To check the distortion for yourself anything with a lot of saxophones and trumpets, e.g. Charles Mingus - Blues & Roots - Moanin'.
A quite less complex recording A Trace of Grace Michel GodardGavino Murgia with a single trumpet the Dynaudio still sounds like a trash can with a trapped homeless musician inside while W6 is acceptable with some stretch.
Thank you @PlanarGianca for this post. My lack of experience and lack of attention caused me a lot unnecessary grief with my W6 2313 driver. There was a clear distortion somewhere I could not figure out, spinning various iterations of parallel XOs and serial crossovers from #430. I thought that either my driver is defective beyond the return window, or I need to step away from W6 altogether. Quite a disappointment.
But after reading your post I finally guessed I am facing the woofer cone breakup.
To verify this I ditched my last XO, disconnected the tweeter completely, and started from scratch gradually targeting the cone breakup with 4th order XO.
This is the naked woofer:
I guessed that the 4.3 kHz peak is the cone breakup resonance, thanks to post by @PlanarGianca. The iterations I did:
Ignore the shape below 2k, I don't have proper parts, and this is an experimental box that I am rebuilding. I was testing with various trumpet and saxophone music, some of it I mentioned a page ago. The results:
1. The naked driver sounds in between harsh and trashcan-like, at par or worse than my Dynaudio Emit, which is a classic trashcan
2. Green line - I tried to test if all this noise around 7k is causing the trouble, I removed just it. No difference
3. Bluish line, I moved the 4.3 kHz peak from 74 to 62 Db. Definitely a progress! I concluded I need to keep moving. It moved from trashcan to regular bad, no more rattling the cover lid but very harsh, not acceptable
4. Orange/turd lines, moved down to 57 and 54 Db. Definitely much better than Dynaudio, at least I've made a huge step forward from my current speaker. And it's easy to compare, because I'm running my DIY left speaker and Dynaudio right speaker. Dialing the volume balance I can switch between trashcan and annoying at will. The theory that the Aegir amp is overloaded doesn't hold because the defects discussed are present in both drivers even at very low volume, like 60 Db. Can the amp be broken? No, many good records without trumpets and saxophones and other challenges sound good at Dyn and very good at W6
5. Marine color 51 Db. Some tracks where I heard the distortion are already fine, some few remain but it's very close to acceptable. I would love to try below 51 Db but ran out of parts. My plan is to stick with 4th order XO, cross at 2.5 kHz and have the breakup at -40 Db or so
So I beg anyone running W6 - please revisit your XO and make sure the 4.3 kHz peak is at least -25 Db or better. You will be much happier.
And if you ask if the distortion is measurable - no. This is the naked woofer distortion chart:
REW found nothing interesting in the 4.3 kHz area, but the listening test did. Maybe there are some measurement tricks that can reveal this problem? I challenge everyone to show your skills torturing W6!
And can someone do me a favor? Try the disk Saxnbass, it still sounds quite bad on both speakers, not sure if this is a recording problem, I don't have a way to easily swap the spare amp, my workspace is very tight and cluttered with 55 wires.
Last edited:
I apologize what I've posted might be incorrect because I might be experiencing amplifier clipping. Different variations of XO might have different power dissipation, which affects results. Trumpets and saxophones that trigger the clippings have a lot of harmonics at once and may require a lot more of power to reproduce. It may or may not be cone breakup, no idea at this point.
I have an old spare amp Arcam Delta 290 and I may try biamping without any extra parts. Went quite crazy but I am so pleased with W6 sound overall (when there is no trumpets) that no matter what happens it's already a substantial upgrade.
I have an old spare amp Arcam Delta 290 and I may try biamping without any extra parts. Went quite crazy but I am so pleased with W6 sound overall (when there is no trumpets) that no matter what happens it's already a substantial upgrade.
