Announcement

Collapse
No announcement yet.

Let's post something interesting: the ultimate small speaker build thread.

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #46
    Got it, thanks!
    Don't waste your money on a new set of speakers, you get more mileage from a cheap pair of sneakers. Next phase, new wave, dance craze, anyways it's still rock and roll to me!

    Comment


    • #47
      Wow, that is a crazy small cabinet. Keep it up!

      It will be interesting to see how the "equal loudness compensation" works out. I don't dispute the equal-loudness curves but I think my brain tries to compensate a bit for the sound level. My perception has been that cranking up the volume causes the tonal characteristics of a speaker to become exaggerated, whatever they may be. (With a clean and well-balanced speaker, it "just gets louder," to some extent.) I think distortion (both from the speaker and the ear) plays a big role in listening fatigue, in addition to the ears' sensitivity to treble. So I would really like to try this myself sometime--how'd you implement it?

      Comment


      • #48
        Originally posted by Michael P View Post
        Wow, that is a crazy small cabinet. Keep it up!

        It will be interesting to see how the "equal loudness compensation" works out. I don't dispute the equal-loudness curves but I think my brain tries to compensate a bit for the sound level. My perception has been that cranking up the volume causes the tonal characteristics of a speaker to become exaggerated, whatever they may be. (With a clean and well-balanced speaker, it "just gets louder," to some extent.) I think distortion (both from the speaker and the ear) plays a big role in listening fatigue, in addition to the ears' sensitivity to treble. So I would really like to try this myself sometime--how'd you implement it?
        Equal loudness compensation is a real issue for pro sound. At such high volumes, the treble must be toned way down relative to the midrange and bass in order not to sound fatiguing. Our ears are not flat, and they are most certainly not linear. Unfortunately few products actually do them right, which is why the "loudness" feature has a bad rep. You do not simply take the curve and implement it. That will sound bad. Only the bass and treble should be touched, no midrange. Also, properly done equal loudness curve needs a calibration in order to know the exact SPL at the listening position. Most don't, and throw a curve that may not be appropriate for the level at all.

        I did it by splitting the tweeter band into two bands at 5000Hz, and then I used a compressor on the 5KHz and up band to reduce its level.

        While distortion certainly causes listening fatigue, the tonal balance is a greater reason why at less extreme volumes. Most speakers are designed with a treble roll off in order to combat this. When you turn the volume up, the ear's treble sensitivity become higher, and the speaker sounds more balanced because the effect of the treble roll off is less severe. This is why many speakers sound better when played louder. However, a speaker like that will sound dull at low volumes. Without equal loudness compensation, it is all about shaping the frequency response of the speaker to maximize the volume range where the ear still find the speaker to have a pleasant tonal balance.


        Comment


        • #49
          Originally posted by bcodemz View Post
          Equal loudness compensation is a real issue for pro sound.
          Nice project, and good point about different volume levels being helped by different frequency balances. Just a suggestion, but basing the equalization on the volume of the program material dynamically will sometimes cause a distortion from realism, even though it might be more enjoyable.

          The dynamic algorithm will need to identify an averaging time to measure the level, and will set EQ based on that. The averaging time can never be set long enough to avoid deviating from being realistic with program material that has wide dynamic ranges and long periods of lower levels. As an example, while listening to a symphony, the volume control might be set so it's a realistic representation of live levels but the dynamic eq will take the long quieter passages and boost the lows and highs. That doesn't happen in real life at the concert hall. A small suggestion, consider placing a switch on the back to defeat it (if you have gpio's to work with on the electronics). It's a pita but much more realistic for those times you do have realistic levels. The holy grail would a remote with switchable fixed eq, but that would be complex to add.

          It looks like you're targeting maximum realism, so just something to consider.

          Comment


          • #50
            Originally posted by DDF View Post

            Nice project, and good point about different volume levels being helped by different frequency balances. Just a suggestion, but basing the equalization on the volume of the program material dynamically will sometimes cause a distortion from realism, even though it might be more enjoyable.

            The dynamic algorithm will need to identify an averaging time to measure the level, and will set EQ based on that. The averaging time can never be set long enough to avoid deviating from being realistic with program material that has wide dynamic ranges and long periods of lower levels. As an example, while listening to a symphony, the volume control might be set so it's a realistic representation of live levels but the dynamic eq will take the long quieter passages and boost the lows and highs. That doesn't happen in real life at the concert hall. A small suggestion, consider placing a switch on the back to defeat it (if you have gpio's to work with on the electronics). It's a pita but much more realistic for those times you do have realistic levels. The holy grail would a remote with switchable fixed eq, but that would be complex to add.

