Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: 24-bit audio proposed for iTunes (Read 44618 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

24-bit audio proposed for iTunes

Reply #100 –
Quote from: NwAvGuy on
Quote from: saratoga on
But are those really so useful to know?  Pretty much every player in existence has an output level with 3-4 dB of every other player, since they're all running off the same supply voltage.  Yeah its interesting to know if your player is marginally louder then another player, but theres so little difference between players I'd hardly consider it critical.  Likewise, if a player a sufficiently flat frequency response, is it really that important to know the precise output impedance?

Yes they are important! The load interacts with the output impedance in a large ways that have little to do with the power supply voltage. And some amps simply don't have the ability to drive low impedance loads to typical levels before the distortion goes off the scale. For example, most portable headphones these days are 16 ohms. And into 15 ohm loads, I've measured anywhere from 179 mV RMS at 1 % THD up to 800+ mV. That's a range of 2.1 mW up to 42.7 mW--a 13 dB difference not a 3-4 dB difference as you suggest.

For example, here's what happens with a 6 ohm output impedance device into typical balanced armature (Ultimate Ears SuperFi in this case) headphones:

(img)

Does 4+ dB of variation matter? Some devices have 20 - 200 ohm output impedance--far greater than the 6 ohms show above. And, obviously, cause far greater deviations from flat frequency response.



Note that I said, "if a player a sufficiently flat frequency response, is it really that important to know the precise output impedance?".  I'm aware that lots of devices don't have a perfectly flat spectrum when driving low impedances.  But if RMAA tells you one does, what additional benefit does output impedance get you exactly?

Furthermore, I don't know what you mean by "some devices have 20 - 200 ohm output impedance".  Certainly not digital audio players, as they would waste so much battery power just heating their output buffering stage.


Quote from: NwAvGuy on
Quote from: saratoga on
Wouldn't DAC nonlinearity show up in the distortion tests though?  Yeah you won't get the exact number, but do you really need that?

That's a fair question, and based on what I know and have seen, the answer is: "not usually". It's really tricky to take accurate distortion measurements at very low levels as many of the harmonics are lost in the noise.


Does that matter?  If a DAC is nonlinear at low levels, and you measure nonlinear mixing between two near full scale sin waves, then you will get a fairly large amplitude peak at sum or difference frequencies.  I don't think you actually have to generate a low intensity signal to measure this effect.


Quote from: NwAvGuy on
And when you test at higher levels, other sources of distortion usually dominate the measurement. So DAC nonlinearity can be lost either way.


Err, if distortion is already generally negligible on most DACs, and this effect is so small that its masked by other distortion, then what purpose does measuring it serve?  What you actually want to check for is audible levels of nonlinearity.  If nonlinearity is so low you cannot practically measure it, then it is clearly inaudible. 

Quote from: NwAvGuy on
So far I've been concentrating on portable equipment, and for that, I mostly agree with you. But I don't have a dScope just to test $29 portable players. That said, even the iPod Touch 3G will give many sound cards a good run for their money in terms of noise and distortion. So it's useful to get real numbers and not be flirting with the limitations of the PC's sound hardware.


Yes, but you specifically complained about people using RMAA to compare MP3 players.  I'm saying that this is unfair, its a perfectly adequate tool for doing so.

Quote from: NwAvGuy on
RMAA is great in that it's easy to use. But, as I've explained, it's very much "garbage in/garbage out" and you're still missing lots of important things (including what we just discussed above). To get the rest of the numbers with a soundcard, meter, etc. is considerably more work. And even with all that there are still typical specs you can't measure (or measure very well). Like I said, RMAA and soundcards are certainly useful. But I don't think it's accurate to imply an audio analyzer has little advantage making real world measurements.


I don't think you've explained how its "garbage in/garbage out" nor how its "missing lots of important things".  You've listed things it missing, but you haven't made a convincing argument that any of them are important.

24-bit audio proposed for iTunes

Reply #101 –
Quote from: Arnold B. Krueger on
Quote from: NwAvGuy on
It also doesn't measure the DAC linearity (unless you count the -60 dB spectrum plot but virtually any DAC has near zero error there) or square wave response.


Sigma Delta DACs (the industry standard for about the past decade or more) can't possibly have linearity problems. That's one reason why virtually every DAC has low error at FS-60 dB .

Of course if every DAC actually had near zero error at -60 dB then they would all give the same results in a standard dynamic range test. But they don't.

