Thanks... I appreciate that. One of my goals is to explain why I do things in a certain order, and how I make the important decisions. This may help others who are getting started in speaker design. I certainly have benefitted from the knowledge that others have shared. It also helps me to have so many knowledgeable people reviewing my work, giving me ideas, pointing out errors and opportunities to do things a different way. I have always believed that progress is a team sport... meaning that we can all help each other advance our skills and knowledge by coordinating and working together. Its not a contest to see who can be the smartest guy in the room... we all know that kind of guy... "Boring (yawn), please annoy someone else with your lectures and tales of greatness"
=== ===
The SB29SDAC might make a good tweeter for a 2.5 way with twin 8" woofers. It can be crossed low, and it has a nice 93 dB sensitivity. This kind of speaker will never be a highly revealing highly accurate transducer, but it can be fun, and loud, and rock'n. A possible future project.
Yeah nice, this is only way to actually make good sound, prototype and wonder about stuff, listen and learn from the process. I mean, having two options and being able to pick the better one is 100% success rate, compared to randomly choosing something without ever making a comparison is game of luck. Even if there was only one set of drivers, comparing how the construct affects thing makes the end result better. Technically better by measurement, but also perceptually, there is high chance perception doesn't quite correlate with measurements which piques interest and now listening skills get better, ability to connect graphs into sound, learn how room affects things and so on. Guaranteed progress with prototyping.
As a general statement, I don't follow, because that correlation is actually pretty high.
Unless you mean subjective taste in a case when someone is unaware. Than, yes
.
I just reflected personal experience that sometimes it seems perception doesn't correlate with data for what ever reason, which then leads to some kind of investigation and interesting observations.
For example, my DSP is otherside of room than the speakers and sitting by the DSP I can use a trimmer to change delay of my tweeters and hear no difference until change in frequency response is noticeable, this is many milliseconds. This is contrary to lot of projects optimized for "time alignment" or what the data seems to indicate in a simulator.
Observation is, time alignment accuracy is not critical unless one listens close up. So, correlation of measured data, perception and words on the internet depended on listening distance in this case.
And point is, had I no DIY speakers or no DSP to tweak and keen interest to listen things like this I could have never noticed such thing. Now that I have I can put this particular aspect of speaker design in perspective because I know how does this particular aspect sound like and in which context, how to listen to it, how it relates to room and position of things and how it relates to my project at hand.
If someone else wrote this post and I was reading it without personal experience on it, I would not have understood it because I couldn't have put information like this in a context. I do not know if you others can, I'm on my own with my own perception and system and learn from that, by prototyping and wondering about things, listening and learning how to relate visual concepts like text and graphs into audio perception in my setup and context.
For example, my DSP is otherside of room than the speakers and sitting by the DSP I can use a trimmer to change delay of my tweeters and hear no difference until change in frequency response is noticeable, this is many milliseconds. This is contrary to lot of projects optimized for "time alignment" or what the data seems to indicate in a simulator.
Observation is, time alignment accuracy is not critical unless one listens close up. So, correlation of measured data, perception and words on the internet depended on listening distance in this case.
And point is, had I no DIY speakers or no DSP to tweak and keen interest to listen things like this I could have never noticed such thing. Now that I have I can put this particular aspect of speaker design in perspective because I know how does this particular aspect sound like and in which context, how to listen to it, how it relates to room and position of things and how it relates to my project at hand.
If someone else wrote this post and I was reading it without personal experience on it, I would not have understood it because I couldn't have put information like this in a context. I do not know if you others can, I'm on my own with my own perception and system and learn from that, by prototyping and wondering about things, listening and learning how to relate visual concepts like text and graphs into audio perception in my setup and context.
Last edited:
Yes, but not in contrary with what most literary concludes.
Although even a more practical argument can be brought up.
The vast majority of speakers with passive filters don't have time alignment, yet they sound wonderful.
BTW, I am sorry, but I am not gonna contribute into a time alignment discussion anymore. That flogged horse is less than a pile of bones at this point.
Yeah, that was just an example. Another would be directivity, accuracy of which is related to listening distance and what you like to hear, how much you want to optimize stereo image, direct sound and envelopment taking consideration of room acoustics and if speaker position is limited or not. Or which graph is more relevant PIR or axial response in CTA2034 set, which one to optimize if you have to choose due to less than ideal DI. Listen too far and any of this is much more loose, listen close up and suddenly things matter more.
In general, I don't think people mind listening distance too much, other than often noticing comments "as long as it's far enough", but for me listening distance makes night and day difference in my space with my speakers. Considering listening distance removes confusion I've got from various advice (online) and should be always included in any loudspeaker discussion where perception is involved. It is important to know whether one is listening with small enough listening triangle so that effects of early reflections are largely suppressed in perception, or not, in order to relate my context to theirs.
I agree with you that crossover time alignment is not audible, at least not when listening in typical hifi listening situation "too far" so that early reflections mess up phase information anyway.
ps. I'm lumping various effects to listening distance here, which affects many things, but is the common denominator and can be adjusted easily in realtime, so a practical way to change and listen effects of early reflections.