            It looks like you're targeting maximum realism, so just something to consider.
            Thanks DDF. A remote with switchable fixed EQ is exactly what this speaker will have. I can have up to 4 preset DSP settings to choose from, which is not enough.

            I agree on the realism. I've been listening with the equal loudness contour lately, and I find that the speaker just never sounds loud anymore. This is fantastic for detail, because the details just oozes out at higher volumes, but it doesn't seem loud, so the speaker just sounds hyper detailed, which is awesome. However, I won't get that jolt from a loud passage. I'm thinking of lowering the strength of the curve so the speaker will subjectively sound "louder" when the volume is increased. The current setting leads to too high of a listening volume because it sounds so good and doesn't seem loud at all.

            Accuracy is definitely a big part of this speaker, but that was simply a starting line. The real goal is to start with an extremely accurate speaker, and then adjust it accordingly to produce the best subjective sounding speaker. "Realism" is never a goal for me for many reasons. My goal is for music to sound pleasing to the ear, and that's quite different from sounding real.

            Comment


            • #51
              Originally posted by bcodemz View Post

              Thanks DDF. A remote with switchable fixed EQ is exactly what this speaker will have. I can have up to 4 preset DSP settings to choose from, which is not enough.

              I agree on the realism. I've been listening with the equal loudness contour lately, and I find that the speaker just never sounds loud anymore. This is fantastic for detail, because the details just oozes out at higher volumes, but it doesn't seem loud, so the speaker just sounds hyper detailed, which is awesome. However, I won't get that jolt from a loud passage. I'm thinking of lowering the strength of the curve so the speaker will subjectively sound "louder" when the volume is increased. The current setting leads to too high of a listening volume because it sounds so good and doesn't seem loud at all.

              Accuracy is definitely a big part of this speaker, but that was simply a starting line. The real goal is to start with an extremely accurate speaker, and then adjust it accordingly to produce the best subjective sounding speaker. "Realism" is never a goal for me for many reasons. My goal is for music to sound pleasing to the ear, and that's quite different from sounding real.
              Sounds good to me, should be really nice. What loudness contours are you using? Fletcher Munson was derived with headphones, Robinson Dadson with speakers so probably more accurate to your application (or even better, the latest ISO 226 spec). So changing the curve target to the ISO spec might help.

              Other thing I like to add is in-room boost below 80 hz per the JBL curve. Works well IME and is supposedly a better reflection of what the recording environment was like. Sounds funner too.

              I get out to the symphony often enough and it never sounds "loud" there until it's stupid loud (almost stressing the ear). My system has that same effect, I personally like it, but it eats allot of power.

              I can't explain why, but sometimes even the amp type can make it sound louder. I just bought a Nuprime IDA-8 for the home office and level matched it to within 0.1 dB of the Bryston 2B LP Pro it replaced. Using this in small room, so it gets loud, but the Bryston sounds subjectively ~3 dB louder, level matched and more dynamic (but much less clean).

              Comment


              • #52
                Originally posted by DDF View Post

                Sounds good to me, should be really nice. What loudness contours are you using? Fletcher Munson was derived with headphones, Robinson Dadson with speakers so probably more accurate to your application (or even better, the latest ISO 226 spec). So changing the curve target to the ISO spec might help.

                Other thing I like to add is in-room boost below 80 hz per the JBL curve. Works well IME and is supposedly a better reflection of what the recording environment was like. Sounds funner too.

                I get out to the symphony often enough and it never sounds "loud" there until it's stupid loud (almost stressing the ear). My system has that same effect, I personally like it, but it eats allot of power.

                I can't explain why, but sometimes even the amp type can make it sound louder. I just bought a Nuprime IDA-8 for the home office and level matched it to within 0.1 dB of the Bryston 2B LP Pro it replaced. Using this in small room, so it gets loud, but the Bryston sounds subjectively ~3 dB louder, level matched and more dynamic (but much less clean).
                I used my own of a combination of the ISO 226 spec, the JBL/Harman curve as well as the Audyssey implementation as a starting point as none of the current implementation curves are ideal to me. A straight Fletcher Munson curve implementation sounds quite bad. Audyssey's implementation is the best I've seen yet. They pretty much nailed the bass and most of the lower midrange, but I feel their treble cut started too high at 7KHz. I started mine lower and the curve is flatter.

