|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
4.30.66.254
In Reply to: RE: JA doesn't routinely perform those tests posted by Brian H P on June 18, 2014 at 10:09:42
>It would certainly be useful if he did!
I don't see the point with non-minimum-phase systems, which is the case
with 99.999% of loudspeakers. The step response that I do publish gives
the readers the same information in a more readily accessible form, I feel.
>It would also be nice to see harmonic distortion vs. frequency, a test
>much more relevant to how a speaker will actually sound.
As I have written in the magazine, it is difficult to get meaningful and
consistent distortion measurements without access to an anechoic chamber.
I also don't think it as important as you do - see link below.
John Atkinson
Editor, Stereophile
Follow Ups:
Hi John,
I have to politely disagree with you about distortion tests on loudspeakers. After 15 years of our testing, we find it to be extremely telling.
First, distortion is typically much higher on a loudspeaker than an amplifier. Why would it be relevant to test on an amplifier then and not a loudspeaker where it's bound to be more noticeable?
Second, it is quite telling of sound quality, particularly if the distortion rises markedly at frequencies that are quite audible. For instance, we often seen distortion rise significantly in the crossover regions, which often does show up in the listening tests.
Finally, the impact of our distortion tests has been so significant, that I've seen no less than 20 manufacturers (that I know of) take the results of those tests when significant distortion occurs (and shouldn't be there) and go back and either improve the drivers or the components in the loudspeakers that might be causing it. Like you, most of these companies aren't able to test distortion reliably like we can (in NRC's facilities), so it's been a useful tool for them as well.
Overall, I see the distortion tests that we perform every bit as valuable as the frequency-response and electrical tests -- and sometimes more so.
Doug Schneider
SoundStage!
. . . D.B. Keele used to show the measured HD of EVERY speaker he tested at E-41.2 Hz and A-55 Hz. Occasionally he used B-61.7 or A-440 if he suspected problems in those areas, occasionally graphed THD vs frequency across the whole spectrum, and occasionally showed intermodulation distortion between the low E and A. I don't remember details of his testing regimen, but I do remember some speaker reviews showing up to 40% second harmonic and 53% (!) third harmonic of the low E. No way THAT won't be audible!
Funny thing, a VERY high percentage (as in 20-40%) of 2nd harmonic can be inaudible or even pleasant sounding.
Starting with the Abby road album, 2nd and even harmonics were added to 'warm" the sound and guitar amplifiers and even the guitar loudspeakers are set up to have preferably even harmonics.
With natural or machine sounds which are not harmonic in structure (unlike music) , both even and odd harmonics change the timber of sound.
Odd harmonics are dissonant sounding and with both harmonics, it is how far they are from the fundamental which partly governs how audible they will be per masking.
" ... it is how far they are from the fundamental which partly governs how audible they will be ... "
I'd still be interested in the numbers, e.g. how many DB down is the 3rd, 5th harmonic audible, or mess with the sound?
Never trust an Atom, they Make Up everything!
I have explored how low a pure tone I can get out of my system and still have it be relatively undistorted (with no obvious doubling). I do this by experimenting with volume and sine wave pitch and exploring this two dimensional space. At a certain frequency I can adjust the volume to hear a (relatively) undistorted tone with no obvious doubling. By increasing the volume above this point, I begin to hear obvious doubling. At a lower frequency I eventually reach the point where turning up the volume sufficiently to hear anything is already at the point where I hear the doubling. I don't recall exactly where this point was, but as I recall it was around 30Hz with my setup.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
I don’t know John’s reasons but given that is one of the most difficult areas in audio, I would be hesitant to use “a” number here. It’s a LOT more complicated than just a number like THD, a graph of fundamental and first five is better but needs interpretation. That was fine for RF where it was the amount of out band signal was bad and still ok when amplifiers pretty much all had similar limitations and behaviors. For low frequency loudspeakers, since woofers and simple alignments generally misbehave in similar ways maybe there too.
Our ears are not equally sensitive to all frequencies, in fact while we are unaware of it as we have never experienced anything else, out ears are FAR from flat in response and that is why THD is only very loosely related to what we hear, it is a ‘flat frequency sensitivity” system, like a measurement mic, perfect recorder etc.
What hearing studies have shown is that our ears have a response curve which also varies with level and frequency. If you examine the current equal loudness curves, one see’s the graph end at 20 to 20K but our hearing continues essentially following the basic trend for several octaves each way.
http://decoy.iki.fi/dsound/pictsnd/robdad1
So far as sensitive to harmonics, look at the curves and then invert the shapes and think sensitivity curve.
