induced phase wrap so it then matches the bad sides delay, effectively you make the good side bad but in doing that the phase then matches and sums correctly despite having a small dip

Sometimes even if you can correct a dip but it still doesn’t sum 100% you can correct the summation with an all pass filter, I have never used a first order filter to achieve this, but second order, generally higher Q all pass filters are the ones that get used

 

Summation isn’t everything, if the delay that’s added on the bad side is moving the centre one way caused by a dominant reflection , and you add an all pass on the good side so it matches and sums, it’s very possible to push the centre further away, we can only add delay with all pass filters (how much is freq dependant, the higher the all pass the less delay it adds… a second order adds one wavelengths delay) so a very small amount at 1.6khz vs 160hz)

every effect has a side effect whether it’s desired or not is another matter 👍🏼

 

Summation isn’t everything, if the delay that’s added on the bad side is moving the centre one way caused by a dominant reflection , and you add an all pass on the good side so it matches and sums, it’s very possible to push the centre further away, we can only add delay with all pass filters (how much is freq dependant, the higher the all pass the less delay it adds… a second order adds one wavelengths delay) so a very small amount at 1.6khz vs 160hz)

every effect has a side effect whether it’s desired or not is another matter [emoji1360]

So in my midbass example above, you would simply leave the phase-related dips there? Or would it just depend on whether you hear any "bad" side-effects after using allpass filters to fix them?

Obviously, I listen to the system both before and after I make changes. If I hear anything "off" afterwards, I wouldn't leave it in place. However, that hasn't been my experience - at least with my system. I have not yet noticed any "bad" side effects from allpass filters that I've used - but I do get better summation.

In some cases, the dips that are corrected aren't necessarily easily audible via your ears - but that is kind of the reason why we use RTA's to begin with - to identify and correct issues to a point that is just not possible using your ears alone. Otherwise, we wouldn't use an RTA at all - we'd just tune everything by ear. However, all of those little "imperfections" add up and may be audible in certain situations. As long as they don't cause any audible (non-wanted) side-effects, I see no reason why wouldn't want to get the phase to match between L and R and get as much summation as you can.

I might get abuse for this, but - all-pass filters are not really for amplitude ("RTA") issues. Those dips on the graph above would not really be noticeable in themselves.

All-pass is for coherence, imaging, soundstage. Amplitude dips are a good way to see the problem, but they are not the problem that you are trying to fix. You are trying to keep phase the same L to R, and through a crossover region. This is best done by measuring phase. I struggle with OpenSoundMeter for hitting a target curve, but that is the type of tool (or a more serious/paid for alternative) that is best for setting up all-pass.

I completely understand that you are better off actually measuring phase directly. However, as you stated - those dips are just another way to identify phase issues. Using an RTA to identify phase issues like this may not be as accurate as actually measuring the phase, but why not make the phase match between L and R and get rid of those dips if you can (as long as there are no audible "bad" side effects)? In my case (and probably lots of others), I really don't have the correct equipment to properly measure phase and correct phase 100% - but in cases like this, it's pretty obvious that these dips are phase-induced. I am simply using the tools that I have to identify and fix what issues that I can. If you can use an allpass filter and get the L+R responses to sum better, without any audible side-effects, that seems like a "win" to me.

That is kind of like saying that you shouldn't fix an issue with your car because a mechanic can fix it better/faster with better tools. At the end of the day, if you can fix the summation without any bad "side effects", then why not? Sometimes, you use what you have and do what you can with what you have.

 

I might get abuse for this, but - all-pass filters are not really for amplitude ("RTA") issues. Those dips on the graph above would not really be noticeable in themselves.

All-pass is for coherence, imaging, soundstage. Amplitude dips are a good way to see the problem, but they are not the problem that you are trying to fix. You are trying to keep phase the same L to R, and through a crossover region. This is best done by measuring phase. I struggle with OpenSoundMeter for hitting a target curve, but that is the type of tool (or a more serious/paid for alternative) that is best for setting up all-pass.

 

So you guys are talking about measuring phase using mics/acoustics, could you also measure phase on the electronics side? For example, using a scope and measure input vs. output and comparing the phase difference there? Or is that different? I mean i know it is different, but how different, relative to this conversation?

 

At least with boxes that have group delay, the accoustic phase lags the electrical phase.
And many speaker dips in amplitude have phase shifts happening there.
So it is usually best to measure the total ensemble effect.

On the crossover regions one could make a case that fixing the phase on the electrical side might make sense, before confirming it acoustically with a measurement.

Once one goes through the mic, then it sort of confirming it electrically... just electrically after the speaker

 

@jtrosky , I’ll take a stab at post #2 and post #8. Basically those small anomalies in the response you see, you probably can’t actually hear those. Many of the dips and comb filtering peaks and valleys you see on an RTA with a resolution of say more than 1/3 octave, and I would say definitely not more than 1/6 octave; you just cant hear those so don’t waste the DSP resolution and time to try and correct those.

