Ah, the pizza's LostAngeles' idea, not mine. Not a bad idea, though.
I summarised the post by using every 500th word.PianoTeacher said:Explain high it is I CDs subjectivists miss and there exist their and 2005
Winny said:I summarised the post by using every 500th word.
I think this improves it no end.
Winny
PianoTeacher said:Those of you who read my explanation of the improvements of the resolution of digital recording may have noticed rather bad mistake in typing, repeated by me stupidly.
When describing the sampling rates, I typed
96/24
when I should have typed
24/96
Thus, we progressed in quality of resolution from 16/44 (16 bits/44.1 kHz sampling rate) to 24/96 (24 bits/96 kHz sampling rate); I reversed the parallel structure with my error.
I missed that in my proof-read. Surely I am not responsible! A cosmic particle must have hit my brain.
PianoTeacher
When describing the sampling rates, I typed
96/24
when I should have typed
24/96
My mistake.
PianoTeacher said:Those of you who read my explanation of the improvements of the resolution of digital recording may have noticed rather bad mistake in typing, repeated by me stupidly.
When describing the sampling rates, I typed
96/24
when I should have typed
24/96
Thus, we progressed in quality of resolution from 16/44 (16 bits/44.1 kHz sampling rate) to 24/96 (24 bits/96 kHz sampling rate); I reversed the parallel structure with my error.
I missed that in my proof-read. Surely I am not responsible! A cosmic particle must have hit my brain.
PianoTeacher
The Central Scrutinizer claimed
I've won this debate.
PianoTeacher said:LostAngeles crafted a hilarious contribution:
>
BPSCG is the nit-picker, Timothy can be the suspicious one and Beleth can be the more suspicious one, Piano can be the wise, rambling man, Lisa Simpson can be the evil witch (I remember your CD suggestions, woman.), KRAMER can be the hot, but bubblehead daughter, and webfusion can be the wacky neighbor.
<
I showed my wife "Piano can be the wise, rambling man", LostAngeles. She agreed in part.
Except for "wise".
PianoTeacher
PianoTeacher said:Some quick observations based on about 2 reads:
>You need eleven bit-by-bit identical disks.
We need to have everybody stipulate that there is no such thing as "bit-by-bit identical disks" since there are actually huge error fluctuations (I can't begin to recount how many CDs I have with occasional ticks). Obviously we have to use a little scrutiny here. Some disks have been re-pressed and actually re-digitized from analogue masters, without your having been aware of it. I cited an example of at least one famous recording earlier; I know of others. I think it might be best to set cutoffs:
1. It would be best to stick with DDD material. It eliminates the chances that a title had been remastered at different points in time from an analogue source tape by a not absolutely identical process (and not so indicated on the jacket) to correct problems. The first run of Reiner's Don Quixote went thru that process but you can't tell from looking at the disks.
2. By using modern recordings, even 24 bit mastered ones, we tend to reduce the audible tape hiss that is on transfers of old analogue tapes. I remember Wellfed commenting about S/N ratios, and wondered if he had been trying to identify alleged GSIC effect by listening to the noise platform. The less chance of audible hiss in the recording, the more we can concentrate on hearing actual musical sounds: if they are better, or similar.
3. *Probably* the best way to tend to make certain that the disks are from the same optical master and pressing run, is if they are found together in the bin at the record store. I have argued that this proves nothing (it does) but it could suggest that the disks are more likely to have manufactured equivalently than disks acquired at different places, months or years apart.
4. I just pulled two identical titles out of my wife's cabinet, copies we bought to give to students. The record number on the spine was the same (Laserlight 15 630); the jackets were identical; but the disks themselves were very different (though it took more than just a glance to notice this.) For one thing: the numerical string engraved in the plastic around the center, which requires you to hold the disk under strong light and sometimes even use a magnifier, was ENTIRELY different.
One said: 15630 S9C09E
The other said: 15630 L4A26F
The numbers were not printed with the same style.
There was one other subtle difference. The first disk had the word "digital" under the trademark printed in yellow. The second disk had it printed in black.
I have no idea what this signifies without knowing Laserlight's supplier's optical code but I very much suspect that this indicates two different suppliers, or perhaps two very different runs that *may* have been made by different optical masters.
So, this shows how elusive "bit by bit identical" is; even how hard it is to control for the same manufacturer.
