Time and phase align with REW: i need help

[Pieter1267]

Im in a few rabbit holes and days and days trawling youtube mccarthy Mery etc only confuses this neo autistic guy.

All these help align videos dont adress the 'show phase traces' second window. Because aligning spl gives different or worse phase traces my question is: wich one is more important?

Even if spl equal as much is more important: do i then only look at the phase tracing at the frequency wich gives problems or not? Because moving the time slider up or down makes phases go closer at one spectrum of frequencies while going wider at others.

For clarity: equaling the spl wich produces a delay number(or not) is time aligning?

And like @Keith_W said: he prefers phase aligning, that means giving more weight to that lower phase trace graph?

Thanks in advance!


EDIT I thought I couldnt fix the problem, but apparently the new measurement is okay. XO at 70. 2 subs (I aligned them first, added .4 to the left. Then PEQd them). The aligned sum said add 22.75. I put 22.75 on top of the .4 (total 23.15ms delay on left sub, and then 22.75ms delay on the right sub)wich apparently went wrong. I now run both at the by REW suggested 22.75 delay and that gives this

original problem

aligned sum from the alignment tool

 

[terryforsythe]

All these help align videos dont afres the 'show phase traces' second window. Because aligning spl gives different or worse phase traces my question is: wich one is more important?

It depends. For a midrange/tweeter crossover I would say phase is more important. If they are getting the flattest frequency response on-axis with them not phase aligned, the off-axis response (vertical off-axis for a typical midrange/tweeter orientation) is not going to be very good, and the in-room frequency response will suffer a bit. Also, the suckout at the crossover frequency will be worse in one direction (up or down).

For subwoofer integration, especially when using multiple subwoofers and trying to cover a large area, that may not be the case.

What crossover frequency are you experimenting with?

Do you have DSP? Is your setup all active?

 

[Keith_W]

Not enough information in your post. I had to search some of your older posts to find out what system you are using, and what DSP. So far I have learnt: Revel F208 speakers, subwoofers, and MiniDSP.

Okay, a bit of explanation is in order.

Time alignment: the impulse of the subwoofer and woofer are aligned.

Phase alignment: you will find that the subwoofer and woofer have different phase slopes. The goal is to align them at the crossover frequency, and maintain that alignment over as broad a frequency range as possible. Preferably through the entire overlap region if you can manage it. To get the phase to align at one frequency, the delay for the entire subwoofer or woofer needs to be shifted. This will mess up the time alignment, and my argument is: don't worry about it. To get the different phase slopes to match you will need all-pass filters.

McCarthy explains it beautifully. If you can't understand McCarthy, I don't think you would understand any of us because that guy writes so clearly and elegantly. I doubt if any of us regular ASR schmucks can do any better. Persist!

To answer your question about SPL equalisation also changing the phase - you have to realise that the MiniDSP uses IIR biquads. Each EQ you put in is one biquad with a centre frequency f0 and Q. Since biquads are minimum phase, altering the SPL also alters the phase.

To me, this whole minimum phase IIR biquad thing is a complete mess. I don't like it because i'm not smart enough to juggle so many balls at once.

 

[Pieter1267]

It depends. For a midrange/tweeter crossover I would say phase is more important. If they are getting the flattest frequency response on-axis with them not phase aligned, the off-axis response (vertical off-axis for a typical midrange/tweeter orientation) is not going to be very good, and the in-room frequency response will suffer a bit. Also, the suckout at the crossover frequency will be worse in one direction (up or down).

For subwoofer integration, especially when using multiple subwoofers and trying to cover a large area, that may not be the case.

What crossover frequency are you experimenting with?

Do you have DSP? Is your setup all active?

Thanks!

source into pre amp into minidsp 2x4 HD. Minidsp output 3 and 4 into passive subwoofer set (2 Dayton 15"), driven by PA amplifier and the main speakers fed by minidsp output 1 and 2(Revel F208) driven by a Quad 606. I use REW with UMIK1.

