IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
How much stereo crosstalk to degrade imaging?
mikenet
post Nov 2 2005, 05:28
Post #1





Group: Members
Posts: 39
Joined: 15-June 02
Member No.: 2313



How much crosstalk does it take before the stereo imaging effect brakes down? Out of curiosity, how much exists in the vinyl world?

I recently got an outboard Dac, and for mono and early stereo(i.e. hard panned) recordings, it blows my iPod out of the water. But real stereo recordings just sound bad on it...the mix collapses, and the sound comes directly from the speakers, instead of from in between them like with my iPod.

iPod 4G/Dac-AH => Sonic Impact T-amp => Athena AS-B1

I just set this up in my dorm room, and was blown away with the iPod, but underwhelmed with the Dac. In fact, I'd much rather listen on the iPod, even though I can now notice slight distortion in it. Double blind tests are hard when you're switching cables around from sources with different, non-adjustable levels, but a few of my friends commented on it. And things like high frequency EQ can affect imaging, too, so I didn't know where to start.

I measured a few different sets of cables with a multimeter(looked good, no obvious shorts). Then I played a tone through one channel, and measured the voltage on each. I got the following values for stereo crosstalk:

-53dB @ 10Hz
-53dB @ 100Hz
-59dB @ 1kHz
-59dB @ 10kHz

My iPod comes in at -104dB(found online), woah, that's many orders of magnitude different!

The opposite channel was toward the low range of the multimeter(it read 0.001 or 0.002 volts), but these numbers match around what was audible(at 1KHz, I didn't repeat all the tests by ear).

Since I don't trust my multimeter at the absolute low end of its scale, I did a couple quick measurements with my ears and a SPL meter. The room was noisy (maybe ~40dB ambient...this is a dorm room with a computer running), so these are very rough estimates. I unplugged the left channel of the dac from the amp(thus the crosstalk of the amp was not a variable), and generated tones on the left channel. I then turned the volume of the amp up until I could hear the tone from the unconnected channel. I then played the tone on the right channel, and measured the level with the SPL meter. So between the Fletcher-Munson curves(of my ears), my hearing ability, the ambient noise, and my speaker's response, these figures are probably optimisitic by ~30-45dB. I just did this to verify that my multimeter wasn't crazy at it's low end.

Level @ Could perceive tone on opposite channel in noisy room
N/A @ 10Hz
-95dB @ 100Hz
-84dB @ 1kHz
-82dB @ 10kHz

Add that 30-45dB of optimism back, and my multimeter measurements were in the ballpark. I wish I had a good soundcard to run an RMAA test with, but my Audigy has pretty bad crosstalk on its own.
Go to the top of the page
+Quote Post
cliveb
post Nov 2 2005, 10:11
Post #2


WaveRepair developer


Group: Developer
Posts: 855
Joined: 28-July 04
Member No.: 15845



QUOTE (mikenet @ Nov 2 2005, 05:28 AM)
How much crosstalk does it take before the stereo imaging effect brakes down? Out of curiosity, how much exists in the vinyl world?
*

You can actually get away with quite a lot of crosstalk before the stereo illusion is destroyed. Vinyl can give pretty good imaging, and typical channel separation figures for a good vinyl setup would be about 25-30dB at 1kHz, and much less at higher frequencies (eg. 15dB at 10kHz).

QUOTE (mikenet @ Nov 2 2005, 05:28 AM)
I recently got an outboard Dac, and for mono and early stereo(i.e. hard panned) recordings, it blows my iPod out of the water. But real stereo recordings just sound bad on it...the mix collapses, and the sound comes directly from the speakers, instead of from in between them like with my iPod.
*

It could simply be that your speakers are too far apart relative to the listening position.

The fact that your iPod doesn't exhibit the same characteristics (I presume played back over the same speakers) is evidence that it has rather poorer channel separation than the DAC. Is it possible that the iPod has a deliberate partial mixing of the channels to compensate for the fact that most listening will be via headphones (where some reduction in the image width can be beneficial)? Anyone around here know about this?

