Keith_W DSP system

[ernestcarl]

REW has an RTA, but no built-in convolver. I can send REW's pink noise through a convolver, but that introduces considerable latency and it won't be an "RTA" any more. In the end I decided it would be easier to simply do a whole bunch of sweeps.

I agree that sweeps may be quicker for you. But, why do you even need the convolver at this point since the uncorrected bass response should already be fine. You are only looking for the optimal available position with the least energy loss/highest overall amplitude gain, right?

 

[Keith_W]

I agree that sweeps may be quicker for you. But, why do you even need the convolver at this point since the uncorrected bass response should already be fine. You are only looking for the optimal available position with the least energy loss/highest overall amplitude gain, right?

Because I want to measure up to the transition zone, and that frequency range is covered by the sub + mains. I could measure them independently, but I would rather do them together.

 

[ernestcarl]

Because I want to measure up to the transition zone, and that frequency range is covered by the sub + mains. I could measure them independently, but I would rather do them together.

Okay, but if you're moving the mic at a significant differing distances, there is still a chance there would be energy loss from non-optimal sub-mains xo. Your speakers and subs after all aren't co-located.

*Room reflections could skew each speaker box's phase response -- which you will have the benefit of "correcting" optimally afterwards anyway.

 

[Keith_W]

I thought other Acourate users might be interested in how I organize my folders and workflow. Acourate recommends the following steps in this order when making a speaker correction:

1. Nearfield driver linearization and phase adjustment. I posted the method here.
2. Crossover creation + convolution of driver linearization with the crossover.
3. Time alignment.
4. MLP sweep of overall speaker followed by overall room correction and generation of filters for listening.

If you look at the steps, you will see that as you proceed down the chain the corrections become more specific for each particular filter. This means you only need to do driver linearization once, and the result can be convolved into any crossover that you like. So my folders are organized like this:

Anything "00" is a basic measurement that can be generalizable to all filters. This contains subfolders for Subs, Woofers, Mids, and Tweeters. Each subfolder contains a nearfield measurement with a correction filter. The correction filter goes to Step 8 in this post, and all that is needed is to convolve the correction filter with the raw crossover.

Anything "10" is a basic crossover that can be used with any target curve. Each contains subfolders for Subs, Woofers, Mids, and Tweeters. Within each speaker subfolder are more subfolders, arranged like this, that shows all the steps. First, the raw crossover is created. Then it is convolved with the correction filter. Then time alignment is performed. Once that is done, the results are placed in "Finalized". After this, all that is required is a room sweep + convolution with a target curve.

Anything with a date on it is a completed filter which is ready for listening.

When I started using Acourate, I kept everything in the same project directory which made it difficult to find things. This way, with separate project directories, I can easily find what I need, and retrace my steps if I have made a mistake somewhere.

 

[Keith_W]

I mentioned in a previous post that I was concerned about the directivity mismatch below 500Hz (woofers) and above 500Hz (horns). This is the crossover I was using, Uli's proprietary "UB Jpol11" crossover. Uli describes the advantages of this type of crossover in this thread.

I mentioned earlier in this thread that this steep XO may not blend the different directivity of the woofer and horn very well. So I made this:

Yes, the XO extends way beyond the limit which I measured and linearized the drivers. But it costs me no money to try this experiment, and I can switch between the two filters to compare what they sound like. Now the woofer extends well into horn territory, and the horns go well down into the bass. I know that the horns will not have a narrow directivity pattern at this lower freq range and they rapidly lose efficiency (as seen in the falling frequency response), but there is no harm in trying.

The other major consideration when pushing drivers so far out of their comfort zone is distortion. I do have distortion measurements, but these were not calibrated to SPL so I need to redo them. I will post those measurements when I have them.