In general, I don't think people mind listening distance too much, other than often noticing comments "as long as it's far enough", but for me listening distance makes night and day difference in my space with my speakers. Considering listening distance removes confusion I've got from various advice (online) and should be always included in any loudspeaker discussion where perception is involved. It is important to know whether one is listening with small enough listening triangle so that effects of early reflections are largely suppressed in perception, or not, in order to relate my context to theirs.
I agree with you that crossover time alignment is not audible, at least not when listening in typical hifi listening situation "too far" so that early reflections mess up phase information anyway.
ps. I'm lumping various effects to listening distance here, which affects many things, but is the common denominator and can be adjusted easily in realtime, so a practical way to change and listen effects of early reflections.
Last edited:
Context is paramount.
My general approach is just very pragmatic, if I can fix it without to much effort, I will fix it.
If it means I need to go trough all kinds of hoops, I will pay less attention to it.
The point is, these kind of things are just not as dramatic of a deal breaker as some people make it seem.
If that would be the case, the vast majority of speaker systems would sound terrible, which they don't.
If you watch Eric's reviews, I have also never seen him complaining about it.
I wouldn't flag it as "not audible", but on the priority list of things, not very high up.
My general approach is just very pragmatic, if I can fix it without to much effort, I will fix it.
If it means I need to go trough all kinds of hoops, I will pay less attention to it.
The point is, these kind of things are just not as dramatic of a deal breaker as some people make it seem.
If that would be the case, the vast majority of speaker systems would sound terrible, which they don't.
If you watch Eric's reviews, I have also never seen him complaining about it.
Yeah, it is weird stuff perceptually and emotionally as listening to music is much fun with any equipment, from poor to great and in general the music is there and unless it's painful the quality doesn't matter that much. This would be important to remember, that what matters to someone might not for someone else Understanding perspective / context of others, and our own, is crucial to be able to consider some advice/concept relevant or not.
And to get back to Hifijim thread, if one is interested about what makes good sound then DIY loudspeaker building and prototyping is great way to do it as the theory and practice meet the perception, suddenly there is audible perspective on things and one can navigate towards a goal in mind by comparing to some other perspective, like changing tweeter delay, or comparing two different baffles on same tweeter or two tweeters on same baffle. Learning in general, experience, curiosity. Fun stuff
Understanding perspective / context of others, and our own, is crucial to be able to consider some advice/concept relevant or not.And to get back to Hifijim thread, if one is interested about what makes good sound then DIY loudspeaker building and prototyping is great way to do it as the theory and practice meet the perception, suddenly there is audible perspective on things and one can navigate towards a goal in mind by comparing to some other perspective, like changing tweeter delay, or comparing two different baffles on same tweeter or two tweeters on same baffle. Learning in general, experience, curiosity. Fun stuff
Update to post 316 https://www.diyaudio.com/community/threads/compact-low-cost-active-3-way-speaker.402812/post-7472906
I now realize that I should have used a smoothing option when I plotted the harmonic distortion graphs. This smooths the fundamental at 1/3 octave smoothing, and then calculates the distortion percentage based on that. This is a more correct way to express the distortion as a percentage.
Originally, I thought the SB29SDAC had a small advantage in terms of distortion performance. However, when the data is processed properly, the two tweeters essentially become equal. Neither has an overall advantage, and both are very good performers for the price.
I also show the SB15NBAC mid driver. I found excellent distortion performance from 500 Hz up. I expect the upper crossover to be somewhere between 1.8k and 3k, and in this range, the performance is first rate.
I do not trust my measurements below 500 Hz. I noticed several cabinet resonances at low frequencies, basically the whole range from 100 - 500 Hz. The foam box buzzed and moaned with either a swept sine or a stepped sine signal. At 450 Hz the resonance was strong enough that it shows up as a +3 dB hump in the power curve. When I use the polar response data for VituixCad simulations, I will have to ignore the power curve hump at 450 Hz.
It is nice to be able to do real distortion evaluations. My old USB interface (Behringer UMC202HD) had a THD+N of about 0.5%, which was good enough for making frequency response measurements, and I could certainly measure distortion down to 0.5% and declare a driver to be "good enough"... but now I have a Motu M2 and the THD+N is orders of magnitude better than my old one.
j.
I now realize that I should have used a smoothing option when I plotted the harmonic distortion graphs. This smooths the fundamental at 1/3 octave smoothing, and then calculates the distortion percentage based on that. This is a more correct way to express the distortion as a percentage.
Originally, I thought the SB29SDAC had a small advantage in terms of distortion performance. However, when the data is processed properly, the two tweeters essentially become equal. Neither has an overall advantage, and both are very good performers for the price.
I also show the SB15NBAC mid driver. I found excellent distortion performance from 500 Hz up. I expect the upper crossover to be somewhere between 1.8k and 3k, and in this range, the performance is first rate.
