or Connect
AVS › AVS Forum › Audio › Audio theory, Setup and Chat › Simplified REW Setup and Use (USB Mic & HDMI Connection) Including Measurement Techniques and How To Interpret Graphs
New Posts  All Forums:Forum Nav:

Simplified REW Setup and Use (USB Mic & HDMI Connection) Including Measurement Techniques and How... - Page 233

post #6961 of 9616
Quote:
Originally Posted by jlpowell84 View Post
 
Quote:
Originally Posted by kbarnes701 View Post

See my reply to Jerry. I don't know how to use the averaging thing so I never have. I naively thought that as I listen to L+R+C+Subs when listening, that is kinda what I should be showing - I now see why that isn't correct though, so progress has been made.


But when measuring final response what combo do we use? Single left or right +subs?

 

I have NFI. How about C+Subs? ;)

Gear mentioned in this thread:

post #6962 of 9616
Quote:
Originally Posted by kbarnes701 View Post

I have NFI. How about C+Subs? wink.gif

lol too funny.

especially since I got "scolded" for NOT posting my L+C+R+SW when posting my waterfall and final FR.

I demand an apology wink.gif
post #6963 of 9616
Quote:
Originally Posted by kbarnes701 View Post

Yes, I've seen that article before, thanks. But these are lab conditions using tones or pink noise. On real material can we really detect these small changes in FR, such that we really need to be aiming for +/-1dB graphs? I know you are not trying to dictate to anyone how they should proceed, Jim, but you have a huge amount of experience and I value your opinion.

To say we aware of small changes doesn't mean we can realistically make +/-1 db graphs. On the other hand, if you see a 2-3db spike or null, and its correctable, then why not fix it? And to do that, you must be aware of it in the first place. High and upper mid aberrations are pretty easy to mitigate once their known and pointed out. What I am aiming for is to distinguish what we can discern from what we can not. To aim for something isn't the same as expecting to obtain it.

I am at +/- 2.5db @1/6th and +/- 4db @ 1/24th octave smoothing (ignoring my purposeful bass elevation of 4db). The main reason I got there was not settling for prescribed targets, but rather, dealing with what could be dealt with and putting aside those that cant.

Lastly, what we are able to accomplish depends on the tools we have. Improve ones tools, and one can surpass previous horizons. In this case, knowledge and technique is our tool set(s). And understanding things more deeply allows possibilities previously unavailable to us.
post #6964 of 9616
Quote:
Originally Posted by kbarnes701 View Post


Are there any others that the panel would like to see?

TIA, Keith

Can you do a FR graph and put L+sub (in black) and R+sub (in red) on the same graph? Then another with the previous adding C (in green) ?
Edited by jim19611961 - 11/15/13 at 8:54am
post #6965 of 9616
Quote:
Originally Posted by jim19611961 View Post
 
Quote:
Originally Posted by kbarnes701 View Post

Yes, I've seen that article before, thanks. But these are lab conditions using tones or pink noise. On real material can we really detect these small changes in FR, such that we really need to be aiming for +/-1dB graphs? I know you are not trying to dictate to anyone how they should proceed, Jim, but you have a huge amount of experience and I value your opinion.

To say we aware of small changes doesn't mean we can realistically make +/-1 db graphs. On the other hand, if you see a 2-3db spike or null, and its correctable, then why not fix it? And to do that, you must be aware of it in the first place. High and upper mid aberrations are pretty easy to mitigate once their known and pointed out. What I am aiming for is to distinguish what we can discern from what we can not. To aim for something isn't the same as expecting to obtain it.

I am at +/- 2.5db @1/6th and +/- 4db @ 1/24th octave smoothing (ignoring my purposeful bass elevation of 4db). The main reason I got there was not settling for prescribed targets, but rather, dealing with what could be dealt with and putting aside those that cant.

Lastly, what we are able to accomplish depends on the tools we have. Improve ones tools, and one can surpass previous horizons. In this case, knowledge and technique is our tool set(s). And understanding things more deeply allows possibilities previously unavailable to us.

