or Connect
AVS › AVS Forum › Audio › DIY Speakers and Subs › CEDIA 2010 and Distortion
New Posts  All Forums:Forum Nav:

CEDIA 2010 and Distortion

post #1 of 68
Thread Starter 
Folks around here and other places have commented about the audibility of distortion, especially in the lower frequencies. The CEA 2010 standard, which Ilkka uses for 'max clean output' measurements, allows the driver to reach large % distortion levels. For a recap, here's how much harmonic content the standard allows compared to the fundamental tone:

2nd Harmonic: -10dB
3rd Harmonic: -15dB
4th Harmonic: -20dB
5th Harmonic: -20dB
6th Harmonic: -30dB
7th Harmonic: -30dB
8th Harmonic: -30dB
9th Harmonic: -40dB

10th Harmonic and above: -40dB

So, I put together a sound clip. It is a 30Hz tone for the 1st 10 seconds, and at 10 seconds, it adds in 2nd Harmonic at 10dB below the fundamental, 3rd Harmonic at over 20dB below the fundamental, and 5th Harmonic at over 20dB below the fundamental. This makes it agree 100% with the spec above.

I generated the file with Audacity, and a Harmonic Generator plug-in.

Play the file at a level which your sub can play an ABSOLUTELY CLEAN 30Hz tone (less than 1% THD in the first 10 seconds). Then wait until the harmonics are added. This TEST ONLY WORKS IF YOU HAVE A REASONABLY FLAT FREQ RESPONSE IN ROOM

Is this distortion not audible? Would you consider this 'clean' output? CEA 2010 standards do...


Here's the spectral breakdown of the last 10 seconds:



Okay, I am unable to attach .mp3 files...how can I do this?

JSS
LL
post #2 of 68
Host the MP3 at someplace like Rapidshare.com or other free hosting place.
post #3 of 68
Thread Starter 
post #4 of 68
Big difference. I'd call the last 10 seconds "not so clean" by a good margin.
post #5 of 68
The wave is not setup right. You lowered the amplitude of the primary signal for the last half of the wave by several dB Peak and RMS.



If you zip the mp3 you can attach it here.
post #6 of 68
I would also back off to -6db or -10dB to leave a little wiggle room.
post #7 of 68
I think a better way to conduct this test would be to generate harmonic distortion with a musical test signal instead of a sine wave...
post #8 of 68
Quote:
Originally Posted by mbentz View Post

i think a better way to conduct this test would be to generate harmonic distortion with a musical test signal instead of a sine wave...

+1
post #9 of 68
Thread Starter 
If I knew how to add the harmonics without reducing the level of the signal, I would, but I just found the plug in, and it does add the harmonics to the levels specified below the fundamental...

Even though the overall level has dropped in the signal, it is the RELATIVE LEVEL to the fundamental that matters here.

Also, I know sine waves are a torture test, but I wanted to demonstrate that even at lower levels (way before actual driver distortion would set in), the amounts of distortion allowed by the standard are what I would consider excessive, by a LARGE margin.

You need considerably more power, driver area and excursion to generate low frequencies, so my guess is that's why they are given so much leeway distortion-wise. But calling that output 'clean' is just not right...

Attached is a zipped .wav file, much cleaner than the .mp3, only with the second harmonic added...

Over the next few days, I'll provide more examples. You'll see that as the fundamental drops in frequency, harmonic distortion is even MORE audible, contrary to popular belief.

JSS

PS - The whole reason I am bringing this to light is because through my own testing, I found that I was highly sensitive to low freq distortion, probably in no small part due to equal loudness curves. Certain movie scenes when played back at higher volumes would sound different than at low volumes, and I wanted to know why...so I tested, and found that distortion is clearly audible at much lower levels than the only distortion standard I know about CEA/CEDIA 2010 allow....

 

Dist1.zip 440.2626953125k . file
post #10 of 68
Quote:
Originally Posted by MBentz View Post

I think a better way to conduct this test would be to generate harmonic distortion with a musical test signal instead of a sine wave...

I agree that real world signals would be preferred to sine waves. However, it may not be easy to generate a distorted version of a complex (multiple frequencies) signal that would be the same as output from a subwoofer with some harmonic distortion performance.

