Channel separation...how much is enough...how much is not?

[Blumlein 88]

I will posit that 30 db or more channel separation is enough. More than that is not audible. In fact 17 db or so is enough.

So why is a device with only 70 db of channel separation described as "poor channel separation"? Poor versus what is possible? Yes. Poor versus what is audible, not in the least.

 

[fpitas]

Good question. I doubt anyone mixes with tremendous separation. Frankly it sounds kind of weird.

 

[fpitas]

Channel separation is also the excuse audiophiles give for monoblocks. Once again, that seems bizarre.

 

[sergeauckland]

I will posit that 30 db or more channel separation is enough. More than that is not audible. In fact 17 db or so is enough.

So why is a device with only 70 db of channel separation described as "poor channel separation"? Poor versus what is possible? Yes. Poor versus what is audible, not in the least.

Exactly this. 20dB at mid frequencies gives decent enough stereo.

I suppose a modern electronic device having less than 70dB separation indicates poor engineering rather than poor audible performance. Much like distortion levels of -100dB or more compared with a more than adequate -80dB or an adequate -60dB.

S.

 

[Matias]

From what I understand, the perceived effect of crosstalk should be a "narrowing of the soundstage", as in the extreme of mixing both channels 100% together the end result is a mono reproduction, everything centered on the soundstage.

The threshold ASR uses for distortions below the main signal is around -115 dB to be considered audible. In that same logic for crosstalk, anything above -115 dB from the one channel "bleeding" into the other channel should be audible too, an infinitesimal narrowing of the soundstage.

The cool thing is that this is easily reproducible: just get a stereo music file of choice, and bleed -115, -100, -80, -60, -40 dB of one channel into the other and check how much the soundstage narrows, where is the point of where this is significant.

But like distortion, the less the better. And if a similar priced gear has better specs in crosstalk, then why not choose it over the worst one all else being equal. Maybe it should not be a determinant main factor, but should have a lower weight when comparing and evaluating options imo.

 

[MaxwellsEq]

I will posit that 30 db or more channel separation is enough. More than that is not audible. In fact 17 db or so is enough.

So why is a device with only 70 db of channel separation described as "poor channel separation"? Poor versus what is possible? Yes. Poor versus what is audible, not in the least.

That 30dB is enough, is probably true. I've not looked for any peer-reviewed work to confirm this, but anecdotally it's the case (given it's probably the best one can achieve with vinyl).

For me, separation better than 30 but less good than 70 indicates potential sloppiness in design. With lower values I would question signal routing and layout thoroughness. This would then make me question other design choices.

 

[fpitas]

(given it's probably the best one can achieve with vinyl).

True. Once again, hard to see why audiophiles go on about fantastic channel separation.

 

[fpitas]

I could be a complete iconoclast and suggest that vinyl preference is because it has tons of crosstalk

 

[AnalogSteph]

From what I understand, the perceived effect of crosstalk should be a "narrowing of the soundstage", as in the extreme of mixing both channels 100% together the end result is a mono reproduction, everything centered on the soundstage.

It can also go the other way, i.e. out-of-phase crosstalk. That results in a widening of soundstage, potentially up to the point where only L-R is left. This is typically a symptom of a shared ground return with not-so-negligible resistance (can happen in 3-conductor headphone wiring; the rule of thumb is you want <1% of nominal driver impedance to keep separation above 40 dB), or an RCA cable with a faulty ground connection.

Typical xtalk values AFAIK:
Vinyl - ~26 dB
FM stereo - 40-50 dB if you're good, can drop to little over 30 dB in the highs at lower bandwidth

20 dB of channel separation is painful on headphones. "30 dB is enough" is probably right.

 

[Sokel]

The threshold ASR uses for distortions below the main signal is around -115 dB to be considered audible. In that same logic for crosstalk, anything above -115 dB from the one channel "bleeding" into the other channel should be audible too, an infinitesimal narrowing of the soundstage.

I can't find the research about it and I'm curious:
Under what conditions -115db of distortion (not noise) is audible?And with what material?

I'm asking because I have to go to 60's to start hearing it with PKharmonic and that with simple material,if it's dense I have to go to embarrassing levels

(both headphones and speakers)

 

[MaxwellsEq]

I can't find the research about it and I'm curious:
Under what conditions -115db of distortion (not noise) is audible?And with what material?

I'm asking because I have to go to 60's to start hearing it with PKharmonic and that with simple material,if it's dense I have to go to embarrassing levels

(both headphones and speakers)

From a purely anecdotal point of view from experience gained while commissioning studio electronics, measuring gear could spot distortion at levels lower than I could hear it.

Noise, on the other hand, I found a lot easier to spot. So I've tended over the years to prioritize low noise over low distortion. This is why, for me, a single SINAD number is less interesting than noise and distortion numbers separated out.

 

[restorer-john]

You have channel separation and crosstalk. Crosstalk may be either L into R or R into L or both. And that crosstalk is usually proportional to frequency, but not always.

How much separation does our head and ears really have anyway when listening to loudspeakers in a typical room?

 

[Blumlein 88]

How much separation does our head and ears really have anyway when listening to loudspeakers in a typical room?

17 db with loudspeakers.

 

[pma]

I will posit that 30 db or more channel separation is enough. More than that is not audible. In fact 17 db or so is enough.

30dB is fine. Then we have numbers games, similarly as with distortion. ASR is, unfortunately, concentrated on simple, easy to understand, but unimportant numbers, speaking about amplifiers.

 

[Matias]

@amirm could you please share your point of view on this?

 

[BR52]

30dB is fine. Then we have numbers games, similarly as with distortion. ASR is, unfortunately, concentrated on simple, easy to understand, but unimportant numbers, speaking about amplifiers.

Can you give us a list of measurements that for your opinion is important? Maybe @amirm or we can make it understandable.

 

[BR52]

No answer can be an answer. Nice style after no explanation of the reason for this statement.

 

[Bleib]

Although with stereo speakers there might not be a need for superhigh separation, how about headphones?

 

[DVDdoug]

With phono cartridges I used to see specs around 20dB. That's good enough. 20dB is a power factor of 100. If you have 100W on the "good side" you'll have 1W leaking into the "bad side".

In the digital age I don't even think about it and in analog days I didn't worry about it in electronics. I never heard a problem in the vinyl days but I used to read the specs...

Although with stereo speakers there might not be a need for superhigh separation, how about headphones?

I've never done any experiments but I'd bet your brain won't allow you to hear the slight-quiet leakage. Especially, with a regular stereo mix where there is rarely silence on one side and sound in the other. And with open headphones there is some acoustic leakage to the other ear.

P.S.
With an audio editor (like Audacity) you could make files with varying amounts of "blend". And/or you can make a file with different tones, or different songs, in each channel. Or you can make one side silent and then try various amounts of panning toward the other side.

 

[Justdafactsmaam]

You have channel separation and crosstalk. Crosstalk may be either L into R or R into L or both. And that crosstalk is usually proportional to frequency, but not always.

How much separation does our head and ears really have anyway when listening to loudspeakers in a typical room?

5db - 10db