Audyssey XT32 vs REW comparison

[robh]

I have a question about those who use Audyssey and measure their results afterwards with REW. In short, are my results normal (below)? I'm a newbie on all this, forgive any dumb questions.

I have a Marantz SR6015 receiver with three identical Ascend Sierra-LX speakers. (The center is on its side, but this post is just about left and right of TV speakers to make the post shorter). The famous Klippel measurements from Ascend:
https://ascendacoustics.com/collect...roducts/sierra-lx-pair?variant=40080757162038

My room is a normal 60's Ranch house living room. Fairly lively, 8' ceilings, minimal treatment. Speakers were measured about 8 feet away on a tripod.

I ran Audyssey MultEQ XT32 on my 3.1 speakers, which includes an SVS Sub. Target curve is High frequency Roll Off 2, no MRC, Full range compensation. No DV or DynEQ for these measurements. Sub crossed at 60Hz for the Ascends (which plays into some of these graphs...I think??). I made a couple slight adjustments to the curve, which is +1dB here and there, but it should be fairly spot on to normal.

I use the MultEQ app on my iPad for measurement control. I used 6 of the 8 positions for measurement. For REW, I just used the single MLP. My laptop has HDMI out to the Marantz, with a calibrated UMIK-1 mic for input. For Audyssey, I used the included microphone with the Marantz.

First, the Audyssey ideal "wishful thinking" curve from the iPad.

Before and after room correction measurements...and yet even more wishful thinking tm. Left and Right.

Here are the REW results at 1/6 smoothing. It includes the sub (and center), because in the Ratbuddyssey graphs below I believe the sub is included in the L&R results since I set them as small.

Here are the Ratbuddyssey graphs of the Audyssey after output, so includes the sub.
My last question for everyone: For Audyssey 1/6 smoothing compared to the REW 1/6, is this about what you get too? Concerned only for 200Hz and above since my sub affects the bottom. It looks very uneven.

Left speaker:

Right speaker:

 

[peng]

It looks normal to me, but I would suggest for the bass range you should use no smoothing, or 1/12, 1/24 smoothing so you can see if there are room for improvements especially when using Rat, that does allow you to tweak the result for smoother bass.

 

[peng]

Here are the REW results at 1/6 smoothing. It includes the sub (and center), because in the Ratbuddyssey graphs below I believe the sub is included in the L&R results since I set them as small.

The Ratbuddyssey graphs for the L, R should represent just the L, R, not the sub. You need to select the sub to see the graph for the sub.

 

[robh]

It looks normal to me, but I would suggest for the bass range you should use no smoothing, or 1/12, 1/24 smoothing so you can see if there are room for improvements especially when using Rat, that does allow you to tweak the result for smoother bass.

Although I didn't focus on the sub, the resulting Audyssey calibration should have smoothed that out right? It doesn't seem so.

The REW results (I assume) are the most important, regardless of the graphs Audyssey spit out. I bought the UMIK-1 mic just for the purpose of crosschecking Audyssey, and the main reason for wanting to do that was the horrible resulting ratbuddyssey graphs. But for some reason, it looks like it came together in the end.

 

[peng]

Although I didn't focus on the sub, the resulting Audyssey calibration should have smoothed that out right? It doesn't seem so.

It should have, but in order for us to see the effects, you have to sweep both with Audyssey on and Audyssey off. Audyssey typically can do a good job flattening the room gain. I think you probably should re-run Audyssey, but this time follow instructions to the letter, assuming you didn't last time, but obviously I shouldn't assume anything.

 

[peng]

Here's an example:

No smoothing sub1+sub2:

 

[robh]

I should be able to measure it later today, or tomorrow. Will update!

 

[chych7]

Few other comments:
- I like using Var smoothing, gives a good representation of the whole FR range and leaves bass unsmoothed.
- Since you have a pre 2022 AVR, you need to multiply the computed distances by 0.88 to fix the distance bug. Or better yet, take measurements in REW at the MLP with acoustic timing reference on, and see how well the speakers are time aligned. Then adjust distances until they are correctly aligned.
- Audyssey often does not blend sub to speaker well. Take L+R with sub sweeps, and tweak the sub distance until the response looks the smoothest over the crossover.

