Crosstalk performance of A90 and THX789 and subjective 'soundstage' feeling.

[tomtrp]

The crosstalk of A90(-105db) measures worse than the THX789(-127db), though both over -85db, should be inaudible. so I understand why Amir doesn't specify the crosstalk issue.
But from Amir's measurement, we see A90's channel difference rises significantly after 16ms compared to THX789. Is this an issue at all?
And from subjective side, lots of subjective reviews in both western and Chinese hifi forum report that the soundstage of A90 is mediocre, worse than THX 789 (though I do not trust such subjective reports at all but you will find this phenomenon if you read some forums and reports about A90.)
And I notice that the Topping's preamp Pre90 has much better crosstalk figure at -140db.
Is this a potential argument to validate the untrustworthy subjective reviews that Topping A90's soundstage feeling is not that good???

 

[Pritaudio]

There is soundstage and there is imaging.
there is also separation of instruments during complex passages.
its hard to exactly define these as each component in the audio chain has a signature, and hence can be considered an combined algorithm which can be tested with the variety of music recordings.

 

[maverickronin]

Why did you post graphs of the pot balance when asking about crosstalk?

 

[levimax]

Crosstalk does not have much of anything to do with imaging. Many people talk about the "better" depth and width of imaging with LP playback yet a stereo cartridge is lucky to have -25 dB of separation. No one is going to be able to hear the difference between -85 dB and -140 dB of crosstalk when music is playing.

 

[companyja]

Audibility of crosstalk is easy to test because you can easily add it digitally. I'm certain you can find a crossover network that will crossfeed the channels. You'd be surprised how low the treshold of transparency for crosstalk can be. You might start hearing it anywhere from -30 to -40 dB. A lot of phones and smaller devices hover around -60 to -80 dB. The cult classic Sandisk Sansa Clip+ has something funny like -35dB into 16ohm. I've used it for years before I got into technical measurements and never once thought the soundstage was smaller than on my desktop.

In terms of subjectivists and relation of subjective impressions and crosstalk, like levimax mentioned, very well-respected phono cartridges like the Shure M95HE advertised -25dB of channel separation. You might go to -30 or -35 with something very exotic. Yet you will find an infinite amount of subjectivists raving about analog instrument separation or whatever the hell they love saying. I wish amir tested crosstalk more because I fear the boutique subjectivist darling gear like audio-gd would have crosstalk measurements all over the place

 

[3125b]

Excessive crosstalk can definitely have the mentioned effect of "narrowing the soundstage". However it has to be extreme, the effect will absolutely not be noticable at levels of something like -80dB.
I tried a 30% (something like -10dBV) "copy channel" filter in Equalizer APO (if you do that, don't forget to add a negative pre), and even with that, the impact really depends on the recording. With some it is extremely noticable, with some, especially older ones that have hardly any channel separation to begin with, barely if at all.

But from Amir's measurement, we see A90's channel difference rises significantly after 16ms compared to THX789. Is this an issue at all?

Those graphes just show channel imbalance vs. volume position, has nothing to do with crosstlk or time.

 

[AnalogSteph]

The cult classic Sandisk Sansa Clip+ has something funny like -35dB into 16ohm. I've used it for years before I got into technical measurements and never once thought the soundstage was smaller than on my desktop.

I would expect it to get wider, given that the part resulting is degraded crosstalk is shared ground return impedance which will attenuate L+R but not L-R. I am pretty sure the cable on my Soundmagic E10 is worse though... I did some measurements and calculated about 1 ohm shared at one point. -34 dB of crosstalk would mean about 2% of 16 ohms = 0.32 ohms shared. Thank goodness for the stereo width control in Rockbox, 0.95 is roughly in the right ballpark.

 

[tomtrp]

Excessive crosstalk can definitely have the mentioned effect of "narrowing the soundstage". However it has to be extreme, the effect will absolutely not be noticable at levels of something like -80dB.
I tried a 30% (something like -10dBV) "copy channel" filter in Equalizer APO (if you do that, don't forget to add a negative pre), and even with that, the impact really depends on the recording. With some it is extremely noticable, with some, especially older ones that have hardly any channel separation to begin with, barely if at all.


Those graphes just show channel imbalance vs. volume position, has nothing to do with crosstlk or time.

