Good to see you have given the averaging a go and I will be interested to hear if you think it produces a better sounding correction. Given the differences in the graph are quite slight it could be a subtle change.
I would recommend you try the latest REW, the time align function gets the measurements quite close and as your impulse peak is cleaner the auto align should do a pretty good job which you can tweak to perfection if you want
The vector averaging is exactly the same as averaging the individual pairs by hand. It doesn't matter which order you do them in the result is the same but it does allow you to use odd numbers like the left right pairs with the centre added.
The other cool feature is the recording of the noise floor by default and the stepped sine measurement which makes the distortion analysis better and also shows which distortion products make it above the noise floor of the room. I find that useful to view.
The damping in the mid to high frequencies is really doing a good job of reducing the room contribution. You have seen my measurements so you would appreciate it but in a non treated room the averaging tends to reduce output steadily beyond the transition region so the results of single channel vs averaged can look a bit different.
From my own average measurements I am also convinced of the improvements I can see over what I have seen from others when averaging a more standard speaker. There are position related differences but they are much smaller than a standard speaker has, the averaging of the drivers over the length does work. It doesn't eliminate the need to get the speaker to co-operate with the room but if you can't or don't want to treat the room the results seem better to me than with other speaker types I have tried from that standpoint.
Most of the discussion I have read debating the validity of averaged measurements is due to wide variance skewing the results, particularly in favour of dips in the response. The arrays in room seem to avoid this quite nicely
Hank, dipole bass in my room is quite similar to what I am getting from the arrays so your plan to cross the CBT's to dipole woofers could well be a good way to go as the drivers in the PE kit could not be EQ'd to get the same low end response as shown by wesayso (at least not at any kind of reasonable listening level from their specs).
I would recommend you try the latest REW, the time align function gets the measurements quite close and as your impulse peak is cleaner the auto align should do a pretty good job which you can tweak to perfection if you want
The vector averaging is exactly the same as averaging the individual pairs by hand. It doesn't matter which order you do them in the result is the same but it does allow you to use odd numbers like the left right pairs with the centre added. The other cool feature is the recording of the noise floor by default and the stepped sine measurement which makes the distortion analysis better and also shows which distortion products make it above the noise floor of the room. I find that useful to view.
The damping in the mid to high frequencies is really doing a good job of reducing the room contribution. You have seen my measurements so you would appreciate it but in a non treated room the averaging tends to reduce output steadily beyond the transition region so the results of single channel vs averaged can look a bit different.
From my own average measurements I am also convinced of the improvements I can see over what I have seen from others when averaging a more standard speaker. There are position related differences but they are much smaller than a standard speaker has, the averaging of the drivers over the length does work. It doesn't eliminate the need to get the speaker to co-operate with the room but if you can't or don't want to treat the room the results seem better to me than with other speaker types I have tried from that standpoint.
Most of the discussion I have read debating the validity of averaged measurements is due to wide variance skewing the results, particularly in favour of dips in the response. The arrays in room seem to avoid this quite nicely
Hank, dipole bass in my room is quite similar to what I am getting from the arrays so your plan to cross the CBT's to dipole woofers could well be a good way to go as the drivers in the PE kit could not be EQ'd to get the same low end response as shown by wesayso (at least not at any kind of reasonable listening level from their specs).
Hello Wesayso,
Are the subwoofers addition to your arrays still in the pipeline or did you give up on that?
I am really looking forward to know how much a high pass filter can improve the arrays performance. I know you said you plan to use the subs just to fill in the dips and all, but still that would take some load off the arrays.
Are the subwoofers addition to your arrays still in the pipeline or did you give up on that?
I am really looking forward to know how much a high pass filter can improve the arrays performance. I know you said you plan to use the subs just to fill in the dips and all, but still that would take some load off the arrays.