Another example: EMI changes mfr's often. I've purchased EMI pressings that say, near the label, "pressed by Nimbus"; or "EMI UDEN" or "EMI JAX". Nimbus is a well known company in the UK hinterlands; and the others are factories elsewhere (I learned via Google that JAX is the Capitol plant in Jacksonville, Florida and UDEN is an EMI plant in Uden, Holland.) Chances are, an actually different optical master disk has created the copies. One can also wonder if the digital tape was from an identical copy medium (U-matic, or other format) our how it might differ. We tend to think "bits is bits", but there are error- correction algorithms all over the place and they introduce (generally imperceptible) uncertainties.
If pressed by different factories in different parts of the world, HOW SIMILAR ARE THEY?
I remembered, while writing this, that I had two copies of an EMI disk of Mahler, by Simon Rattle. My original old copy from c.1990 had failed: the label side became horribly sticky. Fearing losing the album I bought a new one last year. The old one still worked so I put it in the car. I just got them out and examined them: different codes, different makers: one Nimbus (the old one), the new one JAX.
Control for these kinds of variables by trying to get identical codes, identical makers, and labels that have no differences that show up when you compare them.
----------
Now, you see how just this ONE element becomes important, especially when one million dollars is riding on the outcome.
>
It's okay if a rep for the applicant is present, so long as this rep does not touch the equipment nor the coin, nor contacts the applicant in any way between that point and the time of the test. If any of these things happen, the test is invalidated.
<
I'll have to break that down; but I was uncomfortable about some of it. The applicant's representative is there for what reason? To prevent JREF from fudging? But how do you control for the rep lying later? I suspect that two neutral referees who don't know each other would be more foolproof; how far does JREF -- or the applicant -- wanna go?!
>The treatment machine is borrowed/provided by JREF, set once prior to the initial treatment, and not adjusted until the treatment is done.
<
Treatment, according to what I've seen only in the web articles and reviews (NOT the actual copy of instructions with the product) implies that all you do is put in the disk and press play for 2-3 seconds. "Not adjusted until the treatment is done" may not be necessary to stipulate. I'd examine the instructions (everybody of course realizes that I don't put any credence in them at all because of my bias against the device; but that's irrelevant. They MUST be followed PRECISELY!) and make SURE of the exact method for treatment, and time it precisely.
An aside:
Now, I get red flags here (not because of the protocol, but because of the logic of the treatment process.) How in hell can you "treat" the CD so casually, with such looseness, not taking into consideration the effects of the metal case, EXACT criticality of the GSIC placement, internal structures, etc. As I showed a while ago in my speculative analysis of what this might have as a power supply, the "field" that is alleged to be involved by the treatment, would have to be extremely strong so that it always had enough effect on the CD, under varying conditions. And how could you put that kind of power supply under the "dot"? It's silly.
But, no harm in trying to do the protocol testing honestly; at least no one can claim that JREF people were as forgetful of "reality" as the people who wrote the instructions for the GSIC, assuming it does (as they say) involve some kind of field: they forgot inverse-square law!
On the other hand, we give them the benefit of the doubt that it is not a fake, but that the clod who was asked to write up the instructions made a blunder. I've seen it happen, believe me!
>...transports the CDs to the room where the applicant awaits with their preferred equipment setup. (This can include packaging and mailing, in terms of a remote test.)
<
Would we really want to consider the possibilities of a remote test? (Well, maybe so: and have Ingo Swann remote-view it for us.) I should think this introduces dangerous complexities and chances for chicanery, confusion, and mistakes. Avoid remote tests.
>The applicant then receives all eleven CDs, and determines by any means they desire which CDs have been treated and which have not.
<
Aye, there's the rub!
I started parsing the ABX material, but stopped cold when I realized they are discussing tests done over time; there is also no consistency (in this first rough draft) about the number of disk titles that you are testing (how does one find out if you being reasonable and not using too few, or too many?), WHAT is being played, for how long, with what settings, who makes the settings, who can change the settings; so many other things.
I wrote an objection to Wellfed's proposal in which I suggested a formalized play list. Then I checked the next day and could not find it; I think that through a blunder I never posted it.
If so, I summarize here that the type of test alternations I did were with a formalized list of musical selections, and with defined time windows: i. e., "track two, 0:00 to 0:37."
