View RSS Feed

mitchco

Hear music the way it was intended to be reproduced - conclusion

Rating: 1 votes, 5.00 average.
If you have followed this series on a quest for proper timbre, I have reached a conclusion. From wikipedia, in psychoacoustics, timbre is also called tone quality and tone color. No question, music source, electronics, interconnects, power, etc., all have impact on timbre. However, the biggest factor on reproducing proper timbre, by orders of magnitude, is the speaker to room interface which is limited by small room acoustics: http://www.gcmstudio.com/acoustics/acoustics.html

I was lucky to have worked in several recording studio control rooms. The LEDE control rooms are custom designed to produce the absolute best possible sound quality (i.e. timbre). Certified LEDE studio control rooms costs hundreds of thousands (and into the millions) to design, build, and certify. After logging over ten thousand hours of recording/mixing time in these rooms, every other rooms sounds (very) poor in comparison, including my current room.

As much as I would like to, building a properly designed LEDE room, outfitted with the state of the art diffusers, absorbers and bass traps, is not feasible. When I designed my computer audiophile system, I looked into Digital Room Correction (DRC) software. These series of posts have been about calibrating the speaker to room interface using well known industry standards and measures (i.e. B&K house curve, equilateral triangle, room mode calculator, RT60, etc.).

Before we look at some of the final measurements and analysis, a small aside on the measurement setup. Audiolense measures speakers and room. Based on those measurements, DRC filters are designed and hosted in an external to Audiolense program called a Convolver. I have a Convolver in JRiver as my default music player and a standalone Convolver when I am listening to MOG or Netflix.

Normally, I will perform the measurements using Audiolense on my audio computer with a Lynx L22 sound card. However, I wanted to measure sound in 3D (i.e. time, energy, and frequency like my old TEF computer), sometimes called waterfall plots. Ideally you want the calibrated frequency response (i.e. B&K house curve), so once the sound stops, it decays away at the same rate and ends at the same time across the frequency range. The easiest way to display this visually is using the waterfall plot.

Because Audiolense does not have a waterfall plot function, I used another fine piece of measurement software called REW: http://www.hometheatershack.com/roomeq/ I also used a different computer (laptop) and its internal sound module to perform the measurements. I was concerned that using different software and computer/sound module would introduce too much variability to the measurements.

I took a measurement using Audiolense and then the exact same measurement using REW. I cut n pasted the output of both and superimposed the 2 over each other as best as I could. While I was able to get the frequency scales to line up, my inability to line up the vertical scale shows a little variability within 2db. If I could match the vertical scales,the variability would be even less. The important point is that as is, our ears would hear the two as being identical.



The measurement is a high resolution picture of the frequency response at the listening position. I have zoomed the vertical scale and used 1/12th octave smoothing to show a detailed view. We can see a variation of 22db between the lowest dip and highest peak. This article on Small Room Acoustics explains the theory behind why that is: http://www.gcmstudio.com/acoustics/acoustics.html What I am impressed about is the tight tolerance between the two measurements but using different software and computer/sound card combinations. It is good to see this level of consistency.

That variance of 22db is typical in most small room acoustic listening rooms. Let's compare that variance to something in the electronic domain like my Lynx L22 sound card. It is ruler flat from 15Hz to 50Khz.



Here is the onboard sound module in my Dell Precision M6400 laptop. Its frequency response measures 20Hz to 20Khz +0.3 to -0.3db



This is exactly what I mean by the speaker to room interface having orders of magnitude more influence on timbre than anything else in the audio signal chain. Because of small room acoustics theory and measurement, most of our critical listening rooms have peaks, dips, honks, booms, and resonances. While generally speaking, electronics, cables, everything matters, have orders of magnitude less variance in frequency response, but also less distortion than speakers.

Here is a waterfall plot of my left speaker with the measurement mic in the listening position. Vertical scale is energy, measured in decibels. The horizontal scale is the frequency range from 20Hz to 20Khz. The Z scale is time in milliseconds. Starting at 0 time (i.e. measurement mic) and the decay to 300 milliseconds or a 1/3 of a second or given that sound travels a foot per millisecond, the sound has traveled roughly 300 feet in the room. That's a lot of reflections in my 30'L x 16'W x 8'H listening room.

