Channel Separation?

[Blumlein 88]

They can be designed to make crosstalk irrelevant, but it takes careful and considered design, something seemingly lacking in a lot of modern, single board, super-compact designs these days.

CD players in the 1980s achieved >125dB crosstalk with very careful design, so it was essentially 'nailed' in the source department 30+ years ago.

Many stereo preamplifiers achieved around 90-100dB in the early 1980s, so that was deemed done and dusted too I guess.

I'm not confident many of these currently available single PCB mini components can approach similar levels. Maybe Amir can do a few L into R and R into L tests, just for fun, to see where the current gear sits. It's easy enough to shut off one input and measure the breakthrough from the driven channel into the undriven channel.

The recording interfaces I have measuring each other have around 120 db channel separation. A couple db better at low frequencie and a couple worse in the treble. The March DAC 1 I think had 115 db. The Marantz 7701 pre/pro has about 86-90 db depending upon which channels you measure between.
Just as an example.

 

[abm0]

Surprisingly 20 db is enough.

I would be very surprised if that were true. Based on what's happening with other unwanted content and the corresponding audibility thresholds I'd expect something more on the order of 85 dB being "enough for everyone". It should be easy enough to test by creating the crosstalk digitally and ABX-ing against the original, and I plan to do just that, now that you said that.

 

[sergeauckland]

I would be very surprised if that were true. Based on what's happening with other unwanted content and the corresponding audibility thresholds I'd expect something more on the order of 85 dB being "enough for everyone". It should be easy enough to test by creating the crosstalk digitally and ABX-ing against the original, and I plan to do just that, now that you said that.

Be surprised!

LP pickup cartridges have some 25dB of crosstalk at best, less at low and high frequencies, yet stereo is well presented. FM radio has at very best 40dB of crosstalk, depending on the perfect alignment of the decoder, yet live concerts on FM radio are amongst the very best audio source. Semi Pro analogue tape machines have around 45dB of crosstalk and a lot of very good albums have been recorded and mastered on such.

20dB across the frequency spectrum is perfectly adequate for good stereo. More does no harm, clearly, but isn't necessary.

S.

 

[Frank Dernie]

Absolutely. More elements in the room = less wall reflections = more fidelity.

Yes, the more untidy the room the better the sound quality.

 

[pkane]

Yes, the more untidy the room the better the sound quality.

Can I use you as an expert witness for my wife?

 

[SIY]

It should be easy enough to test by creating the crosstalk digitally and ABX-ing against the original, and I plan to do just that, now that you said that.

I wish I could "like" this statement repeatedly. People tend to handwave and not bother to do the experiments themselves...

I think you're wrong and that separation much past 20-25 dB isn't relevant, but if the evidence points out the contrary, you will have reduced my ignorance.

 

[Frank Dernie]

I wish I could "like" this statement repeatedly. People tend to handwave and not bother to do the experiments themselves...

I think you're wrong and that separation much past 20-25 dB isn't relevant, but if the evidence points out the contrary, you will have reduced my ignorance.

Many years ago a friend of mine did some experiments simulating the shortcomings of LPs in order to see if he could work out why LPs didn't sound so bad given how much lower the crosstalk, noise, distortion and speed stability were than CD.
I don't remember exactly what values he put in for crosstalk but it was based on a normal cartridge so probably 25dB. It was completely indistinguishable from the CD level of crosstalk. The one I remember most though was adding noise increased the impression of stereo depth.
So much for the blacker backgrounds etc. and probably why people like those very expensive "grounding boxes", which are nothing of the sort but are aerials connected to the shield side of the signal and add noise.

 

[Werner]

I would be very surprised if that were true.

This has been tested ages ago and is part of accepted psychoacoustics. In a speaker setup based on an equilateral triangle the sound is fully located in one speaker at about 30dB of channel separation. More is indeed not necessary, and from 20-25 dB on it is adequate for
most music, most people, and most systems.

However, what does matter is the cleanliness of the crosstalk signal. It must not be distorted (too much).

 

[abm0]

Welp, the first idea was to check stage size with "large" sounding things, so I went KSC75 and Aero Chord - "Surface". Cross-mixed both channels into eachother at -40 to -20 dB sloping attenuation (not as steep as I've seen on some graphs here, but I didn't want to overcomplicate the discussion of "where exactly is that -20 dB supposed to be inaudible"). Loaded original and modified in the ABX tool and... never got to the ABX part. Because the first step of "listen to A and B carefully first, just to get your bearings" already failed: there were no bearings to get. ) I couldn't hear any difference, at least not in stage size or how far left or right the most eccentric sounds seemed to go.

Next up: HE-400i, Yosi Horikawa - "Letter", same filtering, to check primarily for changes in imaging.

 

[solderdude]

Try it with older Beatles stereo recordings as well.
These have extreme L-R panning.
When you don't hear differences there you won't in any other recording either.

 

[Hypnotoad]

Try it with older Beatles stereo recordings as well.
These have extreme L-R panning.

Elanor Rigby comes to mind there.

 

[MZKM]

I believe the $10,000 amp challenge had parameters stating 35dB of separation at least.

