xpert said:
Hmm ... I'm not too sure. Laurie Fincham was one of the first to use FFT's, but he was also closely collaborating with the BBC since KEF made custom-spec drivers for some of the BBC monitors (like the LS3/5A). The BBC and KEF shared a mutual interest in time-decay colorations, and this started with D.E.L. Shorter's publications in Wireless World in the Sixties.
I would put the earliest use of CSD-like measurements in Shorter's Wireless World articles of what he called "buried resonance" and their relation to audibility. Shorter found that resonances that were too small to have much effect on the FR curve (less than 0.5 dB ripple) could still be audible as a coloration - and that the "buried resonance" chopped-sinewave technique would reveal their frequency and Q. Once revealed, the driver could be modified to remove them (dustcap resonances et al).
Ten years later, Laurie Fincham at KEF was using the more direct FFT method - and both the BBC and KEF were interested in minimizing time-storage artifacts. This led directly to the BBC's work on polypropylene cones in the late Seventies.
mige0 said:
True, it was your effort in processing the data. Thank you.
"hair splitting".
As freq.s increase hair splitting is "where it's at".
..as far as the full bandwidth is concerned, there are significant limitations by not having comparisons to the compression driver without waveguide.. however:
The driver appears to have its fundamental resonance just below 2 kHz (..and frankly looks like the DE 10, not the 250). Of course the waveguide could be "pushing" that resonance farther up in freq.. and in fact I think at the very least it is as you go off-axis. There you can see the ripple prominently near 5 kHz 15 degrees off axis.
It's also interesting that the waveguide appears to be suppressing (in amplitude), and extending (in time) the fundamental resonance of the driver as you go farther off-axis. It also extends (in freq.) the ripple, while again marginally suppressing it (in amplitude).
IMO it rather clearly is intended to be listened to near its 30 degree axis. It not only represents a better low-pass response to be filtered, but it also suppresses the fundamental resonance to a more significant level for its freq. while incurring less objectionable ripple (in comparison to the 15 and 45 degree positions). (..and yes, I realize that "significant level" and "less objectionable" aren't defined.. but that would take a bit more writing.)
The real problem with all of this is:
1. The response shown does not include its CD compensation low-pass filter. When you apply it those upper freq. ripples will become more pronounced.
2. The display is set for 25 db down, not 30 db or 36 db. Again, additional amplitude will display the resonances better.
Rick Miller said:
Hi Rick thanks, but its not only Wilson Audio that is interested in getting a handle on that subject.
Below a link to an early reserch done by Friedemann Hausdorf from VISATON published in the ELEKTOR magazine 1988:
http://www.exdreamaudio.de/?Akustik:Geh%E4use
The big advvantage is that Hausdorf also documents his measurement set up. Sadly the scaling of the CSD Y-axix isn't clear.
Scroll down the pix to
"Bild 12b" whis is for marble
or to
"Bild 16b" where you see the benefits of bracing to the opposite wall for simple 19mm / 3/4"chipboard
"Bild 3b" shows the same simple 19mm / 3/4"chipboard *without* bracing for comparison
Michael
ScottG said:
True, it was your effort in processing the data. Thank you.
"hair splitting".
......
The real problem with all of this is:
1. The response shown does not include its CD compensation low-pass filter. When you apply it those upper freq. ripples will become more pronounced.
2. The display is set for 25 db down, not 30 db or 36 db. Again, additional amplitude will display the resonances better.
Yes, but I wouldn't go too deep into the very details as we don't have data of other horns to compare with done under same measurement circumstances and – even more important – the ripples are small - not really worth to complain IMO as long as there are no pronounced decay tails that can be seen..
But - I did some more processing on Earls IR files - as said - Acourate software is pretty powerful stuff:
What I did, is to explore what we would see on axis with a XO optimised for 30deg response.
An externally hosted image should be here but it was not working when we last tested it.
Well, as the 30deg response gets close to perfect (as should be)
An externally hosted image should be here but it was not working when we last tested it.
the on axis response gets more rugged also showing some decay tails in CSD.
I don't wanna show this pix' as they might be a little bit "overcooked" and also – not knowing the exact circumstances of measurement - there might be (over) compensation for diffraction effects not related to the OS wave guide .
But all in all I feel save to say from the IR files got, that even OS isn't constant directivity (be this good or bad) when it comes to a veeeery close look.
Compared to my OB measurements – they both are pretty close regarding decay time range and smoothness - not that much better in this regard one or the other.
Though OS does *not* show the clear shift in directivity as does OB.
Maybe we could put it that way:
in the macro view OS is CD
in the micro view OS isn't CD (I doubt any horn is anyway)
Michael
The measurements Michael shows feel like they have a realistic level of detail. I could be off base, but I'm skeptical of trying to dig fine detail our of these type of measurements. I'm no mathematician, but I have spent a lot of time doing analytical work in biochemistry. And it strikes me that the measurement errors for the typical set up, have to obscure much of what can be read into some of these detailed CSD profiles. Most of what I see here are single measurements of two or more different conditions. Looking at the comparison that BudP cited, for instance: how much different are they really? Would you see as much difference by moving the mic a few cm? Can you duplicate the measurement on a different day to that level of detail? How much is measurement/calculation artifact, how much represents real differences?