QUOTE (mikenet @ Nov 2 2005, 05:28 AM)
....I got the following values for stereo crosstalk:

-53dB @ 10Hz
-53dB @ 100Hz
-59dB @ 1kHz
-59dB @ 10kHz

My iPod comes in at -104dB(found online), woah, that's many orders of magnitude different!
*

I simply don't believe that figure for the iPod. It's ludicrously high. Sounds like something a marketing man invented while lying in the bath one day.
Go to the top of the page
+Quote Post
mikenet
post Nov 3 2005, 10:24
Post #3





Group: Members
Posts: 39
Joined: 15-June 02
Member No.: 2313



Okay, I ran the Dac through RMAA with my Audigy...not knowing if the Audigy would see anything. It did:

Dac: -66dB
iPod: -85dB(measured value found online elsewhere from more reputible source)


So the iPod has *more* separation than the Dac! I tried adding crossfeed to the dac(under the assumption that the iPod may have had less separation...to try to simulate it), but it didn't help. I tried moving forward and backwards from the speakers(it's hard to move them closer), but it didn't help.

-66dB is very very high for a Dac...I wonder if something's wrong with it. If the channels are leaking, I'm not sure of their relative phase, which could be causing the issues.

This post has been edited by mikenet: Nov 3 2005, 10:37
Go to the top of the page
+Quote Post
KikeG
post Nov 3 2005, 12:56
Post #4


WinABX developer


Group: Developer
Posts: 1578
Joined: 1-October 01
Member No.: 137



Maybe it's your analog cables that are broken.

This post has been edited by KikeG: Nov 3 2005, 12:58
Go to the top of the page
+Quote Post
Tom Servo
post Oct 14 2010, 09:23
Post #5





Group: Members
Posts: 168
Joined: 11-January 02
Member No.: 981



Sorry for bumping this thread, but it's quite on topic for my question.

I'm wondering how it's possible that a device can go from ostensible -90dB crosstalk when attached to an amplifier to something like -40dB when using high quality headphones (high quality according to the tester, they're some AKGs).

I guess the different impedance(/resistance?) affects it a little, but this much? This sounds a little like software narrowing of the image, wouldn't it? Can DACs these days sense the difference in resistance and change processing (or have the host do that?) Seeing how the bass frequencies are more attenuated when attached to an amplifier instead of headphones almost suggests that.

Or did they just mess up the test?
Go to the top of the page
+Quote Post
greynol
post Oct 14 2010, 09:48
Post #6





Group: Super Moderator
Posts: 10339
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Perhaps this was intentionally designed into the headphones. I've read about similar things on and off throughout the years but haven't had the best of luck searching for the sake of this reply.

I did come up with these, however:
http://www.hydrogenaudio.org/forums/index.php?showtopic=7763
http://www.hydrogenaudio.org/forums/index....showtopic=11617


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
Tom Servo
post Oct 14 2010, 11:01
Post #7





Group: Members
Posts: 168
Joined: 11-January 02
Member No.: 981



They've tested other devices, too, I hope with the same headphones. There wasn't such a degradation in crosstalk there.

Well, we're talking about cellphones here, actually. The device in question is an upcoming Samsung. An earlier similar model exposes the same issue, while others don't.

There's also a difference on how the low frequencies respond depending on the device hooked up. Is this really just a resistance thing, or actually DSP tampering?

Go to the top of the page
+Quote Post
udauda
post Oct 14 2010, 13:21
Post #8





Group: Members
Posts: 101
Joined: 19-August 05
From: CA, USA
Member No.: 24009



Maybe this is worth mentioning:
http://en.wikipedia.org/wiki/Crosstalk_measurement
IOW, don't worry about it too much.