Our ears are most sensitive around 3-4 Khz where much of the information in our voice is AND where our head size and pinna information gives us maximal location of voices.
Now, to use an example from an old Dolby labs paper, in a silent background, if one produced a 20Hz tone just above the threshold of detection, that requires about 75dB spl at your ears. The third harmonic or fifth are the generally the highest level and worst sounding harmonics on a woofer so looking back a the graph, one can see that to be the SAME apparent loudness as the 20Hz tone, the third harmonic (60Hz) has to be about -35dB down and at 100Hz has to be about -50dB down. The worst part is for the number obsessed, these distortion levels are essentially impossible to achieve.
On the other hand if one were producing 4Khz to have the same apparent level the 3rd and 5th need to be about +12dB and > 20dB HIGHER than the fundamental. With pure tones our ears are very sensitive to distortion where it falls in a more sensitive range and less sensitive to it when it falls in the less sensitive range (above 4K).
There is more to it however as we never (except testing) listen to a pure tone and even changing the level of a pure tone, broadens the bandwidth the pure tone occupies. We are dealing with a signal’s which can change 10,000:1 in level and can cover 10 octaves or more in steady state bandwidth.
Now one adds auditory masking with another view of our hearing and the concept of the mp3 is born.
Here one finds another odd behavior of our hearing system. If one has a single tone, there is a span on either side where one cannot hear another tone IF it is below the threshold level. Examine fig B in the link to see a set of those masking bands on either side of the test tone.
http://en.wikipedia.org/wiki/Auditory_masking
http://en.wikipedia.org/wiki/Auditory_masking#mediaviewer
/File:Maskingpatterns_sp11.jpg
The idea of the mp3 and other early digital signal compression strategies was that since you can’t detect anything below the masking curve, one didn’t need to encode that part.
Bottom line, can you hear distortion?
Sure you can but does it relate to what one sees with an analyzer, it might and might not AND distortion is VERY hard to measure with a dynamic musical signal which is more like what we use the stuff for.
Also it depends what kind it is, with music even harmonics are musically related, odd are icky, with natural sounds and machine sounds which are usually not harmonic, both even and odd harmonics can alter the sound (I use a lot of that kind of stuff for evaluation in addition to music).
Tying what we measure to what we hear has been the dream of a number of people and how we hear all by itself is fascinating.
For me at work, what was very useful in the development process was making generation loss recordings using a musical track, 24/96 recorder, measurement mic and loudspeaker up on a small tower.
Whatever wasn’t faithful turned in to an exaggeration, a caricature of what is wrong. That doesn’t tell you what is wrong but it helps to hear what it is clearly and also hear when it’s fixed.
So far as being faithful to the signal, Loudspeakers are by far the weakest link, any modern electronic gear like an amplifier or pre-amp, can go many generations before becoming contaminated but even good loudspeakers can rarely go more than two or three generations and most sound bad at one.
Best,
Tom Danley
Danley Sound Labs
I agree that there are different "layers" of it that, from a design point of view, are needed.But for consumer-oriented publications such as SoundStage! and Stereophile, a THD graph can be quite telling, and can be indicative of something wrong or not.
For example, YG's original Anat Main Module had very high distortion for a $30,000/pair speakers.
http://www.soundstagenetwork.com/measurements/speakers/yg_anat_ref_main_module/
That, alone, spells an issue and I believe is something that YG worked in subsequent iterations of their speakers to address.
Conversely, when THD is extremely low, as it was for the KEF R500, then for 1st, 2nd, 3rd, etc., does it really matter, particularly for the consumer? Not in my opinion -- it's extremely low and that's what really needs to be conveyed. Now, for the designers, they might want to see what minor components are cropping up between 200-500Hz when we cranked it up to 95dB, but that becomes a design issue.
http://www.soundstagenetwork.com/index.php?option=com_content&view=article&id=793:nrc-measurements-kef-r500&catid=77:loudspeaker-measurements
Doug Schneider
Edits: 06/22/14 06/22/14
"Second, it is quite telling of sound quality, particularly if the distortion rises markedly at frequencies that are quite audible. For instance, we often seen distortion rise significantly in the crossover regions, which often does show up in the listening tests."
If it (the rising distortion) often doesn't show up in listening tests then how it is quite telling of sound quality?
Never trust an Atom, they Make Up everything!
Hello,
Perhaps my sentence wasn't clear enough, but to clarify: it is audible at certain frequencies. For example, significant distortion around, say, 2kHz, which is where many tweeters are crossed over, is quite easy to hear.
Doug Schneider
nt
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: