REW Alignment tool - Guides or Manual?

[jtalden]

SW to SW Timing:
Was the cursor located at a frequency where the where the room modes were quiet? A room mode impacting one of the 2 SWs at the cursor frequency can cause a problem with the alignment tool delay result. Try setting the cursor at frequency below the XO frequency at a location where the SPL and phase are relatively smoothly changing. Avoid a frequency where there is a significant dip in SPL and sudden variation in phase of either of the 2 SWs.

XO timing - SW's to Mains:
The issue is similar, but the cursor must be placed within the XO range. If the modes disrupt the SWs or mains too much in that range it is probably better to set a different XO frequency.

Post the measurements if you want a second opinion on the timing or want us to better understand the problem.

 

[fattire]

So I did some experimenting. I tried manually implementing a sub delay where the SPL looked good in the alignment tool and set not too far from the distance determined by YPAO (about 5 feet higher). But then if I click Align phase slope at cursor, it gave me a much larger distance of 18 feet above the original YPAO distance I think due to reflections in the room, as when I did the same with FDW to 6 cycles the number was about 5 feet higher. Then when I measured the actual response with both distances, the larger one looked better in the crossover region. What would be the downside of using a high delay like that on the sub if the frequency response looks the best? I'm not really noticing a lipsync error and the bass sounds good to me. Just confused because in Jeff's video I believe he alluded to not wanting the distance to be too far from the actual distance, but if the frequency response looks the best in one's room with a long delay and one is not seeing a lip sync error, what is the problem?

At these frequencies, there really isn't much of a downside. As John mentioned above in #196 subs don't really have an impulse response the way that we think of them for higher frequencies. The best analogy I've seen is that sub frequencies are like a train. When does the train arrive? When the engine gets there? When the caboose passes? At the mid point of the train? Compare that to high frequencies that are a Porsche passing the same listening position at highway or racing speeds.

In the low frequency range, our hearing is just much better at detecting an uneven frequency response than it is a seemingly large delay in timing. I'll happily make the same trade as John - delay I can't really perceive in favor of frequency response that I can absolutely perceive.

 

[gregw]

Thanks for the offer jtalden. I am away from home for 2 weeks but when I get back I may share some measurements. Another question though : in this thread you recommend taking individual L and R measurements and then use the Arithmetic trace to sum them and align that arithmetic trace to the sub in the alignment tool. Why not just take a L+R measurement and align the L+R to the sub? Forgive me if this was answered previously.

And thanks Jeff for the answer on the delay question.

 

[jtalden]

It can be done that way also. I just tried to provide a method to get all the information needed for a setup using the fewest measurements. It is also the best way to measure if someone wants me to help with the analysis as it is the method at provides me some additional ways to check if the measurements are properly timed and the mic is properly centered.

 

[gregw]

Replaced my yamaha with a denon x4500h that has multieq XT32. The subwoofer got nicely smoothed out by Denon. Trying to pick the right delay now. Using one sub. I ran some measurements. Using the alignment tool, I came up with a distance of 17.5 feet. Note that the sub alone and L+R alone measurements were taken with the distance already set to 17.5 ft by me. Tried other distances but nothing looks better in the xover region. Ran a few other distances all shared. Any suggestions for improvement. Getting foci of 10 db drop between 80 and 100 Hz.

 

[jtalden]

As you determined, a 17.5 ft. setting is favorable compared to other settings with these particular settings.

Room effects are responsible for the SPL irregularity through the 80 - 100 Hz range. Changes to the location of speakers and/or the LP may help this significantly, but often it just moves the SPL issues to other frequencies.

Also, There is some indication that reversing the SW polarity prior to running Audyssey may help reduce the SPL variation in XO range with this current setup. Audyssey EQ will account for other SPL changes that would otherwise occur if it is done now, after Audyssey is run. The SW distance/delay setting will also be changed a little.

