To achieve 0.001% THD doesn't that imply that you also need to address , layout , power supply ripple or power supply rejection ratio , and RFI by default.
Another way of putting it about not enough resolution and dynamic range is that the math is really only good for linear waveform portions of a curve -- or very small sections of a curve (distortion) as to approximate a linear sampled range... Thus higher resolution and speed is needed for accurate measurment of very low distortion. Do I have that right? Thx-RNM
Last edited:
This is all old hat eg The Application of Digital Techniques to the Measurement of Loudspeakers. You might want to have another look at your copy of Oppenheim & Schafer. Averaging is NOT inherent in FFT techniques. (Mr. Marsh, the KEF paper has the answer to your question on transients.)
Today, it is possible to get around 'averaging' to whatever degree you wish and still not inconvenience yourself compared to old type analogue methods. eg MLS & Prof Farina's method.
Angelo's method also measures THD to the zillion'th order, has HUGE dynamic range and is now incorporated in the latest AP. It certainly isn't just limited to 'small' distortions but is in fact ideal for huge evil speaker distortions too.
I developed the theory for this circa 1990 for production testing of speakers for response & THD in the theoretically shortest possible time but the computing power and codecs weren't cheap enough then. When I emerged from the bush this millenium, even the cheapest laptop has more than enough horsepower and good codecs are 2 cents.
__________________________
On why speakers can be used to hear very small amounts of amp. distortion, see Intermodulation Distortion Listening Tests. Note the bit where he says, "it doesn't sound like a speaker distortion at all but more like an amplifier distorting."
That's not to say, speakers don't introduce huge amounts of intermod & other distortions but when we listen to them, they seem more innocuous.
__________________________
This fashion of proposing a 'new' distortion mechanism, usually with fancy test & name and incomprehensible (and usually ?? wonky) maths has been extant ever since the advent of recorded sound and probably before.
"When my virginal daughter plays Herr Beethoven's music, my cuckoo clock falls off the wall. Zis does not happen with my older (non-virginal) daughter. From zis, I conclude Herr Ludwig's music has Curl Induced Distortion zat intermodulates with ze Div inherent in virgins.
Zis can only be measured with Hirata cuckoo clocks carved from solid Unobtainium. Za is no need to conduct Blind Listening Tests cos Herr Ludwig is deaf.
JC Maxwell" - nationalities have been changed to protect the guilty.
__________________________
IM not so HO, everything Dick Heyser published in JAES except for his TDS paper is liquid BS and should be consigned to the Don't Recycle Bin. I spent more than a decade deciphering it and there is nothing, no ETC etc nothing useful, practical or worth keeping. TDS is important cos if you stretch the truth a little, Farina's method can be considered a more elegant implementation of TDS.
__________________________
Yes. But if you are a hair shirt minimalist like me, just doing layout bla bla and allowing eg 10x the distortion, say 0.01% may give you all the audible benefits.
Indeed, if you pay attention to layout bla bla, you might better 0.001% without further complexity.
IMHO, layout bla bla need this scrupulous attention anyway .. far more important than Golden Pinnae components cos they have audible (via Blind Listening bla bla) significance.
__________________________
BTW, John V, half the Dynamic Duo (L & V) presented an invited paper on just this subject of distortions at the recent San Francisco AES. Unlike the pseudo prophets, the DD, not only pontificate with triple integrals but also explain in words of 1 syllable for us unwashed masses AND check their assumptions with Blind Listening Tests.
I only have a few pics taken of the screen at his slide presentation but may have to lurn to reed & rite to know more.
Last edited:
Thanks, Scott, it verifies what i had often measured: some distortions adds, some subtract when you chain electronic parts. That explain a little why some preamp can be better with a given amp ?
Between electronic and Loudspeakers, they always add (according to few measurements), i believe because the delay of loudspeakers.
kgrlee, thanks for this nice moment. Will add your sister's cuckoo to my tests.
Last edited:
3-D would be pretty nice to have; until then I use plain 2-D observing changes with signal level.
Nonsense, unless "average" has a different meaning in your language than in English. Same thing with the "valid only for periodic signal components"; how many times do I have to knock down that old wives tale before it sinks in? Any non-infinite sequence may be formally treated as periodic- you know that very well, why keep repeating stuff you know is wrong?
Again, nonsense. A short impulse has lower energy than (say) an MLS, but not only is there no reason to avoid other stimuli, even a short impulse can be detected and transformed at any arbitrarily high resolution. You know this very well, why repeat stuff you know is wrong?
Again, exactly so. The most mundane set of electronics can be dramatically improved from the subjective point of view if all these types of issues are dealt with; it's not the obvious, easy to measure with an expensive instrument on a one time basis, distortion that's the problem , it's the scurrilous, now it's here, now it's gone variety that are the real headache, literally at times.