I don’t know if any studies have been done, it most likely is less audible than THD, whose hearing threshold with music is around -40dB in the treble (much more forgiving in the bass). Keep in mind that it stacks across components too, but around -60dB in the treble is the worst you usually see.

 

[andreasmaaan]

I believe the $10,000 amp challenge had parameters stating 35dB of separation at least.

I don’t know if any studies have been done, it most likely is less audible than THD, whose hearing threshold with music is around -40dB in the treble (much more forgiving in the bass). Keep in mind that it stacks across components too, but around -60dB in the treble is the worst you usually see.

Trained listeners have been able to hear certain kinds of nonlinear distortion below -40dB under double blind conditions, although the results are a bit all over the place and therefore hard to make sense of. Compounding this is the fact that thresholds of audibility for added nonlinear distortion are in many cases lower than the levels of nonlinear distortion produced by the transducers used in the testing. There is a summary of some of the major studies' findings in post #2 of this thread. Clearly, the type of distortion is critical in terms of audibility.

There's also an interesting discussion in this thread.

Anyway, I agree broadly speaking that 1% nonlinear is unlikely to sound distorted, but I think there's plenty of evidence that it may nevertheless be audible - depending of course on the listener, the program, the setup, and importantly the type of nonlinear distortion.

 

[Jukka]

Okey people, I just read a review of headphone amps on a magazine. They published crosstalk measurements, which were mostly in 70-80 dB range, but the lowest was 30 dB for a 1100 € device. The need for channel separation in mixing a record is one thing, but when talking about playback, one would assume crosstalk, among other things, to be very low (good). Which would not always be the case. I would suggest that this measurement be added to measurement suite.

If you think it's a non-issue, I beg to consider an active speaker setup where channel separation is between bass and treble drivers and as such a matter of tweeter life. One can build these trusting that a device at dac-level (haven't touched amps yet) would have good channel separation, but end up smoking a tweeter because of it.

 

[SIY]

Okey people, I just read a review of headphone amps on a magazine. They published crosstalk measurements, which were mostly in 70-80 dB range, but the lowest was 30 dB for a 1100 € device. The need for channel separation in mixing a record is one thing, but when talking about playback, one would assume crosstalk, among other things, to be very low (good). Which would not always be the case. I would suggest that this measurement be added to measurement suite.

If you think it's a non-issue, I beg to consider an active speaker setup where channel separation is between bass and treble drivers and as such a matter of tweeter life. One can build these trusting that a device at dac-level (haven't touched amps yet) would have good channel separation, but end up smoking a tweeter because of it.

30dB at what frequency?

 

[Jukka]

30dB at what frequency?

The following was printed on the magazine for the device in question:
Channel separation
33 Ohm, 1/20 kHz: 45/30 dB
330 Ohm, 1/20 kHz: 45/29 dB

 

[SIY]

That doesn't look like much tweeter danger, but I was honestly unaware that headphones could be biamped.

 

[NiagaraPete]

I guess channel separation is irrelevant then, because no amp or DAC is so bad to have <20dB of separation?

To be honest, part of the reason I am asking is because I'm itching to try a dual mono amp setup for the sake of experimentation just to see (hear) if there's an audible difference. I've been running integrated amps since I was a teenager. Eager and curious to try something different.

Bi / tri amping will give a bigger impact.

 

[Jukka]

That doesn't look like much tweeter danger, but I was honestly unaware that headphones could be biamped.

My assumption here is that if low power, high finesse devices like HP amps can have it that bad, it must be much worse with power amps. Many DIY people use regular stereo/home theater amps for amplification and if the tweeter protection cap is not in place, the source of overheating power can be unexpected and hard to find.

 

[Frank Dernie]

Okey people, I just read a review of headphone amps on a magazine. They published crosstalk measurements, which were mostly in 70-80 dB range, but the lowest was 30 dB for a 1100 € device. The need for channel separation in mixing a record is one thing, but when talking about playback, one would assume crosstalk, among other things, to be very low (good). Which would not always be the case. I would suggest that this measurement be added to measurement suite.

If you think it's a non-issue, I beg to consider an active speaker setup where channel separation is between bass and treble drivers and as such a matter of tweeter life. One can build these trusting that a device at dac-level (haven't touched amps yet) would have good channel separation, but end up smoking a tweeter because of it.

Crosstalk is a mixing of signal between channels and has no similarity with crossover performance.

Headphones normally get the full frequency range signal in any case, why would a bit of the left channel leaking into the right cause any technical problem for it? There is no mechanism by which it could.
How it may influence the integrity of the stereo image is a different story, sound from one channel leaking onto the other channel definitely has potential to influence here.

Good pickup cartridges have crosstalk in the 30dB region at middle frequencies, worse at the frequency extremes, but many listeners find the stereo image from LPs to be OK to very good. The listening tests I was involved in 25 years ago evaluating the impact on the shortcomings of LP compared to CD showed that we couldn't hear a loss of stereo with a crosstalk at 30dB rather than 80dB, so not a big problem, particularly since stereo separation is a pretty random artifact anyway.

Don't worry about the effect on the sound - but it is indicative about the quality of the engineering which should be much better on a DAC. Obviously bad engineering in onwe aspect of performance always concens me about what else may have been not done very well.
Like seeing a round section drive belt on a belt drive record player