This has been my experience anyway so often I tune using 1/3 octave and sometimes I flip back and forth from 1/6 to 1/3 for the lower octave FR response just to gain some additional insight at the larger dips and peaks. I think this is a trap that many people fall into when tuning and they fuss too much over that small stuff when it may not actually matter.

 

See, this is where I disagree. We go to great lengths to squeeze every last bit of performance out of our car audio systems. We spend insane amounts of money on speakers, amplifiers and DSPs. We spend insane amounts of money, time and effort to "sound deaden" our cars. We spend insane amounts of time tuning them to perfection.

So why shouldn't we spend 15-30 minutes to apply a few allpass filters to correct 5dB+ dips in the L+R combined response? It doesn't cost us anything. Those bands on the EQ would probably go unused otherwise (at least on a Helix since it has 30-bands per channel). I mean we're talking about (3) 5dB+ dips that each "span" ~50hz - and all three are in the 200hz - 600hz (approx) range. To me, those dips can certainly be audible - especially when all three are combined. As long as the allpass filters don't cause any "unwanted" side-effects, to me it's a no-brainer. Why purposely leave those dips there if you can easily fix them without any unwanted side-effects?

RTA's are specifically designed to make it easier to identify and correct things that we can't easily identify with our ears. I feel that it's useful to correct all of these small "anomalies", that when combined, actually do make an audible difference. If we just say "well, you probably can't hear that anyway" for everything we see on an RTA, but can't easily identify with our ears, then we might as well just not use the RTA at all, right? Again, I think it's the combined effect of correcting all of these minor issues that makes a difference.

That doesn't mean that I'm actually right though.

 

Following

 

As a general point to this thread, nobody is saying don't use All Pass. Nobody is saying those dips are not candidates to fix and improve.

What is being said is a) don't worry about the audibility of the dip as such b) measure phase directly.

As has been said - we go to great lengths to extract performance from our system. It is not a huge leap to get the capability to measure phase directly, if you want to tackle that issue as well. It has also been said, we use RTA to measure things we can't hear easily. If you want to fix phase, use phase measurement, so you can correct it more accurately than you can by ear.

 

Understood - and I agree that actually measuring phase would be the best way - but at the same time, if you have a few dips like this and don't really have the ability (or desire, for that matter) to properly measure phase directly, in real-time, I feel that it's worth using a few allpass filters to address issues that are pretty obvious just using frequency response measurements. In my case, I don't yet have the ability to properly measure phase in real time. Until I get to that point, I think doing it this way is still beneficial - and may even be "good enough" for most people.

I actually learned about this from one of Skizers videos on how to verify phase through a crossover while remote tuning. He was responding to criticism that you can't possibly identify and address phase issues via remote tuning with a simple USB mic and frequency response measurements. His point was simply that you can actually identify and correct phase issues - even when tuning remotely, without actually measuring phase directly - based solely on the measured frequency response with multiple speakers playing. In the video (which has since been marked private), he used an allpass filter to address a big dip in the crossover region - that was identified with simple frequency response measurements. He simply applied a 2nd-order allpass filter and the dip was corrected. That was his "evidence" that you can use frequency response measurements to identify and fix some phase-related issues, quickly and easily. Is it the best way? No. Is it the most accurate way? No. However, it is still useful and good enough in a lot of cases.

It works for me until I have a better way to do it.

Hell, I used simple frequency response measurements to get both of my subs phase-aligned as well. I didn't use allpass filters, but just by altering the time delays, I could get huge differences in summation - and frequency response measurements allowed me to get the best possible summation. I initially tried doing it via phase measurements via sweeps - and that was U*G*L*Y. I feel that I got much better results, quicker - just by using frequency response measurements instead.

 

Here is another phase-related thread where we discussed this stuff. In fact, I actually used the same midbass measurements in that thread - and even got some feedback from Andy at AudioFrog and others. Good read.

The "All about phase" thread

 

Great post here. Learning a lot that I’ve not been able to figure out about tuning

 

How do you actually apply an all pass filter on a Helix Ultra? As in what steps in the software on which screens/settings/buttons.

 

Once you "Enable" allpass filters in the DCM->PC Tool Configuration screen, then there will be an "Allpass EQ" option in the filter type selection box for each EQ slider. So your choices will be "Fine EQ", "Parametric EQ" and "Allpass EQ" (plus Low or High Shelf EQ if you select the first or last slider).

Once you select "Allpass EQ", you can then specify the filter type (1st order or 2nd order) and then specify the Frequency and Q (if 2nd order) for the allpass filter.