 

[peng]

After tweaking with the app and Ratbuddyssey

With 1/24 smoothing, it looks quite smooth, the graph is for front left, front right and two subwoofers together, with XO 80 Hz.

 

[peng]

Few other comments:
- I like using Var smoothing, gives a good representation of the whole FR range and leaves bass unsmoothed.
- Since you have a pre 2022 AVR, you need to multiply the computed distances by 0.88 to fix the distance bug. Or better yet, take measurements in REW at the MLP with acoustic timing reference on, and see how well the speakers are time aligned. Then adjust distances until they are correctly aligned.
- Audyssey often does not blend sub to speaker well. Take L+R with sub sweeps, and tweak the sub distance until the response looks the smoothest over the crossover.

To me, the 0.88 fix made very little difference, in terms of audibility I think none, but sure why not?

 

[robh]

Few other comments:
- I like using Var smoothing, gives a good representation of the whole FR range and leaves bass unsmoothed.
- Since you have a pre 2022 AVR, you need to multiply the computed distances by 0.88 to fix the distance bug. Or better yet, take measurements in REW at the MLP with acoustic timing reference on, and see how well the speakers are time aligned. Then adjust distances until they are correctly aligned.
- Audyssey often does not blend sub to speaker well. Take L+R with sub sweeps, and tweak the sub distance until the response looks the smoothest over the crossover.

I did the .88 multiple thing on my AVR after I uploaded the calibration. I looked up several guides and ran it several times. Most results look the same as I've posted. On your last point, do I simply change the sub distance in my AVR and re-run REW sweeps until it looks better?

Here are my REW and Audyssey files. REW is "after" Audyssey on Reference. I'll try to run straight sub this weekend. If anyone can give me some advice to tweak...
My files

 

[chych7]

I did the .88 multiple thing on my AVR after I uploaded the calibration. I looked up several guides and ran it several times. Most results look the same as I've posted. On your last point, do I simply change the sub distance in my AVR and re-run REW sweeps until it looks better?

Here are my REW and Audyssey files. REW is "after" Audyssey on Reference. I'll try to run straight sub this weekend. If anyone can give me some advice to tweak...
My files

The frequency sweeps will look the same with the 0.88 fix, but the time alignment will be different. If you measure each speaker individually (with acoustic timing reference on) and look at the impulse response on an overlay plot, you can see how well they are time aligned. The 0.88 fix should better align them. In my system I can hear this in terms of how precise the image is. Also look at this thread and the impulse response plots shown: https://www.avsforum.com/threads/au...ff-according-to-rew-impluse-readings.3243684/

For the sub, yes just slightly adjust the sub distance, run REW sweep (L+R+sub) and find the sub distance that makes the sweep look the best.

 

[peng]

I did the .88 multiple thing on my AVR after I uploaded the calibration. I looked up several guides and ran it several times. Most results look the same as I've posted. On your last point, do I simply change the sub distance in my AVR and re-run REW sweeps until it looks better?

The frequency sweeps will look the same with the 0.88 fix, but the time alignment will be different. If you measure each speaker individually (with acoustic timing reference on) and look at the impulse response on an overlay plot, you can see how well they are time aligned. The 0.88 fix should better align them. In my system I can hear this in terms of how precise the image is. Also look at this thread and the impulse response plots shown: https://www.avsforum.com/threads/au...ff-according-to-rew-impluse-readings.3243684/

For the sub, yes just slightly adjust the sub distance, run REW sweep (L+R+sub) and find the sub distance that makes the sweep look the best.

That AVS thread has 71 pages, it would take me days to read most of them. Anyway, we all know the mic measures delays, not distance. Logically, D+M wouldn't want the $20 app to measure delays, then convert the delays using the incorrect speed of sound, and then calculate/configure the filters, that would be extra work for nothing.

I would think that they would just use the delays measured by the mic and calculate/configure the filters without the redundant conversions.

Now, the app does, like the AVR/AVP display distance, so if someone tweak the distance, then I would assume the app would have to do the distance to delay conversions first and then calculate/configure the filters and then send the info to the unit.

Most users probably don't tweak the distance post Audyssey, may be that's why D+M never say anything about the speed of sound error. Someone posted the response from D+M (Europe) in which D+M customer support said it was not a bug, and basically all was fine (not their exact words but you get the idea..).

