janneman said:
Could you kindly quote that 'many here' ??
Do not forget you speak about polypropylene, so please do not quote mylar or ceramics.
Johnloudb said:
That is true, but meters do that in a predictable and consistent way 😀
But seriously, I sort of get tired of all those who artificially try to divide us in two black-and-white camps. There are no strict meter readers or golden ears. JC also said he uses a combination of test equipment and his ears. I myself have a tendency to concentrate on measurements, but I've been surprised more than once by my own designs once I started listening seriously.
Trying to separate us forcefully and artificially only serves some unknown (and sometimes known) personal agenda by those who do it, but is pretty unrealistic to begin with.
Jan Didden
The question that should be asked is how do we hear it? Because, even if you can't hear it, there are too many audiophiles who know that cheap ceramic capacitors sound much worse than polypropylen caps.
alansawyer said:
I heard Jethro Tull once head at the Delta Center (now Energy Solutions center) the sound was horrible, despite what looked like a really nice sound setup. Acoustics in that place are just horrible. My ears would have been ringing but I put a bit of ear foam in my ears to attenuate the sound. I also much preferred the outdoor concerts I went to.
I'm sure that any half decent audio designer relies on the instruments and measurements, but the ears guide trough design process and tell if what we measure as better has positive influence on sound.
PMA said:
Pavel, I was maybe slightly generalizing in the interest of brevity. I assume that my meaning is clear.
Jan Didden
vuki said:
Ahh, now we're getting somewhere!
First of all, it is clear that ceramics do much worse, and DiffMaker readily brings that to the open.
Secondly, that 'too many audiophiles know...' is in fact 'too many audiophiles have been told ...'.
A guy in France once said: if 50 million French believe a wrong thing, it is still a wrong thing. So, while I agree that ceramics are bad, it's NOT because 'too many audiophiles know...'
Jan Didden
Jan,
maybe I am confused or did not read everything. IMO, it is difficult to distinguish polypropylene from direct connection, in case that the cap is not heavily loaded, and does not change freq. response.
Mylar or non-linear ceramic is audible, that was my point.
maybe I am confused or did not read everything. IMO, it is difficult to distinguish polypropylene from direct connection, in case that the cap is not heavily loaded, and does not change freq. response.
Mylar or non-linear ceramic is audible, that was my point.
Hi all,
Jan emailed me about this discussion, so I thought I'd chime in with what (little, in some cases) I may know.
On DiffMaker and linear distortions: For frequency response differences, it depends on how the DiffMaker test was done. One of the options is to use "EQ", which basically first measures the frequency response of each path using a low level chirp, and then modifies the "Compared" track to compensate for the differences (the Reference track is not modified at all). If all that worked well (which depends on noise, avoidance of clipping during recording, other differences) then frequency response effects should be removed. I used this on the example capacitor files, since any difference in the capacitance values would change the frequency response. The EQ process is meant to take out any effects that could be characterized with a measured impulse response.
The Sousa band thing was added at about 50d to 60dB down from its original recorded level. (As for whether it's 'really' there, well, to hear it you had to do the extraction yourself, so where else would it have come from?? Try using the same track for Reference and Compared and see that the band is gone). I agree that it does sound surprisingly good, apparently from the "power of dither" as someone mentioned. Of course I carefully adjusted the level to make sure the result was dramatic (why else a Sousa band?) while still being very much masked before "Diffing" was applied.
Residuals: DiffMaker, although a digital deterministic algorithm, operates usually on recordings of analog signals. So there will always be noise included. Also the DM process is done by successively narrowing down to the best choices of gain and time position to get the deepest null after subtraction. How well it can narrow-in usually depends on how much noise is involved (which it can't cancel, but will still "try" and in so doing, the best match for removing recognizable sound will be slightly corrupted).
Which is all to say that there will usually be a weak version of recognizable signal left after Diff-ing, which you can hear if you turn up the gain enough. Even if there was no relevant difference other than the noise. That's why the emphasis on comparing "at normal listening gains" -- the fair comparison is to play the Reference track at the level/gain you'd normally listen to it with, and then without changing the gain, see if you can hear anything in the difference. That seems easier to do with headphones (I did a demo at AES last fall and played a difference file through the speakers that I thought would be clearly audible, but it turned out no one could hear it that way...).
I can't say for sure what the Z5U cap is doing. I suspect a voltage-sensitive capacitance effect, but it might be piezoelectric or DA, too, I guess. I'm fairly certain it isn't just operator error or noise-induced, since the results both with Z5U and polypropylene were repeated several times.
