[bigshot]

Have you had a chance to look at my questions yet?

Please note that this discussion did not start with the need for > 16 bits. The argument started with Bigshot complaining that we don't even need to know have good response in the rightmost bits of a 16 bit sample!

Actually I was saying that noise at -90dB wasn't something that would be a problem for someone listening to music on a home stereo. (I fixed your spell checker error there.) And I wasn't intending to complain or argue. I was just putting specs in perspective.

 

[amirm]

You trash the Behringer at one point in here that I use at my desk all the time, and while it certainly has a high noise floor, it disappears once I'm actually playing music.

I don't trash "Behringer." I said that the $29 one doesn't perform as well as I like. Here is the full review: https://www.audiosciencereview.com/...eview-and-measurements-behringer-uca222.2036/

Here is a measurement from that comparing Behringer UCA222 against my laptop's built-in headphone out:

My laptop outperforms both in broadband noise and jitter/distortion spurs. It does the same on linearity of the rightmost 2 bits of 16 bit data (don't be confused by the "24 bit" moniker in this test):

On top of that my laptop has far higher output than the anemic one put out by the UCA222.

This is the data. It can only be challenged with new data not being grumpy that your DAC of choice is looking bad. That can't be helped. I buy these products on my own, review them on my own dime and publish them at no cost or advertising to anyone. You can ignore them but you can't complain about them. The UCA222 is old news, uses knock-off DAC chip and is highly limited to 44.1 and 48 Khz sampling with no support for async USB. Based on all of this I cannot recommend it.

And oh, until recently the Behringer UMC-204HD at just $79 was my recommended DAC. Here is that data which started this argument:

So I have nothing against Behringer at all. That unit by the way has been unseated by the Topping D30 which is $120 and produces most excellent results:

Again, this shows how clean the rightmost 2 bits of 16-bit data are. So no argument for > 16 bits.

 

[amirm]

Amirm, I am NOT disputing your ABX results, I achieved similar results many years ago in DBTs in my studio.

Well then it would have been nice to have some support on them rather that stay on the side letting folks create doubt about them.

I don't think anyone here is disputing that the digital noise floor (dither) can be heard given the correct circumstances. You highlighted "this may be audible to some listeners at elevated listening levels in a very quiet listening environment". I would add; "with some parts of some content". For example, a fade out to digital silence, which is often required in music (particularly acoustic music genres) if the final note decays for longer than is judged to be aesthetically desirable or if the note decays into the environmental noise floor of the recording, so all we're left with is that noise floor which obviously has to end at some point, either abruptly when the track ends or potentially less jarringly if we fade it out. However, if we're talking about "fidelity", then it is not expected/desired that consumers will actually hear the fade out all the way to the digital noise floor, it is expected that the fade out will at some point disappear beneath the noise floor of the consumers' listening environment, significantly before the digital noise floor is reached.

Fade out is but one condition of audibility here. There are lots of others such as notes decaying into background. And recordings that are not at 0 dbFS and hence are listened to at elevated levels, leaving you much less dynamic range.

Ultimately remember, as I keep saying, the argument started with not needing 16 bits. Are you saying all of this in the context of that too?

To be absolutely certain of this, even in the most extreme "real life" consumer circumstances, we have the option of decreasing the digital noise floor by using noise-shaped dither and indeed, the use of noise-shaped dither has been pretty standard practice when distributing 16bit music for roughly 20 years....

It has been known that many years and more. But standard in use? No way. As you note below, people doubt it is needed and do without dither altogether let alone using noise-shaping. Do ahead and examine your library of music and report back on what percentage of them use noise shaping.

... not withstanding the fact that even with good equipment, in a decently quiet environment and at somewhat/slightly elevated levels, listeners have been shown to not even be able to hear the difference between truncation and TDPF dither, let alone between TDPF dither and noise-shaped dither. Of course, even noise-shaped dither is audible if we have a very quiet listening environment and elevate the listening level enough. But we are talking about test conditions, conditions designed to enable the detection of dither NOT "real life", because in real life those conditions do not occur unless we are testing! In real life, in "a very quiet listening environment" we do NOT elevate listening levels, we do the exact opposite! And, all this only really applies to music content, in film (which you quoted extensively) we have a manufactured noise floor and that's what we typically fade into, not even into the environmental noise floor of the consumers' listening environment (the populated cinema), let alone the digital noise floor.

I am not here to put restriction on music, where it is listened, or how it is produced. I am here to advocate excellence in engineering and what must be there to achieve transparency in all cases.

And no, there was next to nothing about "film" in my article. I wrote that for Widescreen Review Magazine so there is a tiny reference. But the rest of the article and references within have little to do with film. Here is one graph from my article again:

Davies hall in San Francisco is a concert hall. Even skywalker sound is used for musical scores devoid of manufactured sound you mention. Here is an example:



The lesson there is that we cannot use single number SPL values. But rather we need to analyze the spectrum of noise and compare it to Hearing Threshold. Once we do, we see that recording venues can hit that threshold of hearing and as such be totally silent. So the argument that no recording is silent enough exists.

Furthermore, we can hear below noise floor of a room --- both in recording and playback. I can go on but you really short changing what I wrote and underlying research which has been peer reviewed and published in prestigious audio journal. Not sure what we are doing in this subform if the first thing we do is dismiss such research.

Referring to what I've quoted above, while in my last paragraph (and previously) I disagree with the level at which you've defined "golden" in real life, I agree entirely with what you've stated and I certainly appreciate your published measurements and want to know of products which fall below "golden". But, what about the other end of the spectrum? While we disagree on what is real life, even taking your criteria rather than mine, what about products which don't just achieve your criteria but exceed them, what is there beyond "golden" and how could it benefit anyone except marketers or those interested in a purely academic engineering exercise? You seem to be avoiding this question, no matter how I ask it!

G

My criteria is two fold:
1. For every system out there, I like to see clean, unadulterated reproduction of 16 bits. This is not hard to do and is dirt cheap as per my $120 DAC recommendation. The Topping D30 is a Chinese product and I assure you, no money is spent on marketing it in that cost which is less than 10 CDs you may buy.

2. For high performance systems, I like to see full 20 bit reproduction as to provide complete transparency. The latest Benchmark HC3 gets to 21 bits and it has my praise for that. And at $2K they charge for it, it is "reasonable" to me in the context of a high-end, ultimate DAC you could get that you can keep for years and years. Again, you are not rewarding any "marketing." You are rewarding excellence in engineering in a low-volume market which drives the costs up.

As you see, I have no patience for people who try to dumb things down below #1. Who are they protecting with that??? Saving people from spending $120 instead of $29? Let's get real.

 

[bigshot]

I’m going to reduce my questions down to one in hopes you’ll answer it. Did you gain ride the test, or did you have a fixed listening level throughout the track?

 

[reginalb]

I don't trash "Behringer." I said that the $29 one doesn't perform as well as I like. Here is the full review: https://www.audiosciencereview.com/...eview-and-measurements-behringer-uca222.2036/

I didn't say you trash Behringer, I said you trashed the one that I use. And don't worry, I don't care. I bought it for features (the ADC) at a price point that I was willing to pay for something that I do as a hobby. I only started feeding my amp with it because it was sitting around, and I had an amp with no DAC built in. That it outputs at a lower level than a computer is of no consequence when I'm just using the line-out off the DAC.

 

[amirm]

I’m going to reduce my questions down to one in hopes you’ll answer it. Did you gain ride the test, or did you have a fixed listening level throughout the track?

You are not going to learn anything by asking me to help you with your argument. Why don't you run a blind test, ride the gain, horses or even a donkey and see if you can pass them. Then you will speak from experience.

 

[castleofargh]

...I think it's best to define what the purpose you're testing for is. If you're testing to find a way to prove you hear things no one else can, then you might design a different sort of test than if you're just testing to see what the threshold is for a piece of equipment you plan to use for normal listening in your home.

oh I'm very clear about the question I'm trying to answer when I set up a test. my mistake is more about assuming I know what someone else is talking about, because I think he's trying to answer the very same question I had. but sometimes, he's not and it takes me a lot of time to see it.
this results in taking experiences out of their initial context by mistake.
I don't know if I'm clear or digging even deeper ^_^.

 

[bigshot]

You are not going to learn anything by asking me to help you with your argument. Why don't you run a blind test, ride the gain, horses or even a donkey and see if you can pass them. Then you will speak from experience.

OK. I gave you a chance to explain. The caginess of your replies says a lot more than your cut and paste plain text ABX test results. I think you cut the tracks down to a tiny sliver of fadeout, then boosted the volume to an unGodly level so you could tell them apart. That doesn't take "experience" or "training" to discern. Even with with her hearing aid on low batteries, my grandmother could tell the difference between 16 and 24 if you turn the volume up high enough. If you're trying to impress us to get us to follow your review site, you're doing it the wrong way. If you're trying to prove you can hear stuff that no one can hear at normal listening volumes, you might as well the math yourself and edit the text of the test results before you post them. That would save a lot of time and accomplish pretty much the same thing.

I also re-sampled the files myself

Yes.

 

[amirm]

OK. I gave you a chance to explain.

Well, in that case, thank you very much!

The caginess of your replies says a lot more than your cut and paste plain text ABX test results. I think you cut the tracks down to a tiny sliver of fadeout, then boosted the volume to an unGodly level so you could tell them apart. That doesn't take "experience" or "training" to discern. Even with with her hearing aid on low batteries, my grandmother could tell the difference between 16 and 24 if you turn the volume up high enough. If you're trying to impress us to get us to follow your review site, you're doing it the wrong way. If you're trying to prove you can hear stuff that no one can hear at normal listening volumes, you might as well the math yourself and edit the text of the test results before you post them. That would save a lot of time and accomplish pretty much the same thing.

You have incredible imagination there and assume people have awfully low ethical standards. Be that as it may, the results of those tests were intensely argued about on multiple forums going on for days and weeks by people far more aggressive and knowledgeable than you. To your earlier suggestion of donating myself to AES, the results were also shared with chair of high-resolution audio group, Vicki Melchior and the rest of the team in that working group. I have to tell you, folks are tired of constant tune you sing that fidelity doesn't matter. This is why latest papers by Stuart et. al. were welcome (see my digest and article: https://audiosciencereview.com/forum/index.php?threads/high-resolution-audio-does-it-matter.11/). This work also generated a lot of buzz in forums and industry resulting in me writing another follow up which answers your question (partly) about how ABX tests work: https://audiosciencereview.com/forum/index.php?threads/statistics-of-abx-testing.170/.

I suggest that instead of this kind of personal tone, you spend your energy learning about blind tests. Google is your friend. As is ASR Forum, Hydrogen Audio, etc. Until such time that you actually put yourself in such a situation, i.e. controlled blind testing, I suggest not going after people so aggressively as you have been doing. Being an objectivist require more than spending time on forums and reading its ten commandments. If you think I am using tricks, try to duplicate my results. Don't assume those techniques work or work all the time. I have presented plenty of tests for you to run and try to duplicate -- cheating or not.

On issue of cheating, later versions of ABX plug-in for foobar have checksums to guard against manipulating the files in any way. Here is an example of me passing 30 Hz jitter (very hard to do due to high level of masking) showing these safeguards:

foo_abx 2.0 beta 4 report
foobar2000 v1.3.5
2014-12-09 14:24:40

File A: 30 Hz jitter strong level .025.flac
SHA1: 54719c17fd29d0546b79f50bd7e3c61de1dd025d
File B: no jitter.flac
SHA1: 262cd6c4d4c73502a0142f867b00aae013fd13ce

Output:
DS : Primary Sound Driver

14:24:40 : Test started.
14:25:00 : 01/01
14:25:06 : 02/02
14:25:16 : 03/03
14:25:21 : 04/04
14:25:27 : 05/05
14:25:34 : 06/06
14:25:39 : 07/07
14:25:45 : 08/08
14:25:51 : 09/09
14:25:56 : 10/10
14:25:56 : Test finished.