At the moment i dont have a fixed XO in mind. I want to understand how the phase alignment works. How to deal with any misaligment issues.

Im trying to learn as much as possible. Ive watched the mccarthy videos, merlijn van veen. I understand maybe 40% of what they say. Getting more grip on various REW tools. The impulse response is another that slowly starts to make sense.

In this post im focussing on trying to get some clarity on the terms time alignment and phase alignment. Some youtube guys dont use those phace tracers at all, just focus on spl.

I also suspect many also maybe focus much on home theatre solutions, with 4 subs. Perhaps the needs for watching movies is different than my user case, music reproduction. Then again, you dont want phase issues watching movies either.

But boy did i spent time the last 7 days on deep, deep level of audio engineering. Its very fascinating.

Not enough information in your post. I had to search some of your older posts to find out what system you are using, and what DSP. So far I have learnt: Revel F208 speakers, subwoofers, and MiniDSP.

Okay, a bit of explanation is in order.

Time alignment: the impulse of the subwoofer and woofer are aligned.

Phase alignment: you will find that the subwoofer and woofer have different phase slopes. The goal is to align them at the crossover frequency, and maintain that alignment over as broad a frequency range as possible. Preferably through the entire overlap region if you can manage it. To get the phase to align at one frequency, the delay for the entire subwoofer or woofer needs to be shifted. This will mess up the time alignment, and my argument is: don't worry about it. To get the different phase slopes to match you will need all-pass filters.

McCarthy explains it beautifully. If you can't understand McCarthy, I don't think you would understand any of us because that guy writes so clearly and elegantly. I doubt if any of us regular ASR schmucks can do any better. Persist!

To answer your question about SPL equalisation also changing the phase - you have to realise that the MiniDSP uses IIR biquads. Each EQ you put in is one biquad with a centre frequency f0 and Q. Since biquads are minimum phase, altering the SPL also alters the phase.

To me, this whole minimum phase IIR biquad thing is a complete mess. I don't like it because i'm not smart enough to juggle so many balls at once.

Thank you!

So its a 2.2 setup. 2 subwoofer cabinets passive, driven by behringer Pa amplifier. The 2 mains indeed, are Revel F208, driven by a 1986 Quad 606mkI.

I use Minidsp 2x4 HD. 4 outputs. 1 and 2 for the mains, 3 and 4 for the subwoofers.

What I do is the following schedule:

First I make a measurement of individual subwoofers

Then I make a measurement of them both.

I then run the individual measurements in the alignment tool to see if and if so, how to align the subwoofers. Usually the better curve is adding some delay on one or the other.

Then I measure both again, and make a PEQ for both. I tried individual peqs, a unique peq for each subwoofer but that doesnt give half as good results as one peq for both.

Then I measure each main speaker, then measure them both. Right now I only run one filter to get a huge node at 90hz down.

Then I set the XO.

Then I measure both subs at said XO

Then I measure both mains at said XO

Is this so far okay?

Then I run a measurement of both subs and both mains at said XO.

If there is a problem (wich there is) I then open the alignment tool, put the mains in A and the subs in B and also let the Phase Traces option run.

I got confused by trying to align as much phase in the graphic as possible. This was the problem.

However, one more question, also for clarity: If I already added a delay to one of the subs to make them play nice together, if I then later in this process use the aligment tool again, and REW suggest to add delay to the subs, do i set both the subs at that delay number, or do I add the earlier added delay on top of this new REW suggestion?

As far as I can see, keeping the .4 delay that was the result of the first process of aligning the indivial subs to play together, and adding the 22.75 for the new aligment on that already time delayed sub (so total 23.15ms ons sub A, and 22.75ms delay on sub B), was a problem. Or do you think differently?

Because if I understand your explanation correctly: in order to have the sub and the mains play nice together, the sub(or in my cases subs) ALL have to get delayed to some extent.

This is new because I thought i some cases you dont need to delay both of them. Thats why these youtube videos confuse me. I thught one of them explained you only need to delay one sub in a multiple sub setup.

