Kvalsvoll

A simple plugin in audacity. I have tried different approaches, but the results are not worth the trouble. The problem is that it is not possible to restore the signal that is lost when clipping occurs - there will be a small timespan where all sound is muted, and this affects higher frequency content.

I checked the differences between dBZ-dBC-dBA for typical program material once, but I don't have the number here now. What I remember is that the diff was much smaller than expected, so around -5dB for C, and some more (-10dB may be too much) for A. It is the frequency distribution that matters, not the dynamics. The dynamics matters for the diff between peak and the dB-z/c/a. Average spl >100dB is loud, but frequency distribution and dynamics determines how it is perceived - more dynamic material will have louder peaks, and will be perceived as louder, heavy bass is a lot more tolerable than loud mid/high. When I play Panda Dub at 125dB peaks, it may be perceived as comparable in loudness to the 123dB dubstep/EDM. But it is not comparable to 130dB+. Did an experiment once on a drum track with Marilyn Mazur. Found that the drums were clipping heavily on the initial attack, used a restorer plug-in to gain around 10dB more dynamics. Now the hypothesis was that the track could be played at +10dB more volume, and still sound equally loud, but with more powerful and realistic drums. Not the case. The drum hits got painfully loud, the overall experience did not improve much. The surprise here was that the louder very short in time peaks had a very significant impact on the perceived overall loudness.

And what makes sense, how loud do you need. Assuming one wants to explore the full dynamic range of human perception of sound, my suggestion is: - 130dB peak below around 100hz. - 120dB peak is enough for trad/acoustic/rock music. - Full frequency range, quality and tactile feel is mandatory. But then I do not listen much to EDM or dubstep, with those continuous droning bass tones. When sound becomes so loud it is no longer pleasant, it does not makes sense anymore. And very loud spl - even in the bass range - is believed to cause hearing damage. In the end it all comes down to personal preference - one persons too loud is exciting and pleasant body massage for another.

Whether this is insane depends on where the measurement is taken - if this is the mix console out where the audience is, as in what we would describe as the "listening position", then it is VERY LOUD. I found this, a spl log playing at 0dB and then +6dB. The mic clips around 120db peak, the +6dB level is around 125dB peak max. Occasionally I will listen to music at this level, but not for very long, maybe a song or two.. or three. Because it is kind of addictive. But this is LOUD. Not loud as in "oh I can actually feel the bass"-loud - this is like a wall of sound hits you, and every time the heavy bass drum hits, you are blasted with a shock wave that makes it feel like everything is moving - the house, the floor, the seating. But this is also dynamic music, it is only the peaks that reach above 120dB. dB-z (full-range, dBC is around 5-6dB lower, dBA maybe 10dB lower) is in the range 105 to 110dB.

I think I have the perfect recipe for selling uber-expensive speakers in the Room2 now: - First, play some dynamic and full-frequency content in the Moderate Cinema, on the big system, at full volume (that is in the 0dB to +8dB range). - Then move the customer into the sweetspot in Room2, and put on some quiet vocal music like Susan Wong, play at modest volume, make sure to point out how "real and present" the vocal is. - The potential customer has already a memory of the transient response and dynamics and realism from the other system, and now experiences this audiophile vocal qualities - all those qualities are now mixed up together, in an assumption that this is what will create the magic sound in their own home. - The customer brings the uber-exp speakers home, and.. disappointed - realism and dynamics are lost, it does not sound like the demo. But the vocal sure is nice, at modest volumes, after placing the speakers out in to the room far away from the front wall. - Since the magic sound did not manifest, something must be wrong with the gear or the room. Perhaps a larger bass system is what it takes, or some new amplifiers.. perhaps a new dac.. - And there you have it, we have a business - this customer will never be satisfied, and will continue upgrading and changing gear forever or until the money runs out. Installing the F1 speakers in Room2 confirms the room is not that bad now, though there are still some problems with cancellations. However, I now believe it was a huge mistake to do this project with the new prototype F2 speakers. Those speakers are still in development, lots of issues remain. They do not image like the F1, they do not have that sense of effortless transients at moderate volume, they do not have that realism and explosive impact on louder volumes, the high frequencies sound kind of muted and there is a distinct color to the sound in the upper midrange, they do not sound pleasant and comfortable at tear-the-house-down volumes. While the F2 in the current state sounds more like a typical hifi-speaker on steroids. the F1 is a different animal. As an example; put on Flashbulb - We are the dispelled (find it on bandcamp). Play this at 0dB in the Room2 - sounds loud, but not too loud, still not pleasant, and it sounds like the resolution in the recording is not very good, the bass is very colored, this is not a fault of the F2 rather than the bass system with 2 excellent but toooooooo small S6-14 in combination with the room. Now fire up the Moderate Cinema system, and enjoy. The difference in resolution and realism and dynamic impact is huge. That the bass is much better should not come as a surprise, but there is more to it than that - the midrange is more see-through and dynamic, the highs have more clarity and reveals more of the instruments individual character, the whole presentation is a lot more pleasant to listen to, and it sure makes sense to listen at this quite loud volume. I think I will put another set of F1 speakers in to the Room2. The F2 speaker was something I wanted to make because I Iiked the idea, and I needed to try out a new HF section for higher output and ability to go lower than the horn loaded ribbons. Other larger speaker designs are coming up, and those will use the 8" LF/mid drivers from he F1/C1, but I had hoped to be able to use the higher output AMT horn for HF, but as this turned out that is really not an option. The room is important, but it is the speakers the determines what is possible to achieve.