                Not sure why your amp sounds like that. Never cared much about amps. I'm still waiting for a chance to hear an amplifier that I can tell a noticeable difference under a level matched condition. Although I must admit the amount of experience I have with high end amps pale compared to my experience with high end speakers.

                Comment


                • #53
                  Thanks, I'll have to try that myself or listen to one of these speakers someday. Can you describe any tapering of the bass response that you're doing in high SPL situations? (Or maybe I just need to figure out what the Audyssey scheme is.)

                  Comment


                  • #54
                    Originally posted by bcodemz View Post

                    I used my own of a combination of the ISO 226 spec, the JBL/Harman curve as well as the Audyssey implementation as a starting point as none of the current implementation curves are ideal to me. A straight Fletcher Munson curve implementation sounds quite bad. Audyssey's implementation is the best I've seen yet. They pretty much nailed the bass and most of the lower midrange, but I feel their treble cut started too high at 7KHz. I started mine lower and the curve is flatter.

                    Not sure why your amp sounds like that. Never cared much about amps. I'm still waiting for a chance to hear an amplifier that I can tell a noticeable difference under a level matched condition. Although I must admit the amount of experience I have with high end amps pale compared to my experience with high end speakers.

                    If you're game, it would be interesting if you could share your final curves, a very interesting topic. We're sort of mixing targets here a bit, so to clarify just mentioning that the ISO curves are free field quasi anechoic (no room contribution) where the Audyssey and JBL curves are in-room long-term average (as you know). For the in room curves, many folks (including Bob Katz) hinge at 1 kHz and ~-5 db at 20 kHz with a straight line.

                    Sean Olive and his guys ran a really interesting set of experiments deriving optimum in-room curves with expert listeners (grain of salt, we all like different flavours of sound) and then showed with typical but well designed speakers, what on axis is required to achieve this in their typical room. I think this may be documented on the InnerFidelity site. You're plowing new territory adding the cardiod into the mix and of course everyone's room will be different. Do you find you have to increase the bass response (when looked at quasi anechoic) with the cardiod relative to a point source, to get the same perceived in room bass balance?

                    For the amp differences, I'm not a tweak-o-file and am scientific/skeptical about things, but for example I built an N-Core NC400 and its sound is markedly better than other good amps (eg Bryston) I have, even if it measures just the same in frequency response (I measure and level match all my stuff when auditioning).

                    Comment


                    • #55
                      I absolutely STRONGLY agree that the cabinet walls should be curved to increase stiffness and use some type of constrained layer damping - such as a sheet of "roofing felt" paper between two curved metal plates as the enclosure. Ideally, the inner wall would be cast with integral bracing, then a layer of damping material, then finally covered with another sheet of metal. Combining two different metals with significantly different Q's (with a thin damping layer between) yields such spectacularly superior performance over a wood based cabinet that I would be hesitant to share that information had I intended to do further commercial speaker work myself.

                      Originally posted by billfitzmaurice View Post
                      IME the panel resonant frequency is of little or no consequence. Speaker panels are subjected to high air pressures by the excursion of the driver cones. If the panel has a combination of sufficient mass and stiffness it won't vibrate. If it's not heavy and/or stiff enough it will vibrate. If you have a panel that vibrates and you sufficiently brace it so that it doesn't vibrate the panel resonant frequency goes higher. It's not the shift in that frequency that stops the panel from vibrating, it's the increased stiffness of the panel. As far as what stops the panel from vibrating is concerned the panel resonance is anecdotal.
                      While I mostly agree with this post and your last post, I want to point out one experiment that I did. I had a pair of speakers that were build with high quality 1/2" baltic birch plywood (the cabinet grade that's the same hardwood through all the layers, denser ans stiffer than the so called "hardwood plywood" sold at "home improvement" stores). I built another pair of the same speakers out of the same type of wood, except 3/4" plywood this time (the size was slightly increased to maintain internal cabinet volume).