----------
Total: 10/10
Probability that you were guessing: 0.1%

-- signature --
ba16bda939028d34d8b131283f9d46709dab36f9

There is a tool where you feed these to it and it will verify if you have or have not messed not only with the audio files but also the text output of foobar ABX plugin: http://www.foobar2000.org/abx/signaturecheck

When I feed the above text to it, it outputs:

Signature matches; the log appears to be valid.​

You can try it yourself and verify the same. Cheat and this kind of measure comes along and hangs you by your shorts!

Again, go and see if you can pass all of these tests using whatever method you choose. Regardless of outcome, you will gain a new appreciation and firsthand feel for these topics. There is a great path here to becoming a critical listener. Once you get there, you will see how there is a gap between you and many others. It doesn't come easy or cheap though. You must put in the time.

 

[spruce music]

The masking demonstration that clinched it for me was the music with the sousa band mixed in at about a -60 db level. (Anyone have the link? I can't find mine.)

http://www.libinst.com/diffmaker_example_files.htm

Listener challenge down near the bottom of this page. Guy who made Diffmaker.

 

[amirm]

Actually I was saying that noise at -90dB wasn't something that would be a problem for someone listening to music on a home stereo. (I fixed your spell checker error there.) And I wasn't intending to complain or argue. I was just putting specs in perspective.

Well, we have what you said:

I don't understand why someone would say that one DAC is better or worse than another because it measures higher or lower while still being well beyond the threshold of transparency. I judge equipment by its ability to do the job of playing music for human ears. I haven't run across any player or DAC or cable or amp in the past 20 years that doesn't fit the bill on that score. Sound Science people can sometimes be just as prone to focus on stuff that doesn't matter as audiophools. If they enjoy the mental exercise, that's fine. But I sure hope no one takes this stuff into account when they make buying decisions. It's a waste to spend money on things you can't hear. Better to focus on features and usability.

You were selling people on anything that produces music, by definition must be good and better than what we hear. That assertion has been made by you repeatedly but with nary of a reference, research study, nothing really. Just an assumption that this must be the case. And hence, no one should bother or care about the work I am doing. That is an insult besides lack of understanding of audio fidelity and requirements for channel transparency for all listeners and content.

You talk about "20 years" of DAC experience. Exactly how many DACs have you used and how do you know they meet the threshold of detection? What formal tests did you do? Do you know what test would be revealing of such?

I have had at least three DACs that easily transmitted computer noise through their output! I have had two other DACs that clip samples near 0 db causing buzzing. I have had DACs that perform worse than what is in your computer already. Currently I probably have 10-15 DACs with more coming. So please put aside your assumption that you already know what you need to know. You don't. We have fed you are PR lines and you are running with them without the proper knowledge and experience to defend them.

There is so much garbage here because both objectivists and subjectivists have given a pass to equipment makers to release products with nary of a measurement. The is more information on a tire sidewall than there is for audio gear! Subjectivists say measurements don't matter. Extreme objectivists say the same because they are all perfect for our ears. Well guess what, neither is correct. There is less or more competence, science and engineering in some products or others. The only way to get there is to examine them on our nickel which is what I am doing. Last thing we need is someone running interference to stop people from either measuring or paying attention to the results.

So please think through what you are doing. You are not doing any favor to the community. You are protecting manufacturers and allowing them to build subpar equipment. Spend the energy learning what is being measured, why it may matter, and advocate excellence in engineering. Don't belittle it because it makes for less rosy chanting like for objectivism.

 

[spruce music]

Well, we have what you said:



You were selling people on anything that produces music, by definition must be good and better than what we hear. That assertion has been made by you repeatedly but with nary of a reference, research study, nothing really. Just an assumption that this must be the case. And hence, no one should bother or care about the work I am doing. That is an insult besides lack of understanding of audio fidelity and requirements for channel transparency for all listeners and content.

You talk about "20 years" of DAC experience. Exactly how many DACs have you used and how do you know they meet the threshold of detection? What formal tests did you do? Do you know what test would be revealing of such?

