high resolution listening test

[Blumlein 88]

Sure but keep in mind this is not dbx or level matched. Oppo105, Emotiva Xda2 xda1 dc1

Well there are levels of credibility. Not blind, but also not level matched is pretty far down the credibility totem pole. Level matched there is some chance biases won't color your perception and you'll hear reality. Non-level matched and differences would need to be very, very large to reliably discern. Like speaker level large. I've heard two of those you list. If you have them available try a pair and level match them. Let us know what you think then.

Even sighted you'll usually find differences you thought you heard either disappear or become teeny, tiny little differences once you level match.

Rule #1 for any comparative listening is level matched within .1 db of each other. We might let you by with matching voltage to within 2% (.17 db is 2%).

Believe me we know how it seems, but you really, really need level matching.

 

[garbulky]

Well there are levels of credibility. Not blind, but also not level matched is pretty far down the credibility totem pole. Level matched there is some chance biases won't color your perception and you'll hear reality. Non-level matched and differences would need to be very, very large to reliably discern. Like speaker level large. I've heard two of those you list. If you have them available try a pair and level match them. Let us know what you think then.

Even sighted you'll usually find differences you thought you heard either disappear or become teeny, tiny little differences once you level match.

Rule #1 for any comparative listening is level matched within .1 db of each other. We might let you by with matching voltage to within 2% (.17 db is 2%).

Believe me we know how it seems, but you really, really need level matching.

I don’t . But I don’t really expect subjective listening to pass the sniff test on this forum. But the post wasn’t really about level matched comparisons.My post was really about dynamic range and how that relates to dac requirements

 

[Blumlein 88]

I don’t . But I don’t really expect subjective listening to pass the sniff test on this forum. But the post wasn’t really about level matched comparisons.My post was really about dynamic range and how that relates to dac requirements

Well, your post queried about tests of DACs at higher levels where people can hear the signal vs levels most probably don't actually hear down low in level like -90 db.
Perhaps those tests have been shown to be useless. Gear that is pretty nice near max level typically only gets nicer as the signal level goes down. So your asking (or it seemed to me) what mid-level testing is done or can be done to show why DACs sound different. The answer is likely to be they don't sound different in those ranges. Your experience of differences in sound non-leveled matched you rightly understand don't pass the sniff test on this kind of forum. Now maybe before going on goose chase looking for measures that explain different sound, it would be worthwhile to do a level matched comparison and see what your experience is then. You may no longer have anything you need to look for in some measurements.

 

[garbulky]

Well, your post queried about tests of DACs at higher levels where people can hear the signal vs levels most probably don't actually hear down low in level like -90 db.
Perhaps those tests have been shown to be useless. Gear that is pretty nice near max level typically only gets nicer as the signal level goes down. So your asking (or it seemed to me) what mid-level testing is done or can be done to show why DACs sound different. The answer is likely to be they don't sound different in those ranges. Your experience of differences in sound non-leveled matched you rightly understand don't pass the sniff test on this kind of forum. Now maybe before going on goose chase looking for measures that explain different sound, it would be worthwhile to do a level matched comparison and see what your experience is then. You may no longer have anything you need to look for in some measurements.

Well I would but I don't consider it worthwhile.
I'm not too interested in debating DBT. I've had that debate quite a few times, never changed anybody's mind on either side. So I'm not interested in repeating it.

But I would be interested in figuring out different tests.
For instance a tone at -10 db and a tone -40 db is just a tone. But...what about dynamics? A tone changing its volume level. And then intersperse that with another tone perhaps changing its dynamic. Music dynamics don't vary massively during a piece of music - at least not compared to -the kind of snr's we are talking about.

What if we calculate dynamic variances in usual listening. And then try to simulate it. And see what info can be gleamed. I don't know enough to figure out how to do such a test and what the interpretation would mean. But I feel there might be something in it worth looking in to.

 

[RayDunzl]

But...what about dynamics? A tone changing its volume level.

A sine wave is constantly changing its volume, and its rate of change, and at a precisely predictable rate.

And then intersperse that with another tone perhaps changing its dynamic.

The expected value of two sines would also be predictable, if you wanted to analyze that result.

What if we calculate dynamic variances in usual listening.

That sine wave - flattening of the peaks raises odd harmonics, if extended, maybe it raises the even harmonics (not sure on that one*)...

Dynamic variance - the sine sweeps the range, what ever power limit you set on it, and remains easily calculable and comparable...

What would you suggest for a test excitation? You could certainly record the output and compare for difference.

---

As far as I know, all sound can be created as combinations of sines, wavy, straight lines, combinations, all are combinations of sines at various frequencies and intensities.

So, I suppose, if a sine can't be reproduced properly at all frequencies, that's about enough test to reveal faults (or at least their intensity - nothing is perfect). What error would show somewhere else but not there?

