Keith_W DSP system

[OCA]

Funny -- seasoned (maybe), not exactly "expert" as I don't even have a good grasp of some of its other features. I've actually been using the more "stable" standard release version recently in my old Win 10 laptop which I prefer to use for quick REW measurements over Linux. Now, I am a little piqued by the beta changes so will take a peak. To be honest, it took me quite a few months to finally get used to some of the UI changes like the new PEQ window interface so I've kept two software versions installed... it's always jarring when you can't find buttons in their "original" locations.

True that! It took me with quite a bit of a shock at first but I am slowly getting a grasp of it I guess.

 

[Keith_W]

Well, I have come across something interesting. Some of you may recall from a few pages back that I was confused as to why phase linearisation is audible, compared to no phase linearisation ... when phase is supposed to be inaudible. I said that drivers that were phase linearised seemed to have the dynamics sucked out of them, they become more polite, but also sound lifeless.

I was wondering whether everybody is wrong about the audibility of phase, but what are the chances of some hobbyist in Australia being right and the luminaries of audio are wrong? I do not have that kind of arrogance, but I can not deny the obvious audible difference. The far more likely explanation is that said hobbyist has made a mistake or is misinterpreting information. So I convinced myself that the phase correction must have done something to the signal, and the difference I am hearing must be something else, and not phase.

Then I hit upon the idea of looking at the impulse response. Voila!

I dug out an old measurement. The intention of the measurement was for time alignment of the midrange horn, so it contains the tweeter as well as the horn, and the sweep was taken from the MLP. I then normalized the volume of both sweeps so that the amplitude can be compared. I switched to the impulse window, and this was the result. The red measurement shows the driver with no phase correction, and the green measurement has the phase linearisation procedure applied. The amplitudes are much higher with the red measurement, indicating that the dynamics are better.

I breathed a sigh of relief when I saw this. Finally, proof that I am not imagining it.

But this raises even more questions.

1. Do you think this is a valid way to measure the subjective phenomenon of dynamics? I have never seen anybody describe this method before. Have I invented a method, or has it been described before (and I have missed it in my reading), or is it totally wrong?

2. Why should phase linearisation kill the dynamics like that? I still have no explanation! I have been discussing this with a fellow ASR member, and he has subjectively noticed a similar phenomenon in his system - that phase linearisation kills dynamics. Of course it is possible that two ASR members using different software packages are making the same error, producing artefacts as a byproduct which kills the dynamics.

Then there are other subjective phenomena which are unrelated to dynamics which I can not explain. Like, the phase linearised version has noticeable effects on the soundstage. It seems to "lift it off the floor" and compress the vertical height (bad), and it pushes the soundstage further back and seems to make it wider (good). I am starting to wonder if a less aggressive correction will provide some desirable qualities (improved depth and width of soundstage) whilst minimizing the negative qualities (vertical compression and killing dynamics). There is no "phase target" that I am aware of, because nobody seems to think that phase is audible.

 

[Tim Link]

My take on it is that the green response is "faster" in the sense that it dies down faster. Am I looking at it right. At the same time, if the initial pulse is as loud on both of them, might the phase corrected version come across as lower in volume overall? What's the scale at the bottom? Is that seconds? If so, I think up to 0.2 is within the time fusion window, so it'll all just sound louder.

 

[Keith_W]

My take on it is that the green response is "faster" in the sense that it dies down faster. Am I looking at it right. At the same time, if the initial pulse is as loud on both of them, might the phase corrected version come across as lower in volume overall? What's the scale at the bottom? Is that seconds? If so, I think up to 0.2 is within the time fusion window, so it'll all just sound louder.

Yes, the x-axis scale is seconds. Good point about the fusion window, I didn't think of that.

 

[Music1969]

The amplitudes are much higher with the red measurement, indicating that the dynamics are better.

i dont think this true

i ask similar years ago here and this was said to be false interpretation

seems this needs to be clarified first

 

[Keith_W]

i dont think this true

i ask similar years ago here and this was said to be false interpretation

seems this needs to be clarified first

Good. I hope to have mistaken beliefs corrected. Do you know what measurement corresponds to "dynamics" then? Seems as if we all hear it, but I don't know what measurement it corresponds to. I thought it might be the height of the step response, but I am not sure.

 

[Tim Link]

Actually .2 would be 200ms. That's beyond the fusion window.

 

[3ll3d00d]

1. Do you think this is a valid way to measure the subjective phenomenon of dynamics? I have never seen anybody describe this method before. Have I invented a method, or has it been described before (and I have missed it in my reading), or is it totally wrong?