Just adding harmonic distortion to a signal (adding spectrum components at double and triple frequencies and so on) won't do as a non-linear system's response to some input will contain more than just harmonics. Some modeling of the underlying input-output transfer function would be needed and even then there may be several different transfer functions resulting in same HD performance. There may be some typical shapes of transfer functions though which would allow a reasonable model for subwoofers.

I think Klippel has some research in this area with auralization of loudspeaker non-linearities. I remember a website with a listening test for different kinds of distortions in loudspeakers, may be there's some low frequency stuff too.
post #11 of 68
The problem here is that one doesn't listen to sine waves, one listens to music. The same harmonics that are the distortion components of a clipped sine wave are present in a music source signal, so the difficulty lies in deciding where to draw the line between which harmonics should be there, and which should not, especially since the source media was mixed by engineers listening to the source over speakers, and how they mixed it depended on how they heard it, distortion and all.
The 10% THD benchmark for xmax via Klippel was arrived at somewhat arbitrarily, but they had to decide on a figure, and 10% was it. Measuring that 10% with sine waves is logical for the pure purpose of arriving at a specific xmax measurement. But where finished speakers are concerned sine waves are moot, program material is what counts, so that, or pink noise to most closely duplicate it, is what speakers should be measured with.
And don't assume that higher than 10% THD will necessarily sound bad. Those same second and third harmonics of the first two octaves of fundamentals that comprise the bulk of the distortion component are responsible for what's often referred to as 'punch'. More than a few newbies to the clean output of folded horn subs actually miss the presence of midbass punch that they're used to from direct radiators.
post #12 of 68
Thread Starter 
I am not doubting in any way shape or form that program material is what counts...in fact....

The impetus for me not only to build more capable subs, but for doing this experiment was that I heard problems while playing back program material, and I was trying to understand what was going on. The easiest way for me to quantify distortion with the equipment I had on hand was to use sine waves....and compare a clean sinewave to one with harmonics added. There is a definite and audible difference. With sinewaves, the difference is more exagerrated than with program material, but the difference is there in program material as well, as it's the reason I looked into distortion in the first place.......

Maybe my ears are just too damned sensitive to it...who knows.

I never said this test was perfect. I am only trying to say that the CEA2010 spec leaves a lot to be desired, and if someone were to listen to program material at the CEA2010 spec's limits, it would sound bad, and therefore it is not a good measure of 'clean' output....

As far as a real world signal, Audacity will add the harmonics to any signal. Gimme a signal you'd like compared with and without distortion (you specify the magnitude), and I'll create the files...I am limited to about 8 second clips due to the attachment size restriction...


JSS


PS - Folks, basically what I am trying to say is that if you listened to a sub playing at the CEA limits, you would think something is amiss, and you would turn the volume down....that's why I think the CEA limits are too high, and in no way represent a 'clean' sound...

PPS - Oh, and the .zip file attached above has only 2nd harmonic distortion in it, which should be less objectionable. When compressing to .mp3, the 3rd and 5th harmonics were added as a result of the compressing process, so I attached the .wav file in the .zip archive.
post #13 of 68
Thread Starter 
Quote:
Originally Posted by breez View Post

I think Klippel has some research in this area with auralization of loudspeaker non-linearities. I remember a website with a listening test for different kinds of distortions in loudspeakers, may be there's some low frequency stuff too.

Does anyone know where I can find this research or the website described above?

JSS
post #14 of 68
Quote:
Originally Posted by maxmercy View Post

Does anyone know where I can find this research or the website described above?

JSS

Klippel website
Listening test

There is a huge amount of excellent info on the site.
post #15 of 68
Quote:
Originally Posted by maxmercy View Post

If I knew how to add the harmonics without reducing the level of the signal, I would, but I just found the plug in, and it does add the harmonics to the levels specified below the fundamental...

Even though the overall level has dropped in the signal, it is the RELATIVE LEVEL to the fundamental that matters here.

No it isn't. You are trying to show how audible it is. You can't do that when half the signal is playing back at a different volume. The change in SPL is enough to throw the tests off. Try it with just a sine wave, and adjust nothing but the SPL change.


Quote:
Originally Posted by maxmercy View Post

Over the next few days, I'll provide more examples. You'll see that as the fundamental drops in frequency, harmonic distortion is even MORE audible, contrary to popular belief.

Actually, that is the common belief. The closer the fundamental gets to the in-audible range then more the higher frequency harmonics stand out.
post #16 of 68
Thread Starter 
Quote:
Originally Posted by soho54 View Post

No it isn't. You are trying to show how audible it is. You can't do that when half the signal is playing back at a different volume. The change in SPL is enough to throw the tests off. Try it with just a sine wave, and adjust nothing but the SPL change.


Actually, that is the common belief. The closer the fundamental gets to the in-audible range then more the higher frequency harmonics stand out.

OK, then. Let me re-phrase. CEA2010 Allows a certian level of harmonics relative to the fundamental, regardless of spl level. In fact, it is only when the level of the harmonics relative to the level of the fundamental reach a certain point, that CEA2010 calls the output 'distortion limited'. CEA2010 does not specify different acceptable harmonic distortion levels for different spl levels (maybe it should?).

That being said, as far as the technicality of the overall level changing: you are free to increase the level of the latter part of the clip to equal the former part of the clip to make sure everything is on an equal playing field in this horribly flawed test.....If I can, I will adjust the clip so that this volume adjustment will not be necessary. The point I am trying to prove is that sub manufacturers touting 'clean' spl levels using CEDIA2010 specs may not play as 'clean' as advertised....

Isn't the whole point of things to have high fidelity? How does the CEA2010 spec represent anything that would be considered hifi?

If common belief is that the lower the fundamental gets the more the harmonics stand out, why the abundance of statements from people that THD does not matter as much at the lowest frequencies?

JSS
post #17 of 68
Quote:
Originally Posted by maxmercy View Post
Attached is a zipped .wav file, much cleaner than the .mp3, only with the second harmonic added...
Same problem as before.

Here is the second wave, amplitude matched within . 1dB.

 

Dist1a.zip 483.9775390625k . file
post #18 of 68
The way I see it the CEA2010 limits are not representing what the "clean" limits of a system are, where you will get high fidelity uncompromised playback, but rather close to the maximum useable limit where going any further is ill advised both from the perspective of the systems health and any semblance of fidelity. I don't know if that is true or not. (Yes I understand that many would not want to use or allow that much distortion and I'm not advocating for it either.) I don't think that CEA2010 provides much more than some form of standard test so that bass systems can be compared to each other. The best way would obviously be with real music or content, but I don't know how you'd standardize that. It's flaw'd, but it's a heck of a lot better than manufacturer spin or nothing at all.

I do agree with Maxmercy about the audibility of the CEA2010 distortion levels at least with sine waves. How could you not notice 90 or 150hz mixed in as compared to the pure tone? Especially since these are the 3rd and 5th? I haven't even listened to the mp3 BTW.
post #19 of 68
Quote:
Originally Posted by maxmercy View Post

OK, then. Let me re-phrase. CEA2010 Allows a certian level of harmonics relative to the fundamental, regardless of spl level. In fact, it is only when the level of the harmonics relative to the level of the fundamental reach a certain point, that CEA2010 calls the output 'distortion limited'. CEA2010 does not specify different acceptable harmonic distortion levels for different spl levels (maybe it should?).

What you aren't factoring in is that the change in amplitude is more noticeable than the harmonics. You could have the first part quieter then the last, but it would still be flawed. That is just the way the human ear and perception works. The only way to really hear a difference in sound and not pressure levels is to match the two signals.

Quote:
Originally Posted by maxmercy View Post

If common belief is that the lower the fundamental gets the more the harmonics stand out, why the abundance of statements from people that THD does not matter as much at the lowest frequencies? JSS

This is two different issues, as I am understanding your wording.
post #20 of 68
I myself, don't think you will have any problems telling the difference between the two sine waves with a pair of headphones, volume matched or not.

I just think you need to cover all the bases.
post #21 of 68
Thread Starter 
Soho, I see what you mean. I would like to cover as many bases as I can...but I have limited tools at my disposal. I'll see if I can level-match at some point....how did you do it?

Basically, I'd like the thread to evolve so that we can come up with a better distortion threshold for good, clean audio to the lowest freqs...I'll post up clips, and people say whether or not they can tell a difference, and if the difference is objectionable...I'd be interested to see what people think when distortion levels reach 5% or more.....

I'll also have to read Klippel's work to make sure I am not re-inventing the wheel here....

Ricci, totally agree with you. It is better than having just a 'freq response range' and 'max spl' numbers in a spec, but I think it paints too pretty a picture. Folks that may not know better will think that they are purchasing a product that will play cleanly to umpteen dBs, but the reality is different...

JSS
post #22 of 68
Quote:
Originally Posted by maxmercy View Post

OK, then. Let me re-phrase. CEA2010 Allows a certian level of harmonics relative to the fundamental, regardless of spl level. In fact, it is only when the level of the harmonics relative to the level of the fundamental reach a certain point, that CEA2010 calls the output 'distortion limited'. CEA2010 does not specify different acceptable harmonic distortion levels for different spl levels (maybe it should?).

I'd think that higher distortion levels would be ok at higher SPL levels. If you are up around 120db it seems like it may be harder to distinguish that H2 or H3 than at a low 80db level that allows you to focus in on the details much more without your face and chest vibrating. Maybe the CEA2010 limits would be ok at live sound levels.

Quote:
Originally Posted by maxmercy View Post

If common belief is that the lower the fundamental gets the more the harmonics stand out, why the abundance of statements from people that THD does not matter as much at the lowest frequencies?
JSS

Probably because the # of systems with real high power output of any kind, distorted, or not below 30hz is rare among non audio fanatics. If you've not heard real sub bass then even a system with high distortion, but that still represents the fundamental will sound better than nothing there at all. That's a guess on my part. The other thing that I'd guess at is usually unless you are severely over driving things the main issues are h2 and to a lesser extent h3 distortion. If you are below 20hz your h2 is below 40hz and just sounds like more bass(=good to a lot of people) and H3 is below 60hz (=more bass=good again). Once into the really deep stuff around 10hz or even below even your 3rd harmonic is below 30hz. And again we are talking with real content not sine waves. That's my half cocked theory anyway.

One more thing I wanted to mention is that they do allow less of the higher order harmonics but what about the odd order harmonics being treated with less tolerance than even order? They do with the 3rd compared to the 2nd but that is it. I've always heard that the odd harmonics are much more offensive, which makes sense because the even order ones could be considered as part of a chord.
post #23 of 68
Quote:
Originally Posted by maxmercy View Post

Soho, I see what you mean. I would like to cover as many bases as I can...but I have limited tools at my disposal. I'll see if I can level-match at some point....how did you do it?

I though Audacity could do it, but it has been awhile since I have messed with it. I usually use Vegas or Cubase.

For this I used Cool Edit. It is a very very useful shareware program. There is an older version Cool Edit 95 I believe, that will allow you to use all of it's functionality for free, but you can only use one tool set at a time. This would be perfect for your endeavors here. I can't find my old copy or I would post it here.
post #24 of 68
Now has everyone read Dr. Earl Geddes papers on speaker distortion? He was able to digitally add distortion to musical test signals using a popular math program.
post #25 of 68
Geddes papers focused on much higher frequency ranges than subs operate at, and he even stated I his papers that things could be different at the lower frequencies. That he just wasn't interested in doing the research at those ranges himself.

He used a speaker emulator like http://www.klippel-listeningtest.de/lt/default.html did to create his distortion. Geddes used Mathlab IIRC. This would be the type of system needed to add distortion to actual musical waves. I don't know where you could find one off hand, and I sure wouldn't want to build one from scratch. I hate Mathlab. Martin J King my have a free version...
post #26 of 68
Here is a nonlinear distortion paper with a large portion of the required math included, if anyone is interested. I just got through with a raytracing program, so I will pass on this at this time.


http://projekter.aau.dk/projekter/fb...061_Thesis.pdf
post #27 of 68
This is from a commentary by John Janowitz:

Quote:


So that all said, what does Xmax really mean? It's a good way to tell how much range of the driver is really useful while keeping distortion inaudible. In speakers, the low end limit for being able to notice distortion is considered to be about 10%. In reality most people can't hear distortion and notice the audibility up until way over 25% even. That seems high, but while at Michigan Tech, the regional AES vice president, Tom Nousaine, came and did a discussion on distortion. Not just subwoofer distortion, but distortion full range. He played a track of Tom's Diner by Suzanne Vega. This was a blind listening test where the track was played over and over again with progressive amounts of distortion added. The trick was to say when you could hear audible distortion, but we didn't know the amounts until later. In a room of about 20 people, all audio geek AES members, we all sat and listened and would say "ok, the 4th time i could hear it". It turned out that only a couple in the room claimed to hear distortion at the point where 10% was added. Most claimed at 25-50% distortion.

Of course, he's talking about THD, not just 2HD, but 2HD is usually the highest.

I've been intrigued by SpectrumLab because it lets you see the playback at the LP and off the disc for direct comparison.

I'll post Irene because I have it handy, but I've been working on others as well. This idea may be something worth discussing that could become a very good test for non linearities using selected high quality source material instead of sine waves/sine bursts.



Worthy of noting is that fact that all of the odd harmonics of the fundamental, through 19HD, are part of the effect. Once I began to focus on the graph, looking at every tiny detail, I have since noticed that many other LFEffects designers use this trick.

With an effect like this, it's impossible to tell what the actual THD is because the odd order distortions would be masked by the effect itself.

I agree with Max. 30% 2HD at 30Hz is absurd. I couldn't force my system to give that, and I'm shocked it's a standard of any kind. Are you SURE those numbers are correct? My math says that list represents 73% THD through 6HD!!

OTOH, John's AES/Nousaine blind listening test story sure makes you wonder when it comes to THD and music vs sine waves.

I know Ilkka didn't retest all of the subs on his list so I assume there's a math formula to translate sine sweep data into CEA 2010, which is obviously way the holy heck more forgiving than the 10% limit I grew up on (TV, Nousaine, Mullen, Ilk, JJ, AVTalk, Wiggins, etc).

Nousaine still uses 10%, no? Does anyone actually use the CEA 2010 standard? I know Ilk did the math and posted the results, but time will render that list useless as the subs on it disappear and news one show up, and I've never seen it anywhere else.

Bosso
post #28 of 68
Thread Starter 
Are you SURE those numbers are correct? My math says that list represents 73% THD through 6HD!!

I have found two other sources that list the same step-wise function for max allowable HD.

In practice, I don't think any sub will hit all of them at the same time...

Here's an example:

http://www.hometheaterhifi.com/volum...06-part-2.html

Check out the graphs with the step-wise function in blue, and you can see that sometimes the sub is amp-limited, other times it reaches one of the higher harmonics first, preserving a lower THD, but most the graphs have distortion that I would consider clearly audible, and not 'clean' audio. 5% Distortion is audible to me in sinewave tests (it just sounds 'different'), and under under 2% is essentially the same to me as less than 1%. Between 2% and 5%, not so sure. Above 10% and the lowest notes in rap and DnB music start 'farting'....How do I know this? I used to calibrate my Omni15TB to give 10% THD at 40Hz at -0dBFS at the max amp gain I was going to give it for outdoor parties (and it was highpassed below 40Hz to protect the driver)....and during some hip-hop tunes, I'd RUN to the volume control to turn it down, because of the distortion that is supposedly 'inaudible' at 10%THD.......go figure.

Bosso,

Jeeeezus. You have <2% Distortion at reference at 6Hz? Wow, I mean WOW. I can only imagine how that sounds....

JSS
post #29 of 68
Quote:
Originally Posted by Ricci View Post

The way I see it the CEA2010 limits are not representing what the "clean" limits of a system are, where you will get high fidelity uncompromised playback, but rather close to the maximum useable limit where going any further is ill advised both from the perspective of the systems health and any semblance of fidelity.

I'm not sure I would look at the CEA2010 limits as a measure of usable output. I believe the intent of the measure is more to provide a means for comparison.

The problem with setting lower distortion numbers is that distortion grows rather slowly at lower levels...such that a 0.5% variation in the distortion measurement might result in a 6-10dB swing in output. However, when you get up near 10% (or more), the distortion starts to rise very quickly such that a 0.5% variation in distortion might only be a 0.5dB swing in output. All that to say, using a higher level of distortion to compare "output specs" ensures that measurement error has minimal effects on the comparisons
post #30 of 68
Quote:
Originally Posted by bossobass View Post

OTOH, John's AES/Nousaine blind listening test story sure makes you wonder when it comes to THD and music vs sine waves.

Blind listening examples like these simply show that musical signals can mask distortion. It does not show that all musical signals will mask all distortion below similar levels.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: DIY Speakers and Subs
AVS › AVS Forum › Audio › DIY Speakers and Subs › CEDIA 2010 and Distortion