To correct the type of dips in my earlier examples, I always use 2nd order - and start with using a frequency that is right around the middle of the dip, with a Q that you would normally use if your were EQ'ing the dip with a parametric filter. Then take a new measurement and see where you stand. You may have to play with the frequency and Q a little to get the desired results. Again, this method is not as accurate as actually measuring phase, but using this method, I usually get the dips corrected in the first few tries.

Just make sure to only apply the allpass filter on one channel. If you apply the same allpass filter to both the left and right channels in my midbass example, they completely negate each other and nothing changes...

Hope that helps.

 

Ok so I have one dip in my tune at around -4db that I cannot get level by boosting. The dip is in the mid or in tweeters crossover slope which I’ve tried equing both. Seems it moves a little more when I play with the tweeters slope in eq. The center of this dip would be at 2.16khz roughly. Which all pass filter order would you recommend and which driver should I apply it to? I have one more dip about the same amount more in the mid range and the center is around 1.38khz and I can’t get it to go away with eqing the crossover slope of tweeter or mid range eq. Any guidance on which all pass to use would be greatly appreciated. @jtrosky Thanks

 

Ok so I have one dip in my tune at around -4db that I cannot get level by boosting. The dip is in the mid or in tweeters crossover slope which I’ve tried equing both. Seems it moves a little more when I play with the tweeters slope in eq. The center of this dip would be at 2.16khz roughly. Which all pass filter order would you recommend and which driver should I apply it to? I have one more dip about the same amount more in the mid range and the center is around 1.38khz and I can’t get it to go away with eqing the crossover slope of tweeter or mid range eq. Any guidance on which all pass to use would be greatly appreciated. @jtrosky Thanks

Can you post graphs showing what you are referring to? Keep in mind that allpass filters are not for correcting dips when a single speaker is playing - only when multiple speakers are playing the same frequencies. I think some graphs would clear up what you are referring to.

 

Well the one at 2200 is now gone and I have one around 650 and it’s more like a 2.5dB null

 

tbh, I wouldn't be using bar-style rta with that low resolution to figure out phase issues. way too low hard to get a proper reading of whats going on.

 

How would you recommend me do it? Any tips and help would be greatly appreciated

 

You want to go into the RTA screen settings (the gear icon) and set the "Mode" to "RTA 1/48 octave" (you can always adjust the response "smoothing" level later while analyzing the responses, but this will capture much more detail in case it's needed).

You also want to disable the "bars" setting. If you are using a newer version of REW, click on the "Appearance" button in that settings window and uncheck the "Use bars on RTA" box.

IMO, you really don't want to use the "bars" anywhere, whether it be the RTA window, the EQ window or any other REW window.

 

Here is the screen measurement with bars off and 1/48 res

 

How is this? Also here is how my eq is set up right side is first 3 left side is 2nd 3 pics. Sub isn’t pictured
Right tweeter

right mid range

right mid bass

 

Actually, images that are smoothed down to 1/12th or even 1/6th should be good for this task. Just open the "Overlays" window and in there, you can select which measurements to display and the smoothing level of them. You basically just capture the data in 1/48th mode - but then you smooth it down to various different levels as needed, depending on what you are looking for. But at least you have the highest level of detail if you need it. Taking measurements at 1/6th, for example, may not always have enough detail, depending on what you are looking to analyze.

The Overlays window in REW is where it's at, IMO - it lets you see multiple measurements at once and the "Predicted" tab will show what the measurements will look like after your proposed EQ is applied.

You may know this already, but just in case you don't. Also, REW has a little camera icon in the upper-left-hand corner to take screenshots of the measurements - much nicer than phone pictures of the screen.

 

This thread has been hugely informative for me. Just want to say thanks for all the great input.

 

Left tweeter

left mid range

left mid bass

 

left mid range View attachment 302946

Just my 2 cents, Take a close look at the most common boost you have in the above pic. It looks like you have about 6 EQ bands boosted up 3db's. If you increased the Gain on that channel up by 3db's, move the 6 bands back to 0db, that would free up those 6 bands on that channel for other adjustments....A very general rule, Cuts are better than Boosts.
And, you would need to drop the other EQ bands on that channel by 3db's. Probably would be easier to 'reset' that entire channel, increase the Gain by 3db's and then re-EQ the whole channel.

If you are EQ boosting the Right Mid Range like the Left Mid Range, you may want to increase the Gain on your Amp, instead of boosting the channel in the DSP.

Edit: I just looked at your pic of the Right Mid Range... and you are doing a bunch of EQ boosting there too.... so, best bet would be to increase the Amp gain on Mids, and re-EQ.

 

ok thanks for the info I’ll try and get more shots of it how recommended when I get back home a little later

 

If it's a Windows 10 computer please use the screenshot utility. Makes it easier to look at.