Keith_W Posted June 4, 2023 Author Posted June 4, 2023 Hahaha, I won't Anyway I have been having some fun with Sketchup: 2
Keith_W Posted June 18, 2023 Author Posted June 18, 2023 (edited) I have my system sounding perfect at the moment with incredible clarity that I haven't heard anywhere else. So what else is there to do but experiment and see if I can make it sound even better? Specifically, this has been bugging me: Brown = no correction with Acourate. Blue = after correction. Note that there is a dip between 3000-4000Hz (I have highlighted it) which corresponds to the horn/tweeter crossover point. Although it sounds fantastic, this dip has been bugging me. Hence the motivation to ask myself if it could sound better if I got rid of it. Note that the reason it is there is because I told Acourate not to make any corrections above 4x Schroder (i.e. 400Hz). Everything > 400Hz is the native response of the speaker. Now, of course I could EQ it out but it would be better if I rethought the crossover and examined the polarity of the drivers. I first asked myself if I should move the crossover points from where they currently are (80, 500, 4000) to get away from distortion. I normally measure the drivers with Acourate, which does not give distortion numbers. So I cracked out REW and did a round of measurements. All of these have no crossover in place (the passive xover has been bypassed) and the digital xover has been switched off. I wanted to see the native response of the driver without any corrections. These are the results: That woofer has quite impressive performance. And so does the horn. As delivered, the horn has a first order Linkwitz-Riley crossover at about 5000Hz, so it has a combination of electrical and mechanical rolloff at about 12dB per octave. This is a very old measurement of the horn when it still had the crossover in place. The tweeter is showing unexpectedly high distortion. It has been a number of years since I replaced the electrodes and the PL519 valve in the tweeter and it is currently making a lot of noise (intermittent whistling sounds) so every now and then I have to get up and reset the tweeter and the noise goes away. Edited June 18, 2023 by Keith_W 2
Keith_W Posted June 18, 2023 Author Posted June 18, 2023 So now I have the data, I pondered where the crossover points should be. It looks as if "getting away from distortion" will not be a reason to redesign the crossover. Re, the first crossover point (80Hz). That woofer is a superb and I can cross it over anywhere I like. I am thinking of moving it down to 60Hz, but for now I don't want to make too many changes. So I will keep it at Neville-Thiele 2nd order. As for the second crossover point (500Hz), current measurements show there is no problem with the crossover at the region (see the first graph in the above post). I did not want to make too many changes at once, so it stays at Neville Thiele 2nd order, 500Hz. The third crossover point is more interesting. It looks as if the dip in the response between 3-4kHz is due to the rapid fall in response of the tweeter below 4kHz. So I decided to move the crossover point to 4500Hz. As delivered, the horn is low passed at 1st order LR at 5kHz, and the tweeter has no crossover. I wasn't game to do that, so I decided on a Linkwitz-Riley 4th order at 4500Hz. This is the new crossover that I generated. Note that the sub and the woofer have driver corrections applied. This is before overall room correction. Acourate allows you to mix and match different crossover types - in this case, the horn has NT2 at the bottom and LR4 at the top. I then did an overall room correction and this is the result: The dip is still there, but it has moved from 3-4kHz to 2-3kHz. So I decided to invert the polarity of the tweeter. This is the result. The dip is gone. I also decided to turn the subs down (but I did not repeat a sweep to show it). The step response shows that I need to do a better job of time aligning the left subwoofer (red). The right sub (green) is perfect). So what does it sound like? Hmmmmmm, I am undecided. The older filter sounded absolutely crystal clear. The current one, not so much. It sounds a bit smoother but some of the clarity is gone. I will have to do more listening to decide if I like it more or not. 1