---

*Test: Adding second harmonic, in phase (positive peaks aligned) with the base tone, causes the positive peak to be extended and the negative peak to be flattened, visually.

1000 and 2000Hz, combined at full value, then combined with the 2000Hz tone at -6dB and at -12dB

 

[Sal1950]

Well I would but I don't consider it worthwhile.
I'm not too interested in debating DBT. I've had that debate quite a few times, never changed anybody's mind on either side. So I'm not interested in repeating it.

Well if your not interested in knowing if you really can hear differences, or are just falling for human weaknesses, why bother with High Fidelity or participating in forums that investigate the details of such? Just put together a kit that sounds good to you and move on. I really don't get where your coming from or why you bother? You say you want to understand certain things, but not if it runs the risk of crashing your base of understanding?

 

[Blumlein 88]

Well I would but I don't consider it worthwhile.
I'm not too interested in debating DBT. I've had that debate quite a few times, never changed anybody's mind on either side. So I'm not interested in repeating it.

I was talking matching levels sighted. No DBT. Just match levels. Honestly, no one of our regulars here will give much credence to what you hear unless you match levels before claiming two bits of gear sound different.

But I would be interested in figuring out different tests.
For instance a tone at -10 db and a tone -40 db is just a tone. But...what about dynamics? A tone changing its volume level. And then intersperse that with another tone perhaps changing its dynamic. Music dynamics don't vary massively during a piece of music - at least not compared to -the kind of snr's we are talking about.

What if we calculate dynamic variances in usual listening. And then try to simulate it. And see what info can be gleamed. I don't know enough to figure out how to do such a test and what the interpretation would mean. But I feel there might be something in it worth looking in to.

If you don't know that is fine. How to explain is another matter. Typically a device of good design has low distortion which might grow some as you approach max level. So if you can do max level pretty clean lower dynamically varying levels aren't going to trip up the circuit. As distortion will go down as level does, and on most devices distortion can be very low at its worst.

Another test that somewhat answers your question is the IMD test. You play two tones and see if they intermodulate due to non-linearities in the gear and cause tones that aren't supposed to be there. One version uses a 19 khz and a 20 khz sine wave mixed together at max level. Looks like this which as you can see has steep transients and the waveform goes from loud to soft and back again. The time to go from soft_loud_soft again is one millisecond btw. Plenty of good devices have no real problem with this. Distortion level is not zero, but have one or two or more zeroes after the decimal point.

Now quoting you:

But I would be interested in figuring out different tests.

That is all well and good. But if you fail to understand the tests normally done, and wish to find one which explains you hearing two DACs as different when measurements show they likely aren't, it would behoove you to determine if you really are hearing a difference or if you being mistaken by your methodology of comparing them. Because you'll never find a test to explain a difference that isn't really there.

 

[garbulky]

Well if your not interested in knowing if you really can hear differences, or are just falling for human weaknesses, why bother with High Fidelity or participating in forums that investigate the details of such? Just put together a kit that sounds good to you and move on. I really don't get where your coming from or why you bother? You say you want to understand certain things, but not if it runs the risk of crashing your base of understanding?

Well I think it's because you are making some assumptions about me that don't really describe
me.

In terms of being interested in learning:
I am not interested in learning what some people in the forum want to teach me.

I am interested in discussing things I'm interested in. Occasionally it involves learning things I'm interested in learning. Point being, it's got to be a two way street of interest.

Why I am here is because I am interested in measurements among other things. There's very few sites that are interested in measurements out there.

I suggest people stop worrying about my opinions regarding things we likely disagree on. Life's too short. Plus it's weird. Let's find common ground and be productive.

 

[garbulky]

I was talking matching levels sighted. No DBT. Just match levels. Honestly, no one of our regulars here will give much credence to what you hear unless you match levels before claiming two bits of gear sound different.

If you don't know that is fine. How to explain is another matter. Typically a device of good design has low distortion which might grow some as you approach max level. So if you can do max level pretty clean lower dynamically varying levels aren't going to trip up the circuit. As distortion will go down as level does, and on most devices distortion can be very low at its worst.

Another test that somewhat answers your question is the IMD test. You play two tones and see if they intermodulate due to non-linearities in the gear and cause tones that aren't supposed to be there. One version uses a 19 khz and a 20 khz sine wave mixed together at max level. Looks like this which as you can see has steep transients and the waveform goes from loud to soft and back again. The time to go from soft_loud_soft again is one millisecond btw. Plenty of good devices have no real problem with this. Distortion level is not zero, but have one or two or more zeroes after the decimal point.
View attachment 11959

Now quoting you:

But I would be interested in figuring out different tests.

That is all well and good. But if you fail to understand the tests normally done, and wish to find one which explains you hearing two DACs as different when measurements show they likely aren't, it would behoove you to determine if you really are hearing a difference or if you being mistaken by your methodology of comparing them. Because you'll never find a test to explain a difference that isn't really there.

