Distortion measurements on loudspeakers _ do they tell anything of value?

[CharlieLaub]

That sort of distortion heuristic should be applied to any and all drivers, not just tweeters...

 

[Juhazi]

Again same driver in a proto 2014. No eq or xo applied. The proto box was large sealed with polyfill stuffing. Differences are in details, general behaviour is similar. This spl is again too low to search for maximum spl, just a sanity check when doing directivity measurements. With REW it is easy to do several succesive sweeps and let the program to calculate means, which reduces noise artefacts remarkably.

Like Charlie said, odd order distortions should be lower than even, and no peaks! Here 3rd exceeds 2nd 1500-3000Hz. Most likely attachment to the proto baffle was not tight enough.

 

[ginetto61]

I have designed speakers to have accurate Square Wave response but this did not result in better sound.

Hi ! thanks a lot and this is unfortunate Another test that cannot be correlated to sound quality Personally i had hope for SW being useful after reading some advertising using a good response to SW as a selling point for speakers
I remember Redheko speakers for instance
But also Quad has a good square wave response

https://soundodyssey.be/en/audio/speakers/quad-esl-2912/

this is very bad because SW test is extremely easy to carry out

Yes. But this does not mean THD tests are not helpful. It's just not a top priority.

Good to know I wonder about priorities now Ok FR is clear than maybe CSD dispersion on H e V directions
As i asked above i would like to check the dynamic abilities/limits of a driver first and a speaker later
because drivers selection comes first

The most important 'measurement' is a Double Blind Listening Test. This is VERY expensive & difficult to do properly so we only test 'good' speakers; ie ones which don't sound poorly.

very interesting and thank you again Some magazines in the past used to carry out shotouts among speakers of a similar price
The results were group test winners and best buys products

They were all good speakers but some which had poor THD, did better than other (good) speakers which had very low THD.

this is a very bad news for me i wonder how a speaker with a poor THD can be accurate And i love accuracy
for instance i have heard a preamp that has a great rendition of details At first i was shocked and hit by the extremely clean sound
Then i listened detalis like never before I was hearing more of the recording Like to see a particular with a magnifying glass Impressive
there is no going back if too sharp i will try to round it in some way

This does not always follow. My design was only the second best selling speaker in Europe for at least seven years. The best selling speaker (which we also made) had better THD, more expensive juju bla bla but wasn't even in the top three small speakers according to our R&D Dept.
My DBLT panel had some of the best ears in the business including some prominent Recording Engineers like "Bat Ears" ********, another who designed his own microphones and insisted on playing his own recordings ... but one of the surprising things (there are many surprises) was, in properly conducted DBLTs, these true golden pinnae, with their own recordings, 'liked' the same speakers as the teenage girl playing Top Ten hits.

very interesting indeed I understand that the last word will be dictated by the listening experience
maybe i should respect listening tests much more In the end the goal of an audio system is to provide music not testing signals

---

Every type of distortion has a Test Signal which is most sensitive for audibility tests. But is often not the most important criteria for better sound/accuracy on music.
If you want low AUDIBLE THD, the best for this is the QUAD ESL 63 for LF and at medium/low levels. It has other good features but is flawed as the mid & HF isn't that good. It's OK in DBLTs but not a top scorer because of this. My ultimate 'conventional' speaker would have 2 x QUAD ESL 63s for LF but moving coil mid & HF 😊

thanks a lot I know the ESL63 because a friend of mine has a pair I was struck but their rendition of sonic detail
this was extremely clear watching a dvd with some special effects I heard a knock on a glass that was real not only realistic I was amazed
like a very good response to impulse ?