Even if it only matches the performance of a single position measurement I'll still favor the averaged measurement. As it would keep me from overcorrecting something placement dependent. While I've always backed off at certain points I can't be sure I got it all right for a broader listening window. I've always wanted to try this.
I would recommend you try the latest REW, the time align function gets the measurements quite close and as your impulse peak is cleaner the auto align should do a pretty good job which you can tweak to perfection if you want
The downloads have moved, right? No longer at Home Theatre Shack? That's the biggest reason for me not running the latest version yet, but I will.
Sounds useful indeed! My noise floor is a moving target with traffic right in front of my house. That part became obvious again when measuring at multiple points yesterday. I should pick a Sunday and kick out other occupants to really see my noise floor.
It does seem that way to me too. I was impressed by this comparison though. As it is a better result than I had when making the decision to stick with a single point measurement about two years ago. At that time I had done some measurements but I was by no means experienced with DRC and DSP yet. I figured to start simple and learn.
Most of the discussion I have read debating the validity of averaged measurements is due to wide variance skewing the results, particularly in favour of dips in the response. The arrays in room seem to avoid this quite nicely
Very nice, I'd say. For me 40 HZ to 11 KHz fits within a plus/minus 2 dB window when viewed with 6 cycle FDW. (aside from one room anomaly)
Agreed...
Subwoofers are still in the pipe line. I just lack the time right now to really get started. I want to be able to make good progress and not an hour here, an hour there...
The arrays being good at the bottom end does not help or press me to hurry up. Though I'm sure there is benefit if I can get them to 'play nice' together.
The first, very raw test was not bad at all. Still needs work but the raw correction showed potential.
Ive used different averages than shown a few posts ago. As it is all very close I decided to maximize the sweet spot and use the measurement points closer to the sweet spot only. So less averages in total and more weighed around the sweet spot. Keeping in mind where both ears are in real life.
Ive used different averages than shown a few posts ago. As it is all very close I decided to maximize the sweet spot and use the measurement points closer to the sweet spot only. So less averages in total and more weighed around the sweet spot. Keeping in mind where both ears are in real life.
It has been a few days since that last raw test. I definitely made some progress with processing based on averages. However there's still so much to try.
Overall I'm pretty pleased with what I've heard so far. But I'm using a highly theoretical way to do the additional processing. Using virtual testing to define my post EQ. I can't seem to find enough time to validate every step as I hardly have enough time to do what I want anyway.
It's vacation time for my son, which means less available 'alone play time' for me. Even though my vacation is coming up, there isn't a whole lot of time available for me to start with building the subs either. More pressing matters take up most of my time right now.
I do miss it, this stuff continues to be on my mind on a daily basis and not finding enough time to try all of it is kind of a bummer.
To not leave this post without adding any clue I'll post the Stereo early waterfall as gathered/created in a virtual test:
It looks pretty clean, right? This is all virtual though, using the averages gained in the last measurement session and running that trough the convolution and additional PEQ edits.
To show the same graph with DRC correction only (same as listened to in the first raw test):
Now I'm not saying the top graph sounds the best as I did nothing to compare them one on one. There's also still work to do on that clean graph, as it may still have some over-correction here and there which might not work out as well in real live measurements.
The top graph does sound better right now, purely based on better tonal balance.
What is clear in listening to both these corrections (days apart) is that using the averages instead of a single point measurement seems to allow me to be able get more clear top end over a pretty wide area.
It's a pretty open sound, though I still have a lot to get right on the bottom end. I was hoping to have the subs ready before doing these exercises.
So no definitive answers from me yet, besides me saying so far it's worth it (for me) to invest more time in it to find out.
The forum seems to slow down in this part of the year too, I'm hoping to gain some free/spare time after the holidays. I'm not going to hold my breath though. It's been a busy year so far for other parts of our family life.
Overall I'm pretty pleased with what I've heard so far. But I'm using a highly theoretical way to do the additional processing. Using virtual testing to define my post EQ. I can't seem to find enough time to validate every step as I hardly have enough time to do what I want anyway.