(edit: I understand another option is to fysically shift the sub and spakers positions. In my sceario, this is not possible, or not enough (I tried a few things).

 

[Keith_W]

What I do is the following schedule:
First I make a measurement of individual subwoofers
Then I make a measurement of them both.
I then run the individual measurements in the alignment tool to see if and if so, how to align the subwoofers. Usually the better curve is adding some delay on one or the other.
Then I measure both again, and make a PEQ for both. I tried individual peqs, a unique peq for each subwoofer but that doesnt give half as good results as one peq for both.
Then I measure each main speaker, then measure them both. Right now I only run one filter to get a huge node at 90hz down.
Then I set the XO.
Then I measure both subs at said XO
Then I measure both mains at said XO
Is this so far okay?

The XO should be set before you equalise SPL and phase. This is because the XO is minimum phase, and therefore rotates phase. The way you are doing it, you will have to go through another round of SPL adjustment after you apply the XO. The order of operations should be:

1. Decide on XO for sub and speaker and apply it.

2. Time and phase align subs to speakers. Here I have aligned the phase of my sub to the woofer at the 50Hz XO point.

3. Check that the XO sums properly. Measure left sub + left speaker and right sub + right speaker individually. Look for cancellation around the XO point which appears in both measurements. Consider changing the XO order, or inverting polarity, etc. to obtain a better summation. Remember that what you do to one side, you must also apply to the other side.

4. Once your XO's sum properly, start tackling room issues. Sweep left and right subs + speakers then sweep both. Apply PEQ's as necessary.

However, one more question, also for clarity: If I already added a delay to one of the subs to make them play nice together, if I then later in this process use the aligment tool again, and REW suggest to add delay to the subs, do i set both the subs at that delay number, or do I add the earlier added delay on top of this new REW suggestion?

The main point of time and phase alignment is to avoid cancellation at the XO point. The secondary reason is to avoid pushing subwoofer group delay into audibility thresholds - this is less important than the main point and some people might argue shouldn't be a reason at all. If there is no cancellation at the XO point, then there is no need to do anything.

As far as I can see, keeping the .4 delay that was the result of the first process of aligning the indivial subs to play together, and adding the 22.75 for the new aligment on that already time delayed sub (so total 23.15ms ons sub A, and 22.75ms delay on sub B), was a problem. Or do you think differently?

Without seeing the measurement, I can't comment.

Because if I understand your explanation correctly: in order to have the sub and the mains play nice together, the sub(or in my cases subs) ALL have to get delayed to some extent.

This is new because I thought i some cases you dont need to delay both of them. Thats why these youtube videos confuse me. I thught one of them explained you only need to delay one sub in a multiple sub setup.

(edit: I understand another option is to fysically shift the sub and spakers positions. In my sceario, this is not possible, or not enough (I tried a few things).

There is no room big enough for us to time align by physically shifting speakers and subs. It is best accomplished by DSP.

The reason Youtube videos might confuse you is because all of us who DSP have different ways of doing things. Also the DSP strategy changes depending on how many subs and where you have placed them.

This is my way, which might be different to someone else's. I make no claims about my way being better. In fact a friend criticised me tonight for being over the top. I think i'm just being meticulous Whichever way you decide to skin the cat - if it works, it's good. Just remember: for bass frequencies, getting a freq response which is as smooth as possible is the most important priority. Everything else is of secondary importance because phase/timing shifts are less audible, or some people might assert - completely inaudible. Nobody disputes that.

 

[Pieter1267]

The XO should be set before you equalise SPL and phase. This is because the XO is minimum phase, and therefore rotates phase. The way you are doing it, you will have to go through another round of SPL adjustment after you apply the XO. The order of operations should be:

1. Decide on XO for sub and speaker and apply it.

View attachment 442167

2. Time and phase align subs to speakers. Here I have aligned the phase of my sub to the woofer at the 50Hz XO point.

3. Check that the XO sums properly. Measure left sub + left speaker and right sub + right speaker individually. Look for cancellation around the XO point which appears in both measurements. Consider changing the XO order, or inverting polarity, etc. to obtain a better summation. Remember that what you do to one side, you must also apply to the other side.