As an aside, all measurements below were taken at the same sound pressure level within +- 0.5 db tolerance.



Classic dips and peaks of small room acoustics below 500Hz. I get 26db difference between the maximum dip at 45Hz and the honking spike at 200Hz. That is a large variance.

Here is the same measurement, but with Audiolense frequency DRC applied. Ignoring the B&K house curve slope, we are about +-3.5db across the listening range. That is a major improvement from +- 13db in the raw frequency response above. Also note the more even decay time across the frequency range.



Here is the waterfall plot with Audiolense True Time Domain (TTD) DRC applied.



Very interesting as the TTD DRC has really smoothed out the bottom end and extended it to 20Hz. Even more so than just with frequency DRC.

Let’s get serious and zoom on in below 500Hz as that's where the listening room has its maximum influence on the sound - i.e. room modes as per small room acoustics.



This room has a few problems for sure, even though it is typical of small room acoustics. Meaning if you take a frequency response measurement of your system at the listening position, you are likely to see something similar. Unless you have a nice large room (+56 feet length) or a golden room ratio or a LEDE room.

Here is the waterfall plot with frequency DRC applied. It is within +-3.5 db from 25Hz to 500Hz with a nice decay.



With TTD DRC applied:



We are seeing +-3db from 20Hz to 500Hz, with a real nice even decay. It can't get any better than that. If you look close, you will see it is labeled as TTD-11. I designed a dozen or so TTD filters, each time adjusting my filters in the Audiolense Correction Procedure Designer. The 11th version of tweaking the filters resulted in the best quality sound (i.e. timbre), both from a measurement and listening perspective.

The sound difference is as dramatic as the graphs depict. And sounds exactly as the graphs depict. With TTD applied, the sound is full, smooth, and crystal clear. A major improvement, bordering on as good as it gets.

I am impressed with Audiolense DRC. I can't listen to my system without it. With DRC in circuit, it sounds similar to the LEDE rooms I used to work in. It does not sound like my room anymore. The peaks, dips, honks, booms, and resonances are gone. The only acoustic treatment is a 10' x 7' throw carpet between the listening position and speakers.

If I looked at what it cost me for Audiolense, the measurement mic and preamp, it is almost criminal as to how good the results are for the investment. In my previous post, I have seen small fortunes spent on diffusers, absorbers, and bass traps. Look at the pictures.

In order to get the level of small room acoustics under control, comparable to Audiolense, I would have to purchase a lot of sound treatment. If I look at the value I am getting comparing the two, the cost of Audiolense and mic/preamp would only buy 2 RPG diffusers. It is likely I would require 6 to 10 or more diffusers for the back wall, depending on diffuser size. Something would need to go on the ceiling, even at a minimum, that would be 2 to 4 diffusers. Then likely 2 or 4 bass traps.

And that is why I have reached my conclusion. With DRC, the frequency response is +-3db throughout the listening range. The honks, booms, and resonances are gone in the time domain. The sound arrives at my ears at the same time, sounds full, super tight, and crystal clear - no more small room acoustics curse!

More acoustic treatment may help, but given the cost and how good the system sounds/measures with Audiolense DRC, I feel I am at the point of diminishing returns. I would need to spend several thousand dollars in acoustical room treatments to match what I can get with Audiolense DRC. Ideally, I would love to have both as then a full-on LEDE room may be possible. But for now, I am really enjoying my un-room. For once hearing proper timbre outside the studio control room.

Now that my system has been calibrated to reproduce the best possible timbre, I can really start hearing music the way it was intended to be reproduced.

I hope you found these series of posts useful in determining the best possible way to achieve proper timbre in the speaker to room interface.

Happy Listening!