 

Good post. Thanks. I have just posted a full set of REW graphs which I believe reflect the current state of the room. One thing that is interesting is that I seem to get different results when using REW to when using OM. My OM always seem to look better - maybe I should go back to OM - LOL!

post #6966 of 9616
Quote:
Originally Posted by kbarnes701 View Post

I have NFI. How about C+Subs? wink.gif

What if you have no center lol! I'll sit back until this one is hashed out smile.gif
post #6967 of 9616
Quote:
Originally Posted by jim19611961 View Post
 
Quote:
Originally Posted by kbarnes701 View Post


Are there any others that the panel would like to see?

TIA, Keith

Can you a FR graph and put L+sub (in black) and R+sub (in red) on the same graph? Then another with the previous adding C (in green) ?

 

Sure...

 

 

post #6968 of 9616
Quote:
Originally Posted by jlpowell84 View Post
 
Quote:
Originally Posted by kbarnes701 View Post

I have NFI. How about C+Subs? wink.gif

What if you have no center lol! I'll sit back until this one is hashed out smile.gif

 

If you have no centre, you can be sure you have t least one channel that isn't causing you problems :)

post #6969 of 9616
Quote:
Originally Posted by sdrucker View Post
 
Quote:
Originally Posted by kbarnes701 View Post

See my reply to Jerry. I don't know how to use the averaging thing so I never have. I naively thought that as I listen to L+R+C+Subs when listening, that is kinda what I should be showing - I now see why that isn't correct though, so progress has been made.

I pity anyone picking up this thread, reading Jerry's Guide, and trying to understand why through 7000+ posts, a fair number of the graphs shown and discussed were corrupted by this comb filtering/short wavelength issue and are henceforth invalid.

So to get closure:
Are L/R+subs STILL valid for waterfalls and spectrograms in the < 300 Hz range?
For accessing frequency response of bass performance < 300 Hz, is it now, as per Nyal and Sanjay, valid to run a combined plot of L/C/R+subs as measured in that way, to capture the complete interaction of speakers for those wavelengths?

 

OMG, I hadn't even considered different parameters for waterfalls and spectros. 

post #6970 of 9616
So, according to Keith's posting, we now need a minimum of 15 meausurements/graphs when posting results of a change?
double eek.gif
Quote:
Originally Posted by kbarnes701 View Post


... 15 measurements/graphs deleted.... eek.gif
TIA, Keith
post #6971 of 9616
Quote:
Originally Posted by kbarnes701 View Post

OMG, I hadn't even considered different parameters for waterfalls and spectros. 

Are any of these graphs useful for determining how well the center and/or surrounds are timbre-matched to the fronts?

I am wondering if, when sdrucker measures his R-972, if there's an "objective" way to determine how well its feature of timbre-matching to the fronts work.
post #6972 of 9616
Quote:
Originally Posted by kbarnes701 View Post

If you have no centre, you can be sure you have t least one channel that isn't causing you problems smile.gif

It's only temporary smile.gif
post #6973 of 9616
Quote:
Originally Posted by artur9 View Post
 
Quote:
Originally Posted by kbarnes701 View Post

OMG, I hadn't even considered different parameters for waterfalls and spectros. 

Are any of these graphs useful for determining how well the center and/or surrounds are timbre-matched to the fronts?

 

 

IDK. All my speakers are from the same manufacturer so timbre-matching isn't too much an issue. And I always choose identical L, C and R speakers - I can't stand it when a different centre speaker is used and you can hear the timbre change when something pans left to right or v-v. I realise that not everyone can accommodate identical LCR speakers but I highly recommend it where possible. I find it difficult enough to optimise all this without having to also cope with speakers from different manufacturers!

post #6974 of 9616
Quote:
Originally Posted by artur9 View Post

So, according to Keith's posting, we now need a minimum of 15 meausurements/graphs when posting results of a change?
double eek.gif
Quote:
Originally Posted by kbarnes701 View Post


... 15 measurements/graphs deleted.... eek.gif
TIA, Keith

 

I hope not. I posted all those to get comments on if I was doing it right (and hopefully to attract some help on how to interpret them and what changes might be beneficial in my room, or at least what is a problem and what isn't). I would hope to not have to submit a full set every time.

post #6975 of 9616
Quote:
Originally Posted by artur9 View Post

Are any of these graphs useful for determining how well the center and/or surrounds are timbre-matched to the fronts?

I am wondering if, when sdrucker measures his R-972, if there's an "objective" way to determine how well its feature of timbre-matching to the fronts work.