Sheldon
Sheldon
Sheldon said:
As for my own measuremets - I feel pretty confident - within the limits outlined.
gedlee said:
Agree - as said several times..
Maybe we should say results are almost stable *if* we would repeat at the exact coordinates for the speaker and the mic.
even one or two cm / 1" is quite a difference if we look into "micro directivity" of a speaker plus whatever diffraction "artefacts".
From that - smoothing isn't necessarily a bad thing when used to get an overview - rather than to hide small but irritating errors by intention.
As said it does not make sense to do equalisation down to the last dB of every ripple by measurement.
To tune to sub dB values is more a matter of good taste to get the tonal balance right..
Michael
soongsc said:
Aren't Earl's files 44.1KHz? It seemed that sample rate based on the text file you posted a while back. If so, what's with the 48KHz sample rate setting here?
I checked with Earl before publishing the processed results - he thought it's rather been 48kHz sampling rate
But anyway - it would only be a mere freqeuncy shift of less than 10% - nothing that interests me in this frame of comparison OB versus horn (OS)
Michael
mige0 said:
My recent analysis *only* concerned your results. Basically, it is what it is - look at it for what's shown or don't look at it at all. (..of course you could say the same about any measurement and data "view".)
IMO all that I discussed would in one way or another be audible to a critical listener in a good system. Are the resonances worth complaining about? To some, yes. Would most of it be a big deal? No.
What would likely be most objectionable would almost certainly be the on-axis waveguide + driver resonance (that 1-2 kHz range, particularly as extends to 3+ kHz on axis). Even if it is 30 degrees off of the listening axis, I still think the on-axis portion would be audible.
Again, the net result isn't bad - but it isn't particularly good (and certainly not "stellar").
BTW, Zaph had the DE10 with ME10 waveguide posted here:
http://www.zaphaudio.com/tidbits/
As far as a comparison is concerned, I'd agree that without CSD at different axis's that it is difficult to make a comparison *except* for the on-axis response. However, note the freq. response plots off-axis and how they don't have a great deal of ripple - which likely means that their CSD probably has fewer problems.
With this You emphazise that You have concerns. But we don't know where they come from. If I don't bring those effects to my perception, what is the reason? Bad system, non critical listening?
Sorry, there is NO resonance. Take it as said, no resonance. None. What You see is a group delay that comes with bandwith limiting. In this case obviously it is a highpass. Group delay is unavoidable with any common bandwith limiting, may be a horn (mechanics) or crossover (electrics). Calculus and nothing else: No resonance. Not bad. No concerns. No perceptibility. No degradation of zound.
Your misconception is quite common with Cumm. Spectr. Decay. You should leave that CSD reading behind.
so long
Interesting aspect. Just to make this comprehensible for my ignorant view - how do I have to understand the above?xpert said:
- CSDs don´t show resonances at all
- in this specific case the CSD doesn´t show a resonance but a GD
- all resonances must be seen as GD
- ?? else.
I would appreciate if you could make this clearer.
Rudolf
ScottG said:
There are good reasons to take CSD with a grain of salt.
First its a highly processed result in order to make obvious what's hidden (well, for the most of us
) and not necessarily telling "the whole truth and nothing else than the truth". And also not telling about limits where calculation is valid or not.
Second any graph has its limits where we have to decide between "truth" and what we can grasp at most. If you look at the trace below it should become clear what I mean – we are lost in "true" details here:
An externally hosted image should be here but it was not working when we last tested it.
Third, if you imagine the time span of sub 1msec decay - and translate this into how many cycles that my be for the frequencies of interest you easily see that it doesn't make sense to crank resolution (and conclusions) too far – no?
Michael
xpert said:
I'm not entirely sure what you are requesting here. (..and I realize that English as a second language isn't exactly easy.)
I do think I understand what you are referring to with group delay (or at least partially). Presumably you mean that phase rotation from filtration (mechanical or electrical) is inducing time decay problems. Even if I were to accept the laughable argument that neither the compression driver nor the waveguide exhibits any resonant behavior; I have *never* seen simply phase-induced time decay that *looks* like what Michael presented. Also, despite what others may report - I do think that phase-induced time delay can be audible (or "perceptible").
So no, I think you are incorrect on this point.
I do think it would be nice though if we had phase and impedance shown as well.. but you work with what you are given.mige0 said:
I don't discount this.. but at the end you have to make a decision (with respect to any measurement):
Do I use it or not?
..and if you do, after a while you will probably start to get a "feel" for the level of accuracy of those results and perhaps even some idea of what those results do subjectively (to the sound).