The motivation for redoing the crossover was mentioned earlier - on some pieces of music, notably a tenor who sings with a frequency range between 100-500Hz (but with harmonics going up to 1kHz or more), the center image is unstable and shifts depending on whether he is singing a high or a low note. I was hoping that a smoother blend between the wide radiating woofer and the narrow midrange horns would make this less noticeable. So I made two filters to compare, using the exact same target curve, keeping all variables to a minimum with the exception of the XO config, and confirmed with verification measurements. And indeed, it works - the central image is more stable. However, it comes at the cost of some muddiness in the midrange. Separation between instruments is less clear in orchestral music.

This is something I have to investigate next. There may be some comb filtering, there may be some distortion, I don't know. I haven't figured out how to demonstrate comb filtering if the sweep at MLP looks exactly the same, but I will think about it.

 

[thewas]

Please keep also in mind that by making the crossover slopes shallower while as you say you make the horizontal directivities blend smoother you possibly create wider and deeper vertical lobes.

 

[Keith_W]

Thanks! Yes, that is true. I am not sure how I can take that measurement. I am thinking of placing the microphone 1m in front of the speaker and taking sweeps at different heights, and then repeat that exercise at MLP. How would you do it?

 

[thewas]

To see the vertical lobing you will need to take anechoic/gated measurements at different angles which in your case won't be easy in a room as due to your low crossover frequency you will need a relatively long gate before the first reflection.

 

[OCA]

the center image is unstable and shifts depending on whether he is singing a high or a low note

I frequently run into this problem. In my view, the main reason for that is phase response differences between speakers around 60Hz-1000Hz (human vocals) which results from asymmetric reflection surfaces effecting each speaker differently. Correcting excess phase of each speaker perfectly will almost certainly cause ringing but correcting the differences between them is a lot easier and usually solves the problem. I had seen in Mitch's video years ago how he could do that quite easily with Acourate, I am sure you also know.

 

[3ll3d00d]

given you've gone to the trouble of moving your system around, it can't be too much more effort to find a way to get some form of quasi anechoic measurement going. Surely will take less time than assessing directivity by ear anyway

 

[dualazmak]

The motivation for redoing the crossover was mentioned earlier - on some pieces of music, notably a tenor who sings with a frequency range between 100-500Hz (but with harmonics going up to 1kHz or more), the center image is unstable and shifts depending on whether he is singing a high or a low note. I was hoping that a smoother blend between the wide radiating woofer and the narrow midrange horns would make this less noticeable. So I made two filters to compare, using the exact same target curve, keeping all variables to a minimum with the exception of the XO config, and confirmed with verification measurements. And indeed, it works - the central image is more stable. However, it comes at the cost of some muddiness in the midrange. Separation between instruments is less clear in orchestral music.

I believe this would be the most critical and difficult issue especially in your system where you use "narrow midrange horns" (highly efficient drivers) crossed with cone woofers.

Just for your reference and info, I use 30 cm cone woofer (50 Hz - 500 Hz) and 8.8 cm Beryllium dome midrange (500 Hz - 6000 Hz) in the same sealed cabinet; and the optimization of XO between them would be easier in my setup than yours, I assume.

Even in my setup, however, I did really intensive objective and subjective try-and-error in XO configuration during the past almost three years mainly in terms of simultaneous establishment of "stable/robust center image (solo vocal: e.g. ref. here, solo trumpet: e.g. ref. here and here, etc.)" and "excellent separation of instruments in full orchestra (e.g. ref. here)". I have tested so many filter type and slope combinations, and finally settled in the seconder -12 dB/Oct mild XO with 0.1 msec precision time alignment between them as well as between midrange and Beryllium dome tweeters.

In my case in my room environments, this XO configuration also gives wider allowance for the best 3D listening position, about plus/minus 70 cm (within about 1.4 m diameter spherical space) of excellent listening position for very stable center image plus nice instruments separation which is more than enough for our (my wife and me) music listening session sitting together on the sofa (capable for 3 people) at the listening position. Of course, I know well that it would be greatly dependent on our room acoustic environments and treatments (ref. here).