IDK, as I wasn't going to look at that L/R frequency response matching option anymore than I would with Audyssey's version. My DefTechs are all timber-matched as a line, so I've never pursued a DSP approach.

FYI, I reached out to Curt Hoyt from Trinnov yesterday about working with powered mains and subs, and how I planned on measuring the results. I may wind up looking into 'professional guidance' if I get far enough into 2D/3D remapping.

However right now we all have enough on our plates...considering that now we're debating just what graphs are appropriate.
post #6976 of 9616
Quote:
Originally Posted by AustinJerry View Post

So what should the guide say?
IF you've placed a certain number of speakers and/or subs so that their interaction results in mode cancelling, then measure the interaction (measure all the speakers and/or subs involved).

You already do this with your subs (i.e., you measure the interaction of all 4 subs, not one sub at a time). Just apply the same concept to your speakers to see what their interaction is doing in the modal range.
post #6977 of 9616
Quote:
Originally Posted by kbarnes701 View Post

IDK. All my speakers are from the same manufacturer so timbre-matching isn't too much an issue. And I always choose identical L, C and R speakers - I can't stand it when a different centre speaker is used and you can hear the timbre change when something pans left to right or v-v. I realise that not everyone can accommodate identical LCR speakers but I highly recommend it where possible. I find it difficult enough to optimise all this without having to also cope with speakers from different manufacturers!

Yeah, i can just see having a Mythos ST as a powered center behind the TV....eek.gif
post #6978 of 9616
Quote:
Originally Posted by kbarnes701 View Post

Sure...






Wow!, that was quick!

Now, lets open another can of worms.....

L + R symmetry.

Image theory: The phase (or timing) of the source signal determines placement of images on the soundstage. That is, the subtle arrival time differences of what comes out the L vs R speaker. Mono = L & R arrive at the same time. Something sounding like its left of center means the left channel signal arrives slightly BEFORE the right one, and so on... (this is oversimplifying the matter a bit, but the illustration is valid). These cues are imbedded in the source material.

Now, for this to work properly, the L & R channels need to be the same loudness (within 2-3db) at any given note or frequency for this to translate properly into a coherent stereo image. If at some given note or frequency, one channel is substantially louder than the other, then the soundstage becomes compromised or augmented from what was intended. The result will often be a "moved" image. That is, the image may not appear where it was intended. It can shift. Likewise, early high gain reflections can cause ghost images to appear and cause blur, image shift, and/or soundstage collapse. A combination of a asymmetric L to R FR and early high gain reflections is what causes poor image resolution and detail.

So, in summation, two things are most important for proper imaging:

1) Proper L + R attenuation levels across the frequency band at least down to the transition frequencies (250hz or so), but down to about 100hz even better.
2) Proper timing of these arrivals (with a minimum of ghost images created by room reflections)

So Keith, what I see more clearly now in your FR graphs are some pretty substantial deviations L to R in the 250hz to 600hz range (the meat of the lower to mid vocal range). What this may translate to is what I call wandering. That is, as a vocalist (for instance) strikes different notes, their is a tendency for the image location of their voice to shift or wander a bit in terms of its perceived location from note to note.
Edited by jim19611961 - 11/15/13 at 9:45am
post #6979 of 9616
Quote:
Originally Posted by sdurani View Post

IF you've placed a certain number of speakers and/or subs so that their interaction results in mode cancelling, then measure the interaction (measure all the speakers and/or subs involved).

You already do this with your subs (i.e., you measure the interaction of all 4 subs, not one sub at a time). Just apply the same concept to your speakers to see what their interaction is doing in the modal range.

Fair enough. But with a L/C/R+subs 'global' measurement in the modal region, how do we get this to work with REW/HDMI? Do a single input and use a weird DSP mode like Mono Movie, to have a single signal source for this purpose? Wouldn't Dolby PLII Cinema focus on the signal to the center channel?
post #6980 of 9616
Quote:
Originally Posted by jim19611961 View Post
 
Quote:
Originally Posted by kbarnes701 View Post

Sure...







Now, lets open another cans of worms.....

L + R symmetry.

Image theory: The phase (or timing) of the source signal determines placement of images on the soundstage. That is, the subtle arrival time differences of what comes out the L vs R speaker. Mono = L & R arrive at the same time. Something sounding like its left of center means the left channel signal arrives slightly BEFORE the right one, and so on... (this is oversimplifying the matter a bit, but the illustration is valid). These cues are imbedded in the source material.

Now, for this to work properly, the L & R channels need to be the same loudness (within 2-3db) at any given note or frequency for this to translate properly into a coherent stereo image. If at some given note or frequency, one channel is substantially louder than the other, then the soundstage becomes compromised or augmented from what was intended. The result will often be a "moved" image. That is, the image may not appear where it was intended. It can shift. Likewise, early high gain reflections can cause ghost images to appear and cause blur, image shift, and/or soundstage collapse. A combination of a asymmetric L to R FR and early high gain reflections is what causes poor image resolution and detail.

So, in summation, two things are most important for proper imaging:

1) Proper L + R attenuation levels across the frequency band at least down to the transition frequencies (250hz or so), but down to about 100hz even better.
2) Proper timing of these arrivals (with a minimum of ghost images created by room reflections)

So Keith, what I see more clearly now in your FR graphs are some pretty substantial deviations L to R in the 250hz to 600hz range (the meat of the lower to mid vocal range). What this may translate to is what I call wandering. That is, as a vocalist (for instance) strikes different notes, their is a tendency for the image location of their voice to shift or wander a bit in terms of its perceived location from note to note.

 

Jim - thanks for this. I can see exactly what you mean. This is the sort of analysis that I find very helpful. I have to say I haven't been aware of any wandering of the image, but do remember that I only use the system for movies, and voices are rarely just dead centre as one might expect from a vocalist on a music track. In fact, I have always thought (post treatments) that my imaging was really very, very good. That is not to say, of course, that it could not be better. But for a movies-only system, do you think I would benefit from the no doubt considerable time expenditure required to improve things?  If I used the system for music I would definitely adopt a different approach as any 'wandering' then would drive me nuts. But with movies, the actor is for ever (usually) moving about the screen, and visual dominance of our senses being what it is, the voice appears to follow the actor, even though we know it is actually (usually) rooted to the centre channel. So there is an element of wandering built in. I do think is perhaps easier to work with a room used exclusively for movies than it is for one used with music, or hardest of all, a room used for both.

 

Do you see any other problems in the 15 graphs I uploaded?

 

Quote:
 Wow!, that was quick!

 

One of the benefits of REW. Once the measurements are made I can present them any whichway...

post #6981 of 9616
Quote:
Originally Posted by sdrucker View Post
 
Quote:
Originally Posted by kbarnes701 View Post

IDK. All my speakers are from the same manufacturer so timbre-matching isn't too much an issue. And I always choose identical L, C and R speakers - I can't stand it when a different centre speaker is used and you can hear the timbre change when something pans left to right or v-v. I realise that not everyone can accommodate identical LCR speakers but I highly recommend it where possible. I find it difficult enough to optimise all this without having to also cope with speakers from different manufacturers!

Yeah, i can just see having a Mythos ST as a powered center behind the TV....eek.gif

 

LOL. Well, yeah... it's why I wouldn’t choose speakers like that for a movies-system. One of the influential factors in choosing my M&K S150s is that they are an identical set across the front. As were my last speakers and the speakers before them and the speakers before them.... centre channel is so important for movies of course. Perhaps THE most important channel of them all.

post #6982 of 9616
Quote:
Originally Posted by kbarnes701 View Post

Jim - thanks for this. I can see exactly what you mean. This is the sort of analysis that I find very helpful. I have to say I haven't been aware of any wandering of the image, but do remember that I only use the system for movies, and voices are rarely just dead centre as one might expect from a vocalist on a music track. In fact, I have always thought (post treatments) that my imaging was really very, very good. That is not to say, of course, that it could not be better. But for a movies-only system, do you think I would benefit from the no doubt considerable time expenditure required to improve things?  If I used the system for music I would definitely adopt a different approach as any 'wandering' then would drive me nuts. But with movies, the actor is for ever (usually) moving about the screen, and visual dominance of our senses being what it is, the voice appears to follow the actor, even though we know it is actually (usually) rooted to the centre channel. So there is an element of wandering built in. I do think is perhaps easier to work with a room used exclusively for movies than it is for one used with music, or hardest of all, a room used for both.