As far as square wave response goes, this sort of thing should really be done with an ADC that is sampling at a fir higher rate then the UUT or an actual oscilloscope.

Square wave testing has gone out of favor because any related sonic problems would show up in a frequency response test.

I should probably give it up, but there are plenty of examples of Delta-Sigma DACs with linearity error. It can be caused by lots of things including the power supply/voltage reference used in the implementation or noise sources. If you'd like, I'll point you to some references measured with Audio Precision gear. I've also seen it personally.

And why do you care about square wave information on an audio device that's beyond 96 Khz? That's the only difference between viewing the square wave of a faster scope vs the dScope. In fact, because most faster scopes have only 8 bit DACs, once can argue the 24 bit DACs in the dScope actually yield more useful information--especially if you want to zoom in on a particular vertical anomaly (like say ringing, etc.). I have lots faster scopes (including a high-end 4 Gs/sec Agilent) but I don't use it for audio use when the dScope does a better job.

Square waves can tell you a lot about any amplifiers output stability (including feedback loop design) and also the type of digital/analog low pass (reconstruction) filtering used after a DAC--i.e. is it a linear phase design, etc. Neither of these show up in any consistent or conclusive way in frequency response tests.

24-bit audio proposed for iTunes

Reply #102 –
Quote from: saratoga on
Does that matter?  If a DAC is nonlinear at low levels, and you measure nonlinear mixing between two near full scale sin waves, then you will get a fairly large amplitude peak at sum or difference frequencies.  I don't think you actually have to generate a low intensity signal to measure this effect.


Some noise levels don't seem to scale linearly with output power, even when the DAC chip itself has good linearity, for example due to improper shielding/grounding/power supply/etc, and they are much more audible at lower levels. The measurement of those artifacts mostly relates to the recording device's (or headphone's) sensitivity and not to the loaded impedance. So you can get cases where RMAA looks excellent while your IEMs and better measurement equipment say otherwise.

24-bit audio proposed for iTunes

Reply #103 –
Quote from: saratoga on
Note that I said, "if a player a sufficiently flat frequency response, is it really that important to know the precise output impedance?".  I'm aware that lots of devices don't have a perfectly flat spectrum when driving low impedances.  But if RMAA tells you one does, what additional benefit does output impedance get you exactly?

If you're talking about one person using their current headphones then yes, you can verify the device is sufficiently flat with RMAA. But what good is that information for all the other people trying to use the posted RMAA data that have different headphones? It's especially useless unless the person also posted an impedance plot of the headphones--which by the way is yet something else RMAA won't do that the dScope does. So what you suggest is really useless in terms of a general result that others can fairly use for comparison.

Quote from: saratoga on
Furthermore, I don't know what you mean by "some devices have 20 - 200 ohm output impedance".  Certainly not digital audio players, as they would waste so much battery power just heating their output buffering stage.

You would think that would be true, but it's not. I've measured output impedance of ~50 ohms on portable gear and my HP PC's headphone jack is 170 ohms. The iPod Touch 3G is 7 ohms.

Quote from: saratoga on
Does that matter?  If a DAC is nonlinear at low levels, and you measure nonlinear mixing between two near full scale sin waves, then you will get a fairly large amplitude peak at sum or difference frequencies.  I don't think you actually have to generate a low intensity signal to measure this effect.

I guess the short answer is it matters to some people. We can turn this into the endless complex discussion of what levels and kinds of distortions are audible, but that's not what you asked. I thought the issue was if DAC non-linearity was accurately reflected in THD measurements. The answer is no. If you want to ask if measurable DAC non-linearity matters, that's a different question.

Quote from: saratoga on
Yes, but you specifically complained about people using RMAA to compare MP3 players.  I'm saying that this is unfair, its a perfectly adequate tool for doing so.

I maintain it's far from "unfair". I can point you to all sorts of misleading RMAA results on the web of portable players. A perfect example is I have an older Sansa player (or the new Cowon i9 for that matter) that rolls off in audible ways at bass frequencies when driving real 16 ohm headphones. But with no load, they're both flat to 20 hz. Many (most?) RMAA tests are run with no load but that's never disclosed. So it's easy to have misleading information.

And you're completely ignoring the 13 dB of power output difference not measured by RMAA. Or the fact that just because it shows reasonably flat response with dynamic headphones, the results could be wildly different with balanced armature headphones.