4. Once your XO's sum properly, start tackling room issues. Sweep left and right subs + speakers then sweep both. Apply PEQ's as necessary.



The main point of time and phase alignment is to avoid cancellation at the XO point. The secondary reason is to avoid pushing subwoofer group delay into audibility thresholds - this is less important than the main point and some people might argue shouldn't be a reason at all. If there is no cancellation at the XO point, then there is no need to do anything.



Without seeing the measurement, I can't comment.



There is no room big enough for us to time align by physically shifting speakers and subs. It is best accomplished by DSP.

The reason Youtube videos might confuse you is because all of us who DSP have different ways of doing things. Also the DSP strategy changes depending on how many subs and where you have placed them.

This is my way, which might be different to someone else's. I make no claims about my way being better. In fact a friend criticised me tonight for being over the top. I think i'm just being meticulous Whichever way you decide to skin the cat - if it works, it's good. Just remember: for bass frequencies, getting a freq response which is as smooth as possible is the most important priority. Everything else is of secondary importance because phase/timing shifts are less audible, or some people might assert - completely inaudible. Nobody disputes that.

This is great info! It makes much more sense to first make the XO up and running, and only then go to the room issues, peq etc. Like i said. I went into way too many rabbit holes at once. Glad I finally admitted defeat and posted this SOS.

In all consternation i forgot to say i changed the 2.2 to a 2.1 setup - recently decided to run the subs dual mono but I get the concept and right work flow schedule now. Many thanks!

 

[Pieter1267]

So far it all works as expected and much more predictable. At step 3 im comparing the left main speaker and dual mono subs vs the right main speakers and dual mono subs.

With the alignment tool I get further in the 52XO experiment and REW now suggest a 37.20ms delay.

My question: is this still okay to listen to? And minidsp only goes to plus 30ms. Should I instead peq and or play with gain levels, or move the XO a bit up or down for different results.

This looks great but REW suggest adding 37.20ms delay



These 2 measurements were put into the alignment tool. This is the right main speaker plus dual mono subs crossed at 52hz.

This is the left main speaker plus dual mono subs, also crossed at 52hz.





mind you, this XO setting doesnt need to be perfect, Im just experimenting.

 

[ernestcarl]

Are you using a loopback acoustic timing reference channel to anchor your measurements when doing the alignment adjustments in REW?

 

[Keith_W]

Are you using a loopback acoustic timing reference channel to anchor your measurements when doing the alignment adjustments in REW?

Genuine question: He has a UMIK-1. If he is going to do a loopback measurement, it will have to be via WASAPI. And that probably means software cables, such as VAC, ASIO4ALL, etc.

Loopback measurements measure the delay of the entire signal chain from REW downstream including WASAPI latency, software cable latency, USB microphone timing variability, etc. It is therefore sensitive to sampling rate, buffer size, and so on. I suspect that attempting a loopback measurement without an interface would give you timing measurements with poor repeatability. I personally think acoustic timing is more reliable, and if you are going to do loopback, you need an interface to avoid all these variable latencies. Is this wrong?

 

[ernestcarl]

Genuine question: He has a UMIK-1. If he is going to do a loopback measurement, it will have to be via WASAPI. And that probably means software cables, such as VAC, ASIO4ALL, etc.

Loopback measurements measure the delay of the entire signal chain from REW downstream including WASAPI latency, software cable latency, USB microphone timing variability, etc. It is therefore sensitive to sampling rate, buffer size, and so on. I suspect that attempting a loopback measurement without an interface would give you timing measurements with poor repeatability. I personally think acoustic timing is more reliable, and if you are going to do loopback, you need an interface to avoid all these variable latencies. Is this wrong?