Mitch

Updated 05-09-2012 at 01:57 AM by mitchco

Categories
Personal Blogs

Comments

  1. TimDH's Avatar
    I thoroughly enjoyed your approach to the topic. It seems crazy to spend too much time obsessing about cables when the room provides so many first order things to fix. It sounds as if your experience with Audiolense was very satisfying? I went to their website to try to learn more. It looks as if the end result of the correction is that the program writes a "corrected" .wav file that is then playable by any software player? That sounds like a simple solution, but I guess you end up needing more storage space in the end to hold the original ripped music and then the corrected .wavs? Thanks filling me in if I'm wrong.



    Tim
  2. mitchco's Avatar
    Hi Tim, thanks. I am very satisfied with Audiolense. It works as advertised and has great support from Bernt and members on the Audiolense user forum: http://groups.google.com/group/audiolense



    Yes, Audiolense produces a .wav file that is used in real-time playback of your music. You do not need to re-rip your music. http://en.wikipedia.org/wiki/Digital...tion#Operation



    The "try before you buy" feature indeed lets you rip 90 seconds of corrected music so you can hear the difference yourself and should play in any software player.



    But once you purchase, you can save the filters for any sample rate and "host" the .wav filters in a Convolver like http://convolver.sourceforge.net/vst.html that plugs into popular music players like JRiver and does the real-time processing as your music is playing. No need to re-rip.



    You can also use a standalone Convolver like Savihost: http://www.hermannseib.com/english/savihost.htm This means that it can convolve any music source like Spotify, MOG, Netflix, whatever you are listening to in real time.



    Hope that helps,



    Mitch
    Updated 05-09-2012 at 10:59 AM by mitchco
  3. TimDH's Avatar
    Thanks very much for the encouraging information. Sorry to be a little slow on the uptake: is the use of .wav files in a convolver the only way to play the corrected music, aside from the 90 second demo file that can be played back as is? I ask because many music players don't support a convolver and it would be interesting if Audiolense also does support re-ripping files for playback too, so that one could compare one's favorite player.



    Thanks,

    Tim
  4. mitchco's Avatar
    Tim, I don't know. I suggest you post your q on the Audiolense User Group: http://groups.google.com/group/audiolense Bernt (Audiolense creator) and members are very responsive and helpful.



    Cheers,

    Mitch
    Updated 05-09-2012 at 11:00 AM by mitchco
  5. Nyal Mellor's Avatar
    Hey



    Good series of articles...whilst I don't agree with all your points you certainly covered a lot of ground!



    Looking at your waterfall it's clear there is quite a lot of midrange room ring where the midrange frequencies are taking a long time to decay. Likely a result of your sparse furnishings and all the hard surfaces...room correction can't absorb any energy that's been out into your room by the speakers...you need things that absorb sound to do that...but you know that already I'm sure!



    The audiolense guys certainly know their stuff! I did have some debate over with them over at hi fi zine on one of the articles they wrote



    If your getting more into the analysis of the measurements I might suggest taking a look at a white paper I just published with another acoustician...http://blog.acousticfrontiers.com/whats-new/2011/10/13/acoustic-measurement-standards-for-stereo-listening-rooms-pu.html
  6. mitchco's Avatar
    Hey Nyal, thanks for your comment.



    re: ...quite a lot of midrange room ring... Well, I would say there is some for sure, but not quite a lot. Given that I like concert sound, I enjoy longer RT60 decay times. My room measures 0.6 to 0.7s, which is a little more than what you mention in your article, but still falls in the accepted range as described by Jackson, Leventhall, in the Acoustics of Domestic Rooms.



    I may put in a touch more dampening to bring the midrange RT60 down by 0.1 sec or so, but I am pretty happy with the decay time now. I have a strong aversion to overly dampened rooms.



    With respect to ETC, I meet your spec and my ETC plot shows pretty good smoothness and decay. As mentioned in my article, I could use some diffusers on my back wall, and some touch ups on the early reflections (i.e. ceiling) but given the cost and WAF...



    I enjoyed your article. I noticed no waterfall plots? For small room acoustics, I feel waterfall plots are essential to visualize sound in 3D. It would be cool to see the 3D plots of the very nice rooms you have highlighted in your article.