I have had at least three DACs that easily transmitted computer noise through their output! I have had two other DACs that clip samples near 0 db causing buzzing. I have had DACs that perform worse than what is in your computer already. Currently I probably have 10-15 DACs with more coming. So please put aside your assumption that you already know what you need to know. You don't. We have fed you are PR lines and you are running with them without the proper knowledge and experience to defend them.

There is so much garbage here because both objectivists and subjectivists have given a pass to equipment makers to release products with nary of a measurement. The is more information on a tire sidewall than there is for audio gear! Subjectivists say measurements don't matter. Extreme objectivists say the same because they are all perfect for our ears. Well guess what, neither is correct. There is less or more competence, science and engineering in some products or others. The only way to get there is to examine them on our nickel which is what I am doing. Last thing we need is someone running interference to stop people from either measuring or paying attention to the results.

So please think through what you are doing. You are not doing any favor to the community. You are protecting manufacturers and allowing them to build subpar equipment. Spend the energy learning what is being measured, why it may matter, and advocate excellence in engineering. Don't belittle it because it makes for less rosy chanting like for objectivism.

I think you make a very important point. I have assumed mostly you can forget about most modern gear as having a sound. I think probably more than any other time in history that is true. However some people design for a sound which isn't about fidelity. And some seem intent on shearing the sheep just because conditions have made it possible in the subjectivist camp. I think Schiit products are a good example. Almost all have some oddity to their operation, they don't have terrific measured specs, their reputation is pretty good because they put a human face on their company in forums like this one. That is good, well except for some of the product.

I do remember when Audio (and Stereo Review) did testing of things they reviewed. Mostly things met spec or exceeded it or came close. Amps adhered more or less to the FTC guidelines. With the rise of subjective audiophiles measurements being meaningful were sneered at. Thankfully JA at Stereophile and people at Soundstage.net still do measurements, but so many publications don't. I understand it somewhat with the proliferation of products who could keep up. And hell must be the guy who measures every aspect of AVR's for a living. Yet some products are taking liberties with the situation. They feel safe making claims. Using chip specs as listed specs without showing they actually managed that or just making things up or saying they don't believe in your usual measurements. Oh well, long as it doesn't hurt sales nothing is likely to change. Seems strange to disagree with someone trying to somewhat dot the I's and cross the T's. Providing info on products actual performance.

 

[bigshot]

You have incredible imagination there and assume people have awfully low ethical standards.

You ask me to believe you when I offered to set up a test for you that you weren't in control of, and you ignored me. I think you're creating your own test so you can control the outcome. That's why DBX has the DB in it.

I'm down to reading the first line or two of your posts now. No need to type out long justifications.

 

[gregorio]

[1] Well then it would have been nice to have some support on them rather that stay on the side letting folks create doubt about them.
[2] Fade out is but one condition of audibility here. There are lots of others such as notes decaying into background.
[3] And recordings that are not at 0 dbFS and hence are listened to at elevated levels, leaving you much less dynamic range.
[4] It has been known that many years and more. But standard in use? No way.
[4a] As you note below, people doubt it is needed and do without dither altogether let alone using noise-shaping.
[5] And no, there was next to nothing about "film" in my article.
[6] But rather we need to analyze the spectrum of noise and compare it to Hearing Threshold.
[7] I am not here to put restriction on music, where it is listened, or how it is produced. I am here to advocate excellence in engineering ...
[7a] For high performance systems, I like to see full 20 bit reproduction as to provide complete transparency. The latest Benchmark HC3 gets to 21 bits and it has my praise for that. And at $2K they charge for it, it is "reasonable" to me in the context of a high-end, ultimate DAC you could get that you can keep for years and years. Again, you are not rewarding any "marketing." You are rewarding excellence in engineering in a low-volume market which drives the costs up.
[7b] Saving people from spending $120 instead of $29? Let's get real.

