What Geddes says.....

[Notorious_AK]

according to Dr.Gedlee hd measurements are irrelevant.

http://www.gedlee.com/distortion_perception.htm

IMO klippel tests at diymobileaudio tell much more about driver's performance.

EDIT by moderator. This new thread was created after people insisted in hijacking Paul's thread....

http://www.htguide.com/forum/showthread.php4?t=32222

[mazurek]

according to Dr.Gedlee hd measurements are irrelevant.

http://www.gedlee.com/distortion_perception.htm

IMO klippel tests at diymobileaudio tell much more about driver's performance.

I believe that doing HD testing at various levels, up to really loud, should capture elements of what klippel tests show.

[ThomasW]

according to Dr.Gedlee hd measurements are irrelevant.

http://www.gedlee.com/distortion_perception.htm

IMO klippel tests at diymobileaudio tell much more about driver's performance.

The thread isn't about what Earl Geddes thinks....

FWIW "Gedlee" is the name of his company, it's a fusion of Earl's and his wife's last names...

[Dennis H]

according to Dr.Gedlee hd measurements are irrelevant.
http://www.gedlee.com/distortion_perception.htm

That's not what he says. He says THD (a single number that describes a driver) doesn't tell you much about how it sounds because the higher orders sound worse than the lower orders. Nothing new there -- Olson did all that back in the 60s and 70s -- but Geddes seems to assume in most of his papers that we're all idiots and he's the only smart one. Kind of annoying but 'hello my name is Earl'.

IMO klippel tests at diymobileaudio tell much more about driver's performance.

Klippel and Dumax are pretty good at predicting LF distortion but tell you little about distortion at higher frequencies. Clearly a sweep that gives you real (as opposed to theoretical) numbers at all frequencies and has separate curves for each order tells you much more. How you interpret the curves is another matter.

[seattle_ice]

-- but Geddes seems to assume in most of his papers that we're all idiots and he's the only smart one. Kind of annoying but 'hello my name is Earl'.

Well, in his paper on compression driver distortion, he does point out the radical new concept that distortion in compression drivers is more noticeable to listeners at higher playback levels and greater delay times than low ones.

I have to say, a lot of his stuff reads like every speaker builder in the world simply takes straight line distortion readings of the drivers at one fixed level, then constructs their crossovers.

But what do I know, I am a mechanical and software engineer, not an EE.

[ThomasW]

Once again, this isn't a thread about Earl Geddes

[Dennis H]

Once again, this isn't a thread about Earl Geddes

Okay, how about Olson? :B

Here's his original graph from 40 years ago.



And here's the same thing replotted (by someone else) as dB the harmonics are below the fundamental. I added the % distortion numbers along the left.



The interesting thing to me is how MUCH more obnoxious the higher orders are than the lower orders. At 80dB, it might take 20% 2nd order to sound really bad but only .0006% 9th order to sound bad. At 110dB, it might take 30% 2nd order to sound bad but only .01% 9th order to sound bad. It sorta validates some of the tube amp guys -- they generally have high 2nd order distortion but very low higher order.

Bringing it back to speakers and testing in ARTA, we really need to look hard at the highest orders it will plot: 5 and 6+.

[Ray_D]

Is this frequency dependent?

Okay, how about Olson? :B

Here's his original graph from 40 years ago.



And here's the same thing replotted (by someone else) as dB the harmonics are below the fundamental. I added the % distortion numbers along the left.



The interesting thing to me is how MUCH more obnoxious the higher orders are than the lower orders. At 80dB, it might take 20% 2nd order to sound really bad but only .0006% 9th order to sound bad. At 110dB, it might take 30% 2nd order to sound bad but only .01% 9th order to sound bad. It sorta validates some of the tube amp guys -- they generally have high 2nd order distortion but very low higher order.

Bringing it back to speakers and testing in ARTA, we really need to look hard at the highest orders it will plot: 5 and 6+.

These are for a "single tone", but it does not sat what frequency. I suspect this is important.

[Geddes]

First, those are very interesting curve of the ears nonlinearity, but I suggest caution. While they may be plausible I would be suspicious on two accounts. First H.F. Olsen was know to "extrapolate" a lot of data. Much of his work was incorrect and has been shown to be so. Second, I don't know how he could even have gotten this data, even today, but especially 50 years ago. I'm not exactly sure what it means or how it was obtained.

Now, I would like to defend myself since there has been a lot of "Earl" bashing going on. When I write a paper I always assume as little knowledge on the readers part as is plausible. This only makes sense since the knowledgeable reader can always skip the material that they already know. I find it most curious why someone would be offended by this practice.

As to my own work it is being misquoted and needs to be corrected. While it is true that many people had hypothesized that higher harmonics would be more audible - and this was acknowledged by us - no one had actually done the work to show this. We did. But further, no one else had ever hypothesised that the location of the nonlinearity was important, which we did. Howard talks about this is in Stereophile article, but he uses the concept of the phase of the harmonics, which ends up being the same as the location of the nonlinearity along the transfer characteristic. So those who claim to have know or that others knew all that we said in our papers are simply being dishonest.

We showed that THD and IM, as numbers, do not correlate with the subjective perception - not that they always wrong, they can be right, but they can also be very very wrong. Quite simply they can't be trusted to reflect the perception of any arbitrary case.

It is also true that a complete map of harmonic structure with frequency and SPL could actually be used to construct all of the measurements done by a Klippel setup, but its not trivial. Thats why it hasn't been done, but it could be. In fact this technique would probably work better because, as I found out, the Klippel data is not very stable above the third order and its these higher orders that matter. A reconstruction based on actual radiated data would be more stable because it would only be based on sound that actually radiates, not a physical measurement of cone motion and a calculation of the radiated sound.


Earl Geddes

[Dennis H]

Thanks for replying, Earl. I apologize for my snotty tone (it had been one of those days) and I appreciate all the research you have done.

It is also true that a complete map of harmonic structure with frequency and SPL could actually be used to construct all of the measurements done by a Klippel setup, but its not trivial.

The price of software to measure distortion has come way down, to the point it's reasonable for hobbyists to own and use. STEPS, included with ARTA, will plot individual 2-5 and 6+ harmonic curves. Last I checked, it was about $120 (varies with the euro). The free demo version will do everything except load and save files.

http://www.fesb.hr/~mateljan/arta/index.htm

[cjd]

I didn't remember there was much bashing of you Earl (re-reading of course I see Dennis's comments, which he's apologized for - I tend to just filter that kind of stuff on the internet any more). I did remember some response to someone that came in suggesting we were all both wrong and dumb (exaggeration, I know), and your work was used as supporting evidence. At least, that's how it came across to me. And of course I'm applying my own bias and whatever else, so... who knows.

I've rarely read a white-paper or technical document that did what you're doing - they all assumed the reader had a whole boatload of knowledge already and simply built on it. So for anyone with that background, you filling in the gaps might come across as a bit patronizing.

Then again, we're not your usual batch of hobbyists.

I also suspect you've now gone and prompted a few folks to perhaps TRY to take that next step.

Anyhoo... Welcome to HTGuide Pull up a chair. It gets interesting sometimes.

C

[mazurek]

It is admirable for any researcher to work towards pinning down anything involving human perception. I used to do research on biomechanics, people are a pain to measure.

I appreciate your participation in a public forum where not everyone has done their requisite literature review. Now more than ever, I believe we have a lot of tools to help us assess psychoacoustic in loudspeaker reproduction. Hopefully that knowledge will be published rather than remaining business proprietary or in the hands of eccentric experimenters without the resources to produce professional level documentation.

[AJINFLA]

Welcome aboard Earl. Your input here will certainly be welcomed by me..and I suspect, many others. Beware of the Mod. He's a crusty, ill-tempered old man who doesn't take kindly to manufacturers peddling their wares here. You don't want to join the Fight Club list like a few others have.
But at least he won't try to sell you lipstick and other makeup for pig drivers here as opposed to other forums.
Are you planning any audibility studies for your amp tests ?

cheers,

AJ

[Dennis H]

Just a further thought on Klippel, I think those measurements are most useful if you're a driver designer. They need to know what's causing the distortion (Bl, Kms, Le, whatever), not just how much distortion there is. Things are much easier for us hobbyists buying drivers that already exist. We don't need to care about what causes the distortion, just how the distortion actually measures.