No, this is nothing to do with dynamics, it's just the strength of the overall impulse at different point in time. Were these two measurements taken with the "freeze gain" option selected? if not, they're not directly comparable to each other, only relative to the initial impulse of the same measurement, which may also be misleading. I don't think "dynamics" is a quantifiable thing as it's never been defined precisely by anyone that I've ever seen (i.e. you'll get n different answers to what it might mean)

for what you are looking at (comparing measurements over time), probably you'd be better off looking at a wavelet spectrogram (e.g. in rew).

 

[Music1969]

Do you know what measurement corresponds to "dynamics" then?

maybe compression test like Erin does

i started a thread about it long ago, and it is not conclusive how to quantify dynamics but maybe compression is one part

I thought it might be the height of the step response, but I am not sure.

as mentioned before, it is not that

 

[Music1969]

I don't think "dynamics" is a quantifiable thing as it's never been defined precisely by anyone that I've ever seen (i.e. you'll get n different answers to what it might mean)

correct there is already a thread and lots of discussion over the years

Very good video in the opening post to watch - amir noted it was a good watch also

Dr. Klaus Heinz of HEDD Audio (ex ADAM Audio) - measuring speakers, in particular speaker dynamics

Darko had a great interview with the main man of Hedd Audio (founder of Adam Audio before he left Adam to start Hedd). Really interesting discussion (especially 2nd half of interview) about speaker measurements and dynamics and how to measure. Link: Klaus Heinz talks HEDD Audio Tower Mains

www.audiosciencereview.com

 

[3ll3d00d]

Assuming freeze gain wasn't selected then the relative height difference (between initial and later) could be due to a sharper/tighter initial step which is what you may get from a different phase correction. Compare the step responses to see if that is the case.

 

[Keith_W]

I had a gathering over the weekend where some friends came over and listened to these target curves blind.

All the filters were generated with Acourate using the exact same measurement and all processing with the same settings. I selected 15/5 windowing so the upper frequencies were mostly smoothed over. I loaded the filters into Acourate Convolver, then relabelled them "A, B, C" and so on. Acourate Convolver has a web interface, so they can cue up any music they like and change filters on the fly using my Android tablet.

Sadly, only two of them could make it (the others had family members with COVID). Here were some comments from the gathering:

- The Toole target was felt to be too bright and lacking in bass by both of the listeners. I did not feel this way, I thought it sounded linear.
- The Brueggemann target was felt to be "shouty" by one listener. The other was neutral about it. This is my preferred target, it has a nice midrange fullness and enough bass without sounding excessive.
- The HATS target was agreed on by all listeners to have too much bass.
- The Rtings and Harman target were preferred by 2 of the listeners who felt it was the most balanced. I disagree, I thought that the bass was bloated.
- Some types of music showed up the differences between the curves more dramatically than others. One of my friends decided he wanted to torture test my subwoofers by playing bass heavy tracks. He generally preferred the targets with more bass. These same targets sounded unbalanced with my usual choice of music.
- Target curve burn-in is a thing. After some time listening to a target curve, it is preferred and suddenly changing tonality with another target causes initial dislike of the new target.

There were even more targets available for listening (including the personal target curves used by my friends), but by this time listening fatigue had totally set in.

I believe that preferred target curves can vary because of:

1. Different directivity characteristics of different speakers + room reflections. If your speaker has wonky directivity characteristics like mine, some target curves just don't work.
2. Different volume of initial measurement will impose a target curve for that particular volume. We know from Fletcher-Munson that listening volume changes tonal characteristics. So if the initial measurement was very loud, and music is played back softly, it will sound bass deficient.
3. Different taste. I prefer less bass, some people prefer more. We know this from Olive's research on headphones, yet it is not generally accepted for loudspeaker correction.

 

[Duke]

I had a gathering over the weekend where some friends came over and listened to these target curves blind.

All the filters were generated with Acourate using the exact same measurement and all processing with the same settings. I selected 15/5 windowing so the upper frequencies were mostly smoothed over. I loaded the filters into Acourate Convolver, then relabelled them "A, B, C" and so on. Acourate Convolver has a web interface, so they can cue up any music they like and change filters on the fly using my Android tablet.

Sadly, only two of them could make it (the others had family members with COVID). Here were some comments from the gathering:

- The Toole target was felt to be too bright and lacking in bass by both of the listeners. I did not feel this way, I thought it sounded linear.
- The Brueggemann target was felt to be "shouty" by one listener. The other was neutral about it. This is my preferred target, it has a nice midrange fullness and enough bass without sounding excessive.
- The HATS target was agreed on by all listeners to have too much bass.
- The Rtings and Harman target were preferred by 2 of the listeners who felt it was the most balanced. I disagree, I thought that the bass was bloated.
- Some types of music showed up the differences between the curves more dramatically than others. One of my friends decided he wanted to torture test my subwoofers by playing bass heavy tracks. He generally preferred the targets with more bass. These same targets sounded unbalanced with my usual choice of music.
- Target curve burn-in is a thing. After some time listening to a target curve, it is preferred and suddenly changing tonality with another target causes initial dislike of the new target.

