John,
I wrote several comments here:
http://forum.stereophile.com/forum/showflat.php?Cat=0&Number=35333&Main=35279#Post35333
I wrote several comments here:
http://forum.stereophile.com/forum/showflat.php?Cat=0&Number=35333&Main=35279#Post35333
The question that seemed most important was how 'imaging' is achieved. Well, the BIGGEST mistake with the Levinson JC-2 was the asymmetrical crosstalk in the original motherboard layout. Mark just would not fix it.
We take care to have real channel separation.
We take care to have real channel separation.
thanks JC,
Now how difficult to organise a group buy of the JC2 and what level of discount on the retail price might be available?
Now how difficult to organise a group buy of the JC2 and what level of discount on the retail price might be available?
Was the ML crosstalk really that high? You know how bad is the vinyl crosstalk (-20 to -30dB), and people speak about superb imaging.
john curl said:
Do you have some figures we can see ?
Very bad is obviously not good, but exactly how much asymmetrical crosstalk is acceptable?
Magura 🙂
IF left talks to right, but right doesn't talk to left, and visa versa, then it's bad. The other thing is that many crosstalk specs start to collapse after 1KHz or so. This is not too good either, but not as bad as asymmetrical crosstalk.
Yes, this is obvious and clear. We were just interested in a MEASURE of assymetry. I guess the answer would be "as low as possible" 😉
Imaging is a curious thing. On the surface, it seems as though it would be simple--just keep the channels separate. Only that doesn't work. It's a necessary condition, but not sufficient in and of itself. If channel separation were the whole deal, then everybody would build dual mono stuff and all pieces would image equally well. But they don't. So clearly there's more to it.
Just as no one has ever managed to come up with a measurement to quantify imaging, no one has managed to invent one that reduces detail retrieval to a number. It's apparent that imaging has some relation to the direction of an echo and its duration. A moment's thought will show that since the level of the echo falls over time, anything that truncates that signal at some arbitrary level will deprive the image of some of its information. You might be able to point to the instrument by using first arrival wavefronts, but the less detail available, the more it sounds as though the instrument is playing in a smaller, deader space. The more detail, the larger the room.
Some posters appear to be adopting the all-or-nothing mindset that you see in, say, discussions about opamps. Their thought process appears to be that if opamps were as bad as some people claim, then the presence of even one opamp in the signal chain would render the music unlistenable. Since the music is obviously listenable, then opamps aren't bad. Er, if that constitutes logic where they come from, I don't want to live there. It's a question of deterioration, not a binary on/off process. Each opamp takes a little away. A decent discrete circuit takes some, but less away. If you're trying to preserve low level information, then everything counts. It's a no-brainer to use discrete circuits. But if you're going to attempt to resolve things down to -80dB or whatever the lower limit of imaging information happens to be on a given recording, then you need to do as little damage to the signal as possible. If you're attempting to reproduce a -80dB signal and the circuit in question attenuates that signal, or it drops beneath the noise floor (think not just of noise as a single figure, but also noise spectrum), or it gets destructively phase shifted, or it gets buried beneath distortion products that happen to occur in that same band of frequencies, then the imaging information is effectively lost. Attention to detail (play on words is intentional) is essential if you're going to image well. And noise. And phase problems. And distortion products...ad infinitum...
Grey
Just as no one has ever managed to come up with a measurement to quantify imaging, no one has managed to invent one that reduces detail retrieval to a number. It's apparent that imaging has some relation to the direction of an echo and its duration. A moment's thought will show that since the level of the echo falls over time, anything that truncates that signal at some arbitrary level will deprive the image of some of its information. You might be able to point to the instrument by using first arrival wavefronts, but the less detail available, the more it sounds as though the instrument is playing in a smaller, deader space. The more detail, the larger the room.
Some posters appear to be adopting the all-or-nothing mindset that you see in, say, discussions about opamps. Their thought process appears to be that if opamps were as bad as some people claim, then the presence of even one opamp in the signal chain would render the music unlistenable. Since the music is obviously listenable, then opamps aren't bad. Er, if that constitutes logic where they come from, I don't want to live there. It's a question of deterioration, not a binary on/off process. Each opamp takes a little away. A decent discrete circuit takes some, but less away. If you're trying to preserve low level information, then everything counts. It's a no-brainer to use discrete circuits. But if you're going to attempt to resolve things down to -80dB or whatever the lower limit of imaging information happens to be on a given recording, then you need to do as little damage to the signal as possible. If you're attempting to reproduce a -80dB signal and the circuit in question attenuates that signal, or it drops beneath the noise floor (think not just of noise as a single figure, but also noise spectrum), or it gets destructively phase shifted, or it gets buried beneath distortion products that happen to occur in that same band of frequencies, then the imaging information is effectively lost. Attention to detail (play on words is intentional) is essential if you're going to image well. And noise. And phase problems. And distortion products...ad infinitum...