Jan emailed me about this discussion, so I thought I'd chime in with what (little, in some cases) I may know.
On DiffMaker and linear distortions: For frequency response differences, it depends on how the DiffMaker test was done. One of the options is to use "EQ", which basically first measures the frequency response of each path using a low level chirp, and then modifies the "Compared" track to compensate for the differences (the Reference track is not modified at all). If all that worked well (which depends on noise, avoidance of clipping during recording, other differences) then frequency response effects should be removed. I used this on the example capacitor files, since any difference in the capacitance values would change the frequency response. The EQ process is meant to take out any effects that could be characterized with a measured impulse response.
The Sousa band thing was added at about 50d to 60dB down from its original recorded level. (As for whether it's 'really' there, well, to hear it you had to do the extraction yourself, so where else would it have come from?? Try using the same track for Reference and Compared and see that the band is gone). I agree that it does sound surprisingly good, apparently from the "power of dither" as someone mentioned. Of course I carefully adjusted the level to make sure the result was dramatic (why else a Sousa band?) while still being very much masked before "Diffing" was applied.
Residuals: DiffMaker, although a digital deterministic algorithm, operates usually on recordings of analog signals. So there will always be noise included. Also the DM process is done by successively narrowing down to the best choices of gain and time position to get the deepest null after subtraction. How well it can narrow-in usually depends on how much noise is involved (which it can't cancel, but will still "try" and in so doing, the best match for removing recognizable sound will be slightly corrupted).
Which is all to say that there will usually be a weak version of recognizable signal left after Diff-ing, which you can hear if you turn up the gain enough. Even if there was no relevant difference other than the noise. That's why the emphasis on comparing "at normal listening gains" -- the fair comparison is to play the Reference track at the level/gain you'd normally listen to it with, and then without changing the gain, see if you can hear anything in the difference. That seems easier to do with headphones (I did a demo at AES last fall and played a difference file through the speakers that I thought would be clearly audible, but it turned out no one could hear it that way...).
I can't say for sure what the Z5U cap is doing. I suspect a voltage-sensitive capacitance effect, but it might be piezoelectric or DA, too, I guess. I'm fairly certain it isn't just operator error or noise-induced, since the results both with Z5U and polypropylene were repeated several times.
That's partly what the Brahms Choir/Sousa Band was meant to be -- the diff part is the added-on-purpose Sousa Band part. I was trying to show DiffMaker pulling out of a pair of recordings something wild that couldn't be heard otherwise.
janneman said:
Of course, not. That part was for you 😀 I thought you don't believe in audible difference between caps, so I guessed you must believe in those mass GE-audiophile disillusions. 😉
Sorry Bob, it was Jan that I hoped would 'get up to speed'. I only wanted to add to your statement about Bob Pease, because I know that he did not personally develop that model. I am touchy about credit giving.
Jan, your question answers itself, if you just listen to your test. Dielectric absorption creates an error signal that tracks the music, usually with a bass boost. As I have said repeatedly, Walt Jung and I learned this more than 25 years ago by using analog differential comparison. This is what we normally call 'linear distortion'. Of course, there can be non-linear distortion there also, BUT it will be at a lower level, for the most part, and almost unmeasurable with polystyrene, polypropylene, or Teflon.
Jan, your question answers itself, if you just listen to your test. Dielectric absorption creates an error signal that tracks the music, usually with a bass boost. As I have said repeatedly, Walt Jung and I learned this more than 25 years ago by using analog differential comparison. This is what we normally call 'linear distortion'. Of course, there can be non-linear distortion there also, BUT it will be at a lower level, for the most part, and almost unmeasurable with polystyrene, polypropylene, or Teflon.
The noise is in both tracks (the Reference and the Compared), so it isn't "noise" in that case but part of the overall signal. The Sousa part is only in one (still no one has hazarded a guess which, BTW). So the noise common to both tracks cancels out but the band plays on.
Hi Bill
Would it be possible to make the mixer section in the diffmaker so that we could listen to the original track and the diff signal at the same time and be able to adjust the volume of the diff signal?
Cheers
Stinius
Would it be possible to make the mixer section in the diffmaker so that we could listen to the original track and the diff signal at the same time and be able to adjust the volume of the diff signal?
Cheers
Stinius
bwaslo said:
I don't understand: if the band is buried in the noise of Brahms it cannot be audible. Where am I wrong?
I'm asking because I can hear the band when boosted by >9dB.
- Status
- Not open for further replies.