1. I'm on both sides! I support your (and my) DBT results which demonstrate it can be detected under test conditions but I also support those "folks" who are effectively saying it can't under normal listening conditions. Can't you see/admit that both sides can simultaneously be correct?
2. Notes decaying into what background? The background you quote of an empty hall with no HVAC or the background of a hall full of musicians and commonly therefore a legal requirement to have the HVAC on? And obviously, if it's the former, who is playing the notes which decay into that background?
3. True but of course if the recording peaks at say -6dBFS and you compensate by elevating the gain, then you do not have a 16 bit recording, you only have 15 bits because the MSB in not being used.
4. Yes it's at least a standard recommendation and many commercial engineers apply noise-shaped dither as standard. I agree though that it's not always applied ...
4a. Isn't this statement effectively an argument against your assertion? Those "people" are content creators, they're doing their creating in reasonable to excellent listening/monitoring environments and at least some of the time at somewhat elevated levels and yet they're not aware of even truncation distortion, let alone noise-shaped dither! Or, if they are aware they choose not to apply dither, why do you think that is?
5. Come on, do you really need your own article quoting back to you? "Before diving deep let’s cover some basics. Movie soundtracks come in digital form on the Blu-ray disc.". "If you have a dedicated theater or listening space". "you can convince yourself that we could go back to cassette tapes and still have sufficient dynamic range for our movies!". "Let’s review the measurements he took with respect to noise floor for a sampling of live halls and a film recording studio". "When you stand outside of your theater..."
6. Agreed BUT (!) if you are going to quote real life then that means not ONLY analysing the noise of empty concert halls and recording studios (with the HVAC turned off)! You have to also look at those halls with musicians in them, you also have to look at the noise/signal ratio of the mics + pre-amps and the "real life" usage/positioning of them, plus what the creators/producers wish to achieve! And thanks for your example video. You did notice that there was a piano and pianist in the studio didn't you? Did you also notice the mics used, where they were positioned and considered the impact on noise/dynamic range that "real life" situation would have? Apparently not, but why? Do we only need to "analyse the spectrum of the noise" which supports your argument or should we analyse the spectrum of ALL the noise which exists in "real life"?
And BTW, yes, I know what Davis Hall is, I've worked there, as well as many, many others. I've also worked extensively with the BBC, their recording rooms and the concert venues where they record. I've never been to Skywalker Sound but I know very well what it is.
7. Exactly, that is exactly my point, a point which it's getting difficult not to assume you are now wilfully ignoring/avoiding ...
7a. AGAIN! If 20 bit reproduction provides "complete transparency" or as you described it before, is "golden", what is there beyond complete transparency/golden? What audible benefit does the 21 bits of the HC3 give you that a "completely transparent" and "golden" 20 bit DAC does not give you? How is that extra bit "reasonable" except in terms of a bigger number for marketing or academic engineering excellence? ...

7b. Amirm you are obviously an intelligent man and apparently well informed. So why are you playing the game of mis-direction and omission, unless it's to fulfil a some personal agenda? You are omitting many of the real life practicalities of recording and the production/mastering goals and you are mis-directing with your statement quoted here. Yes, no one is disagreeing that it's very important to have measurements, to identify under-performing units but the issue here is also the other side of the equation, the side which you CONTINUALLY AVOID, what about the over-performing units. Assuming your quoted $120 D30 provides your 20bit "completely transparent"/"golden" performance requirement and using your also quoted $2k HC3, then what about the other side of the equation: "saving people from spending $2k instead of $120"? What audible benefit does that extra $1,880 provide, if $120 already gives you "golden"/"completely transparent"?

G

 

[Darren G]

A 90 degree point of view -

I was a first generation CD adopter. A few hours of WOW, followed by a long period of 'this sucks'. Okay, well that was long time ago, and digital audio quality is amazing today. I write that only to point out, I agree that just because it's digital does not mean flawless.

Still, the harsh truth I realized about myself... nothing can cure obsessively looking for flaws. If someone is dead set on finding flaws, that is all they are going to do, and it won't matter how objectively correct (or incorrect) the gear is. They are going to find a flaw. They will convince themselves there is a flaw, and there are no objective arguments that will sway them. That is not to say there is no objective truth, just that the only thing that can fix an obsessive focus on flaws requires a change in personal focus. Learning to enjoy the music first, and focusing on flaws second (if even that), is the only way to really get to that point of the gear is 'good enough'.