Note:
For timing work it is necessary to measure the setup as it will function in usage. The posted measurements show a larger SPL overlap range than expected if an 80 Hz XO was active. That is fine if this is how it will actually be operated. Normally main speakers are set to 'small' and measured with the XO active. The mains here do appear to rolloff as expected with a normal 80 Hz XO frequency setting. The SW is then measured using a front channel with that main speaker disconnected so that the redirected bass is measured. The SW SPL thus will rolloff quickly above the XO frequency. Here the rolloff starts somewhat higher than expected. Possibly the LFE channel was measured by mistake, but if not, it is okay. I have seen this wider XO range occur sometimes even with the correct settings. It is not necessarily a problem.

 

[gregw]

Good eye. Yes you are right I thought that I needed to measure the sub with the XO set to 250 Hz, and I had the mains set to Large. The mains are SVS Prime Satellites so that is why they appeared small on the graph. The sub is an SVS SB1000 Pro. If you think it is necessary, I can rerun those initial measurements with the XO set to 80 for both subs and mains. But like you said, I dont know if it will make a difference in the analysis. I don't really have any options for moving the sub with most floor space in my apt living room taken up by furniture. The mains could be slightly moved on the furniture they are resting on but not by much.

I am intrigued by your suggestion to reverse the polarity. I actually did try that (without rerunning audyssey) and made a measurement of the sub alone. I have attached it. The freq response almost exactly matches the one with positive polarity so are you sure I need to rerun Audyssey? I toyed around with neg polarity and various distances but nothing looked better than what I already had. What distance would you suggest trying?

 

[jtalden]

See the note in Post-206. Any analysis for distance adjustments will not be correct without the XO active and proper measurements:

• L+R (alone)
• SW (alone with redirected bass)

Small changes to the position of the mains locations alone are not likely to have much impact on the bass SPL.

 

[gregw]

Measurements correctly done with xovers at 80 and 90 for comparison. I also did L alone and R alone. Comparing to my sub alone measurement at 250 Hz, I didnt realize how much the sub was being attenuated by the filter really starting at 60.

 

[jtalden]

There did not appear to be any significant difference between the 80 and 90 Hz XO options. So I only looked at the 80 Hz measurements. Using the first 3 measurements I found 2 optional SW distance settings based on improved phase tracking through the XO range. They smoothed out the SPL in the XO range, but left a large sag in the SPL through that range instead. Overall, the SPL of those 2 options did not look as favorable as the current distance settings.

These timing options may be an improvement if the SW level was increased several dB and the EQ could be adjusted accordingly, but that probably is not possible using Audyssey. You are thus best served by just using the current settings and adjusting the SW level to your preference.

 

[gregw]

Thanks for your analysis and the graphs.

Audyssey would allow me to make changes to the speakers and SW target curves in the multieq editor app. I was thinking I could boost the mains and sub by a few db from around 60-120 to see if there is any improvement.

I was wondering is applying 1/24 smoothing your standard?

 

[jtalden]

We would normally use 1/48 smoothing for SPL in the bass range. I simply picked 1/24 smoothing as I thought it would be a little easier to follow the close traces without losing any significant detail. This choice is not better than other settings depending on what we are trying to learn.

Attached is my mdat so you can change smoothing or review other charts of the data.

 

[gregw]

Thanks so much for sharing your data. I looked through it and was using the alignment tool to duplicate what you did. Super helpful. But then I took my L+ R measurement and tried to align with the sub using the settings you recommend, and the SPL over the XO region doesn't look as smooth as it does when comparing the arithmetic trace of L+R vs sub. I am confused - shouldn't these look the same? ie, the arithmetic trace of (A+B)/2 should have the same phase as a L+R measurement? Also, why not use the arithmetic trace of A+B rather than (A+B)/2? The SPL of A+B exactly matches my direct L+R measurement whereas A+B/2 is a few db lower.

I also reviewed some of your previous posts in this thread. Now that John has released a stable 5.2 version of REW, maybe some of your summary posts can find their way into the help pages for REW. Especially the summary idiot's guide steps!