Yep, I get it. It was a wrong choice of graph.
But still, is there anyone in this forum owing both A90 and THX789 and provide some insights about the actual ''soundstage'' difference? And if there is an actual audible difference(its not that easy to conduct double blind tests though) which cannot be reasonably explained by the sets of measurement provided by Amir now, maybe we can dig further and include more measurements. A90s are quite popular in the Chinese community in the beginning, but there are just so many of them in the second hand market now and some of the sellers report the same ''small soundstage''. Though the phenomenon in the second hand market may have nothing to do with sound quality(we know hifi peole like to trade gears), it is concerning for someone who thinks the amp is near perfect from the given measurement and want to buy it based on the measurement.

 

[companyja]

It's very easy to hear whatever you're conditioned to hear. If someone says the Topping A90 has a smaller soundstage, your brain will fill in the rest even if you know there is no reasonable way for this to happen. It has happened to me, it has happened to audio engineers, it's just the way the brain (NOT ears, brain specifically) works.

I'm going to guess nobody has posted any evidence of a blind test to confirm their suspicion. In trying to be pragmatic I think people often leave that little disclaimer of like "yes these are specs known to affect audio performance but hey we can't measure EVERYTHING" or something like that. I think it's time to push back against this thought a little further because it invites a lot of subjective impressions that are backed up by absolutely nothing to creep back in. Remember, people have sworn and will swear that cables, table surfaces, USB purifiers, etc. make a big audible difference.

If someone can provide a measurement that can indicate audio quality aside from the usual test-fare, nobody is stopping them. Just saying however that "well we don't measure everything" is no more scientific than saying measurements are irrelevant

Soundstage is a function of speaker/headphone placement and design. In a transparent DAC or amplifier, there is no such thing as "soundstage". It is created when the stereo signal is sent to a multiple discrete speakers placed in a way to create an illusion of a spread-out sound. Whether you're using speakers or headphones, it is always an illusion since 100% of the sound is coming from two boxes (or more if you have surround). A lot of people prefer speaker soundstage, I'm in the smaller camp that loves headphone soundstage, but in the end none of it is contingent on the DAC or AMP as long as the source is transparent enough and the crosstalk isn't badly broken.

 

[jasonhanjk]

A better design headphone amp (or DAC Amp) is capable at -70dB crosstalk or better (lower).
With 16 ohm load, 5mW power and near 0 ohm output impedance.

OP didn't mention the headphone use.
High impedance HP (300 ohm) will get better crosstalk while low impedance one (16 ohm) are likely above -70dB.

 

[tomtrp]

It's very easy to hear whatever you're conditioned to hear. If someone says the Topping A90 has a smaller soundstage, your brain will fill in the rest even if you know there is no reasonable way for this to happen. It has happened to me, it has happened to audio engineers, it's just the way the brain (NOT ears, brain specifically) works.

I'm going to guess nobody has posted any evidence of a blind test to confirm their suspicion. In trying to be pragmatic I think people often leave that little disclaimer of like "yes these are specs known to affect audio performance but hey we can't measure EVERYTHING" or something like that. I think it's time to push back against this thought a little further because it invites a lot of subjective impressions that are backed up by absolutely nothing to creep back in. Remember, people have sworn and will swear that cables, table surfaces, USB purifiers, etc. make a big audible difference.

If someone can provide a measurement that can indicate audio quality aside from the usual test-fare, nobody is stopping them. Just saying however that "well we don't measure everything" is no more scientific than saying measurements are irrelevant

Soundstage is a function of speaker/headphone placement and design. In a transparent DAC or amplifier, there is no such thing as "soundstage". It is created when the stereo signal is sent to a multiple discrete speakers placed in a way to create an illusion of a spread-out sound. Whether you're using speakers or headphones, it is always an illusion since 100% of the sound is coming from two boxes (or more if you have surround). A lot of people prefer speaker soundstage, I'm in the smaller camp that loves headphone soundstage, but in the end none of it is contingent on the DAC or AMP as long as the source is transparent enough and the crosstalk isn't badly broken.

You are definitely correct about how we our brain is affected by the opinions and then affect our listening. Its a valid explanation.But is it sufficient?
On paper, without the confirmation of blind tests, I think it is also not a scientific attitude to simply attribute all the repeated subjective opinions to phychological effects. The key test could be the blind test between A90 and THX789 with a set of music with great spatial effect. It will reveal things more than soundstage. Maybe people cannot distinguish between these 2 transparent amp at all with volume carefully matched, haha. I hope somebody can do this test.