Keith_W Posted June 28, 2023 Author Posted June 28, 2023 A number of people have visited me now, and I have secretly been deploying a distortion adding VST plugin. I don't tell them what I am doing, I simply ask if they prefer A or B (single blind test). Without exception, everybody has voted for the added distortion as sounding better. About 10 people so far (names not mentioned!). Their reactions range from disbelief to disgust when I tell them what I am doing, but hey - they voted for the distortion as sounding better! The idea of adding distortion to improve the sound came to me when I realized from multiple audio discussions that people like the sound of vinyl. I do, too ... despite knowing that its technical performance is inferior to digital. Therefore, there is something that vinyl is adding that is aesthetically and sonically pleasing. It can be summed into six types of distortion: harmonic distortion, altered frequency response, increased crosstalk, reduced dynamic range, increased noise floor, and speed stability. In addition, some vinyl systems I have heard sound noticeably more dynamic than digital. My theory is that it is some type of distortion which is apparent only with transients which causes the signal to overshoot (I may be quite wrong on this, because I have never seen any authority say so, nor seen any published waveforms to show this is the case). There are some that can be dismissed out of hand because they are unlikely to improve the subjective appreciation of sound - i.e. reduced dynamic range, increased noise floor, and speed stability. This leaves three types of distortion which I might be able to simulate by using a VST plugin: harmonic distortion, altered frequency response (I can choose a different target curve), and increased crosstalk (Acourate Convolver is able to do this). The result of my single blind test administered to about 10 fellow SNA'ers is quite interesting, because 10/10 (maybe more, I am not keeping count) preferring the sound of added harmonic distortion in a single blind test is statistically significant. This is consistent whether I apply overall distortion (with PKHarmonic) or targeted distortion (with Thrillseeker XTC). In particular, Thrillseeker allows targeted distortion, so I can choose to only add upper frequency harmonic distortion (which produces the illusion of clarity) or lower harmonics only (which makes the sound seem fuller). Of course, I adjust the added distortion to taste, because overdoing it will result in screechy or bloated sound. Both outcomes are certainly possible. Conclusion: adding harmonic distortion to my system is subjectively more appealing to people in a single blind test. 3 1 1
Satanica Posted June 28, 2023 Posted June 28, 2023 33 minutes ago, Keith_W said: A number of people have visited me now, and I have secretly been deploying a distortion adding VST plugin. I don't tell them what I am doing, I simply ask if they prefer A or B (single blind test). Without exception, everybody has voted for the added distortion as sounding better. About 10 people so far (names not mentioned!). Their reactions range from disbelief to disgust when I tell them what I am doing, but hey - they voted for the distortion as sounding better! The idea of adding distortion to improve the sound came to me when I realized from multiple audio discussions that people like the sound of vinyl. I do, too ... despite knowing that its technical performance is inferior to digital. Therefore, there is something that vinyl is adding that is aesthetically and sonically pleasing. It can be summed into six types of distortion: harmonic distortion, altered frequency response, increased crosstalk, reduced dynamic range, increased noise floor, and speed stability. In addition, some vinyl systems I have heard sound noticeably more dynamic than digital. My theory is that it is some type of distortion which is apparent only with transients which causes the signal to overshoot (I may be quite wrong on this, because I have never seen any authority say so, nor seen any published waveforms to show this is the case). There are some that can be dismissed out of hand because they are unlikely to improve the subjective appreciation of sound - i.e. reduced dynamic range, increased noise floor, and speed stability. This leaves three types of distortion which I might be able to simulate by using a VST plugin: harmonic distortion, altered frequency response (I can choose a different target curve), and increased crosstalk (Acourate Convolver is able to do this). The result of my single blind test administered to about 10 fellow SNA'ers is quite interesting, because 10/10 (maybe more, I am not keeping count) preferring the sound of added harmonic distortion in a single blind test is statistically significant. This is consistent whether I apply overall distortion (with PKHarmonic) or targeted distortion (with Thrillseeker XTC). In particular, Thrillseeker allows targeted distortion, so I can choose to only add upper frequency harmonic distortion (which produces the illusion of clarity) or lower harmonics only (which makes the sound seem fuller). Of course, I adjust the added distortion to taste, because overdoing it will result in screechy or bloated sound. Both outcomes are certainly possible. Conclusion: adding harmonic distortion to my system is subjectively more appealing to people in a single blind test. Hi, did you randomise the play order of the two options for your preference tests?