Yes, start the target-curve /house-curve thread, @SME. I see some interest for this topic around, and there is much knowledge learned between the Bruel&Kjær-experiment and todays situation.

I have mentioned before some of the consequences of radiation pattern - mainly, general differences between horn - small direct radiator. One aspect is that the horn tend to sound brighter/louder, for same spl and frequency response. This is due to differences in radiation pattern. The small direct radiator (dome and very small midrange cone driver) has a wide radiation pattern where the sound slowly decreases in level as you move further off-axis. The controlled radiation horns in my speakers have fairly flat response across a defined angle, and then drops very fast when moving further off-axis. The consequences are that reflections from the room behave very different. The horn focuses the sound more towards the back of the room, and actually creates louder reflections, after some time. The dome/small speaker will have stronger reflection level overall, especially very early reflection level will be louder, but the reflections coming from the back of the room will be smaller. I tried to make some graphs to help illustrate this: ETC with smoothing so you can see the individual plots - horn gives overall lower reflection level, but some around 10ms are louder: IR, no smoothing, show that the peak is quite a lot higher in level on the horn speaker: ETC/IR plots are normalized, so differences in loudness between the speakers have no significance. This can also be seen in the waterfall plot - the horn will tend to have more late energy. It is this late energy that adds to the perception of tonal balance. One consequence of the horn radiation is that it makes it easier to create the sound I want in Room2. Less early reflections even with the speakers close to wall boundaries give improved clarity, and the strong focused sound towards the back makes it possible to use small diffusors to create lots of ambience.

Pictures of the cylindrical diffusors while we wait for the acoustics to "burn in". Left side diffusor: The back of the room with 2 side, one small back, one larger top back: Cylindrical diffusors are simple and should work well with the small distances in this room. Reflection level from the left speaker - reflecting from the right corner with the fireplace - is quite high.

I found it, it is exactly like you describe it, @SME. DRC is enabled by default in ffmpeg, and this causes disaster for sound quality - hard, harsh sound, too loud dialogue, loss of the impact I imagined I would get from the massive basseq. The impact on sound quality is much more than just a little louder background effects and a little less brutal booms and thunders. And when it is applied, there is no way to repair it. I had made 3 movies, with brutal basseq, it took me a while, and sat down the next day to enjoy a real beating in a battle scene. And then.. WHAT IS THIS! I had prepared to maybe adjust the level a little to compensate for the dialnorm now missing, but clearly something more was very wrong. Solution is to use ffmpeg command line to extract the audio. Audacity can not take arguments for ffmpeg, at least I was not able to find out how. All ffmpeg decoding of dolby ac3/eac3 defaults to using drc, I assume controlled by metadata, so some files can be good, others not. This problem is caused by Dolby defaulting all playback software and devices to the worst possible sound option, so that people can have the worst sound experience out of their already quite bad soundbars and alike.

Some minor acoustic fixes, and repositioned the speakers - with the corrected crossover. Starting to sound quite nice now, can easily enjoy music for several hours in there. If I can fix the remaining issues, this will be something quite special. Now there is much more defined body on instruments in the lower midrange, it is fun when it starts to sound like real things are playing. So far I believe the listening experience correlates well with measurements. The better it measures, the better it sounds. But it can sometimes be a challenge to analyze and find what exactly caused a specific difference in perceived sound. The only deviation from ideal, clean, uncolored I have found to be good for sound is the slightly tilted response with more level at lower frequencies - perhaps also the added reflections from sides, because they do actually compromise the IR. Clarity down through the whole midrange has improved. One interesting observation is that even with this very dry front stage it seems to amplify room information from the recording - it is much more like you move the whole listening room in to the concert hall, or jazz club, or studio - depends more on the recording, and the acoustics of the listening room is simply not there. And there is so much of this room information, it is like there is much more room than you get in a listening room with no acoustic treatment. QSound sounds better than a surround system you can buy at a local shop. There is sound from the sides, and even above, and from behind. And not only ambient effects, there are objects with physical size and location, like you get from a very good surround setup. But of course, this only works in the sweetspot. The work with this room confirms that focusing on a flat/smooth frequency response alone is not enough, what happens in the time domain is what makes the real differences.

?: So, I have this "movie" with EAC3 audio. It is encoded for playback with -7dB dialnorm/offset. Decoding for processing, fixing the 40hz cut-off, making for an improvement that only bass-eq addicts can believe. But. Levels are off, and there is something strange going on with dynamics. Inspection show dialogue is 7dB louder on center channel, music/l+r is 3dB louder. Listening to different scenes suggests the center could be reduced -4db to -7dB. Dialogue is far too LOUD when played at 0dB. But then I notice background effects are also louder on the fixed decoded version. Something strange is going on here. It looks like the decoder did not do what the processor does on playback of the original eac3 track. If the whole thing was just 7dB louder, that is not a problem, just reduce level to -7dB. Or simply ply the whole thing louder - my original intention. But that does not work well if dialogue level is too LOUD and some kind of dynamic compression is applied in the decoding process. I used ffmpeg(Audacity) for decoding. Any ideas?

Interesting thoughts about the hf roll-off and how flat to 20k makes a difference. My experience correlates well with your findings - flat all the way up makes for better transients and texture. And it does sound a little brighter. I decided that this has to do with changes inside the audible range - up to 16k for me now, and not only level, but also radiation and power response. If the hf is rolled off, and also beams above say 10k, there will be very little energy left up around the very highest you can hear. Working on the F2 with this huge amt driver with very bad radiation in the upper range I notice high frequency sounds are less realistic and kind of appears smaller in size. Increasing the dispersion in the upper end and also eq to flat all the way up improved this a lot, but it is still not up with the smaller horns with the ribbon drivers. It sound analytic and very refined, but lacks realism.

Solved one part of the problem - the peak around 1k: Turns out the last crossover version on the F2 speaker had a fault in the correction circuit for the lf driver. A quick-fix improves things a lot, much smoother response, now it sounds kind of ok without additional eq. At 500hz+ wavelengths are already getting small enough to be possible to find by moving a smaller absorption panel around, while observing changes in the response on rta, I use this technique a lot, both for placement, room acoustics and also to help identifying problems with speakers. But I could not get rid of this peak, no matter where and what I placed around in different locations. The only thing that worked was blocking the direct sound from the speaker. Dragged a different speaker into the room and placed it at the exact same location and there it is - no peak, just horrible response below 500hz. There is still some gain above 500hz, but the level is acceptable. So now the problem with the dips from 500hz and down remains.

Velocity measurements can only tell the direction of the sound, what is actually happening can be a mix of reflections from many directions. From the plots I presented here it is not even possible to see if the sound comes from front or back, only if the direction is front-back. Just listened to several tracks, comparing the 2 systems. I still believe there is a difference in transient attack in the midrange. Like the drums loose some of the initial snap. Had the spectrum analyzer running, to see what i am actually listening to. Frequency distribution of signals are often much wider than you think. Example: Hadouk Trio, Moussa. I also believe the Room2 system has improved. EQ to remove the loss of energy in the 200-500hz range, adjustments to acoustic absorption. I find it fascinating to observe properties of sound, and then try to find what is causing what I hear. Everything can be measured, but it is not that easy to tell how to analyze the measurements, and see how they correlate with what is actually heard.