QUOTE (Tom Servo @ Oct 14 2010, 03:01) *
There's also a difference on how the low frequencies respond depending on the device hooked up. Is this really just a resistance thing, or actually DSP tampering?


The graph shows your headphone's resonance frequency is @ ~50Hz, and it is slightly underdamped.

This post has been edited by udauda: Oct 14 2010, 13:24
Go to the top of the page
+Quote Post
saratoga
post Oct 14 2010, 15:54
Post #9





Group: Members
Posts: 5163
Joined: 2-September 02
Member No.: 3264



QUOTE (Tom Servo @ Oct 14 2010, 04:23) *
I'm wondering how it's possible that a device can go from ostensible -90dB crosstalk when attached to an amplifier to something like -40dB when using high quality headphones (high quality according to the tester, they're some AKGs).


Because when its hooked to an amp, no current is flowing from the DAC, and so theres nothing to generate crosstalk. When you hook up to headphones, suddenly a lot of current flows, and that generates cross talk if the design isn't good.

QUOTE (Tom Servo @ Oct 14 2010, 04:23) *
Can DACs these days sense the difference in resistance and change processing (or have the host do that?)


Theres nothing to change. Crosstalk is analog. You can't "process" out analog.

QUOTE (Tom Servo @ Oct 14 2010, 04:23) *
Or did they just mess up the test?


No, they messed up the design of the DAC wink.gif
Go to the top of the page
+Quote Post
Slipstreem
post Oct 14 2010, 16:02
Post #10





Group: Members
Posts: 966
Joined: 7-July 06
Member No.: 32660



QUOTE (saratoga @ Oct 14 2010, 15:54) *
No, they messed up the design of the DAC wink.gif

Or far more likely, IMO, the design of the analog buffer stages (or the power supply running them) after the DAC.
Go to the top of the page
+Quote Post
Ed Seedhouse
post Oct 14 2010, 16:15
Post #11





Group: Members
Posts: 175
Joined: 19-May 09
Member No.: 69959



The very assumption behind the thread title is faulty. In fact, crosstalk can be and often is used to enhance stereo imaging, not degrade it. One of the major problems with headphones is too little crosstalk, which is why many headphone amplifiers deliberately introduce it.

This post has been edited by Ed Seedhouse: Oct 14 2010, 16:25


--------------------
Ed Seedhouse
Go to the top of the page
+Quote Post
saratoga
post Oct 14 2010, 16:24
Post #12





Group: Members
Posts: 5163
Joined: 2-September 02
Member No.: 3264



QUOTE (Slipstreem @ Oct 14 2010, 11:02) *
QUOTE (saratoga @ Oct 14 2010, 15:54) *
No, they messed up the design of the DAC wink.gif

Or far more likely, IMO, the design of the analog buffer stages (or the power supply running them) after the DAC.


Well yeah, but on a phone they're adjacent transistors on the same IC, so its all pretty much the same thing.
Go to the top of the page
+Quote Post
dhromed
post Oct 14 2010, 16:42
Post #13





Group: Members
Posts: 1339
Joined: 16-February 08
From: NL
Member No.: 51347



QUOTE (Ed Seedhouse @ Oct 14 2010, 17:15) *
The very assumption behind the thread title is faulty. In fact, crosstalk can be and often is used to enhance stereo imaging, not degrade it. One of the major problems with headphones is too little crosstalk, which is why many headphone amplifiers deliberately introduce it.


I've found that an experimental binaural component for foobar2K, based on crossfeeding the channels by a small amount, did very little except muddy up the sound a bit, almost equivalent to very gently rolling off higher frequencies— though I haste to add that I didn't do the obvious EQ experiment. Switching back and forth I noticed a difference, but not an improvement in any sense.

A proponent of the component reported that it reduced his personal experience of fatigue when using it for prolonged times, but I think that's small wonder given that it reduced the intensity of frequencies that humans commonly experience as loud.