It's vacation time for my son, which means less available 'alone play time' for me. Even though my vacation is coming up, there isn't a whole lot of time available for me to start with building the subs either. More pressing matters take up most of my time right now.
I do miss it, this stuff continues to be on my mind on a daily basis and not finding enough time to try all of it is kind of a bummer.
To not leave this post without adding any clue I'll post the Stereo early waterfall as gathered/created in a virtual test:
It looks pretty clean, right? This is all virtual though, using the averages gained in the last measurement session and running that trough the convolution and additional PEQ edits.
To show the same graph with DRC correction only (same as listened to in the first raw test):
Now I'm not saying the top graph sounds the best as I did nothing to compare them one on one. There's also still work to do on that clean graph, as it may still have some over-correction here and there which might not work out as well in real live measurements.
The top graph does sound better right now, purely based on better tonal balance.
What is clear in listening to both these corrections (days apart) is that using the averages instead of a single point measurement seems to allow me to be able get more clear top end over a pretty wide area.
It's a pretty open sound, though I still have a lot to get right on the bottom end. I was hoping to have the subs ready before doing these exercises.
So no definitive answers from me yet, besides me saying so far it's worth it (for me) to invest more time in it to find out.
The forum seems to slow down in this part of the year too, I'm hoping to gain some free/spare time after the holidays. I'm not going to hold my breath though. It's been a busy year so far for other parts of our family life.
To add a little more info, there's the above graphs as viewed as an IR:
As you can see, the additional PEQ work does clean up the IR graph. Being more clear of reflections than a single point measurement does help to make it a pretty graph.
Averaging does help reduce the direct/reflected ratio which means I can get away with more on the correction side. However, only measured results will tell me where to lay off. So far it's only an improvement in a theoretical sense. Though the reality was never far behind in previous tests.
A sharp eye might notice an improvement in the pré peak area. I've changed a couple of things in my DRC template to gain an advantage there.
As you can see, the additional PEQ work does clean up the IR graph. Being more clear of reflections than a single point measurement does help to make it a pretty graph.
Averaging does help reduce the direct/reflected ratio which means I can get away with more on the correction side. However, only measured results will tell me where to lay off. So far it's only an improvement in a theoretical sense. Though the reality was never far behind in previous tests.
A sharp eye might notice an improvement in the pré peak area. I've changed a couple of things in my DRC template to gain an advantage there.
Last edited:
Are you using IIR/PEQ on the average of multiple measurements and FIR correction on a single point measurement?
I am unaware if there is any way to use impulse correction for multiple measurements.
It would be interesting to know what technique Dirac uses. Maybe it is as simple as multiple measurement based PEQ and Single measurement FIR or maybe not, given their research strength. But they use low taps in comparison, so maybe mostly IIR and nothing much elaborate.
If you post your impulse response in the multi-way forum, they will ask why you took a measurement without connecting the mid and woofers
I am unaware if there is any way to use impulse correction for multiple measurements.
It would be interesting to know what technique Dirac uses. Maybe it is as simple as multiple measurement based PEQ and Single measurement FIR or maybe not, given their research strength. But they use low taps in comparison, so maybe mostly IIR and nothing much elaborate.
If you post your impulse response in the multi-way forum, they will ask why you took a measurement without connecting the mid and woofers
Wesayso has used the same technique I used for combining the impulses from the multiple measurements (there is more information and a walkthrough by Swissbear in the rephase thread towards the end).
Take the measurements, time align the IR's and then use trace averaging or the vector averaging function in REW.
The averaged impulse is then used as the measurement for DRC correction.
Wesayso has then used PEQ after the DRC correction as well.
There is less difference in the measurements from an array spread across a couch than I have seen from conventional speakers measured in the same way. The acoustic averaging the arrays achieve are probably helping to make the impulse averaging technique work well.