Sorry, to be clear I meant either some loopback or acoustic timing reference. Now, I've only ever used loopback with another pro audio interface with xlr cables and an omni mic a very long time ago... can't even remember how it went. For a USB UMIK, yes, acoustic time reference it is... There can still be issues/glitches, of course -- most often I experience being an interruption in the middle of a sweep.

 

[Keith_W]

My question: is this still okay to listen to? And minidsp only goes to plus 30ms. Should I instead peq and or play with gain levels, or move the XO a bit up or down for different results.

View attachment 442215

This looks great but REW suggest adding 37.20ms delay

Please read this thread. Your vertical scale is too wide, it makes that SPL graph almost impossible to read. Set the Y axis limits to something like 50dB - 100dB then we can see a lot better.

 

[Pieter1267]

Please read this thread. Your vertical scale is too wide, it makes that SPL graph almost impossible to read. Set the Y axis limits to something like 50dB - 100dB then we can see a lot better.

At the moment im very busy experimenting with MSO, learned about all pass filters. (Big shout out to @DaveBoswell for finally making it clear to me what an all pass filter is and how it works, and how to fiddle with it in MSO.)

I have another question:

The MSO feauture is fantastic, but after its done making the subs playing nicely together, if I afterwards implement an XO to mate it with my main speakers it would kind of push the results off again, no? I thought trying MSO subs+ mains would make it work but as @andyc56 said, not many people seem to use that feature and after some trials, I find it very difficult and decided im going to use MSO now only to match my 2 subwoofers.

The question is then, what is the correct procedure for building a 2.1 or 2.2 setup if you are going to use MSO?

First set the XO and let MSO aim for that setting?

Or can you first let MSO have the subs work together for a full sub range from say 10-200hz, and then after inserting the MSO biquads, do REW measurements with the mains and work with the alignment tool? In my head that creates a carcrash because the gain and delay settings after MSO are set, and any alteration because of the REW alignment tool, would make all that MSO optimization, void and null.

On top of that: if MSO only is used for making the 2 subs work together...later implementing an XO - and its inherent time and phase altering...how does one go about it?

 

[Keith_W]

The question is then, what is the correct procedure for building a 2.1 or 2.2 setup if you are going to use MSO?

You are much better off asking Andy instead of me. I don't use MiniDSP nor do I use MSO. I have only played with MSO, so I am hardly an expert on it.

But I can guess. MSO does its optimisation based on the measurements you feed it. If you want to implement an XO with MSO, I would suggest you put your XO's into your MiniDSP first and then take measurements for MSO. Run MSO with the proper settings. You then get a bunch of biquads. Put those biquads into your MiniDSP and then re-measure to confirm that the correction worked. I make no claims that what I am suggesting is the correct procedure, but it is what I would do.

 

[Pieter1267]

Well, about 60 hours later:

I threw out the Dirac and downgraded the DDRC 24 to a 2x4HD

Then I ran new MSO testing

This was the baseline how I measured, MLP from individual sub measurements. The blue line is the MLP. Bit wonky res, sorry I dont know how or what settings PNG needs.

Actually pretty good till about 70, give or take a few db

Then I ran Multistage MSO

Flat with a slight null at 80hz. The reference was set at 78db.

It correlates to a T afterwards with REW, except maybe a drop of 2db in output with stereo subs. This is the measurement with both subs in stereo input

This is in Dual Mono, 3db above the reference.

Afterwards I implemented a 100hz XO in MiniDSP. The full range stereo result ha d aslight dip but the mono config looks pretty good

I knew that for some reason the spike at 92hz is from the main speakers so I decided to implement a slight PEQ (92hz, -3db Q6.5)

The spike went down 2db but immediatly the phase looked upset and above the XO line more nulls

But not bad! I finally figured out how to make adjustments and use MSO and despite a slight difference in actual SPL, it correlates well with little XO marriage problems.

Now im going to relax and later this week see if the Subs+Mains config can make it even better. The above was with an XO set at 100hz. The 60hz setting wasnt so good. Either way, im happy. Good days work! And MSO is real.