Perhaps the component was suboptimal in some way, which I think is likely, but it didn't do much for my idea that crosstalk is a magical beneficial effect, even when applied with care.


Is this getting OT?
Go to the top of the page
+Quote Post
Northpack
post Oct 14 2010, 17:03
Post #14





Group: Members
Posts: 457
Joined: 16-December 01
Member No.: 664



QUOTE (dhromed @ Oct 14 2010, 16:42) *
I've found that an experimental binaural component for foobar2K, based on crossfeeding the channels by a small amount, did very little except muddy up the sound a bit, almost equivalent to very gently rolling off higher frequencies— though I haste to add that I didn't do the obvious EQ experiment. Switching back and forth I noticed a difference, but not an improvement in any sense.

Are you refering to the bs2b component? I think it makes a hughe improvement, especially for those recording which have instruments panned very far to the left/right. However, one shouldn't confuse crosstalk with crossfeed, as the latter feeds in the other channel's signal frequency dependent and with a phase delay.
Go to the top of the page
+Quote Post
Arnold B. Kruege...
post Oct 14 2010, 18:22
Post #15





Group: Members
Posts: 4485
Joined: 29-October 08
From: USA, 48236
Member No.: 61311



QUOTE (Slipstreem @ Oct 14 2010, 11:02) *
QUOTE (saratoga @ Oct 14 2010, 15:54) *
No, they messed up the design of the DAC wink.gif

Or far more likely, IMO, the design of the analog buffer stages (or the power supply running them) after the DAC.


As a rule, cross talk is rarely a problem until it is pretty gross, such as 10 dB. Back in the days of analog recording and playback and tubes we tolerated 20 -30 dB crosstalk in virtually any signal source. With digital, crosstalk can be truely zero. I would be surprised if anybody ABXed the diffference between 60 dB crosstalk and 90 dB crosstalk.

A look at how the crosstalk varies with frequency give clues to the cause.. If the cross talk is more-or-less constant at all audible frequencies, then it is probably a grounding problem. If circuity for both channels is grounded through too much resistance in common, then there will be crosstalk.

If the problem is far worse at high frequencies, then there is probably some capacitive coupling between the channels, probably due to the signal paths for the two channels being too close together and having too high of a high impedance. This used to be more of a problem with tubed circuits, but is usually no problem with good SS designs.

Historically, the worst crosstalk was with vinyl, tape, and FM stereo. IOW, just about anything that we had to listen to music from. Starting with CDs, that all could improve greatly.
Go to the top of the page
+Quote Post
dhromed
post Oct 14 2010, 20:36
Post #16





Group: Members
Posts: 1339
Joined: 16-February 08
From: NL
Member No.: 51347



QUOTE (Northpack @ Oct 14 2010, 18:03) *
Are you refering to the bs2b component? I think it makes a hughe improvement, especially for those recording which have instruments panned very far to the left/right.


That's the one. I have a hard time imagining how people think it's "huge" improvement. I I fiddled with its sliders to every far end and most positions in between, turning it on again, off again, but the difference was minimal every time. Maybe I wasn't using the right kind of music; maybe I missed the sliders' sweet spot. I always preferred uncrossed streams.

Will try again in a bit. Maybe I'll have better luck this time.

QUOTE (Northpack @ Oct 14 2010, 18:03) *
However, one shouldn't confuse crosstalk with crossfeed, as the latter feeds in the other channel's signal frequency dependent and with a phase delay.

Close one. some of the posts threw me off, so I looked up both Wikipedia articles for a quick 101.
Go to the top of the page
+Quote Post
Tom Servo
post Oct 14 2010, 22:28
Post #17





Group: Members
Posts: 168
Joined: 11-January 02
Member No.: 981



QUOTE (saratoga @ Oct 14 2010, 15:54) *
No, they messed up the design of the DAC wink.gif

The DAC's very likely a WM8990, FWIW.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 26th December 2014 - 10:01