Last edited:
Still trying to narrow down the source of distortions as above. As @kgrlee suggested it could be amp clipping. Swapped Aegir to 250W purifi - great sound in general and running cold but nope, distortions are there. Swapped DIY W6 to Dynaudio - nope. Swapped volume control from autoformers to potentiometer - nope. Swapped DAC - nope. Swapped PC to Mac - nope. Swapped Tidal to Youtube - nope.
The only thing I did not swap - cables and my head. I am also thinking about building a balanced stepped attenuator, but if balanced doesn't it will be stupid.
I'll get back to trying to fix the distortions in crossover, with that much power I can run 4th order and narrow down the distortions region.
Is there a Amp / Speaker combination that does NOT have this distortion?
If so, it's worth checking again just to make sure.
It could be the source file. Can you view the file with Audacity and maybe post a picture of how it looks there?
@kgrlee thanks for following up! I am still working on it, and for now I have only preliminary results, too early to make any conclusion. I can only say that I was able to build a 4th order XO that was very enjoyable well beyond anything I had before. At any rate, my TB W6 project is a striking success even if I have to stick with 4rt order. However a 4th order has too many parts on the sound path and is shifting the phase too fast. I would do the community a disservice to recommend it, and it will be misleading to claim it is necessary. At this point I am unable to narrow down which part of the W6 woofer right slope must be suppressed and by how much. I am playing with 250W amp so any clipping is impossible. My current effort is to try every possible 2nd order lowpass on the woofer until I give up. For now, every 2nd order is really bad, I will also try to combine 2nd order with a notch filter.
The distortion cannot be measured, it requires a lot of listening tests. The easiest way to detect it is with trumpets and saxophones. It is there with piano but the keys fade too fast to concentrate. Those instruments produce multiple harmonics that must sound in concert to reproduce a proper sound. When the driver has to play multiple coordinated frequencies, it gets confused. When I cut this chain with 4th order XO and transfer part of it to the tweeter it sounds great. And don't think that the music without trumpets and saxophones is clean, no it's there, it's just my engineering approach to reproduce it with ease.
I've made a playlist specifically dedicated to reproduce the problem. Every single track here has a saxophone and a trumpet that sounds really bad every time it tries to play a low pitch, far worse than acceptable. WIth 2nd order it hits hard in the face and is easy to spot all across this playlist. Please try it for yourself. And to exclude this this is somehow a Tang Band problem, I can say Dynaudio sounds even worse, and that I already proved that it is fixable, at least with a very complex XO.
The distortion cannot be measured, it requires a lot of listening tests. The easiest way to detect it is with trumpets and saxophones. It is there with piano but the keys fade too fast to concentrate. Those instruments produce multiple harmonics that must sound in concert to reproduce a proper sound. When the driver has to play multiple coordinated frequencies, it gets confused. When I cut this chain with 4th order XO and transfer part of it to the tweeter it sounds great. And don't think that the music without trumpets and saxophones is clean, no it's there, it's just my engineering approach to reproduce it with ease.
I've made a playlist specifically dedicated to reproduce the problem. Every single track here has a saxophone and a trumpet that sounds really bad every time it tries to play a low pitch, far worse than acceptable. WIth 2nd order it hits hard in the face and is easy to spot all across this playlist. Please try it for yourself. And to exclude this this is somehow a Tang Band problem, I can say Dynaudio sounds even worse, and that I already proved that it is fixable, at least with a very complex XO.