Throughout my such tests and tuning, I found that the use of consistent and suitable "reference/sampler music playlist" (summary ref. here and here in my case) is really important especially for our final subjective evaluation at listening position. I believe you also have your own consistent "reference playlist", right?

In any way, we need our individual intensive tuning and optimal compromisations in each of our DSP multi-amplifier audio setups.

 

[Keith_W]

I was curious about why the soundstage was now so ridiculously wide. Was it because the subwoofers are now spaced far apart and in the corners? Was it because the speakers are far from the side walls and the reflections are delayed?

So I made two filters - one without subs, and the control with subs. The first, shown below in red/green (upper curve) is a normal filter with the system configured as a 4 way, with subs in the mix. There are two versions of this, one with 50Hz as the XO point, the other with 80Hz. The second filter (blue/brown) has the subwoofers turned off, with the woofers extending all the way down to 20Hz. Once overall room correction is applied, the curves look almost the same, but with the 3 way version without subwoofer showing slightly earlier bass roll-off and some dips in the FR:

Result: not even close. The version with the subwoofer has an enormous sense of scale and the feeling that the walls aren't there, even at low volumes. The 3 way version sounds like a normal speaker in a room. It's good, but nothing special. The version with the sub really is something special. Precision and size of the center image is the same with both filters, which is what you would expect given that everything from 500Hz up is exactly the same. And then there is the impact of the version with subwoofers, there is an authority that the 3 way version does not have. I could confirm this subjective impression by taking sweeps at different volumes, and I am 100% sure I will see the bass dropping off in the 3 way version, but there is no need to since the effect is so powerful.

David Griesinger talks about "bassiousness" - the idea that subwoofers set up in stereo and spaced widely apart contributes to subjective spaciousness. There is a debate somewhere on ASR about this, and it linked to this article. The article mentions that the spatial enhancement of low freqs has a lower limit - with some debate whether it is 50Hz or 80Hz (hence the motivation for trying two different XO points). In this system, and in this room, the 80Hz version seems to create more spaciousness than the 50Hz version, although the effect is pretty subtle.

That article makes it seem as if you need to make a choice between a setup for stereo bass or for proper bass management, and seems to forget that you can have both if you have DSP. I am now a believer - spacing your subs wide apart really helps.

On another note, I did an unplanned experiment yesterday. I washed my laundry and hung it up outside to dry. As is typical for Melbourne, one moment it is sunny and the next moment severe thunderstorms roll in. So I had to take the laundry rack with wet clothes into the house. Here it is, placed in front of the right subwoofer:

I realized that I had made myself a bass trap with wet clothes, so I turned on the system and had a listen before the clothes had a chance to dry. The impression was that the room felt a little smaller with the wet clothes in situ and I was getting ready to move the clothes rack around the room to experiment with various positions when suddenly the power cut out - 300,000 homes were left without power as the thunderstorm damaged critical infrastructure, and my home was one of them. So no more listening and dinner was biscuits, canned fish, and cheese by candle light because all the local restaurants were closed and I could not cook.

I have power and internet back now. It was a crippling feeling, as if you are jailed in your own home.

 

[dualazmak]

I read your above new post with much interests.

My honest impression would be that even though your intensive tuning and approach thereof are excellent reference to us all, any "generalization of your tuning" for other people's setup is almost impossible since your tuning is so greatly dependent on "your" audio rig and "your" room acoustics.

Please understand this is not at all a criticism to your sharing the efforts; let me repeat your efforts are really interesting and very nice reference for us.

"Your" efforts clearly tell us that all of us need similar efforts as "yours" for optimal tuning of individual DSP-based multi-amplifier audio rig, since all of us have different gears and different room acoustics.

 

[ernestcarl]

I was curious about why the soundstage was now so ridiculously wide. Was it because the subwoofers are now spaced far apart and in the corners? Was it because the speakers are far from the side walls and the reflections are delayed?