Well it doesn't have to be listening tests. It could be in measurements of the gear using multiple tones with changing dynamics. I would go so far to say music with different instruments. But I fail to figure out how to make that a quantifiable measurement. So maybe tones first.

 

[garbulky]

A sine wave is constantly changing its volume, and its rate of change, and at a precisely predictable rate.



The expected value of two sines would also be predictable, if you wanted to analyze that result.



That sine wave - flattening of the peaks raises odd harmonics, if extended, maybe it raises the even harmonics (not sure on that one*)...

Dynamic variance - the sine sweeps the range, what ever power limit you set on it, and remains easily calculable and comparable...

What would you suggest for a test excitation? You could certainly record the output and compare for difference.

---

As far as I know, all sound can be created as combinations of sines, wavy, straight lines, combinations, all are combinations of sines at various frequencies and intensities.

So, I suppose, if a sine can't be reproduced properly at all frequencies, that's about enough test to reveal faults (or at least their intensity - nothing is perfect). What error would show somewhere else but not there?

---

*Test: Adding second harmonic, in phase (positive peaks aligned) with the base tone, causes the positive peak to be extended and the negative peak to be flattened, visually.

1000 and 2000Hz, combined at full value, then combined with the 2000Hz tone at -6dB and at -12dB

View attachment 11956

I'm not entirely sure I understood your post correctly. Thanks for the response. Lots of neat graphs! I need to read it again. That looks like audacity? How are you generating the tones with it? Is there a program or someway in audacity where I can do this too? I understand I won't be able to record the dacs output without some more effort. But at least I could listen to it.

 

[RayDunzl]

I'm not entirely sure I understood your post correctly.

Oh well... Some amateur rambling...

That looks like audacity?

Yes.

How are you generating the tones with it?

"Generate" menu.

 

[Blumlein 88]

I'm not entirely sure I understood your post correctly. Thanks for the response. Lots of neat graphs! I need to read it again. That looks like audacity? How are you generating the tones with it? Is there a program or someway in audacity where I can do this too? I understand I won't be able to record the dacs output without some more effort. But at least I could listen to it.

Yes it is Audacity. Warning potentially damaging signal, read my warnings and be careful.

Make a new track. Generate a tone at 20 khz with an amplitude of .5. Make a second new track and make one for 19 khz with an amplitude of .5 just the same.

Now select both tracks and under the Tracks menu choose Mix and Render. This will mix both tracks together.

Now before you go listening to this is a max level tone. I would suggest turning the level down a bunch or dropping the level of the track in Audacity by 20 db before you play it.

Secondly you likely will hear absolutely nothing. But this signal could easily burn out tweeters. Unless you are young with very good hearing you don't hear 19 or 20 khz anymore. It could be making your dog run and hide while you'll hear nothing. This is really not the signal to listen to over headphones (especially headphones) or speakers.

This signal was developed to put maximal stress on non-linear circuits to see how much distortion they generate. Wasn't really meant for listening to for the most part.

I might suggest mixing together 5 khz and 6 khz if you want to hear it. Even then you have a loud signal start with volume very low. I would keep the length to 10 seconds or so.

The signal does not sound as dastardly as it really is.

EDIT to add: Oops you were replying to Ray. I'll leave this here as it might be helpful to figure out making your own tones in Audacity where you combine more than one generated tone.

 

[Arnold Krueger]

Very interesting stuff. Thanks!
1) Where is the "Digital Music Player" subforum?
2) Are these products identified there? If not, what are they specifically?

The Digital music player forum is here: https://www.audiosciencereview.com/...-audio-player-reviews-and-discussions-dap.44/

 

[Arnold Krueger]

Sure but keep in mind this is not dbx or level matched. Oppo105, Emotiva Xda2 xda1 dc1

Here is a helpful tool for level matching in these days of digital volume controls with step sizes optimized for general listening, not lab work:

https://www.ebay.com/itm/Douk-Audio...916711?hash=item3d3ac3d167:g:cQcAAOSwX61ZE7d3

Depending on what you want to do, you might need up to 4 units to get levels and balance controls for 2 DACs.

The way to match levels is measure with a DVM connected say, across the relevant speaker terminals. Play a test file with relevant pure tones created in say, Audacity, switch back and forth between the DACs, and adjust as required to have the same voltage within 1%.

Here's a meter with adequate features: https://www.ebay.com/itm/UNI-T-UT61...492698&hash=item5d7c439cd9:g:4Y0AAOSwdPtawfLs

 

[Arnold Krueger]

Well it doesn't have to be listening tests. It could be in measurements of the gear using multiple tones with changing dynamics. I would go so far to say music with different instruments. But I fail to figure out how to make that a quantifiable measurement. So maybe tones first.