It is strange to see so few measurements from users of these pretty coaxials.Most freeware speaker measurement SW can show spl response and distortion etc. and indoor measurements are pretty good if done properly and conditions/settings are described.
https://audioxpress.com/article/test-bench-w4-2315-home-audio-4-5-coaxial-driver-from-tb-speaker
obviously the woofer has something goin' on 800-1600Hz (resonances) and Bl(x) isn't symmetrical
Distortion around 800Hz is very high
https://audioxpress.com/article/test-bench-w4-2315-home-audio-4-5-coaxial-driver-from-tb-speaker
obviously the woofer has something goin' on 800-1600Hz (resonances) and Bl(x) isn't symmetrical
Distortion around 800Hz is very high
Last edited:
@Juhazi thanks, I still can't wrap my head around how to make this kind of measurement. Apparently I need to capture pink noise into file and somehow process it with spectrum analyzer, I never saw easy to understand instructions and never did it. If I understand how, I will try. I presume I have to repeat those steps with every variation of the low pass filter. My understanding in the link you've posted with W4 the distortion at 5.5 kHz was clearly captured in both 2nd and 3rd harmonics. I presume W4 has it somewhere in the 4 kHz +/- region.
On the flip side, with a listening test I can swap a part on my table and the effect hits me hard in the ears. It's really huge and hard to miss.
On the flip side, with a listening test I can swap a part on my table and the effect hits me hard in the ears. It's really huge and hard to miss.
RoomEQWizard is freeware and easy to use for room and nearfield measurements. REW uses log sine sweep. Noise cannot be used for distortion analysis. With REW's signal generator you can also use many different signals and watch analysis display simultaneously.
One should always measure woofer and tweeter both separately and together which is easy with parallel xo or dsp-xo
https://www.roomeqwizard.com/
W4 tweeter looks okay, distortion is highest around 5-9kHz, but only 3,5% when woofer has 15% at 800 Hz with same signal. Notice the really high fundamental spl because of nearfield measurement. My UMIK-1 has high distortion above 120dB, so I measure usually at 1m which is normal distance for spl-based scale. (Dickason fixes voltage which is even better but too difficult with REW/UMIK) Dickason at Voice Coil always measures drivers without any crossover. Coaxial tweeter needs equalization for horn effect, which dramatically changes distortion% profile.
One should always measure woofer and tweeter both separately and together which is easy with parallel xo or dsp-xo
https://www.roomeqwizard.com/
W4 tweeter looks okay, distortion is highest around 5-9kHz, but only 3,5% when woofer has 15% at 800 Hz with same signal. Notice the really high fundamental spl because of nearfield measurement. My UMIK-1 has high distortion above 120dB, so I measure usually at 1m which is normal distance for spl-based scale. (Dickason fixes voltage which is even better but too difficult with REW/UMIK) Dickason at Voice Coil always measures drivers without any crossover. Coaxial tweeter needs equalization for horn effect, which dramatically changes distortion% profile.
Last edited:
ps. I lost my interest to TB coaxial when I saw those measurements above.
KEF Q100/150 unit is worlds better https://www.erinsaudiocorner.com/driveunits/kef-q100-drive-unit/
KEF Q100/150 unit is worlds better https://www.erinsaudiocorner.com/driveunits/kef-q100-drive-unit/
@Juhazi no, REW swipe in particular, and REW in general is not designed to measure this kind of distortion. I cannot stress enough that the measurement requires playing multiple harmonic multiplier frequencies at once. For example, you play 800 Hz, 1600 Hz, 2,400 Hz and probably even 3,200 simultaneously, because the trumpet requires 4 harmonics to sound right, and expect them to form some sort of proper predefined shape, and observe it on oscilloscope. Maybe a square wave is a good approximation of this but I don't really know.
For starters, watch few videos about how a trumpet sounds at oscilloscope (search link).
Suppose I need to measure just a square wave, this post explains that I need to somehow combine REW, TrueRTA and Audacity to achieve that. However there are no step by step instructions, and the post is from 2012, maybe there is a simpler way now. People are even mentioning using 2 computers concurrently, one to generate a square wave and another to record it. And I'm not even sure the square wave will do. I apologize maybe modern REW can do that but I have no idea how, I'm a noob really.
You can imagine why I'm saying a listening test is a whole lot simpler for me. The human ear can capture and sum C1*800 + C2*1600 + C3*2400 + C4*3200 with ease and that's actually why we are even discussing the speaker building here as being kind of hard (but fun).