    As a constructive comment, it would be great to see a follow-on article with "how to" tips to bring someones acoustic listening environment into spec, something like: http://www.gcmstudio.com/acoustics/acoustics.html but tailored for stereo listening rooms. It needs a professional translation.



    Also, a calibrated frequency response has a direct correlation to timbre. Most people that try frequency correction and target a flat response are dissapointed to hear how "forward" (i.e. bright) the sound is and don't know why. As mentioned in my article, the B&K house curve http://www.bksv.com/doc/17-197.pdf Fig. 5 is an accepted industry standard and produces the most natural timbre from a frequency response perspective.



    Cheers, Mitch
    Updated 05-09-2012 at 11:00 AM by mitchco
  7. Nyal Mellor's Avatar
    Hi Mitch



    Actually the B&K house curve was one of the things I disagreed with. Just because there is correlation doesn't mean there is causality. I think there are many more relevant controlled listening test results to evaluate such as those covered in the Toole book that give better insight than the B&K paper.



    Re the use of waterfalls; Jeff and I prefer to use the 2D spectograms as shown in the low frequency decay section. It's the same data just presented in what I find to be an easier to read format.



    I understand your point about preferring a longer decay time, that's ok, each of us has their subjective preference. T60, like I said in the white paper, doesn't tell us much more than if a room is too live or too dead anyway. Much more instructive with reference to reflected energy in the room to look at an ETC and specifically one third octave filtered versions of that or a one third octave smoothed waterfall or spectogram.



    Thanks for your constructive critique, I did make a call early on that the white paper should not cover the how. There are just too many variables for someone like me (someone who does room design professionally) to be able to write a generic article on 'how to treat your room' and think that I had done a good job from an engineering perspective...maybe I could write a book, since there are so many nuances and if/then issues with the how to treat a room question...if you want simple answers better to get that from the acoustic treatment manufacturers.



    Thanks for writing your series of articles. We need more people who understand the importance of the room to sound reproduction!



    Nyal
  8. hulkss's Avatar
    You posted: "I designed a dozen or so TTD filters, each time adjusting my filters in the Audiolense Correction Procedure Designer. The 11th version of tweaking the filters resulted in the best quality sound (i.e. timbre), both from a measurement and listening perspective."



    Can you tell us more about the filter "adjustments"? Any "lessons learned" for a good solution?



    Thanks for a great series of articles.
  9. mitchco's Avatar
    Nyal for your reply.



    Re: B&K. Do you have a link to the book? I would love to purchase it and know more. Over 30 years of "house curves" I can't tell you how many I have tried and always keep coming back to B&K. Maybe all of those years in the studio has programmed me...



    My ears and measurements tell me that of all the house curves I have tried, this one seems to give the best timbre - i.e. tone quality, It sounds like a real instrument, not too dull or bright, but just right. Just like a real instrument with the location cues intact. Not to forward, not to back in the mix, but right where it should be. Used a lot in the recording industry.



    I would love to try something new out and if it brings me closer to the real thing, great, I am all for it!



    Keep up the good writings!



    Mitch
  10. Nyal Mellor's Avatar
    Hi Mitch



    The book is called Sound Reproduction by Floyd Toole, get a copy it's really THE summary of small room acoustics!
  11. norseman's Avatar
    Tim,

    Audiolense do not require any extra storage, apart from the correction files themselves. Think of this as an advanced software based equalizer with 65.000 taps for adjusting the different parts of the frequency. It's like you play your sound find and then "convolve" the filter on top of the frequencies of your songs.
  12. mitchco's Avatar
    thanks hulkss,



    Yes, it was an interesting experience and a few lessons learned.







    - initially used the "default TTD" in which you see the settings in the Audiolense Correction Procedure Designer above. Sounded good, but Bernt encouraged me to experiment.



    - incrementally increased the cycles in the measurement correction window and the true time domain subwindow (measurement window must be > than time window). Both in the low end and in the high end.



    - on Bernt's advice, added a mid-range filter centered at 1Khz. Basically, this is a pivot point, so that the bass measurement and time windows can be increased without affecting midrange.







    - Again on Bernt's advice, increased the amount of correction from default of 6db to 12db then to 15 db. I settled on 12db seems to work well in my scenario. TTD-4 above sounded better to me than the Default TTD.







    In JRiver ConvolverVST plugin above that hosts the Audiolense generated correction filters, there is a replay gain setting that with 12db of correction, needed to be cranked up to about 9db replay gain. From the Audiolense Help file: "We often get questions about why the filters attenuate the output so much. The short answer is that in order to avoid digital clipping, the average sound pressure level has to be attenuated by, typically 6-10 dB to give enough headroom for the single frequency that needs the most amplification. The frequency response of the correction filters – as they appear in the main form will give a 100% accurate picture of how the filters will attenuate the output. The sound pressure level of the evaluation charts doesn’t give the right picture in this regard."



    - Once I settled on a place to start, I went up and down in cycles in both the measurement and time windows. Each time I adjusted, I would play the same or different songs and listen for a while. Then bumped up or down the cycles, one at a time and repeat above. I would say about 8 to 12 hours of critical listening in 2 to 3 hour sessions over 3 or 4 days before I arrived at the final setting.







    - Once settled on the final settings, I measured TTD-11 and that is the one you see above. The measurement verified where I had ended up with my ears. I have not made any further adjustments. I feel it is as good as it gets. For me it is important to correlate what I am measuring and what I am hearing. The 2 pictures match.



    However, the obsession that being an audiophile is, I am sure I will get into another round of fine tuning the fine tune ;-)



    - One caveat is that using TTD, since it is messing with the time domain, the sweet spot is on the couch and in the 3 seating positions get the correct time and frequency correction. Note that Audiolense supports multi-seat correction.



    But for standing around, casual listening, parties, etc., I switch to frequency correction (select a different filter), which gives me the same tone quality (timbre), but myself and guests can hear that tone quality anywhere in the room (house for that matter) - not just the sweet spot. I lined up the equilateral triangle with a laser distance measure and got it to within a 1/16". The stereo image is point source and rock solid. You can wander around it in the room, like it is in 3D.



    As far as lessons learned, the main one is to experiment. It really helps to be able to listen and measure and correlate the two together. That helped me decide if I should design in less or more correction and add or subtract more time window (i.e. cycles).



    Hope that helps.
  13. mitchco's Avatar
    from: http://www.hometheatershack.com/foru...k-you-rew.html



    -------------

    Mitcho,



    You mention that the decay (waterfall) is better with Audiolense. But I really can't see any improvement in a shorter decay. Can you posts some more graphs that shows this?



    Also would really like to see an ETC of each speaker before and after the use of Audiolense.

    -------------



    Hello omholt,



    Thanks for asking. I think if you re-read my article, I used the phrase more "even" decay, especially in the low end, down to 20Hz with Audiolense True Time Domain (TTD).



    I did mean to say more even decay, but yes, as it so happens, a little bit shorter decay too in the mids, and the extreme lows, but an ETC plot won’t show that. An ETC plot, in this scenario, is likely the wrong tool for the job as Digital Room Correction (DRC) uses time, energy, and frequency correction in a real time convolver scenario.



    Meaning, in a passive acoustic room treatment scenario, diffusers, bass traps, and absorption are added to the physical room, each designed to handle a frequency band. And with the right amount of each type of passive acoustic treatment, a nice sounding room can be had (within the limitation of the room ratio and $'s outlaid) with an even RT60 throughout the frequency range.



    In this scenario, physical objects are placed in the room, but the frequency response coming out of the speakers is a constant. In other words, the frequency response coming out of the speakers does not vary. Using ETC curves in this scenario assists with the placement of acoustic treatment and you can compare the differences. This is our traditional approach.



    Contrast that to a DRC scenario where the situation is reversed. The physical room is now the constant, but the digital filters vary the amplitude, frequency, and time, (65,535 filter taps), So the music stream is convolved with these correction filters in near real time.



    Looking at the ETC's of my left speaker before DRC, with frequency DRC, and with TTD DRC, they look like 3 different rooms and that's what they sound like. Aside from the dominant reflections (which are also different), there is no correlation between the graphs as nothing was physically put in the room. So the ETC's look (seemingly) randomly different, i.e. different rooms.



    I think a better measure to compare in this scenario is RT60 decay http://en.wikipedia.org/wiki/RT60#Reverberation_time as you are asking to see a “shorter” decay. Using TTD I am able to get a more even decay or RT60 in my room and can be seen using these RT60 plots. I will get to those in a moment, but…



    But it is the TTD "focusing" that seems to be the magic of Bernt's Audiolense. I can hear Audiolense aligning or focusing the timing of the low end, midrange, high end and how much correction for each band so it all lines up giving you both the best frequency response and time domain response. Basically, you are sorting through the time domain (i.e. early reflections) to give you the "clearest" response. Too much correction over too long time window sounds like wearing headphones.



    It should be no surprise the clearest response and most natural timbre is also the one that is the most even or flat from an RT60 decay perspective. Here is the RT60 of the room without DRC:







    And here is the same measure, but with TTD DRC. As you can see, the RT60 increased in the 100 to 200Hz range (ponder that) and “shortened” the decay in the extreme low 20 to 80Hz and a bit across the critical midrange (700Hz to 3Khz). As stated, a more even RT60 throughout the frequency range, wihtout any passive acoustic treatment.







    Our ears are the most sensitive in the 2 to 5Khz range http://en.wikipedia.org/wiki/Equal-l..._determination I think Bernt is working on that from a psychoacoustics point of view and that seems to be one of the focal points of his (Audio)lense.



    Further down in the comments below, I was fine tuning the correction filters to apply just the right amount of time correction to make the sound crystal clear in the listening area. It is amazing to me the impact this has on how clear my stereo sounds and how 3D the soundstage is. Using Audiolense generated filters, I can dial in just the right amount to have the time domain, frequency domain, and RT60 decay all line up. Personally, I have never heard anything like it before other than working in a $1 million LEDE recording studio control room that has been designed to produce the best possible sounding room.



    Hope that answers your question.



    Cheers, Mitch
    Updated 05-09-2012 at 11:01 AM by mitchco
  14. ccclapp's Avatar
    Mitch



    I have followed your blog and the room correction series.



    Because you are interested and experienced in this, Im posting a link to a related thread on stereo and multi-ch DRC. We'd welcome your thoughts, comments, experience.



    http://www.computeraudiophile.com/content/Room-correction-Stereo-Surround



    In order to keep hardware out of the post-DAC signal path and to avoid additional a/d d/a conversion, the discussion also relates to multi-ch DAC's and pc-based DRC to minimize what "touches the signal", yielding PC DRC->Multi-ch DAC-> (amp) Speakers.
  15. ccclapp's Avatar
    Mitch



    You said "You can also use a standalone Convolver like Savihost: http://www.hermannseib.com/english/savihost.htm This means that it can convolve any music source like Spotify, MOG, Netflix, whatever you are listening to in real time."



    I looked at this but didn't see anything about its use in this context. Likely because I don't know what I am looking for ;-) Would you please elaborate?



    As is being discussed in the thread I posted above, one of the issues with my preferred approach of PC-based DRC (vs extra hardware) is being able to play things like MOG/Spotify etc and also external sources, e.g. tivo, etc, etc. At this point JRiver MC doesn't support these two types of sources, so I cant use thier DRC engine. It sounds like Savihost could possibly be the answer for at least 1 of these 2 scenarios, and possibly both.



    --How does it work in concept?



    --Does is spupport external sources comming into PC via spdif as well as streams like MOG?



    -- How do you find its results compared to JRiver MC? (NOTE: Im sure you know and are likely using JRiver v17, which greatly improved its DRC capability)



    By the way, if it is helpful to you or others, there is also a plugin for Squeezebox that enables that device to use FIR filters for DRC. Here is the link:



    http://inguzaudio.com/RoomCorrection/



    Ive not tried it yet



    Thanks again for your great write-up, hard work and feedback!!