Quote from: saratoga on
I don't think you've explained how its "garbage in/garbage out" nor how its "missing lots of important things".  You've listed things it missing, but you haven't made a convincing argument that any of them are important.

That's your opinion. There are several examples of "garbage" in my RMAA blog article and I've presented several in this thread. If you're not convinced, that's fine. I guess we have very different ideas of what matters.

24-bit audio proposed for iTunes

Reply #104 –
Quote from: googlebot on
Some noise levels don't seem to scale linearly with output power, even when the DAC chip itself has good linearity, for example due to improper shielding/grounding/power supply/etc, and they are much more audible at lower levels. The measurement of those artifacts mostly relates to the recording device's (or headphone's) sensitivity and not to the loaded impedance. So you can get cases where RMAA looks excellent while your IEMs and better measurement equipment say otherwise.

You claimed this before and never answered back as to if you had tested (your Sansa) again with a load.
I believe you will see the noise in question if you perform the RMAA test loaded.
Creature of habit.

24-bit audio proposed for iTunes

Reply #105 –
Quote from: googlebot on
Quote from: saratoga on
Does that matter?  If a DAC is nonlinear at low levels, and you measure nonlinear mixing between two near full scale sin waves, then you will get a fairly large amplitude peak at sum or difference frequencies.  I don't think you actually have to generate a low intensity signal to measure this effect.


Some noise levels don't seem to scale linearly with output power, even when the DAC chip itself has good linearity, for example due to improper shielding/grounding/power supply/etc, and they are much more audible at lower levels. The measurement of those artifacts mostly relates to the recording device's (or headphone's) sensitivity and not to the loaded impedance. So you can get cases where RMAA looks excellent while your IEMs and better measurement equipment say otherwise.


I'm not really sure how to interpret this response to my statement.  Are you disagreeing with me about nonlinear mixing, or did you just quote the wrong block of text?

24-bit audio proposed for iTunes

Reply #106 –
Quote from: NwAvGuy on
Quote from: saratoga on
Note that I said, "if a player a sufficiently flat frequency response, is it really that important to know the precise output impedance?".  I'm aware that lots of devices don't have a perfectly flat spectrum when driving low impedances.  But if RMAA tells you one does, what additional benefit does output impedance get you exactly?


If you're talking about one person using their current headphones then yes, you can verify the device is sufficiently flat with RMAA. But what good is that information for all the other people trying to use the posted RMAA data that have different headphones? It's especially useless unless the person also posted an impedance plot of the headphones--which by the way is yet something else RMAA won't do that the dScope does. So what you suggest is really useless in terms of a general result that others can fairly use for comparison.


Its not useless at all.  While different headphones have different impedance vs. frequency curves, in practice two people both using 16 ohm IEMs will typically get similar results unless one has particularly weird headphones.  Thats why knowing the output impedance doesn't really get you anything. 

Quote from: NwAvGuy on
Quote from: saratoga on
Furthermore, I don't know what you mean by "some devices have 20 - 200 ohm output impedance".  Certainly not digital audio players, as they would waste so much battery power just heating their output buffering stage.

You would think that would be true, but it's not. I've measured output impedance of ~50 ohms on portable gear and my HP PC's headphone jack is 170 ohms. The iPod Touch 3G is 7 ohms.


While I'm skeptical about those numbers, I'd like to point out that your PC is not a portable audio player, and so clearly not what I was referring to.

Quote from: NwAvGuy on
Quote from: saratoga on
Does that matter?  If a DAC is nonlinear at low levels, and you measure nonlinear mixing between two near full scale sin waves, then you will get a fairly large amplitude peak at sum or difference frequencies.  I don't think you actually have to generate a low intensity signal to measure this effect.

I guess the short answer is it matters to some people. We can turn this into the endless complex discussion of what levels and kinds of distortions are audible, but that's not what you asked. I thought the issue was if DAC non-linearity was accurately reflected in THD measurements. The answer is no. If you want to ask if measurable DAC non-linearity matters, that's a different question.


Wait what?  I'm saying that if your DAC has significant nonlinearity, that will be reflected in THD and IMD measurements.  Thats why nonlinearity matters, it generates distortion.  If its not generating distortion, then why would you care about it?

Quote from: NwAvGuy on
Quote from: saratoga on
Yes, but you specifically complained about people using RMAA to compare MP3 players.  I'm saying that this is unfair, its a perfectly adequate tool for doing so.