So I made two filters - one without subs, and the control with subs. The first, shown below in red/green (upper curve) is a normal filter with the system configured as a 4 way, with subs in the mix. There are two versions of this, one with 50Hz as the XO point, the other with 80Hz. The second filter (blue/brown) has the subwoofers turned off, with the woofers extending all the way down to 20Hz. Once overall room correction is applied, the curves look almost the same, but with the 3 way version without subwoofer showing slightly earlier bass roll-off and some dips in the FR:

View attachment 349510

Result: not even close. The version with the subwoofer has an enormous sense of scale and the feeling that the walls aren't there, even at low volumes. The 3 way version sounds like a normal speaker in a room. It's good, but nothing special. The version with the sub really is something special. Precision and size of the center image is the same with both filters, which is what you would expect given that everything from 500Hz up is exactly the same. And then there is the impact of the version with subwoofers, there is an authority that the 3 way version does not have. I could confirm this subjective impression by taking sweeps at different volumes, and I am 100% sure I will see the bass dropping off in the 3 way version, but there is no need to since the effect is so powerful.

David Griesinger talks about "bassiousness" - the idea that subwoofers set up in stereo and spaced widely apart contributes to subjective spaciousness. There is a debate somewhere on ASR about this, and it linked to this article. The article mentions that the spatial enhancement of low freqs has a lower limit - with some debate whether it is 50Hz or 80Hz (hence the motivation for trying two different XO points). In this system, and in this room, the 80Hz version seems to create more spaciousness than the 50Hz version, although the effect is pretty subtle.

That article makes it seem as if you need to make a choice between a setup for stereo bass or for proper bass management, and seems to forget that you can have both if you have DSP. I am now a believer - spacing your subs wide apart really helps.

On another note, I did an unplanned experiment yesterday. I washed my laundry and hung it up outside to dry. As is typical for Melbourne, one moment it is sunny and the next moment severe thunderstorms roll in. So I had to take the laundry rack with wet clothes into the house. Here it is, placed in front of the right subwoofer:

View attachment 349511

I realized that I had made myself a bass trap with wet clothes, so I turned on the system and had a listen before the clothes had a chance to dry. The impression was that the room felt a little smaller with the wet clothes in situ and I was getting ready to move the clothes rack around the room to experiment with various positions when suddenly the power cut out - 300,000 homes were left without power as the thunderstorm damaged critical infrastructure, and my home was one of them. So no more listening and dinner was biscuits, canned fish, and cheese by candle light because all the local restaurants were closed and I could not cook.

I have power and internet back now. It was a crippling feeling, as if you are jailed in your own home.

From what you wrote, the perceived spaciousness primarily came from the contribution of the subs and their physical spacing. If you can isolate the effect of positioning only for the main stereo speakers (minus subs) between the two room orientations, how much do you think did the actual change in spacing from the sidewalls make any difference/improvement?

Not much difference or about the same?

 

[Keith_W]

I read your above new post with much interests.

My honest impression would be that even though your intensive tuning and approach thereof are excellent reference to us all, any "generalization of your tuning" for other people's setup is almost impossible since your tuning is so greatly dependent on "your" audio rig and "your" room acoustics.

Please understand this is not at all a criticism to your sharing the efforts; let me repeat your efforts are really interesting and very nice reference for us.

"Your" efforts clearly tell us that all of us need similar efforts as "yours" for optimal tuning of individual DSP-based multi-amplifier audio rig, since all of us have different gears and different room acoustics.

I never pretended it was otherwise My gear, my room, this is how I do it, and it may not mean anything to anyone else. This thread has been useful for me though, repositioning the speakers has made a giant improvement thanks to @OCA.

From what you wrote, the perceived spaciousness primarily came from the contribution of the subs and their physical spacing. If you can isolate the effect of positioning only for the main stereo speakers (minus subs) between the two room orientations, how much do you think did the actual change in spacing from the sidewalls make any difference/improvement?

Not much difference or about the same?