Do you see any other problems in the 15 graphs I uploaded?

The first thing id do is make sure your data is good, That is, make sure the mic is EXACTLY the same distance from the L and R channels. Using OM, run track #2 (both channels, short sine sweep) and find the place where the 5-20K part of the graph is flat and maxed out in terms of splDB level. This is also a good way to determine if your listening chair is in the right spot.

Rerun the measurements.

If after this, there are still problems, they may be easy to fix. Lets see what the new data looks like first.

I agree that for movies, pinpoint exacting image locations is not as important (for reasons you pointed out). But getting this more right can only make things better.

It didnt see anything else at a glance with your other graphs that stuck out as bad to me smile.gif
post #6983 of 9616
Quote:
Originally Posted by kbarnes701 View Post

Nyal - does this summarise as 1/6 smoothing >300Hz and unsmoothed <300Hz?  If it does, that is what I have been doing (albeit to the wrong combination of speakers it seems)?

I use 1/3rd mainly above 250 and look at the 1/6

Below 250 I use 1/12th and look at the 1/24
post #6984 of 9616
Quote:
Originally Posted by kbarnes701 View Post

I thought I had a handle on this but the more I read the more uncertain I become .... so playing a mono source on a two channel system is a problem then due to this comb filtering?  I ask because I do this frequently on my separate stereo system and I don't hear anything odd even if I don't sit exactly equidistant from the speakers. 

Tis true, if you read a bit about what some of the detractors of stereo say about mono signals played back in both channels.
post #6985 of 9616
Quote:
Originally Posted by kbarnes701 View Post

LOL. Well, yeah... it's why I wouldn’t choose speakers like that for a movies-system. One of the influential factors in choosing my M&K S150s is that they are an identical set across the front. As were my last speakers and the speakers before them and the speakers before them.... centre channel is so important for movies of course. Perhaps THE most important channel of them all.

Agreed. Of course, when I bought them I was thinking of music first, and tested the powered Mythos supertower speakers on music rather than movies.

In fact, timbre matching of the DefTech line aside, that's why I picked up a powered center channel ultimately, with the idea of having better blending across the three in-front speakers due to the extra amplification for mid-bass. I'd tried the non-powered Mythos Nine and Ten (the latter uses the same midrange/tweeters as the Mythos ST) and didn't like it as much because I felt that the mid-bass and upper midrange was lacking, and the center stood out. It certainly sounds better for music with the CS-8080 as the center. At one point, my bass feel was so good that listening to concert BluRays had an almost "live" feel to how the bass strings were plucked. I lost some of that while getting the center image down better, and with REW, as you know we're never happy with what we have...rolleyes.gif

OTOH, if I knew what we knew now, I might have played more with positioning and/or treatments before switching the Mythos Nine/Ten out. But I should mention that a fair number of DT users like the powered centers, even if they have real subs.
Edited by sdrucker - 11/15/13 at 10:20am
post #6986 of 9616
BTW with reference to the spectrogram / decay charts I think for a full examination you should be looking at it down to the noise floor of the room.

There are often structural or other secondary resonances which are audible and distracting yet you will miss them if you look with a 40dB window on your graph and you are in a low noise room.
post #6987 of 9616
Quote:
Originally Posted by jim19611961 View Post
 
Quote:
Originally Posted by kbarnes701 View Post

Jim - thanks for this. I can see exactly what you mean. This is the sort of analysis that I find very helpful. I have to say I haven't been aware of any wandering of the image, but do remember that I only use the system for movies, and voices are rarely just dead centre as one might expect from a vocalist on a music track. In fact, I have always thought (post treatments) that my imaging was really very, very good. That is not to say, of course, that it could not be better. But for a movies-only system, do you think I would benefit from the no doubt considerable time expenditure required to improve things?  If I used the system for music I would definitely adopt a different approach as any 'wandering' then would drive me nuts. But with movies, the actor is for ever (usually) moving about the screen, and visual dominance of our senses being what it is, the voice appears to follow the actor, even though we know it is actually (usually) rooted to the centre channel. So there is an element of wandering built in. I do think is perhaps easier to work with a room used exclusively for movies than it is for one used with music, or hardest of all, a room used for both.

Do you see any other problems in the 15 graphs I uploaded?