Andythiing Posted June 28, 2023 Posted June 28, 2023 Did you use a randomiser to select the SNA visitors 1
Keith_W Posted June 28, 2023 Author Posted June 28, 2023 (edited) 54 minutes ago, Satanica said: Hi, did you randomise the play order of the two options for your preference tests? Yes. When people come over, I tell them they will be comparing various DSP settings. I have compared different crossover points and slopes, different correction strategies, uBACCH or no uBACCH, and various VST's. I never tell people what I am doing, all I ask is simply "do you prefer A or B". It's an honest question, because to me it's a matter of preference and I want to gauge whether my opinion is in line with what other SNA'ers think (I have said that better measuring is not the same as better sounding!). The effect that has the most consistent response is distortion, it seems as if every visitor has unanimously voted for a small amount of added distortion as the preferred sound. This is in line with my own preference, and I find it quite an interesting phenomenon. 31 minutes ago, Andythiing said: Did you use a randomiser to select the SNA visitors No Even total randoms like you are welcome! Edited June 28, 2023 by Keith_W 1 1
Keith_W Posted July 16, 2023 Author Posted July 16, 2023 I made a new video describing how the system works. 4 1 2
davewantsmoore Posted July 16, 2023 Posted July 16, 2023 On 24/5/2023 at 4:34 PM, frednork said: The blocks are self - explanatory, except COLOUR_STREAM ... and also why is "specific effects caused by the RIAA emphasis" in the non-linear block?
frednork Posted July 16, 2023 Posted July 16, 2023 13 minutes ago, davewantsmoore said: ... and also why is "specific effects caused by the RIAA emphasis" in the non-linear block? No idea but implies something happening at the cutting head, no ?
davewantsmoore Posted July 16, 2023 Posted July 16, 2023 (edited) On 19/6/2023 at 12:25 AM, Keith_W said: The step response shows that I need to do a better job of time aligning the left subwoofer (red). The right sub (green) is perfect). I assume this is the response of the whole speaker. Why do you say it is the subwoofer which is out of alignment? 27 minutes ago, frednork said: No idea but implies something happening at the cutting head, no ? RIAA is a pre-distorted (pre-emphasis) frequency response, and the RIAA filter in a phono "un-does" the distorted frequency response. <shrug> Edited July 16, 2023 by davewantsmoore
Keith_W Posted July 16, 2023 Author Posted July 16, 2023 8 minutes ago, davewantsmoore said: I assume this is the response of the whole speaker. Why do you say it is the subwoofer which is out of alignment? Because the tweeter and horn impulse is aligned, only the great big impulse (from the subs) is not.
frednork Posted July 16, 2023 Posted July 16, 2023 16 minutes ago, davewantsmoore said: RIAA is a pre-distorted (pre-emphasis) frequency response, and the RIAA filter in a phono "un-does" the distorted frequency response. <shrug> Yep sure, but the fact they put it in the non linear box implies something not wanted happening, Overdriven riaa preemphasis circuit, some other unwanted effect at the cutting head due to riaa, just guessing. Maybe the graphics guy just thought it looked better in the other box ? dunno? With Weiss you would have to take it a bit more seriously than a typical hifi manufacturer. Maybe they discovered something?
davewantsmoore Posted July 16, 2023 Posted July 16, 2023 1 hour ago, Keith_W said: Because the tweeter and horn impulse is aligned, only the great big impulse (from the subs) is not. The step response chart you posted shows the HF not aligned. This is, assuming that the step response you showed was for the whole speaker (which I think it is, but). The "big impulse" is not from the sub.... it is the highest frequencies. In a step response chat red is the HF green is the MF blue is the LF
parrasaw Posted July 16, 2023 Posted July 16, 2023 (edited) Bach cello Suites. Beautiful music played on a wonderful system. Having just read @Andythiing's comment, I too didn't follow the technical intricacies of your system, but I too appreciated what you have put together and the sound which is produced, so thank you for posting the video, and for the description and analysis of your system. Edited July 16, 2023 by parrasaw
Andythiing Posted July 16, 2023 Posted July 16, 2023 14 hours ago, Keith_W said: I made a new video describing how the system works. Keith Thankyou for making this - you lost me at hello in terms of technical intricacies of your system but even listening over my headphones left me in no doubt that this is a truly special system - it’s so great to see what you have achieved after so much effort and discovery 1
Keith_W Posted July 17, 2023 Author Posted July 17, 2023 Thanks for your kind remarks guys, but you really can't judge sound quality from a Youtube video I took the video with my phone and did a quick edit and posted it. I listened to it with my headphones as well, and that video is severely lacking in bass and sounds too top endy. There is a lot of richness in the actual sound which is missing in that video. Actually, I do have a professional calibrated microphone, only it is a mono mic and not stereo. I could set that up and make a mono recording and it will be pretty accurate tonally, but that is before it gets mangled by Youtube's compression algorithm.