Grey
I don't think that there is any magic solution for imaging, but good, low to no compromise engineering certainly helps. Most layout people just don't keep their channels away from each other.
I assume it is a complex of all the known parameters (channel separation is not the only one) and some of the uknown. Also, it differs with global NFB and non-global NFB designs.
What would be a good xtalk figure so that the xtalk from the signal is, say, an order of magnitude better than the ears? What is the xtalk from a signal close to one ear to the other? I know there is severe xtalk across a face, but can't remember the number. Was pretty horrible though.
Jan Didden
Jan Didden
h-cat ?
APPARATUS FOR IMPROVED AMPLIFICATION OF AUDIO SIGNALS WO9941831
If this is considered inappropriate or getting to off-topic please delete.
Regards
James
APPARATUS FOR IMPROVED AMPLIFICATION OF AUDIO SIGNALS WO9941831
If this is considered inappropriate or getting to off-topic please delete.
Regards
James
Please give us the rest
and amplified so that they can be projected by ordinary audio equipment to produce a holographic wavefront giving the illusion of signal sources that "hang in the air" instead of being sensed as emanating solely from a speaker.
and amplified so that they can be projected by ordinary audio equipment to produce a holographic wavefront giving the illusion of signal sources that "hang in the air" instead of being sensed as emanating solely from a speaker.
I think *benign* crosstalk is not a hinderance to imaging... if it were, vinyl wouldn't work. Benign now surely means symmetrical, and as true signal correlated as possible -- read: low distortion (ie, *no* crosstalk from halve-wave rectified class B supply currents and stuff like that).
I imagine (sic) a good way of maintaining imaging is to not process the signal in the L/R domain but rather in M/S. This keeps any distortion products and any channel imbalances symmetrical. This should start right from the source, that is starting with the DAC-output (thus maybe even preprocessing the digital datastream before conversion) and it should stop right at the amp's output stages, or even better yet, right at the listeners ears. The last is problematic. At any rate one needs three reproducing speakers to project the 2-channel information into the room. The most convincing method (there are many aspect to take care of) is a semi-M/S-approach right at the point of the acoustic reproduction, known as "Optimum Linear Matrix" or "Trinaural". I strongly believe that all the problem with "imaging", "soundstage" etc come from the (to me) intrinsically flawed attempt to project the 2-channel information with only two playback channels which must combine all but the simplest directional information in phantom images. For me, the 3-channel rematrixing has achieved the biggest step forward in imaging quality, by an order of magnitude. Better electronics in a two playback channel setup can better the system only to the level that the conventional method is capable of by its design limits. And that design is very sensitive to interchannel differences. The M/S processing and the rematrixed playback is less so.
Just 2cts from a M/S and rematrixing believer (this techniques also have some drawbacks, of course. Nothing is perfect).
- Klaus
I imagine (sic) a good way of maintaining imaging is to not process the signal in the L/R domain but rather in M/S. This keeps any distortion products and any channel imbalances symmetrical. This should start right from the source, that is starting with the DAC-output (thus maybe even preprocessing the digital datastream before conversion) and it should stop right at the amp's output stages, or even better yet, right at the listeners ears. The last is problematic. At any rate one needs three reproducing speakers to project the 2-channel information into the room. The most convincing method (there are many aspect to take care of) is a semi-M/S-approach right at the point of the acoustic reproduction, known as "Optimum Linear Matrix" or "Trinaural". I strongly believe that all the problem with "imaging", "soundstage" etc come from the (to me) intrinsically flawed attempt to project the 2-channel information with only two playback channels which must combine all but the simplest directional information in phantom images. For me, the 3-channel rematrixing has achieved the biggest step forward in imaging quality, by an order of magnitude. Better electronics in a two playback channel setup can better the system only to the level that the conventional method is capable of by its design limits. And that design is very sensitive to interchannel differences. The M/S processing and the rematrixed playback is less so.
Just 2cts from a M/S and rematrixing believer (this techniques also have some drawbacks, of course. Nothing is perfect).
- Klaus
- Status
- Not open for further replies.