The mind gets awfully confused and forgetful if you play minutes' worth of music, and then go back and play minutes' worth again. Your falses will go way up.
Furthermore this allows you to control repeatability. What if you wanted to have a non-blind preliminary test, and then do it again, as closely as possible, double blind? You probably would want to use the same musical examples (of course after scrambling the ABX order) but for variety, and not tiring the ear, you might scramble the line items in the playlist.
Again, I don't know what the ABX experts think about this: it's one more detail that I would investigate.
Below, here's where the applicants can have an input, if JREF thinks it acceptable. Wellfed tried to select music he thought could help him discern the effect. I find nothing wrong with that idea at all: because in testing audio, I can hear certain nuances better with certain music. Noise modulation effects will show up well with a solo acoustical guitar, or an acoustical piano in a dry hall. But they are hard to discern with a very complex mix densely recorded. Would not it be best for the claimant to know the music on the CDs? And early on for him to select parts of neutral copies of the CD titles -- never any of the ones in the test; those are kept separate! -- and for all to decide on what the lengths of the passages should be, in consultation with the protocols of tests that show positive ABX discriminations. I haven't thoroughly studied ABX from original sources, only digests and reports on the Net and in magazines. Best consult the authorities about *how long the examples* should probably be.
No mention here, either, of using only one control player or somehow randomizing three ostensibly identical brands and models so that the disks can be pre-cued and ready to go for a quick switch. I am under the impression, from what I've read, that ABX'ing is at least sometimes done that way so that the changeover is fairly fast, keeping the impressions fixed reasonably well. I know from personal experience that if I take a disk out of the player, and put in another copy of the same title, and take the time to cue it up, I lose track of differences...and I've tested for audible differences of CD against CD-R copies; differences between machines with normalized output adjustments having been controlled, and so forth; and in the tests (simple A-B ones) the switchover was quick that way and I could *really* discern differences. Often.
But, I'm jumping the gun. I am offering input, not critiquing. As I said, I don't consider myself qualified; I'm ignorant of the *best possible ABX protocol*; but one can find out. Unfortunately, tonight: no more time. And in this case, I'd rather just critique than craft the protocol. I don't want to be responsible for designing some fundamentals until I know more (as I said, I need to get up to speed on testing post-1980.)
And the last two paragraphs of the ABX material left me very bewildered.
I think that I would advise that Googling ABX, and looking over one or two articles, may be too superficial a method of research. Often the best and most useful articles about testing are NOT free on the Net; the magazines don't post them. Or, they list the articles and a precis in their online sites, and none of the details. I learned a lot from magazines that only use the Net for promotional sales of issues, and subscriptions, putting glossy stuff on their websites only for attracting customers. Sure, you get an impression you are getting their articles, but not necessarily the good stuff! Check Absolute Sound, which has some very fine information at times, and some off-the-wall stuff at others. Occasionally I've read articles and reports of tests that were so scholarly I was amazed.
Other in-depth stuff might be available from high end audio manufacturers of highly respected products. Some are rationalists and use metrics -- and wisely. I've read test reports from manufacturers that told me lots more about real repeatable testing than most of the more superficial articles by audio critics or journalists, who often do a very good job of summarizing things; but they leave out details. Or their editors ruthlessly remove them.
PianoTeacher, the rambler
No Problem!alfaniner said:I'm sure someone can summarize it in a line or two.
alfaniner said:Sigh, I don't even bother anymore. Is there anything worth reading in the above? (Hell, if PT can waste bandwidth, so can I!) I'm sure someone can summarize it in a line or two.
Look at how much effort it takes to bore me.
Yep, I read the whole bloody post. Your summary is pretty much accurate.-42- said:Summary:
People make CDs.
CDs of the same title often have differences, due to different pressings, possible corrections or made at a later date, etc.
An exact track and time should be used in the protocol.
Champ, you really need to take a brevity pill.PianoTeacher said:Some quick observations
PianoTeacher, the rambler
PianoTeacher said:OK. The point is clear.
I shall now cease posting to this forum.
Incidentally, I explained in my essay, posted earlier today, that the article was written for claimants, not the "Rational Logicians" who are on the forum.
PianoTeacher
Concise, but still too long.Winny said:I summarised [Piano Teacher's] post by using every 500th word.