parrasaw Posted July 17, 2023 Posted July 17, 2023 10 minutes ago, Keith_W said: Thanks for your kind remarks guys, but you really can't judge sound quality from a Youtube video I took the video with my phone and did a quick edit and posted it. Agreed. What was posted however, was enough to know that your system is something special. Some time ago I Googled Acapella speakers, and the reviews which are available for them, so perhaps I was predisposed to enjoying whatever you posted - that might cut me out from being a test subject for one of your "A/B" comparisons?? Seriously though, I have been following your system thread for a while now, and once you posted one of Bach's great pieces for the cello, I couldn't help posting an appreciative comment. Looking forward to hearing your proposed next attempt to capture the sound of the system.
tripitaka Posted July 17, 2023 Posted July 17, 2023 21 hours ago, davewantsmoore said: The step response chart you posted shows the HF not aligned. This is, assuming that the step response you showed was for the whole speaker (which I think it is, but). The "big impulse" is not from the sub.... it is the highest frequencies. In a step response chat red is the HF green is the MF blue is the LF Regardless of who is right, this is the value of this forum. Even the most esoteric of topics will bring out someone with genuine knowledge, allowing expert discussion and/or clarification. I assume @davewantsmoore is correct?
Keith_W Posted July 17, 2023 Author Posted July 17, 2023 (edited) 1 hour ago, tripitaka said: Regardless of who is right, this is the value of this forum. Even the most esoteric of topics will bring out someone with genuine knowledge, allowing expert discussion and/or clarification. I assume @davewantsmoore is correct? He isn't. I can turn on and off individual speaker drivers and see what effect is has on the step response. I assure you, it is the subwoofer which is out of alignment. For example, when I was doing the time alignment, I left the microphone in place and did individual sweeps of the tweeter, horn, woofers, and subs. This is the tweeter compared to the horn (tweeter is the small green wiggly line): And here is another step response from when I got the time alignment completely wrong: Time alignment in Acourate is a manual and reiterative process, and error prone if you are not careful. It's not like Audiolense which automagically pops out a number after a single sweep. When I saw that, I was insanely jealous. It took him 2 minutes to get the time alignment. It takes me about an hour. Edited July 17, 2023 by Keith_W 1
frednork Posted July 17, 2023 Posted July 17, 2023 On 16/07/2023 at 6:50 PM, Keith_W said: Because the tweeter and horn impulse is aligned, only the great big impulse (from the subs) is not. 22 hours ago, davewantsmoore said: The step response chart you posted shows the HF not aligned. This is, assuming that the step response you showed was for the whole speaker (which I think it is, but). The "big impulse" is not from the sub.... it is the highest frequencies. In a step response chat red is the HF green is the MF blue is the LF My guess is that everything after the tweeter in the left speaker is delayed for some reason. The tweeter is inverted phase and both sides look close. Then there is a gap (circled in red) on the left side. Get rid of the gap both aligned , all good.
davewantsmoore Posted July 18, 2023 Posted July 18, 2023 (edited) On 17/7/2023 at 7:13 PM, Keith_W said: He isn't. I am correct (in regards to the shape, and frequency composition of the "step response") Having a more detailed look.... Fred has noticed something. However, these is something else not right with these data.... the HF component of the pulse should be the tallest. It should look like the chart I posted. That being said.... there isn't really enough clear information for me to conclude that subwoofer isn't out of alignment..... only that the tweeter and MF look very wrong, and it may be that neither of them are out of alignment either. If the step was arrived at via convolution, and the sizes of the pulses used were mis-sized, or..... (it is very difficult to say). Should look like this: Fred might be correct... that, in combination with the tweeter pulse not correctly sized..... That could make up for it <shrug> Edited July 18, 2023 by davewantsmoore
Recommended Posts