                      I was shocked to find that panel vibrations were a more serious problem in the enclosure with the thicker (and substantially stiffer) 3/4" walls. In hindsight, I suppose it makes sense that equivalent bracing and damping implementations would be less effective at controlling resonances in a stiffer & heavier panel?
                      "...this is not a subwoofer" - Jeff Bagby ;)

                      Comment


                      • #56
                        Originally posted by Michael P View Post
                        Thanks, I'll have to try that myself or listen to one of these speakers someday. Can you describe any tapering of the bass response that you're doing in high SPL situations? (Or maybe I just need to figure out what the Audyssey scheme is.)
                        ​This is the Audyssey curve.



                        Originally posted by DDF View Post

                        If you're game, it would be interesting if you could share your final curves, a very interesting topic. We're sort of mixing targets here a bit, so to clarify just mentioning that the ISO curves are free field quasi anechoic (no room contribution) where the Audyssey and JBL curves are in-room long-term average (as you know). For the in room curves, many folks (including Bob Katz) hinge at 1 kHz and ~-5 db at 20 kHz with a straight line.

                        Sean Olive and his guys ran a really interesting set of experiments deriving optimum in-room curves with expert listeners (grain of salt, we all like different flavours of sound) and then showed with typical but well designed speakers, what on axis is required to achieve this in their typical room. I think this may be documented on the InnerFidelity site. You're plowing new territory adding the cardiod into the mix and of course everyone's room will be different. Do you find you have to increase the bass response (when looked at quasi anechoic) with the cardiod relative to a point source, to get the same perceived in room bass balance?

                        For the amp differences, I'm not a tweak-o-file and am scientific/skeptical about things, but for example I built an N-Core NC400 and its sound is markedly better than other good amps (eg Bryston) I have, even if it measures just the same in frequency response (I measure and level match all my stuff when auditioning).
                        If I do nail it down, I might share the curves, but probably not. It requires much more listening tests in various rooms. Even then, nobody can implement it properly, and if someone simply forces an implementation it'll lead to people getting poor results. The loudness compensation feature has a bad rep in audio because almost all of the implementations are bad. In order for any chance of good results it must have sole volume control over the speaker and a calibration at the listening position to apply the right curves. The native speaker response needs to be taken into account as well. I'd be happy to talk to people over PM about it if they are really determined to do it.

                        As for the cardioid effects, this speaker is not cardioid in the bass region. I don't have nearly enough displacement for that. It reverts back to monopole as it approaches 100Hz. I'm already throwing away up to 12dB just to get cardioid to 110Hz or so.

                        Comment


                        • #57
                          Thanks for the info. I can see that trying to apply a predetermined configuration to a new speaker without the designer's attention could do more harm than good.

                          Comment


                          • #58
                            I am very intrigued by this design. I love seeing new stuff that is outside the norm. Specifically I will be very interested in your results with the thinner walled cabinet.

                            I would like to know more about the BBC research you are referring to as well. Sounds very interesting. Do you know what was the panel size they used to get the 100hz frequency? If the theory of frequency moves up as you add bracing, wouldn't the frequency move up for you as you make your box panels smaller? It was my understanding that bracing was essentially making the panel sectionally smaller.

                            For the dual opposing woofers; IMO I think you might benefit from adding bracing between the two magnets, essentially making them one unit. This would also in effect use the baskets as wall to wall bracing.

                            Thanks for sharing what looks like a very cool project.

                            Comment


                            • #59
                              Originally posted by DDF View Post
                              For the amp differences, I'm not a tweak-o-file and am scientific/skeptical about things, but for example I built an N-Core NC400 and its sound is markedly better than other good amps (eg Bryston) I have, even if it measures just the same in frequency response (I measure and level match all my stuff when auditioning).
                              I agree 100 %. i have several good amps that all sound different when listening. so do you have any ideas or thoughts as to what is going on ?
                              craigk

                              " Voicing is often the term used for band aids to cover for initial design/planning errors " - Pallas

                              Comment


                              • #60
                                Originally posted by DDF View Post
                                For the amp differences, I'm not a tweak-o-file and am scientific/skeptical about things, but for example I built an N-Core NC400 and its sound is markedly better than other good amps (eg Bryston) I have, even if it measures just the same in frequency response (I measure and level match all my stuff when auditioning).
                                To what would you attribute this?

                                dlr
                                WinPCD - Windows .NET Passive Crossover Designer

                                Dave's Speaker Pages

                                Comment

                                Working...
                                X