There were even more targets available for listening (including the personal target curves used by my friends), but by this time listening fatigue had totally set in.

I believe that preferred target curves can vary because of:

1. Different directivity characteristics of different speakers + room reflections. If your speaker has wonky directivity characteristics like mine, some target curves just don't work.
2. Different volume of initial measurement will impose a target curve for that particular volume. We know from Fletcher-Munson that listening volume changes tonal characteristics. So if the initial measurement was very loud, and music is played back softly, it will sound bass deficient.
3. Different taste. I prefer less bass, some people prefer more. We know this from Olive's research on headphones, yet it is not generally accepted for loudspeaker correction.

Very interesting! I assume these are "in-room response" target curves? Microphone position(s) in the listening area?

 

[Ze Frog]

I had a gathering over the weekend where some friends came over and listened to these target curves blind.

View attachment 369217

All the filters were generated with Acourate using the exact same measurement and all processing with the same settings. I selected 15/5 windowing so the upper frequencies were mostly smoothed over. I loaded the filters into Acourate Convolver, then relabelled them "A, B, C" and so on. Acourate Convolver has a web interface, so they can cue up any music they like and change filters on the fly using my Android tablet.

Sadly, only two of them could make it (the others had family members with COVID). Here were some comments from the gathering:

- The Toole target was felt to be too bright and lacking in bass by both of the listeners. I did not feel this way, I thought it sounded linear.
- The Brueggemann target was felt to be "shouty" by one listener. The other was neutral about it. This is my preferred target, it has a nice midrange fullness and enough bass without sounding excessive.
- The HATS target was agreed on by all listeners to have too much bass.
- The Rtings and Harman target were preferred by 2 of the listeners who felt it was the most balanced. I disagree, I thought that the bass was bloated.
- Some types of music showed up the differences between the curves more dramatically than others. One of my friends decided he wanted to torture test my subwoofers by playing bass heavy tracks. He generally preferred the targets with more bass. These same targets sounded unbalanced with my usual choice of music.
- Target curve burn-in is a thing. After some time listening to a target curve, it is preferred and suddenly changing tonality with another target causes initial dislike of the new target.

There were even more targets available for listening (including the personal target curves used by my friends), but by this time listening fatigue had totally set in.

I believe that preferred target curves can vary because of:

1. Different directivity characteristics of different speakers + room reflections. If your speaker has wonky directivity characteristics like mine, some target curves just don't work.
2. Different volume of initial measurement will impose a target curve for that particular volume. We know from Fletcher-Munson that listening volume changes tonal characteristics. So if the initial measurement was very loud, and music is played back softly, it will sound bass deficient.
3. Different taste. I prefer less bass, some people prefer more. We know this from Olive's research on headphones, yet it is not generally accepted for loudspeaker correction.

Interesting. I have long thought that people seem to like too much of a bass boost and it seems to be a rather commonplace thing that people seemingly enjoy. Listened to many systems over the years where I've felt bass is overdone yet others have really enjoyed it. When I was younger I did enjoy more bass, to the uninitiated me back then I felt that bass was kinda the mark of a quality system, I'm guessing that's something that largely is a thing to people who aren't so into Hi-fi so deeply. Should imagine those particular curves probably are really great for the 95% though.

 

[Mnyb]

Nicely done .

I think an explanation for "to much bass" is that we tend to listen at a lower volume than intended hence the equal loudness contour comes into play and the anemic mixes on many tracks if the kick drums does not "kick" even in the mastering what should a poor sod do .

This plays in to the circle of confusion and that there is no standard for music production at least movie sound has some loudness goals and other standards so playback volume and balance can be calibrated .

Did you do this over speakers ?

As dr toole has explained ( yet again and again in another tread ) the preference curve is not realy a target in itself but a results of a speaker behaving well in home acoustics .
Trying to fit these curve by eq of the direct response might be unpredictable at best ?

The low frequency part is probably a very OK tests as there room and speakers can not be separated by two ears and a brain .

 

[Basic Channel]

As dr toole has explained ( yet again and again in another tread ) the preference curve is not realy a target in itself but a results of a speaker behaving well in home acoustics .
Trying to fit these curve by eq of the direct response might be unpredictable at best ?

The low frequency part is probably a very OK tests as there room and speakers can not be separated by two ears and a brain .

I agree with you and dr Toole, it even intuitively makes sense when you think about it. I have issues with the idea of “home acoustics” as if that’s a thing. Is that a new home or an old home? A big gigantic US villa, or a pokey wee flat in Edinburgh?

That is the main confusion (along with people publishing/praising in room graphs) with Toole’s curves (and ER research)

 

[Keith_W]

Very interesting! I assume these are "in-room response" target curves? Microphone position(s) in the listening area?

That's right. Mic in MLP, amplitude cut to match the target.