 

[jtalden]

But then I took my L+ R measurement and tried to align with the sub using the settings you recommend, and the SPL over the XO region doesn't look as smooth as it does when comparing the arithmetic trace of L+R vs sub. I am confused - shouldn't these look the same? ie, the arithmetic trace of (A+B)/2 should have the same phase as a L+R measurement?

Yes, a new measurement using a new distance setting will match the alignment tool prediction of that same situation. When it does not there is a difference somewhere. It could be and error in the analysis or in the settings change or possibly in the measuring conditions (mic position, etc.).

Also, why not use the arithmetic trace of A+B rather than (A+B)/2? The SPL of A+B exactly matches my direct L+R measurement whereas A+B/2 is a few db lower.

We can do that.

[While there are other ways to do measurements for analysis, my default method is to measure L, R, and SW with signal output to a single measurement channel. This way the SW level matches with the L and R individual measurements. Then when doing the L+R+SW combined analysis, the L+R is averaged to sum with the SW measurement. The result will have the correct relative level between the L+R with SW. Since all 3 traces are at the same level it is easy to see the relative differences between them. This method results in a 6 dB lower in overall level from a L+R+SW measurement using a L+R output signal. To compare the analysis result to a new L+R+SW measurement we can just shift one of them by 6 dB.

I have sometimes also summed the L+R and added 6 dB to the SW within the alignment tool to get the same level that a L+R+SW measurement will. For the analysis within REW this level can then be reduced 6 dB to compare it more easily with the 2 individual channels.]

 

[gregw]

Excuse me for being new to all this, but I just want to take a step back on theory here and ask what is maybe a dumb question. I am confused as to why everyone time aligns their SW to main Left and Right speakers in a home theater when 80% of the output in a movie is in the center channel and that center channel has deep male voices that can go down to 80 or 90 Hz right in the xover region. Plus movies have their own LFE track for the subwoofer going up to 120 Hz where crossover should be irrelevant (I think). I suppose many people use their home theaters for music, but for someone like me who use the home theater 95% of the time for movies and TV, should I be focusing my time alignment on the center channel xover rather than the mains?

 

[jtalden]

There is no requirement that FL and FR need to be analyzed and timed first using the alignment tool. The order that the channel timing is done will not impact the timing accuracy of the other channels. I have seen others prefer to start with the CC.

The FL and FR do need to be treated as a group. If they are timed differently then any center panned sounds would appear to be off-center. In many setups the FL and FR channel phase does react very differently in the XO range due to room geometry. I just prefer to first analyze them and choose a good compromise SW timing.

 

[gregw]

But if there is only one subwoofer distance that you can change, how would you have different timings for the center vs for the FL/FR? You would make changes to the FL/FR distances? But dont those distances need to correspond roughly to the actual distance?

It seems like you would have to choose whether to make the center time aligned to the SW or the FL/FR time aligned to the SW but you couldnt have both.

If I am right that you shoudl only be changing the sw distance setting, then you would analyze the optimal setting for FR/FL alignment and then analyze for center channel, and then implement a compromise setting?

 

[jtalden]

We would just change the CC distance setting in the AVR.

Maybe your AVR automated Setup/EQ system does not let you change speaker distances after running it? If it won't let you make changes without loss of EQ settings then that is a problem. Others may know of workaround options, but I am not familiar with your equipment capabilities and limitations. You can always just setup to the FL and FR and just check to confirm if the timing of the CC is really that bad for SPL (or the reverse approach). If there is a notch in SPL near the CC XO, just invert the CC polarity to see if it looks better.

 

[gregw]

Audyssey does let me change distances, individual xovers, and levels while leaving multieq xt32 ON. OK, interesting, I had no idea the implied next step is to look at the alignment of other channels with respect to SW after aligning FL/FR with SW and then change distances of each speaker. Wow this is an enormous amount of work if you start adjusting center and surround channel pair timing. It's hard enough to get FL/FR aligned. hopefully the others wouldnt be too far off. Thanks

 

[gregw]

jtalden I got a chance to do some eq with the multieq editor app and I moved the sub back closer to the wall by about 6 in and then redo the alignment and the results are much better! I want to thank you. I improved the L+R and then aligned the CC. It's not perfection but now the variation is within +/- 3 db. I think the problem is that my Denon Audyssey might have been EQing my SVS Prime Satellite front mains as if they were larger speakers because they were really bringing down their output below 90 Hz. I know this because I measured them without and with Audyssey turned on. So used the multieq editor app to boost them 6 db at 80 Hz and 3 db at 100 Hz in the XO region where I was losing output. I have a control points at 0 db at 70 and 110 Hz. I am attaching this speaker measurement file audy on vs off (sub turned off).

A few questions
1) Is it ok to be boosting these satellites at 80 and 100 Hz? It looks like they start to roll off around 80 on the non-eqed plot, and I never listen at reference - the most would be about 10 db below reference. If I lived alone I would get towers or bookshelves but I have 3 kids who like to knock things over so at this point it's better for me to use the satellites.

2) The other question I have is with regard to my CC. It's an SVS prime center and its response alone showed it would work well with XO at 60 Hz. Its graph looks great in XO region after aligning by distance in AVR, but it does have a null by itself at 90. My sub also has a null there. So I think my only option to fix that would be to change the LP or move the CC closer?

3) For me despite my best efforts to prevent any changes during measuring conditions (ie, nothing in room or mic position changed except maybe temp increasing by a degree while I have to turn off the AC), for L+R+sub alignment the Aligned Sum would always look slightly different than when I change the SW distance in the AVR to the exact number and remeasure. I would adjust by 0.5 feet in either direction and then the measured result would match the aligned sum almost exactly. Whereas with aligning the CC, the optimal aligned sum matched the measured result precisely after making the change the CC distance. Any idea why this might be? It makes it harder to implement the alignment tool.

4) and last - slight variation in mic position by a few inches or slight change in position of pillow or toys in the living room floor from one day to the next affect the measurements. Is this normal?

 

[jtalden]

1) Is it ok to be boosting these satellites at 80 and 100 Hz? It looks like they start to roll off around 80 on the non-eqed plot, and I never listen at reference - the most would be about 10 db below reference. If I lived alone I would get towers or bookshelves but I have 3 kids who like to knock things over so at this point it's better for me to use the satellites.

I would think 6 dB is okay given your listening level.

2) The other question I have is with regard to my CC. It's an SVS prime center and its response alone showed it would work well with XO at 60 Hz. Its graph looks great in XO region after aligning by distance in AVR, but it does have a null by itself at 90. My sub also has a null there. So I think my only option to fix that would be to change the LP or move the CC closer?

Yes, I think it would be very difficult to avoid this null. Many of us have a similar situation.

3) For me despite my best efforts to prevent any changes during measuring conditions (ie, nothing in room or mic position changed except maybe temp increasing by a degree while I have to turn off the AC), for L+R+sub alignment the Aligned Sum would always look slightly different than when I change the SW distance in the AVR to the exact number and remeasure. I would adjust by 0.5 feet in either direction and then the measured result would match the aligned sum almost exactly. Whereas with aligning the CC, the optimal aligned sum matched the measured result precisely after making the change the CC distance. Any idea why this might be? It makes it harder to implement the alignment tool.

I don't.

4) and last - slight variation in mic position by a few inches or slight change in position of pillow or toys in the living room floor from one day to the next affect the measurements. Is this normal?

If this is a 1-2 dB variation in a small number of places particularly near nulls as in your last 2 measurements, that is not unexpected. If the differences are significantly greater or over wider frequency ranges then it is not normal in my limited experience. You could experiment with small mic placement changes to see how much SPL impact occurs. Toys, pillows, temperature and other minor room changes are not significant factors in my experience. Again, you can test this for yourself.

 

[welldun]

What is the difference between the Align Phase Slope at cursor and Align Phase at cursor? For, using the latter gives a cleaner result.

 

[jtalden]

Align Phase Slope at cursor: This virtually shifts the measured data near 0 ms such to provide a clearer view of the relative phase. Both channels are shifted together so there is no change in the phase relationship between them. It is very helpful top clarify the view when the impulse responses located well away from 0 ms. It can be turned on, or off as desired. The actual data is not changed in any way. Manual or 'Align phase at cursor' adjustments can still be made in this view to adjust the relative timing. When using acoustic timing on HDMI 1, 2, or 3 as the timing channel then the impulses are already very near the same position of the impact has little effect.

Align Phase at cursor: This makes a virtual adjustment of the relative timing to suggest a new more favorable XO timing. It often does that well, but room modes in XO range sometimes causes phase disruptions that interfere with this calculation so it is necessary manually evaluate that timing suggestion. It is thus most reliable when the 2 phase traces are smoothly changing through the midportion of XO range.

 

[Duper]

Hey Guys, this is my first post so please don't be harsh. I have been playing with REW for about 2 weeks. I still don't have calibrated mic so I was using the mic that came with the Onkyo receiver. Doing all sorts of measurements and stuff. I was trying to time align my L and R speakers (they do not look too aligned in the graphs). And then time align my only SW to the mains. I have few questions:
1 - Does the Left and Right look aligned to you guys? Crossover on the receiver is set to 80hz. Right speaker is a bit further from the Left one.
2 - When I do "trace arithmetic /2 to get average" it is way lower than the actual signal from L + R played together. Is that normal?
3 - when I was trying to align SW to the Left speaker. I was changing distance for the SW in the receiver. I think I got it close to 1.xx ms at 10.6ft. The question is, if I take multiple measurements without changing mic position or anything, I am getting different result. Sometimes it shows delay in 18ms range and sometimes 1.xx ms why is that? You can see that in 5.6 and 10.6 measurments.

 

[welldun]

Hey Guys, this is my first post so please don't be harsh. I have been playing with REW for about 2 weeks. I still don't have calibrated mic so I was using the mic that came with the Onkyo receiver. Doing all sorts of measurements and stuff. I was trying to time align my L and R speakers (they do not look too aligned in the graphs). And then time align my only SW to the mains. I have few questions:
1 - Does the Left and Right look aligned to you guys? Crossover on the receiver is set to 80hz. Right speaker is a bit further from the Left one.
2 - When I do "trace arithmetic /2 to get average" it is way lower than the actual signal from L + R played together. Is that normal?
3 - when I was trying to align SW to the Left speaker. I was changing distance for the SW in the receiver. I think I got it close to 1.xx ms at 10.6ft. The question is, if I take multiple measurements without changing mic position or anything, I am getting different result. Sometimes it shows delay in 18ms range and sometimes 1.xx ms why is that? You can see that in 5.6 and 10.6 measurments.

I hate to say what you don't want to hear, but if I had to guess, the issue is that you are not using a calibrated microphone.
Are you using a reference speaker for the time alignment?
There is an old thread (2010) on AVSforums from a few members that were trying to use REW using the microphone supplied with their AVRs. They all seemed convinced that the Frequency response graph they generated were pretty accurate, but also agreed that the SPL measurements were not accurate.

Before going any further, I'll ask if you've already read AustinJerry's simplified REW manual? That manual can atleast help you get all of the other step correct if you don't have the proper mic, so that the only error could come from the mic and not necessarily the process that you used. I've attached it to this post in case you need it.

All that aside, if your setup is not very complex, the auto-calibration in your Onkyo AVR will get you very close with both the SPL levels and the speaker distances. The subwoofer distance is usually a little tricky because if it is a self-powered subwoofer, the amplifier that it has can add delay to the signal, which is why the subwoofer distance when measured via a microphone is usually not the same as the actual physical distance of the subwoofer itself.

Having access to both a UMIK-1 calibrated microphone as well as the AVR supplied microphone, I can tell you that the result are different when using them with REW.