Please notice that we all understand the extra difficulties of doing good DBTs as compared to the relative simplicity of level matching. If you don't match levels intentionally, it doesn't generally happen by itself. You have to intentionally do it in most cases!.

If you don't match levels I guarantee that everything will sound different all the time, even two copies of the exact same audio gear.

So, when people refuse to do level matching, what's the use? They obviously want to hoot and hollar about hearing diffenences because they are too proud to try to make proper use of volume controls. Pretty simple, really.

 

[restorer-john]

Seriously, messing around with 20KHz high level tones is a recipe for instantly destroying tweeters and causing amplifiers to oscillate or worse.

Music does not contain high level, high frequency content. Speakers are not designed for that. Most tweeters will expire with 2-8 watts of HF content.

 

[Arnold Krueger]

Seriously, messing around with 20KHz high level tones is a recipe for instantly destroying tweeters and causing amplifiers to oscillate or worse.

Music does not contain high level, high frequency content. Speakers are not designed for that. Most tweeters will expire with 2-8 watts of HF content.

Totally agreed. Because of the prevalence of sighted evaluations which are generally positive for whatever the listener thinks he should hear, we end up doing a lot of extreme things such as playing CCIF IM tones though Digital Audio players and noticing that nothing bad happens.

Every time I've posted potentially damaging files here, I've been pretty clear about their potential dangers.

I take the 19 and 20 Khz comments above as an example, not a recommendation.

But its hard to be too careful with other people's gear. For the recommendations, I've made here about level matching, let me add that level matching at say 1 watt or 10 watts is easily as relevant as trying to do it at 50 or 100 watts, so I'm recommending people think about the former, not the lattter!

BTW many modern speakers can handle a lot of power with their tweeters - there was a sea change in tweeter power level capabilities after they invented ferro-fluid, high temperature adhesives and insulation some decades back.

For example some 2-way speakers have tweeters and woofers with comparable power ratings. Many tweeters can handle much higher power at 20 KHz better than they do at 3 KHz.

 

[garbulky]

A sine wave is constantly changing its volume, and its rate of change, and at a precisely predictable rate.



The expected value of two sines would also be predictable, if you wanted to analyze that result.



That sine wave - flattening of the peaks raises odd harmonics, if extended, maybe it raises the even harmonics (not sure on that one*)...

Dynamic variance - the sine sweeps the range, what ever power limit you set on it, and remains easily calculable and comparable...

What would you suggest for a test excitation? You could certainly record the output and compare for difference.

---

As far as I know, all sound can be created as combinations of sines, wavy, straight lines, combinations, all are combinations of sines at various frequencies and intensities.

So, I suppose, if a sine can't be reproduced properly at all frequencies, that's about enough test to reveal faults (or at least their intensity - nothing is perfect). What error would show somewhere else but not there?

---

*Test: Adding second harmonic, in phase (positive peaks aligned) with the base tone, causes the positive peak to be extended and the negative peak to be flattened, visually.

1000 and 2000Hz, combined at full value, then combined with the 2000Hz tone at -6dB and at -12dB

View attachment 11956

Did another read through. Great that's exactly what I was hoping for.

The mix graphs: I assume first is 1k and 2k combined at full value. Then the second combined with 2k -6db. And the third combined with 2k at -12 db?
And I assume if I want to offset the time (like have one interspersing wave come in slightly out of phase with the other), I use the time align function and use the right, left arrow keys? Or is there another more precise way to align the second wave where it intersects with the first? Like entering the start phase or something?

 

[RayDunzl]

The mix graphs: I assume first is 1k and 2k combined at full value. Then the second combined with 2k -6db. And the third combined with 2k at -12 db?

Yes, but "full value" of the 1 and 2khz tones is actually half-value each - else the combination would clip when (if) the peaks aligned.

And I assume if I want to offset the time (like have one interspersing wave come in slightly out of phase with the other), I use the time align function and use the right, left arrow keys? Or is there another more precise way to align the second wave where it intersects with the first?

There is a tool to drag the wave - the symbol looks like <-> in the toolbox. Zoom in to see where you're at. The selection tool leaves a marker that will show up in the moving wave at the selection point when you move and let go. Click the first wave, move the second, you'll see it.

And there are boxes at the bottom that give details on the selection point and range - by time, or samples.

 

[garbulky]

Yes, but "full value" of the 1 and 2khz tones is actually half-value each - else the combination would clip when (if) the peaks aligned.



There is a tool to drag the wave - the symbol looks like <-> in the toolbox. Zoom in to see where you're at. The selection tool leaves a marker that will show up in the moving wave at the selection point when you move and let go. Click the first wave, move the second, you'll see it.

And there are boxes at the bottom that give details on the selection point and range - by time, or samples.

Very helpful thank you