This plays in to the circle of confusion and that there is no standard for music production at least movie sound has some loudness goals and other standards so playback volume and balance can be calibrated .

Did you do this over speakers ?

Yes, we listened over speakers. The same system discussed in this thread. I do have Dan Clark Stealth headphones, but those are PEQ'ed to my taste. It is slightly more U-shaped than standard. I have no way of measuring headphones, so I did it subjectively.

As dr toole has explained ( yet again and again in another tread ) the preference curve is not realy a target in itself but a results of a speaker behaving well in home acoustics .
Trying to fit these curve by eq of the direct response might be unpredictable at best ?

It should be noted that Dr. Toole's recommendation of not using a target curve relies on the natural behaviour of a speaker which is anechoically flat in a listening room. Suppose you own a perfect speaker which is ruler flat in an anechoic chamber at 1m. Place it in your room and measure at 1m, 2m, 3m, etc. You will see the FR start to fall, with a more pronounced slope the further away you are from the speaker. Therefore, the tonality of the loudspeaker is dependent on listening distance.

I am sure that Dr. Toole himself realizes this, because he also said that "broad tone controls" can be applied to upper frequencies. Choosing a FDW (Frequency Dependent Window) of 15/5 refers to 15 cycles of bass (i.e. fine correction), and 5 cycles of treble (i.e. a lot of smoothing). This is the same as a "broad tone control". If I wished, I could forego high freq correction altogether, but my speaker needs some kind of HF target curve because (1) the passive crossover has been bypassed, (2) the horns have different nearfield/farfield characteristics as the bass cabinet - so an anechoic flat frequency response will not result in a falling Harman-like curve. I know from my own experiments (detailed earlier in this thread) that equalizing the speakers to flat nearfield results in a rising frequency response at MLP. Therefore I need to impose a target curve from the MLP. Don't worry, I know better than to correct every little wiggle that I see.

BTW, I do a lot of experiments with DSP to satisfy my own curiosity, and sometimes this involves deliberately ignoring Toole and doing the opposite of what he says. I can tell you that the result isn't as disastrous as it's made out to be, but there is a niggling feeling of discomfort at the back of my head that comes from ignoring the experts. The reason I choose a FDW of 15/5 is to alleviate that discomfort.

I agree with you and dr Toole, it even intuitively makes sense when you think about it. I have issues with the idea of “home acoustics” as if that’s a thing. Is that a new home or an old home? A big gigantic US villa, or a pokey wee flat in Edinburgh?

It was a new home when I bought it 14 years ago. It's not so new any more. It's a moderate size home in suburban Melbourne with a nice, big listening room for its size.

That is the main confusion (along with people publishing/praising in room graphs) with Toole’s curves (and ER research)

As I have said a few times in this thread, I do what works for me I make no claims as to whether it will work for you!

All I can say about this particular blind test is that it is specific to my system and my room. And I do think that I blinded the participants pretty well, I sat behind them and said nothing.

 

[dualazmak]

As I have said a few times in this thread, I do what works for me I make no claims as to whether it will work for you!

All I can say about this particular blind test is that it is specific to my system and my room.

Thank you as always for sharing your nice and interesting audio experiences and results.

I have been, and I am now, well understanding above your very important point/stance on your own audio system in your own room acoustics!

I agree with you that all of our audio experiments "at home" is always depending on the specific audio setup/gears in the specific room acoustic environment.
Sharing such experiments with other people would be a nice "reference" information, but it would not always work similarly in other people's audio system and room acoustics. Nevertheless, your approach and results are very much worthwhile for other audio enthu people including myself.

I have done (on my project thread), and I also doing now (e.g. ref. my post here and thereafter) my audio experiments based on the same "stance" (or "policy") which you kindly pointed above; really important point in sharing our audio exploration at home with other people.

 

[Music1969]

- Target curve burn-in is a thing.

Yes this is simply brain burn-in.

 

[Basic Channel]

It was a new home when I bought it 14 years ago. It's not so new any more. It's a moderate size home in suburban Melbourne with a nice, big listening room for its size.



As I have said a few times in this thread, I do what works for me I make no claims as to whether it will work for you!

All I can say about this particular blind test is that it is specific to my system and my room. And I do think that I blinded the participants pretty well, I sat behind them and said nothing.

People often distill what Toole says (afaic tell and understand). I was just pointing out I had a problem with "home acoustics" and the notion of it.

If we assume your room is what Toole considers a typical home theater (100m3 - bigger than my flat almost) and assume your speakers and acoustics is good, the effect of EQing to a target would surely be better (or less bad) than if I did the same while sat in the corner of a small room. By definition if you have a "standard" home and good speakers then it'll be close to Toole already.

A similar thing with early reflections. Your room is maybe big enough that you can have an opinion on whether they are beneficial. In mine they are not beneficial.