You can see from Wesayso's measurements that there is not a huge difference between the single position and the averaged response in both frequency and time. The averaging helps to only correct parts of the response that are common between positions.
I have been unable to work out the method Dirac's 'algorithm' uses to work out what to do with the multiple measurements. They do mention using both IIR and FIR so it is possible they use the sweetspot measurement to do a time correction and then add an IIR average based frequency correction on top. There are quite a few papers released by them which have general information on 'mixed phase' correction but nothing that spells out what they do in any detail. Not surprising as it is a commercial product
PS also see that Wesayso posted a waterfall that goes to 20K, he hasn't done that before so he must be quite pleased with the way it looks
Take the measurements, time align the IR's and then use trace averaging or the vector averaging function in REW.
The averaged impulse is then used as the measurement for DRC correction.
Wesayso has then used PEQ after the DRC correction as well.
There is less difference in the measurements from an array spread across a couch than I have seen from conventional speakers measured in the same way. The acoustic averaging the arrays achieve are probably helping to make the impulse averaging technique work well.
You can see from Wesayso's measurements that there is not a huge difference between the single position and the averaged response in both frequency and time. The averaging helps to only correct parts of the response that are common between positions.
I have been unable to work out the method Dirac's 'algorithm' uses to work out what to do with the multiple measurements. They do mention using both IIR and FIR so it is possible they use the sweetspot measurement to do a time correction and then add an IIR average based frequency correction on top. There are quite a few papers released by them which have general information on 'mixed phase' correction but nothing that spells out what they do in any detail. Not surprising as it is a commercial product
PS also see that Wesayso posted a waterfall that goes to 20K, he hasn't done that before so he must be quite pleased with the way it looks
Fluid is spot on with his comments, even an 8 point average is a spitting image of the center measurement when we look at phase. The base of this is a weighted average near the center listening position using only 4 points per side.
Dirac is using IIR mostly and has 'some' phase correction. However as it has virtually no play back delay the number of tabs is very limited.
Personally I'm not that interested in Dirac after seeing some examples and playing with it myself (a long time ago). I'd be way more interested in comparing to Audiolense and Acourate. Or APL_TDA_EQ as I consider those to be on another level when compared to Dirac.
I use a template that is mixed phase, in the sense that the FR gets corrected to minimum phase with Excess phase correction applied between 20 - 1000 Hz.
How perceptive, yes I am! The graph is showing a 25 dB range at 2 KHz. Everything above 2 KHz is of interest. The window isn't long enough to judge anything below that as it hasn't completed the number of cycles yet to show a valid curve. To judge the part below 2 KHz would need a longer window for the waterfall.
Dirac is using IIR mostly and has 'some' phase correction. However as it has virtually no play back delay the number of tabs is very limited.
Personally I'm not that interested in Dirac after seeing some examples and playing with it myself (a long time ago). I'd be way more interested in comparing to Audiolense and Acourate. Or APL_TDA_EQ as I consider those to be on another level when compared to Dirac.
I use a template that is mixed phase, in the sense that the FR gets corrected to minimum phase with Excess phase correction applied between 20 - 1000 Hz.
PS also see that Wesayso posted a waterfall that goes to 20K, he hasn't done that before so he must be quite pleased with the way it looks
How perceptive
, yes I am! The graph is showing a 25 dB range at 2 KHz. Everything above 2 KHz is of interest. The window isn't long enough to judge anything below that as it hasn't completed the number of cycles yet to show a valid curve. To judge the part below 2 KHz would need a longer window for the waterfall.
Just spend a lot more time than I originally planned listening.
I've made my tweaks to protect this current scheme from over correcting and started the listening session.
All I can say is: I started out in serious listening mode and wound up just queuing up more and more songs, singing along and moving to the music. So far for a serious session.
All this tells me: I'm on my way here. Again with measurements and graphs pointing me towards that way. Still have lot's of little tweaks to go. Listening sessions like these is what makes the journey worth while. My smiles per minute gauge is making overtime again.