For starters, watch few videos about how a trumpet sounds at oscilloscope (search link).
Suppose I need to measure just a square wave, this post explains that I need to somehow combine REW, TrueRTA and Audacity to achieve that. However there are no step by step instructions, and the post is from 2012, maybe there is a simpler way now. People are even mentioning using 2 computers concurrently, one to generate a square wave and another to record it. And I'm not even sure the square wave will do. I apologize maybe modern REW can do that but I have no idea how, I'm a noob really.
You can imagine why I'm saying a listening test is a whole lot simpler for me. The human ear can capture and sum C1*800 + C2*1600 + C3*2400 + C4*3200 with ease and that's actually why we are even discussing the speaker building here as being kind of hard (but fun).
Last edited:
Have you tried REWs multitone generator? You can set frequencies as you like. No square wave. But what will be your reference? Traditional sweep and multitones should be enough.
The problem that you describe is strange, trumpet has lots of harmonics yes and it has sliding scale of notes
The problem that you describe is strange, trumpet has lots of harmonics yes and it has sliding scale of notes
@Juhazi Yes I am able to generate various strange signals with REW but that's where I stop, I don't know to to record and analyze them, I really only started speaker building 2 months ago and I am lacking the basic knowledge. No, REW sweep shows no distortion, the sweep can only play one frequency at a time, which is useless, it must be multitone complex signal or it's no starter.
The playlist from post #488 is my oscilloscope and it works really great but slow.
The playlist from post #488 is my oscilloscope and it works really great but slow.
This is how my naked woofer no XO distortion tab in the sweep shows. By my understanding REW doesn't show much distortion here and yet it sounds like a rattling trashcan full of rats if no XO is applied to cut everything beyond 3 kHz. CC @Juhazi I read this "distortion" measurement as "I'm useless".
Please read the thread I progressed with mine. They do not measure that bad. This is page 3 where in box FR/HD were taken at 1m. Or look Monoculus up here on DIY, as the same graphics are presented here in my project thread.
https://diy.midwestaudio.club/discussion/2155/project-monoculus-versabox-project-3/p3
@wolf_teeth thanks I need to stress once more that I can make W6 to sound in order of magnitude better than Dynaudio Emits but so far only with 4rh order XO. It's not the driver defect, I am pretty sure looking to chart #496 that many "full range" drivers that people just measure with REW and see nice flat response and "low" distortion in fact will sound like a trashcan if you give them to me and ask me my honest opinion. There is something that REW alone absolutely cannot capture, you either need an oscilloscope and good reference signal for every frequency or just give them a good listening test as the playlist from post #488.
@PlanarGianca described the problem in #358 and gave a crossover that I'm yet to try given a lack of exact parts, it's a strong move in the right direction.
@PlanarGianca described the problem in #358 and gave a crossover that I'm yet to try given a lack of exact parts, it's a strong move in the right direction.
Last edited:
And I understand you are using the W6. My reply was to @Juhazi as the W4 does not measure as badly as shown in another review. I used Omnimic for my measurements.
That said, i hope you sort it out, because I don't have the W6 in front of me to say or hear what it is you are hearing. Trumpets sound awesome on my Monoculus, and I played in school so I should know how they should sound.
That said, i hope you sort it out, because I don't have the W6 in front of me to say or hear what it is you are hearing. Trumpets sound awesome on my Monoculus, and I played in school so I should know how they should sound.
Amir gave a very conclusive result that KEF LS50 Meta is giving major distortions if made to play any sound below 200 Hz along with music. KEF implicitly acknowledged the problem by crossing R3 Meta at 420 Hz. They clearly signal that LS50 driver must be high-passed at 420 Hz, which makes it strictly midrange driver. I am not engaging with 3-way at desktop DIY, so I trust it's great but requires a large complex build to be played at a distance.