I do not trust my measurements below 500 Hz. I noticed several cabinet resonances at low frequencies, basically the whole range from 100 - 500 Hz. The foam box buzzed and moaned with either a swept sine or a stepped sine signal. At 450 Hz the resonance was strong enough that it shows up as a +3 dB hump in the power curve. When I use the polar response data for VituixCad simulations, I will have to ignore the power curve hump at 450 Hz.
It is nice to be able to do real distortion evaluations. My old USB interface (Behringer UMC202HD) had a THD+N of about 0.5%, which was good enough for making frequency response measurements, and I could certainly measure distortion down to 0.5% and declare a driver to be "good enough"... but now I have a Motu M2 and the THD+N is orders of magnitude better than my old one.
j.
Best advice I can give, is to just check several 3rd party measurements as well and see if they correlate with your own measurements.
If you're down to 0.5% THD+N, there isn't much to worry about anymore in general, especially when it comes down to tweeters.
The masking effect is so freaking high at those higher frequencies.
At this day and age, I mostly find it pixel peeping with very little substantial objective data being shown.
It's getting maybe old, but look at the distortion of the D&D 8C and the general opinion about those speakers.
All that being said, as a measuring tool itself, it's nice to have an additional decimal mostly just to see what is going on and be sure about it! (aka confidence)
What I find FAR more important, is at what lower frequency the distortion goes up (important for crossover) and if there are any nasty left overs from resonances in the dome or other irregularities.
If you're down to 0.5% THD+N, there isn't much to worry about anymore in general, especially when it comes down to tweeters.
The masking effect is so freaking high at those higher frequencies.
At this day and age, I mostly find it pixel peeping with very little substantial objective data being shown.
It's getting maybe old, but look at the distortion of the D&D 8C and the general opinion about those speakers.
All that being said, as a measuring tool itself, it's nice to have an additional decimal mostly just to see what is going on and be sure about it! (aka confidence)
What I find FAR more important, is at what lower frequency the distortion goes up (important for crossover) and if there are any nasty left overs from resonances in the dome or other irregularities.
Yeah, with tweeters, the distortion I am most focused on is 1k - 5k... the 1k-2k region to assess how low the crossover can be, and 2k - 5k because that distortion is more likely to be perceptible. At 6k fundamental, H2 is at 12k and H3 is at 18k... I don't see that as very important, as you said.
@tmuikku , is there also a relationship to which part of the frequency band? Our hearing is in particular very well developed for the human voice range. So a timing variation at say 1200 Hz will probably have a different treshold than say 6000 Hz.
This is a strength of STEPS.
When displaying distortion REW displays the fundamental with a smoothing level of 1/24th octave. It is fixed at smoothing interval and cannot be changed, AFAIK.
One really needs click to various viewing options eg. Hx and fundamental as absolute SPL and analyse it further, like how Jim has kindly demonstrated, to see why distortion percentage might appear to have ripples.
@JohnPM
In the Distortion graph; Would it be possible to have the frequency response smoothed at 1/12 octaves or 1/6 or 1/3 octaves? I find that 1/24 is great for midrange/treble frequencies, but below the Schröeder frequency it does cause fluctuations is apparent bass distortion
When displaying distortion REW displays the fundamental with a smoothing level of 1/24th octave. It is fixed at smoothing interval and cannot be changed, AFAIK.
One really needs click to various viewing options eg. Hx and fundamental as absolute SPL and analyse it further, like how Jim has kindly demonstrated, to see why distortion percentage might appear to have ripples.
@JohnPM
In the Distortion graph; Would it be possible to have the frequency response smoothed at 1/12 octaves or 1/6 or 1/3 octaves? I find that 1/24 is great for midrange/treble frequencies, but below the Schröeder frequency it does cause fluctuations is apparent bass distortion
I have STEPS but haven't used it yet. Did you just read the manual, find a useful resource on the Web, or just figure it out yourself?
I followed that manual. I had to adjust the "audio device setup" control panel since it seems to require a different set of values than ARTA. I also found that with my laptop and my motu audio interface, STEPS requires a 48k sampling rate, while in Arta I can use a 96k sampling rate. When using a 96k sampling rate, STEPs would give me a buffer overflow error and crash Windows. The overall "look and feel" is very similar to Arta, and it is obvious they share a common heritage. I am actually a little surprised I got it to work without much problem... Normally I am barely literate when it comes to PC based audio.
Regarding the smoothing... I was looking for a way to smooth the %distortion plots, and I did not find a way to do it... but I stumbled on an option to smooth the recorded signal at 1/1, 1/2, or 1/3 octave. I tried this, and got the result I wanted... and then thinking about it, it all made sense.
Regarding the smoothing... I was looking for a way to smooth the %distortion plots, and I did not find a way to do it... but I stumbled on an option to smooth the recorded signal at 1/1, 1/2, or 1/3 octave. I tried this, and got the result I wanted... and then thinking about it, it all made sense.
I have finally read through the whole chain and I've got a couple of questions. @hifijim did you consider other shapes, like a trapezoid style of the Troels' SBAcoustics SBA-741? I've been highly intrigued by this design, it looks cool in pics and it's even got bevels!