No on to do the last bit of actual (paid) work I had planned... or should I play just one more song?
I've made my tweaks to protect this current scheme from over correcting and started the listening session.
All I can say is: I started out in serious listening mode and wound up just queuing up more and more songs, singing along and moving to the music. So far for a serious session
.All this tells me: I'm on my way here. Again with measurements and graphs pointing me towards that way. Still have lot's of little tweaks to go. Listening sessions like these is what makes the journey worth while. My smiles per minute gauge is making overtime again.
No on to do the last bit of actual (paid) work I had planned... or should I play just one more song?
I think when one's ears finally find that magic speaker setup, 1 thing happens first: after you listen to 'that track' it resonates in your head the following day. Then you have to come back after a day at work at sit down and listen again, re-evaluate. Maybe at lower volume, maybe louder... If it still impresses you, you know your onto something!
I know what you mean! I even take notice if that happens for people that happen to listen in. When they walk away singing I'm on the right track.
I do make sure this happens to me for many tracks, not just one. I must be able to live with it and still be pleased for weeks on end to make sure I'm on the right path.
2 Things that help with this: tonal balance and good rhythm. (the one you feel as well as hear)
Even way off axis, in the joining kitchen space it still sounds pleasing. If I remember and think back how I started I cannot deny the level of progress I've been able to get.
I should though, with 3700+ posts in this thread, almost 600.000 views and it will be spanning a 4 year period in September this year!
No regrets, that's for sure! Time to start on the subs and take it to the next level.
I'm hoping to find the time for that soon.
I do make sure this happens to me for many tracks, not just one. I must be able to live with it and still be pleased for weeks on end to make sure I'm on the right path.
2 Things that help with this: tonal balance and good rhythm.
(the one you feel as well as hear)Even way off axis, in the joining kitchen space it still sounds pleasing. If I remember and think back how I started I cannot deny the level of progress I've been able to get.
I should though, with 3700+ posts in this thread, almost 600.000 views and it will be spanning a 4 year period in September this year!
No regrets, that's for sure! Time to start on the subs and take it to the next level.
I'm hoping to find the time for that soon.
I just came across a track that sounded way above average. Heavy metal genre. Megadeth - When. Lyrics are very aggressive, but I was mezmerized by the audio mix and sound quality. It really came through on my setup. Was wondering maybe you could give it a listen... Half way through, the quality seems to degrade and bass disappears but then returns in the final minute or 2.
I didn't know that track, nor do I have it. I've listened to it over headphones trough Youtube. But to really give it a shot I'll have to see if I can get my hands on the original.
Did you ever try "Keith don't go" in the live version by Niels Lofgren? It's an audiophile favorite and sounds very good on almost any setup. Basically it's filled with the performance room sound that's enhancing the guitar parts. Let me see if I can upload a snippet of it...
Intro part
Did you ever try "Keith don't go" in the live version by Niels Lofgren? It's an audiophile favorite and sounds very good on almost any setup. Basically it's filled with the performance room sound that's enhancing the guitar parts. Let me see if I can upload a snippet of it...
Intro part
Last edited:
Holy cow! First time my crappy Yamaha active desktop monitors showed their potential!
Thanks, wesayso
BYRTT thought it almost sounded a bit too sweet in my room.
Ehm... well... comparing your arrays to my Yamahas...
I'd really love to hear yours, BTW.
BYRTT thought it almost sounded a bit too sweet in my room.
You learned me and is good at analyzing tracks up against real world sound thanks for that, or was it track Treetop Flyer by Stephen Stills which is much more hot in lows and DC thumbs plus higher DR than Nils Loftgren track.
Ehm... well... comparing your arrays to my Yamahas...
I'd really love to hear yours, BTW.
Had 690 kilometers to go and was worth it
memories is a warm little house where live a fantastic human family and they have some extraordinary great sounding speakers plus good coffee.