 
Quote:
Originally Posted by jim19611961 View Post
 
The first thing id do is make sure your data is good, That is, make sure the mic is EXACTLY the same distance from the L and R channels. Using OM, run track #2 (both channels, short sine sweep) and find the place where the 5-20K part of the graph is flat and maxed out in terms of splDB level. This is also a good way to determine if your listening chair is in the right spot.

 

Ah - first problem. My MLP is not, and cannot be, centred. (: Due to the crappy little room and the way it is, I have two listening chairs and neither can be in the centre spot. I have cheated a little and moved mine closer to the centre than the other (rarely used) chair but that is all I can do. It may be, in general, that the diktats of my room prevent me from ever getting a 'perfect' response (where perfect is defined as something like your own for this purpose).

 

 

Quote:

 Rerun the measurements.

If after this, there are still problems, they may be easy to fix. Lets see what the new data looks like first.

 

I assume that we are stymied by my observation above...

 

Quote:
 I agree that for movies, pinpoint exacting image locations is not as important (for reasons you pointed out). But getting this more right can only make things better.

It didnt see anything else at a glance with your other graphs that stuck out as bad to me smile.gif

 

Well, that is some good news then at least :)  Thanks for your help, as always. 

post #6988 of 9616
Quote:
Originally Posted by Nyal Mellor View Post
 
Quote:
Originally Posted by kbarnes701 View Post

Nyal - does this summarise as 1/6 smoothing >300Hz and unsmoothed <300Hz?  If it does, that is what I have been doing (albeit to the wrong combination of speakers it seems)?

I use 1/3rd mainly above 250 and look at the 1/6

Below 250 I use 1/12th and look at the 1/24

 

Thanks Nyal. What plus or minus dB figures are you after in either case?  Ideally I mean?

post #6989 of 9616

OK, guys, no one has done anything to make my headache any better.

 

This is where I think we are:

 

- Below 300Hz, measure all speakers that contribute to bass response, i.e. left+right+subs, or center+subs.  Always use unsmoothed data.  Use these measurements to produce waterfalls, spectrograms, and decay graphs.

 

- Above 300Hz, measure left, center, and right speakers individually (with or without subs) to avoid undesirable interactions between speakers which gets worse at higher frequencies.  Publish frequency response graphs using 1/24 smoothing to observe any remaining issues, or using 1/6 smoothing to get a general feel of the progress that has been made (whatever that means).

 

BTW, regardless of what some of you think, the subject of the interaction between left and right speakers was raised by me several weeks back.  I was noticing a strangeness above 10KHz in the left+right graph that was not present in the individual left and right graphs.  That precipitated a discussion involving the importance of making sure the mic was placed in exactly the center between left and right speakers.  I believe Jim contributed to that discussion, as well as several others (sorry, don't have the time to dig out the original posts).  Since then, I have experimented several times by taking a left+right measurement, observing the flatness above 10KHz, and adjusting the mic in small increments either left or right, observing the effect on the graph's flatness.  The mic's position can make a significant difference with only a slight movement.  So, in retrospect, someone might be chasing what looks like a high frequency issue, when it is simply the interaction between the two speakers.

 

Make any sense?  Whaddya think?

post #6990 of 9616
Quote:
Originally Posted by Nyal Mellor View Post
 
Quote:
Originally Posted by kbarnes701 View Post

I thought I had a handle on this but the more I read the more uncertain I become .... so playing a mono source on a two channel system is a problem then due to this comb filtering?  I ask because I do this frequently on my separate stereo system and I don't hear anything odd even if I don't sit exactly equidistant from the speakers. 

Tis true, if you read a bit about what some of the detractors of stereo say about mono signals played back in both channels.

 

Interesting. I have a fair number of mono albums and I always play them back through the stereo pair of speakers. I shall have to listen more closely and perhaps play back through just one speaker and compare it with playback through both and see if I can hear a difference. Aaaaagghhh.... so far I have resisted going down the rabbit hole in the room where my two-channel music system lives! :)

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Audio theory, Setup and Chat

Gear mentioned in this thread:

AVS › AVS Forum › Audio › Audio theory, Setup and Chat › Simplified REW Setup and Use (USB Mic & HDMI Connection) Including Measurement Techniques and How To Interpret Graphs