It should not have significance, but more experiments are necessary to remove that possibility. The bass is actually not delayed, the initial pulse starts at 0, it is the room that makes adds group delay and time smearing. The bass is perceived as quite good, but not like the other room, especially when turning up the volume the difference is very obvious. Understandable and acceptable, considering the very huge difference in size and capacity of the bass systems. Some of the differences may be caused by the F2 speaker with the rather generous contribution in the 2k-5k range, this is caused by a fault in the radiation pattern of the horn. The added late energy in the 2k-5k range will affect perception of tonal balance on transients, and can also (as mentioned by @SME) mask. It's like the transients have lost a lot of the addictive realism that the other set-up has, and that is a huge contributor to the addiction you get when listening to it - you always want to hear one more album. Smoothness and clarity and resolution at higher frequencies is a little better on the F2 speakers, but they also sound a little more laid-back. Voices have very good presence, like there is a person inside the room. But there are measurable differences, clearly there is a problem in the lower mids where you get huge cancellations, and the phase and timing errors will still be there even after eq. The peak around 600hz-1k will also contribute to the tonal balance, it is quite huge. .

Yes, ir/etc is not useful at all to see what is really going on, it is also quite easy to get a very "good" ir by adding some thin high frequency absorption in the right places, but this will not improve the sound, only make it more dead at high frequencies. I have the mdat-files, so for me it is possible to view and analyze these measurements in any imaginable way, but when posting measurements it is not practical to present more than a few graphs, with one scaling for each chart type. The idea of making the complete set with spl and velocity is to have information that can reveal where the reflections are coming from both in time and direction. I managed to remove the 200hz dip completely, but the sound did not improve much. Or did it. Sometimes I wish I had golden ears; then I could just swap the speaker cables, sit down and listen, and experience a huge improvement across the whole midrange, with "prat" and "palpable bass". But some of us are not so easily fooled, we need real improvements to be satisfied. I need to find a more reliable way to do listening evaluations, I think I will look into that later today. It's not that the sound is horribly bad, it is just that the other room is so much better. Having this possibility to compare instantly reveals differences quite effectively.

h direction is horizontal, meaning it is measured in the horizontal plane, it shows the amplitude of the velocity vector normal to this plane, so this will then be the velocity of the sound in the vertical direction. Sound waves can be fully described by its scalar pressure - spl - and the particle velocity. Particle velocity is a vector - it has both magnitude and direction. Usually we only measure the spl, and leave it at that. Measuring the velocity in 3 directions gives additional information about the sound, such as direction. In free field with no reflections there will only be velocity in the 0-direction. When the h-direction plot shows velocity with magnitude close to or above the 0-direction something has changed the direction of the sound waves. In a room this directional change is a result of the reflections from the surfaces in the room. By looking at the velocity plots it is then possible to get more information about what is causing a troublesome reflection.

Comparing the Room2 with some eq and The Moderate HT. I measured both rooms, spl and velocity. That they sound different should not be a surprise, but the differences I am trying to analyze now, is the low-mid - mid range. The ModerateHT has better realism and impact in this range. Why is that so. The differences can not be explained by looking at the frequency response alone. With eq on Room2 to boost the 200hz-600hz range, it sounds more balanced, and freg resp does not have the dip. Room2 spl and velocity 0-90-h directions: MderateHT spl and velocity 0-90-h directions: Room2 decay (20ms lines, 10ms pre-window): ModeratHT decay: Room2 spectrogram: ModeratHT spectrogram: The problem is the Room2, the other room can be seen as a reference. The velocity spl plots reveal that the cause for the peak around 800hz is likely to be ceiling or floor. The dips may be cause by reflections from the back, but down at 200hz there may be a sideways reflection, and up closer to 600hz there may be a vertical reflection. If I place the speakers from the ModerateHT room in to Room2, they have the same faults in the response, only worse. So this has to be fixed, and it is a room acoustic problem, it is not the speakers.

Nothing to see so far, still working on it. Absorbers close to speakers improved things, you can see it in the measurements if you look very closely. Which means it did not improve enough. But they look nice. I have to improve this, the room is useless for the intended purpose like it is now. Unfortunately the dips are most likely caused by a combination of reflections from the back corners and side wall reflections, and that is not easy to fix without reconstruction. Lets see.. the measurements are open here, I can find a plot of the situation like it is now: The problems are the peaks around 800hz-1k and the dips in the 200-500hz range. Measurements with a different set of speakers shows close to identical response, this is caused by the room.

@SME: Yes indeed it has. Speaker radiation is one of - maybe the most - important single parameters for sound. Comparing those 2 rooms, the room also contributes, and more for lower frequencies. Decay in low mid - mid is much drier in the new Room2, and that is not only due to differences in speakers, it is because the room eats more energy in that frequency range. And the interesting point is that even though the decay is slower in The Moderate Cinema, the lower midrange is perceived as better - better imaging, more physical, more presence. That may have to do with the spectral balance - frequency response is smoother overall, no huge dips in the midrange, and the slower decay means you get this smooth, sloping decay response. Building SBIR/close to speakers absorbers to fix the low mid issues now, will be finished in a couple of hours, one is already done, one more to go.