It turns out mid accuracy is by far the most important factor for good DBLT score and this holds true even for Heavy Metal fans.

very important conclusion again and thank you I think i will end up selecting tracks good to test the midband accuracy
also because the midband is very crowded by different instruments
i will ask about advice for best music tracks to test the midrange

 

[ginetto61]

https://www.klippel.de/know-how/literature/papers.html

https://www.klippel.de/fileadmin/kl...Diagnosis_and_remedy_of_Nonlinearities_00.pdf

aView attachment 1405390

Hi thanks a lot but this is way above my understanding capability It is too difficult
probably the best solution could be trust the best designers or copying their creations
I also read that some epic designs are not fully optimized usually in the bass reflex design or the xover parts quality and values
This is what i would like to avoid i.e. to use a driver outside its optimal working range
But to do this i must individuate first this blessed optimum working range with datasheets or measurements

 

[ginetto61]

^ do you remember the watt puppy sounding good anywhere in the room, as you were sitting and standing, or just at the sweetspot?

please let me elaborate I was standing on the right side almost in line with the right speaker At a certain point i heard a triangle that was sounding absolutely real from a point above the speaker and behind it Completely detached from the speakers The sound image was extremely sharp
It was a strange track like a mix of simple sound I remember the triangle realism and sound image
But what amazed me was the sensation of sounds completely out of the speakers almost like if the speakers were disconnected from the amps
Wonderful And i remember also a very powerful but controlled sound

If everywhere, the slanting has nothing to do with the sound quality, because any alignment it provides would matter only to a single plane 😉 Perhaps the speakers were positioned skillfully in the room, which made all the difference, and sound had not much to do with any single technical detail of the speaker, but just a nice speaker positioned well, all things combined together.

thanks a lot i understand I said that i do not like slanted baffles also because are more difficult to make
i like using felts around mid and tweeter to reduce diffraction but this will come much later
the fine tuning of a speakers can take months

 

[ginetto61]

This was already all discussed few month ago 🙂 long thread

K

Thread 'Do measurements of drivers really matter for sound?'

2024-02-27 9:40 pm

I have been studying a lot of driver measurements from different sites, and what I can tell is: cheap drivers measures poor, midrange driver can measure excellent or poor, hi0end drivers usually measures great, but can bad too.
My main interest were midrange drivers 3-5 inches, both cone and dome. As a result I saw that:
- Scan-Speak 12MU, one of arguably best midrange on this planet, measures just OK
- SB Acoustics NBAC15 - kills the 12MU in terms of harmonic distortions, especially third, costs 20% of the former
- Accuton - despite the cost - very high distortion on 3rd harmonic (few...

• kapelli
• Replies: 621
• Forum: Multi-Way

• 

I have been studying a lot of driver measurements from different sites, and what I can tell is: cheap drivers measures poor, midrange driver can measure excellent or poor, hi0end drivers usually measures great, but can bad too.

hi very interesting i have to study the thread for sure However i wonder if there is a consensus about what means measuring poor or great
Some measurements are out of discussion They are fundamental like FR and CSD and dispersion
Are these ones the measurements taken into considerations ? if so this means that they do not correlate with sound
there could be other more telling measurements or none at all
if none at all nothing can replace a proper listening test A very important conclusion anyway
Then the discussion could move to the best music tracks to use

My main interest were midrange drivers 3-5 inches, both cone and dome. As a result I saw that:

• Scan-Speak 12MU, one of arguably best midrange on this planet, measures just OK
• SB Acoustics NBAC15 - kills the 12MU in terms of harmonic distortions, especially third, costs 20% of the former
• Accuton - despite the cost - very high distortion on 3rd harmonic (few drivers checked, 90-170mm)

i still think that some efforts should be done to set up an instrumental testing procedure
in particular to establish the best range of work for a driver This is very important
a driver could be very good only within a limited range of frequencies Just saying

 

[ginetto61]

Hi, yeah it is bit daunting since there is lots of details available but if big picture is missing it's hard to know what is relevant and what is not so I feel your pain. Or felt few years ago 😀 sometimes still.

yes this is my situation The need of fixing some general concepts

To check out if 4FE35 would work for 100db / 1m system and for what bandwidth, I'd do this:
Load the driver into VituixCAD enclosure tool.