I never tweaked distances, found no need to worry about the speed error in my case based on my understanding, but in your case, if you can actually see the effect from the impulse responses, and could hear a difference using the 0.88 in terms of more precise image, then that's great. Just that I highly doubt most (some do, I believe, but it depends..) would have the need to worry about this speed error thing, though again, it is easier enough to do, for those who don't mind spending a little time on it. To me, D+M's response is reasonable, and I think Audyssey overrated the issue a little, nothing wrong with trying to be more accurate where possible, so good for them.

 

[chych7]

That AVS thread has 71 pages, it would take me days to read most of them. Anyway, we all know the mic measures delays, not distance. Logically, D+M wouldn't want the $20 app to measure delays, then convert the delays using the incorrect speed of sound, and then calculate/configure the filters, that would be extra work for nothing.

I would think that they would just use the delays measured by the mic and calculate/configure the filters without the redundant conversions.

Now, the app does, like the AVR/AVP display distance, so if someone tweak the distance, then I would assume the app would have to do the distance to delay conversions first and then calculate/configure the filters and then send the info to the unit.

Most users probably don't tweak the distance post Audyssey, may be that's why D+M never say anything about the speed of sound error. Someone posted the response from D+M (Europe) in which D+M customer support said it was not a bug, and basically all was fine (not their exact words but you get the idea..).

I never tweaked distances, found no need to worry about the speed error in my case based on my understanding, but in your case, if you can actually see the effect from the impulse responses, and could hear a difference using the 0.88 in terms of more precise image, then that's great. Just that I highly doubt most (some do, I believe, but it depends..) would have the need to worry about this speed error thing, though again, it is easier enough to do, for those who don't mind spending a little time on it. To me, D+M's response is reasonable, and I think Audyssey overrated the issue a little, nothing wrong with trying to be more accurate where possible, so good for them.

From the thread, I meant to just point out to the first post that shows the impulse response misalignment. But yeah I think what D+M does is "good enough" for most people, although they did fix the issue in the current 2022 AVRs, so there was some acknowledgement that the calculation was incorrect.

In a typical acoustically challenged room it can be hard to hear the differences between the distances - the extra reflections in an acoustically challenged room smears out the image. My room is well treated, so I am mostly getting direct sound and late reflections, which gives high clarity and imaging precision. If I play something with a strong center image, and then start moving up/down the distances slightly on one speaker, it's audible how the image sharpens/smears (this is a good simple experiment to do). For me it was worth it to get the time alignment precise (to within ~2 cm equivalent distance at the MLP). In this hobby we are looking to eek out that last 95%+ of sound quality in our systems; time alignment tweaking is one of those things that can be done for that, and is a "free" upgrade.

 

[robh]

Here are my REW measurements with and without Audyssey enabled, but speaker distances and such were configured at kept from the Audyssey calibraiton. It seems I have something going on around 80Hz that Audyssey itself isn't fixing.

Blue = Audyssey off
Orange = Audyssey on

 

[peng]

Here are my REW measurements with and without Audyssey enabled, but speaker distances and such were configured at kept from the Audyssey calibraiton. It seems I have something going on around 80Hz that Audyssey itself isn't fixing.

Blue = Audyssey off
Orange = Audyssey on

That looks great! Audyssey improved the 80 Hz dip by more than 10 dB !!!!!!!!!, you can't complain about that..

Also, keep in mind Audyssey has to try and integrate the low frequency response with the bass capable speakers so they have to do some balancing act. If they were to flatten the Sub's response, then the combined response (example: with the front bass capable speakers) would suffer.

From what I can see, you have the potential to get the response from15-150 Hz flat to within +/- 2.5 dB easily (even with no smoothing) if you spend a few hours tweaking with the app using Ratbuddssey. Or you can spend less than two hours doing the same and achive +/- 3 to 3.5 dB and then customize your own bass tilt (so called Harman curve), but that's up to you as not everyone likes such a tilt (I don't particurlay feel the need for one), while some prefer to just bump up the sub's level trim by a few dB and be done with it.

 

[robh]

That looks great! Audyssey improved the 80 Hz dip by more than 10 dB !!!!!!!!!, you can't complain about that..

True! But is there anything I can do in the analog space to fix that? I admit I just plopped my subwoofer close to the corner (but not AT the corner) of my living room. I didn't want artificially heightened bass, but could try that, or add a bass trap or something in that corner.