First you need to come to the view that audible distortion doesn't nicely spring to life every time you're in the mood to measure it, and it might just disappear as soon as you hook up the test lead; perhaps for the simple reason that the underlying problem was a static charge somewhere, which you just nicely tamed by touching the gear.
I've only spent 25 years fighting all this sort of "nonsense", and I hear the effects of it in nearly very normal hifi setup ...
Frank
Last edited:
It is obvious, Frank. Think of wires that carry currents as real wires, with inductance and resistance, and of capacitances between hem, and you will be happy designer.
Yep, trouble is, that there is more to it then that again, because all the parts are made of physical materials, which have fun adding their electrically related behaviours. Which can take some human significant time to audibly manifest; in years gone bye I could nominate the number of minutes the sound would remain stable, before the quality began to degrade ...
Frank
Well said Frank!
I'm glad I'm not the only one who has been frustrated in the past by this.Us old fogeys take a belt and braces attitude to stability, RFI bla bla cos we know how difficult it is to track down.
Yep, people working with MOSFETS are paranoid about the slightest arm jiggle blowing up their precious acquisition, yet people are happy draping cables against each other, and other materials, and then revving up the system, meaning air and other vibrations are setting everything off in a fine dance. Any wonder then that different materials used on cables and such make a difference ...
Frank
Last edited:
You make my joy.
I only listen to music when my room is at 21°ct, 75 mm of mercury air pressure, and 50% of humidity. Of course, my loudspeakers are threated with garlic.
I also noticed that, when i change CDs, there is a change in the sound.
By the way, this had been threated ad nausea here:
http://www.diyaudio.com/forums/everything-else/193100-speaker-cable-myths-facts.html
Last edited:
Of course it has, and where the whole thing has got out of control is because silly money is paid for the stuff, and silly explanations abound. Once you get inside of nice, lumpy electronic parts people stop sniggering about materials used and nod sagely about why certain things have to be considered. Yet, once you go outside the box and stretch the component out then all bets are off: anything, in any configuration will behave identically to anything else.
Yes, because cables are components: give me a piece of wire, surrounded by insulating materials, etc, that exhibits null electrical properties in every respect, and I'll give you a winning lottery ticket ...
Frank
Last edited:
yes it is.
Yes. Its in the linear math/calculus being applied to non-linear waveforms... not the appl. of averaging (multiple passes eg) So long as the sampling time is short enough, all is OK. To what accuracy do we need.
Some while back I said I could hear distortion to just below 0.1% -- using speakers with about that same level - so masking prevented me from finding my true lower limit. Had I done that test with the Quads and thier very low distortion, I should be able to detect harmonic distortion lower than 0.1%. So, maybe somewhere between 0.1 and .01% For Me in actual testing. No virgins participated.
Now if we are to design for that as a detectable limit .... I would do it 10X better to be sure no one could ever hear anything. BUT - I am a systems guy and if I naturally think about the entire chain from recording to playback, I would want another X10 reduction, so at the end, through it all, nothing will be heard from harmonics generated.
NOT that I can hear one item at .001% ! Lets be clear on that.
However, if the detectable threshold is lower for others than my own, then we need the lowest we can make at every stage for the system chain to be transparent or 'prefect'. Thx - RNMarsh
Yes. Its in the linear math/calculus being applied to non-linear waveforms... not the appl. of averaging (multiple passes eg) So long as the sampling time is short enough, all is OK. To what accuracy do we need.
Some while back I said I could hear distortion to just below 0.1% -- using speakers with about that same level - so masking prevented me from finding my true lower limit. Had I done that test with the Quads and thier very low distortion, I should be able to detect harmonic distortion lower than 0.1%. So, maybe somewhere between 0.1 and .01% For Me in actual testing. No virgins participated.
Now if we are to design for that as a detectable limit .... I would do it 10X better to be sure no one could ever hear anything. BUT - I am a systems guy and if I naturally think about the entire chain from recording to playback, I would want another X10 reduction, so at the end, through it all, nothing will be heard from harmonics generated.
NOT that I can hear one item at .001% ! Lets be clear on that.
However, if the detectable threshold is lower for others than my own, then we need the lowest we can make at every stage for the system chain to be transparent or 'prefect'. Thx - RNMarsh
Last edited:
FFT techniques, and Farina's method is very much an FFT technique, can be used to investigate non-linear phenomena.
A classic case is our assumption that speakers are Linear Time Invariant. They most certainly aren't. It's just convenient for us to believe that most of the time.
But Farina's method can be used to get a very precise handle on this. Just try doing it with a 1s log sweep and comparing the response & THD results with doing the same with a 14s sweep as you would have with a B&K 2307 chart recorder. Somewhat to my surprise, this is very repeatable.
BTW, though I champion Farina's method for looking at large evil distortions of speakers, AP like it for 1 ppzillion work on their new box.
- Status
- Not open for further replies.