I think that the spacing from the side wall has made a huge improvement, especially with BACCH crosstalk cancellation. This is independent of the subs. With Roger Water's "Amused to Death" the barking dog now sounds as if it coming from 90 degrees to the right from the MLP and from far outside the room. In the old orientation it was about 60 degrees, or just lateral to the right speaker. I have heard reports that people hear the barking dog coming from behind them with BACCH, and this is not something I have achieved.

There is also Billy Eilish's song "Ilomilo" where @Gwreck made a BACCH version with BACCH baked into the .WAV file. When I play it on headphones, the singer rotates to almost behind my head in a 270-300 deg arc. I do not get that effect with my speakers. With the old config, the arc was much narrower. Now it is much wider.

Note that there are 4 interventions here that are potentially responsible for the wider soundstage: (1) main speakers are further from the side wall, (2) overall setup is more symmetrical, (3) subs placed wider apart, (4) BACCH has been retuned*. It is a bit hard to tease out which is responsible for what effect except by playing music which has different freq content. I can easily turn off BACCH and remove it as a variable and the wider soundstage is still apparent.

* I forgot to mention that I retuned BACCH when I repositioned the speakers. @BACCH Labs themselves recommend tuning their VST subjectively instead of entering values based on geometry. I couldn't get BACCH to play pink noise, so I used REW to generate pink noise and sent it through JRiver. I then adjusted the slider until the soundstage was at maximum width. I wonder if I might be able to get a finer adjustment by playing HF clicks (since I find that easier to localize thank pink noise), but I do not have a source that can generate HF clicks that alternate left/right and BACCH recommends pink noise.

I would love to demonstrate this to you but you would have to travel to Australia Anyone who is local - feel free to drop me a line.

 

[OCA]

This is independent of the subs. With Roger Water's "Amused to Death" the barking dog now sounds as if it coming from 90 degrees to the right from the MLP and from far outside the room. In the old orientation it was about 60 degrees, or just lateral to the right speaker.

I wouldn't worry too much about this. It's a tricky track recorded with Q-sound (a method from the 80s which never took off) which basically flips one channel signal's phase 180 degrees and sends that to the opposite channel to create a surround effect. The location of the resulting sound will be determined by toe-in angles and their relative symmetry. If you tilt your head a little bit to the left usually the barking also moves to rear with it. Sound stage width is more meaningful IMO and I'd stick with the toe-in angles optimized for that as in your case.

 

[STC]

I wouldn't worry too much about this. It's a tricky track recorded with Q-sound (a method from the 80s which never took off) which basically flips one channel signal's phase 180 degrees and sends that to the opposite channel to create a surround effect. The location of the resulting sound will be determined by toe-in angles and their relative symmetry. If you tilt your head a little bit to the left usually the barking also moves to rear with it. Sound stage width is more meaningful IMO and I'd stick with the toe-in angles optimized for that as in your case.

With stereo it should be slightly behind your left or right. can’t really remember now. That’s how it sounded during my stereo days with the Harbeth speakers. I tried looking for the article describing how each tracks of the CD should project the sound. With XTC, I think it still about the same. Will listen again this weekend.

 

[Keith_W]

I did some measurements today. The intention was to find out a bit more about the reconfigured system.

First up, a control sweep to see where I am at:

Blue/Brown pair (above) = verification measurement when the filter was created.
Red/Green pair (below) = today's measurement.

Well, will you look at that. The two measurements are using exactly the same filter, the difference is that the first measurement was done without the sofa in place and the mic was not moved from the initial sweep. The second measurement has the sofa pushed back in place, and the mic may be a tiny bit further into the room than the first. We now see a huge bass discrepancy in the lower woofer and a tweeter discrepancy from 7kHz upwards. In both cases, the mic is exactly in the center (aligned using Acourate's mic alignment tool, which alternates left/right tweeter clicks. The mic is aligned by moving the mic until L/R tweeter impulses overlap).

I suppose I knew that this would happen already, but this is the worst verification measurement I have ever seen. In my previous setup I have seen inconsistencies between initial verification measurement and repeat (with mic moved and sofa back in place), but they generally followed the target curve. There is obviously something about this new configuration which is making it extremely sensitive to microphone position.