Also, keep in mind Audyssey has to try and integrate the low frequency response with the bass capable speakers so they have to do some balancing act. If they were to flatten the Sub's response, then the combined response (example: with the front bass capable speakers) would suffer.

I played with my new UMIK-1 mic and rudimentary understanding of REW yesterday to measure a Samsung HW-Q60B sound bar + sub. This is a smaller secondary TV in another room, next to a fireplace. The results look great! Although because of the way they work, every LCR measurement included the sub.

From what I can see, you have the potential to get the response from15-150 Hz flat to within +/- 2.5 dB easily (even with no smoothing) if you spend a few hours tweaking with the app using Ratbuddssey. Or you can spend less than two hours doing the same and achive +/- 3 to 3.5 dB and then customize your own bass tilt (so called Harman curve), but that's up to you as not everyone likes such a tilt (I don't particurlay feel the need for one), while some prefer to just bump up the sub's level trim by a few dB and be done with it.

It sucks that I'd have to do that, but perhaps I will. I'd rather figure out why that 80Hz suckout happens, but as you said, maybe it's intentional when crossed with the front speakers. What is the best way to measure that? (And I guess it's possible one of my earlier graphs with all the speakers illustrates that?)

 

[peng]

True! But is there anything I can do in the analog space to fix that? I admit I just plopped my subwoofer close to the corner (but not AT the corner) of my living room. I didn't want artificially heightened bass, but could try that, or add a bass trap or something in that corner.
View attachment 322169

You can do a subwoofer crawling, and after each move, run REW again to see the effect. The photo shows you have the flexibility to move the sub at least 6-8 inches to the left, 2-3 to the right and 6-8 inches to the front, in any combinations.

I played with my new UMIK-1 mic and rudimentary understanding of REW yesterday to measure a Samsung HW-Q60B sound bar + sub. This is a smaller secondary TV in another room, next to a fireplace. The results look great! Although because of the way they work, every LCR measurement included the sub.

View attachment 322170

I think your room is very good, that's why the graphs look so good.

It sucks that I'd have to do that, but perhaps I will. I'd rather figure out why that 80Hz suckout happens, but as you said, maybe it's intentional when crossed with the front speakers. What is the best way to measure that? (And I guess it's possible one of my earlier graphs with all the speakers illustrates that?)

You don't have to do that at all, if you like the way it sounds now, if you find bass lacking, just try bumping up the sub level a little, say 2 dB at a time, until you are happy.
Before you worry about the 80 Hz suck out, take a look of what REW shows with the front left+right+sub to see how bad it really is with them all playing 80 Hz simultaneously.

I don't think it really edit: doesn't matters much with that kind of a minor suck out. The only reason why the app can fix (at least partially) is that you can lower the humps and bumps, and then elevate the whole range to compensate. Other than that, no room correction can fix room mode dips like that, except in theory, DL's upcoming ART should be able to do a lot of good in that area, but to get the ART pay option, you need to step up to the AVR-X3800H or Cinema 50 or models above that. So your best bet is to use Ratbuddyssey, or just the app but triple, quadrable or more the tweak time. Again, I wouldn't worry about the minor (relatively speaking) suck out. Take a look of what others posted, I have seen far worse.

 

[moonlight rainbow dream]

You should try measuring the sub alone and L + R individually (no crossovers applied) to figure out what the source of the dip is. It might be integration error at the xover (need to play with distance to get it to sum correctly) or it might be room effect and you can compensate by raising or lowering the crossover point. ie. If it's only in the sub response you can lower the crossover point and the mains will fill it in. If it's present in both you might consider relocating the sub. You can use pink noise and REW's RTA feature to easily figure out a better location.

 

[peng]

You should try measuring the sub alone and L + R individually (no crossovers applied) to figure out what the source of the dip is. It might be integration error at the xover (need to play with distance to get it to sum correctly) or it might be room effect and you can compensate by raising or lowering the crossover point. ie. If it's only in the sub response you can lower the crossover point and the mains will fill it in. If it's present in both you might consider relocating the sub. You can use pink noise and REW's RTA feature to easily figure out a better location.

I think it looks so bad because he applied GEQ, prior to running REW. If he just sweeps using stereo with Audyssey on, and XO 80 Hz it would look a lot better even with LRS on together.