Good ! so the SW is needed I will download it soon and study it carefully I do not even know what loading the driver means
i will look for a tutorial

I did nothing else here than increase the bandwidth shown on graphs from 500 > 1000 to see bit better, this is the view:
View attachment 1405648

From here, top left SPL graph the blue line shows maximum SPL in this enclosure. The enclosure is what the system thought would work fine and offered it ready there. You could shrink the box size or change to reflex box if you already know what the enclosure is. But let's just use this data. First I just look the light blue line on SPL window, it seems this driver can actually do 100db.

this is amazing The SW can provide the max SPL obtainable from a driver mounted on a specific cabinet ? this is more than great
i think this could solve any issue i could have Thank you very much

Next I just increase input power to see how much is needed for the driver to make about 100db. I just scroll my mouse on the voltage box to increase input voltage, and look the SPL graph change until it's about at the 100db mark. Now we found out about 30W of power is required. Nice my amplifier is sufficient so this driver seems plausible, it's also the RMS power limit on the datasheet so the driver will survive just fine. As you see the top right graph showing excursion goes beyond xmax, so next thing I want to find out is how low crossover the driver can do 100db. It obviously cannot without highpass filter as the image shows xmax isn't enough so let's add high pass filter.
View attachment 1405650

Here I've opened the filter tab and engaged HighPass filter and just using my mouse scroll wheel I move the high pass filter frequency up, looking at the excursion graph, top right, until excursion goes below the xmax line. Now I can read from the graphs that the driver is capable of 100db of output if I have 2nd order high pass at least at ~150Hz.

View attachment 1405651

Obviously, the high pass filter could be steeper and higher in frequency and the excursion would go down even further, but this here is the minimum based on datasheet TS parameters. In general, the less excursion, the less there is distortion so you'd likely want to experiment how low you can have the xover before it sounds bad. In practice, you would make a prototype box to try it out, where the crossover needs to be. Prepare to adjust the crossover filter from say 150hz all the way up to 300hz or even higher, in order to make sound change with the 100db/1m output. Now if you think the system as a whole, crossover somewhere beterrn 150-300Hz means the bass section on the system needs to sound fine up to 300Hz at least, which is another task you need now to look at and plan for, if you decide to use this particular mid driver.

perfect I think i have understood I am open to other drivers of course This is a very cheap driver
But i understand the procedure much better now This is what to design means I wonder how many Diyers are aware of this SW and use it
i am not an expert but it looks fantastic to me Impressive

Next up, you could look for the frequency response and rest of the datasheet, to try and questimate top end performance to find some hints how high the driver could be used before issues. https://faitalpro.com/en/products/LF_Loudspeakers/product_details/index.php?id=401005150
View attachment 1405655

Datasheet says Le (inductance) is very low, which the impedance plot also shows. Also, datasheet reads there is copper cap demodulation, which indicates the motor has got some attention to it, not the cheapest possible so performance is likely fine. The frequency response together with impedance plot wiggle shows anomalies starting about at 3kHz and get worse the higher the frequency. So, you might be fine to run this at least 2kHz and beyond. Directivity for such small driver is also fine for many kiloherz.

Thanks again I will follow your directions Once i will be able to use this tool i will move to a prototype
I have already some small aluminum cabinets that i could use to host the driver From Teac
I think they should be good after adding some damping sheets inside the box on the walls
they accept up to 4" cone drivers

These simple analysis based on datasheet indicate the driver might work fine for what you have in mind. But since the data is not from your speaker you haven't even build yet, or heard any sounds, you still need to make it happen to actually listen whether it works or not. So, now you would design the whole system around the mid driver and 3way concept you've chosen. When you are done designing, then buy the drivers, build a prototype box, measure all the drivers in the box,

sorry do you mean measuring all the drivers mounted in the box indipendently without crossover ?

load measurement data into vituixCAD and work out few crossovers: make few xo variations with varying xo frequencies, whatever looks fine by the data, then implement them both and listen which sounds better. Remember to measure the whole system after impelmenting the crossover to make sure there is no mistakes, so that the speaker still measures like it simulates. Now you know whether it worked out fine or not!
If the system is good, build final version of speakers with finished surfaces and all, fix any mistakes / problems you encountered with the prototype speaker, and enjoy. If the system doesn't work out fine, you'd just change your design, build another better structure, perhaps change drivers and so on. It can be a process that rinse and repeat many years 😀 Basically your system quality is directly related to how many things you have been able to notice and fix, which kind of relates how many prototypes you've build. If you build no prototypes, you can be certain the speaker has some issues because you didn't take time to spot them and resolve. When you've built several speakers, you can be certain at least some issues are not there, the ones you were able to notice and resolve. At least some cosmetic issues, physical design of the system for less painful building experience and so on, perhaps some audio issues as well, like bad sounding port noises, sound of midrange distortion, and so on 😉

I hope this thought process explanation is helpful. Have fun!🙂

thank you very much indeed for this very kind and extremely valuable lesson I feel like i have got all the needed information now
in order to familiarize myself with the procedure could I use a low-value three-way as a test bench?
the test would be accelerated by having all the components already available
there are widespread opinions that say how various commercial speakers are not perfectly optimized and in fact kits are sold to upgrade them especially in the xover and then in the cables and damping materials

 

[tmuikku]

sorry do you mean measuring all the drivers mounted in the box indipendently without crossover ?

Hi, yeah thats the most flexible way to do it, a modern way to get good amount of data how the system radiates sound to all directions. Search for "spinorama" and CTA-2034-A speaker measurement standard paper.

The software is VituixCAD, you should download the manual and take a look!🙂

 

[ginetto61]

Of course you're getting mad, as you don't specify the steepness of the filter i.e. electrical type of filter, its order, and eventually the response of the speaker in exam. Moreover, the other two entities are unknown

Hi yes of course But anyway i have learned a very important thing i.e. that
once fixed the wanted max SPL sim SWs can predict the usable range for a driver 😵😀👍👍👍
This is what i was actually looking for The FR alone is not enough I can see the flatteness extension but not the optimum band
I have seen very expensive and also famous speakers that did not take into account this issue and when pushed a little some drivers started distorting or even tended to break under pressure Probably different xover points would have avoided any issue Wrong design
Quite surprising considering the brand a case here below with the cone mid being the problem

so those numbers are quite meaningless...on the other hand they can tell a lot of a speaker design!
Ah ah!

i think that all measurements tell something to those who can understand that But speaker design is challenging
if it were not so all the speakers would go from decidedly decent to excellent
My feeling is that the vintage speakers in particular are not fully optimized also because there were no design programs available that there are today
But I decided to leave aside everything that is not brand new
Also because as someone said even cheap drivers can give good performance when used correctly
Now I have discovered a truly formidable SW tool
I will download it, study it and use it Any future questions will certainly be very targeted
I hope there is a thread dedicated to VituixCAD Loudspeaker simulator

 

[ginetto61]

"
a 150mm woven-Kevlar-cone midrange cut at 350Hz if i am not wrong
this means that the energy at 300Hz could be still high ?"

Yes, crossovers don't "cut" usually. By definition they cross over each other, depending on topology/type. But there are very very deep dsp-xo that practically cut 96dB/oct

http://en.wikipedia.org/wiki/Audio_crossover

View attachment 1405411

Hi ! thank you very much indeed for confirming a very important issue
I have to make a consideration that will sound weird
A constructor of the highest fame designs a top of the range speaker that is adopted as a reference monitor by renowned recording studios even if it has serious design shortcomings ? 😵😵😵
and yes, perhaps a different xover or a doubling of the driver would have been enough to avoid the problem ?
this thing shocks me sincerely
limited minds like mine need myths
when a myth collapses they feel a mixture of disappointment and dismay
they feel betrayed
I'll have a double malt beer and hope to recover from the shock

 

[ginetto61]

Audax distortion in % looks quite different... This was done in 2013, REW didn't show noise floor. And these don't tell much, they were done with relatively low spl. Just an example that I picked from my archives.
Best source for distortion behaviour of drivers is hificompass.com!

Hi thanks a lot again ! the idea is to start using sim SW and then measure a speaker i have at hand just to practice with the process

View attachment 1405712View attachment 1405713

And ten years later when I used it in a 3-way, xo LR3 300/3500Hz. Again quite low spl, but obviously no problems. Again measured indoors with quite high noise floor

View attachment 1405718

very interesting I said 100dB/1m as worst case during some peaks But they could be quite less
I have the feeling that a listening can be satisfying also at lower SPLs as long as the response is flat
i come from using small speakers and now i have understood what a more full range speaker can provide i.e. a much more realistic experience
there is no going back without a sub

 

[ginetto61]

To maintain flat frequency response output, volume displacement of a transducer needs to quadruple every octave down. So, if the driver hits xmax at 150Hz with 100db output, it means that 100db output at 300Hz makes only 1/4 of the excursion, is only 1/4 of xmax. So, if assuming that most distortion is due to excursion, crossover at 300Hz is already a great deal better than crossover at 150Hz and if you push it to 500Hz the excursion is 1/10th of xmax (lazy math), with 100db output.
When you move crossover even a little bit higher, the excursion of the driver goes down, distortion of the driver (whole pass band) goes down.

Good morning and thank you very much again for having shown me the way mentioning the sim SW
i am quite done with the questions for now because i am pretty sure that the SW will answer many questions
I have seen that it is quite popular among speakers designers Like almost a standard ?
good Using it is my only task for now
As i said i grew up being fascinated by electronics But lately i have understood how decisive are speakers for the overall performance
I am worried by the challenge but i understand that it is the only way to go now
Thank you sincerely again
Have a nice day
gino🙂👍

 

[Juhazi]

"some efforts should be done to set up an instrumental testing procedure in particular to establish the best range of work for a driver"
- That has been very clear and solved issue. Just look at data from eg. www.hificompass.com or all the others linked earlier.
- There is no direct and simple answer naturally, it depends on the system design.

"I hope there is a thread dedicated to VituixCAD Loudspeaker simulator"
- these (notice plural) have been linked for you many times before. You can also use search engine at diyaudio.com or google etc.
- official homepage https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html
- official thread about updates and help https://www.htguide.com/forum/mission-possible-diy/development-tools-techniques/42827-vituixcad-v2
- some original guides also at Youtube https://www.youtube.com/@kimmosto and several from random users
- main thread at diyaudio.com https://www.diyaudio.com/community/threads/vituixcad.307910/

 

[ginetto61]

Again same driver in a proto 2014. No eq or xo applied. The proto box was large sealed with polyfill stuffing. Differences are in details, general behaviour is similar. This spl is again too low to search for maximum spl, just a sanity check when doing directivity measurements. With REW it is easy to do several succesive sweeps and let the program to calculate means, which reduces noise artefacts remarkably.

Like Charlie said, odd order distortions should be lower than even, and no peaks! Here 3rd exceeds 2nd 1500-3000Hz. Most likely attachment to the proto baffle was not tight enough.

View attachment 1405742View attachment 1405743

thank you very much again
it is important to me to understand the behaviour of a driver in the worst scenario i.e. high SPLs
until now i have never paid attention to the Xmax for instance
i see that for some mid is even 2mm !!! this means that above that value the driver starts to have problem
now my main goal is to use the sim SW an unbelievable tool indeed

 

[ginetto61]

"some efforts should be done to set up an instrumental testing procedure in particular to establish the best range of work for a driver"
- That has been very clear and solved issue. Just look at data from eg. www.hificompass.com or all the others linked earlier.
- There is no direct and simple answer naturally, it depends on the system design.