In the past, I have thought about doing an MMM correction instead of SPS. But when I investigated it, it turned out that the MMM did not look substantially different to the SPS - not enough for me to bother anyway. I forgot to do an MMM today, but it did occur to me to investigate the appearance of the SPS at different points in the sofa. Using Acourate's mic alignment feature again, I moved the mic 30 samples to the left and right, and repeated the measurement.

Here is the left speaker, at LLP, MLP, and RLP. The difference between all 3 measurements is about 60cm, or one body width. There are some little differences here and there.

And the right speaker, at LLP, MLP, and RLP. Again, not much to write home about.

But ... does it sound like that? Honestly, it does not. You would expect overwhelming bass and too much treble, but it actually sounds balanced and smooth.

So it appears that the mic is less sensitive to position in the X-axis, more sensitive in the Y-axis. So what do I do now? MMM a 3D space (i.e. a box encompassing all possible head positions, including leaning forward and back)? MMM in a plane? Do I do the MMM with or without the sofa? I think that MMM in a box shape with the sofa in situ seems to be the most logical solution.

 

[Keith_W]

Next we have some measurements to look at reverb, and to see what can be done to reduce it.

First up we have the "Control" measurement. Mic at MLP, all doors closed.

The RT60 is about 500ms which is a bit on the high side for me. My living room has a couple of doors behind the sofa (i.e. the back wall). One leads to a bathroom, the other to a bedroom. What happens if I open the doors?

Answer: nothing happens. I would have expected this intervention to improve the bass RT60 at least, given that the bass now has additional room sized chambers to escape to. The total length with the doors open is 12m (39ft). Of course, the "R" in RT60 means "reverberant field", and my room is a bit too small to create a reverberant field, so there is a lower limit where this measurement means something. I don't know how to calculate the lower limit of RT60 effectiveness, I would appreciate if someone could tell me. Regardless, there is absolutely no difference, not even in the frequency response.

By now I was pretty surprised. So on to the third experiment. I have some acoustic foam, which I normally stack up on the front wall where it is not very effective. What happens if I redeploy the treatment around the room? I got some dining chairs out and spaced them around the room, and placed the foam panels on it. Each foam panel is 1.2m tall, 60cm wide, and 12cm deep. I repeated the sweep:

Answer: nothing happens. Again. This time I would have expected the high frequency RT60 to at least drop, given that I am now attenuating flutter echoes.

Looking at RT60 graphs individually is a bit difficult to read, so I exported all the sweeps and loaded them into REW and made a spectrogram. I turned it into a .GIF:

I realize you can't see the scale on the GIF, so here is one image where you can read the scale:

I guess if there is one good thing to say about the spectro, the decay time of all frequencies seems to be constant. But this system is really turning out to be a head scratcher. I can certainly hear the difference the room treatment makes, but I can't seem to measure it.

 

[OCA]

aligned using Acourate's mic alignment tool, which alternates left/right tweeter clicks. The mic is aligned by moving the mic until L/R tweeter impulses overlap

I used that thing for a brief 2 weeks during a trial period some time ago and especially when used at 192Hz sampling rate, the mic alignment is more sensitive than the touch of a hand and it took quite some time to perfectly align it but ONCE I did, the L+R measurements I took at that time are the only ones I ever had and will probably ever have with literally zero comb filtering!

The second measurement has the sofa pushed back in place, and the mic may be a tiny bit further into the room than the first

Neither the sofa nor the mic position would make such a difference m8. It looks like it's around 50-60Hz. I don't know what frequency is the AC down under but it might be a problem with your AC lines maybe? I had similar major jumps or dips from one measurement to the next due to phase shifts around that region in a previous apartment with problematic wiring. I have also "rarely" seen such differences between left and right speakers in measurements sent to me from some quite randomly treated rooms.

That same area (60Hz) switches back and forth also in the spectrograms with the doors btw