@SME, I have already tried that. Various eq settings on the F2 speakers, and also 2 other speaker systems. If the F2 was to be used like it is now, and the room was like when those measurements were taken, then eq would be a good solution to improve the overall sound. Because what you say is true, the frequency response of the decay affects perception of tonal balance. However, it has improved some, and this is more of a problem with the speaker - which is not completed - than the room. Decay for other speakers show different decay response, and as a general rule, the decay response tend to be similar in different rooms, this is a result of the speaker radiation pattern. In the other room, the F1 speakers have a smooth sloping frequency response, and what is interesting is that the decay is also smooth and sloping with progressively more tilt for increasing time. Speakers with horn and controlled radiation pattern is interesting. Early decay energy drops very fast, but late energy has a relatively high level. When frequency response is equal, they tend to sound brighter than trad hifi-speakers.

Oh.. long post, but.. I like it, will read it later. Just wanted to clarify about the RT60 issue: I must admit I can not resist to look at the RT60 myself, even though it is "wrong". It tells the story of the room in a very quick glance, you can immediately get an idea of whether the space is kind of dead or more lively, or simply too much room. And you can see of the decay is smooth and even across the frequency spectrum. The problem is when the decay is very short and not necessarily smooth. The RT60 values are calculated from the decay rate within a given reduction in level, and when this reduction is not following a smooth decay line, the number does not represent the actual decay rate. You can see this when the 20dB, 30dB and Topt does not match up.

Thank you for sharing your thoughts, @SME. While waiting for the paint to dry on some mounting brackets for another diffuser for Room2, is a good time to post about the progress, and some general thoughts on sound. The Room2 gets closer and closer to the Moderate Cinema room as things are adjusted and improved, but I can't help it, it sounds like a little version of the other room - soundstage is smaller, instruments are smaller, there is too little body on smaller drums and instruments in the lower midrange, and when you turn up the volume the bass lacks some of the realism and impact. Some of this is caused by differences in the loudspeakers, some if it is room acoustics. @SME, reverb times (T60) are not very useful information for small rooms, in those rooms here the T60 values typically lingers between 0.1 - 0.2s across most of the frequency range, there is simply not enough diffuse and late energy to get a meaningful reverb. But that does not mean the space is dead. People usually think the surround speakers are also playing with 2-channel stereo, in the Moderate Cinema. This is because the sound fills the whole room, like you are sitting in the room where the recording took place. But the precision and placement of instruments are not compromised, everything has a precise location up front, and instruments and sounds take on a physical presence - they get size and body. It is the radiation pattern of the speakers that makes this possible - precise direct sound with little early reflected energy, and lots of delayed reflections from the back of the room, which gets diffused and spread by objects on the back wall and side walls. In Room2 I have added diffusors on the side walls close to the back wall, to create more reflections. I am now adding one more at the back wall - ceiling transition.

I should have a too dry room to test this in before giving an answer, but I can live with that. In any case, the Moderate Cinema is a very dry room compared to what most people have, it's just that there is a significant difference from dry and controlled to too dry. Assume we are talking about movie sound here. A too lively room will not sound good for surround, but too dry will also compromise the experience. Experience shows that it is not necessary to do much additional acoustic treatment for multichannel, such as trying to absorb early reflections from every single speaker - which would need damping on every surface in the room. It turns out that the radiation pattern from the surround speakers determines how the surround sound is perceived. With correct radiation the placement of sound objects and envelopment is not compromised if the room has some reflection and liveliness. And some reflections actually help to make the sound from the different speakers blend in seamlessly. The surround sound in The Moderate Cinema is very good. This is the result of the radiation pattern of the surround speakers and the room acoustics, combining to create this immersive and seamless and at the same time accurate presentation. In a multichannel room the ceiling should have acoustic treatment/absorption spread out across the whole surface, not just in first reflection points like it is often suggested for 2-channel. Doing this will not compromise 2-channel sound, actually I will recommend using the same approach also for 2-channel only rooms.

Yes. It is the sound pressure acting on surfaces and objects in the room that makes the rattles and buzzing and noise. By using common sense and a scientific approach you found out, instead of chasing problems in the wrong direction, which is the common approach still used in hifi. Lots of good information in your thread here, I like it.