"I hope there is a thread dedicated to VituixCAD Loudspeaker simulator"
- these (notice plural) have been linked for you many times before. You can also use search engine at diyaudio.com or google etc.
- official homepage https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_help_20.html
- official thread about updates and help https://www.htguide.com/forum/mission-possible-diy/development-tools-techniques/42827-vituixcad-v2
- some original guides also at Youtube https://www.youtube.com/@kimmosto and several from random users
- main thread at diyaudio.com https://www.diyaudio.com/community/threads/vituixcad.307910/

thanks a lot again I think i have all the info now My understanding of things is limited
i was decent only with simple tasks following procedures Theory has always been hard for me like math
my biggest regrets well the 2nd one actually
thanks a lot again for your kind and extremely valuable support

 

[Juhazi]

I am poor at maths, dumb and lazy too. I have no intention to start learning VCAD2 myself.

I have to read a lot and believe what the more skillfull and experienced people say and do.
Then I make my synthesis and start experimenting and measuring. Multichannel dsp and REW opened a new world for me in 2013. I still learn new things every day.

Happy journey, Gino!

 

[ginetto61]

I am poor at maths, dumb and lazy too. I have no intention to start learning VCAD2 myself.

Good morning ! then we are two at least 😎
when you want to define something that is right, rational, perfect, indisputable, you say it is mathematical
this is fascinating to me

I have to read a lot and believe what the more skillfull and experienced people say and do.
Then I make my synthesis and start experimenting and measuring. Multichannel dsp and REW opened a new world for me in 2013. I still learn new things every day.
Happy journey, Gino!

Thank you very much sincerely
i think that i great start is to begin playing with sim sws Reading their manuals
They have most of the useful equations in them
as i said i am only interested in speakers these days
i have a very strange feeling i.e. some well reviewed speakers have flaws that show up under stress (high SPLs)
but i am not expert enough to be sure
Have a nice day
gino🙂

 

[Juhazi]

Electronics, physics and acoustics can be explained with mathematical functions which are based on "laws of physics".
But I am a visual person, I understand graphical presentation better than formulas.

https://www.linkwitzlab.com/filters.htm

 

[ginetto61]

Hi exactly ! the problem is that math is not easy to understand for anyone Imho there are those who are gifted
however the availabity of SWs that do the math gives me some hope

again my initial point was about measurements
I have understood the concept of Xmax of a driver
the question is still the same i.e. when a driver displacement exceeds its Xmax this should show up in some measurements
i get that SWs can predict at what Hz and SPL this will occur but i would love to see this event with instruments
for instance using a sine generator i could send a signal in and raise the level

when the X will exceed the Xmax will i get an increase of distortion peaks level ?

as an example when an amp starts to distort at clipping the sine top flattens
at some points the predictions of the design SW should be confirmed by specific measurements
i would do this only for the mid driver
the fact that TOTL speakers had flaws in the design is shocking for me
unbelieavable
it would be like Usain Bolt who trips when he runs to catch the bus instead of running in front of it
some speakers should come just perfect
i guess that in reality most speakers are normally used at very low SPLs i.e. with drivers working well below their Xmax
Therefore their limits do not show up

 

[Juhazi]

Driver's Xmax (voice coil overhang) exceeding will not result in clipping, but distortion (nonlinearity of vibration) skyrockets. Coil/magnet constuction is important and there are many ways to reduce distortion near and beyond Xmax (eg. copper ring)

Xmech is often much higher and then behaviour is like clipping of an amplifier. Every loudspeaker driver behaves a little differently, some will break from first hit, some other can take it for some time. Testing of this will get very expensive...

G

Thread 'Maximum excursion (Xmax) of a driver'

2023-04-21 7:19 am

I'm using a 4" woofer driver, which according to its manufacturer has a +/- 3 mm maximum excursion (Xmax).

I also use a DSP for this driver which has a limiter that I could use to limit the maximum excursion in accordance with the above specs.

My concern is, could it be safe for the driver to set the limit at a higher excursion, for example +/-5 mm, if the higher distortion is accepted.

Any experience on this issue?

• gdan
• Replies: 36
• Forum: Multi-Way

https://speakerwizard.co.uk/driver-ts-parameters-xmax/
https://www.klippel.de/fileadmin/_migrated/content_uploads/AN_04_Peak_Displacement_Xmax.pdf

Lars Risbo of Purifi explains https://purifi-audio.com/blog/tech-notes-1