I maintain it's far from "unfair". I can point you to all sorts of misleading RMAA results on the web of portable players. A perfect example is I have an older Sansa player (or the new Cowon i9 for that matter) that rolls off in audible ways at bass frequencies when driving real 16 ohm headphones. But with no load, they're both flat to 20 hz. Many (most?) RMAA tests are run with no load but that's never disclosed. So it's easy to have misleading information.


Yes, if you test without a load the results aren't very relevant.  But so what?  Test with a load and you're good.
Quote from: NwAvGuy on
And you're completely ignoring the 13 dB of power output difference not measured by RMAA. Or the fact that just because it shows reasonably flat response with dynamic headphones, the results could be wildly different with balanced armature headphones.


Which two players have a 13 dB difference?  And I don't doubt that some headphones exist that are extremely hard to drive, but they're the exception, not the norm.

Quote
That's your opinion. There are several examples of "garbage" in my RMAA blog article and I've presented several in this thread. If you're not convinced, that's fine. I guess we have very different ideas of what matters.



The only points I recall you making on your blog were complaints about people doing unloaded RMAA tests, and something about one of your tests rolling off high frequencies inexplicably which I think was actually do to some oddness with your sound card.  Pardon me for finding these unconvincing.  To convince me you would have to have some measurement that tells me something I didn't already know just from RMAA.  Something surprising and useful, not merely something interesting in an academic sense.

24-bit audio proposed for iTunes

Reply #107 –
Quote from: saratoga on
While different headphones have different impedance vs. frequency curves, in practice two people both using 16 ohm IEMs will typically get similar results unless one has particularly weird headphones. Thats why knowing the output impedance doesn't really get you anything.

Knowing the output impedance gets you a lot. A device with a < 1 ohm output impedance will deliver it's measured frequency response within a fraction of a dB to nearly any headphone load. As the graph I posted earlier shows, one with even a 6 ohm output impedance clearly will not. You don't find that useful?

And certainly not everyone uses 16 ohm headphones. And even among those rated at 16 ohms, the differences can run wide. My SuperFi IEM's peak at 80 ohms and dip down to below 10 ohms and peak at about 1200 hz:



And here's a pair of Sony IEMs which fluctuate by only 1 ohm from 17 to 18 ohms and peak at 5000 hz:



My Eytmotics have a totally different shaped curve than either of the above and peak at 100 ohms. Some IEM's have multiple resonances while double and triple driver balanced armature IEMs get even more weird. You apparently haven't measured, or seen, many of the impedance plots?

Can you honestly tell me the two headphones above, will yield "similar results" connected to the 7 ohm output impedance of an iPod Touch 3G? They won't. Not even close. But they will connected to a Sansa Clip+. So your statement above, and the one earlier, are simply wrong. It does matter--at least to lots of people. If you don't care about audible frequency response swings that's your choice but don't mislead others down the wrong path.

And, more generally, the whole point I've been trying to make is what's required to publish data that can be reasonably compared between reviewers and devices. Any engineer worth the title will tell you the best way to do that is to control as many of the variables as reasonably possible to get consistent results. Published data that cannot be compared, or verified, is generally considered of little use. In the medical and scientific communities it's generally considered worthless.

If you test the 7 ohm output iPod with the Sony's IEMs above while the guy reading your results has the SuperFi's, what happens? He reads your review, sees the flat response, buys the iPod, and ends up with weird sound that's not even close to flat, and is disappointed. It's very misleading to do what you're suggesting--especially when both the impedance plot of the headphone used, and the output impedance of the device are complete unknowns. Anyone who understands math will tell you have too many unresolved variables to get anything resembling the final frequency response. If you can't see that I don't know what else to say?

Quote from: saratoga on
[To convince me you would have to have some measurement that tells me something I didn't already know just from RMAA.  Something surprising and useful, not merely something interesting in an academic sense.

Apparently hard facts, ohms law, etc. are not convincing enough for you? If frequency response swings of 8+ db are not "useful" enough, how about 13 dB more volume? Is that only academic to you as well? There are plenty of portable players, lots even, that can't drive plenty more low-sensitivity headphones to sufficient volumes. It's a very common complaint with some portable players. And 13 dB is huge.

But this has moved way off topic. I've been under the impression HydrogenAudio is about getting the facts right and being objective so that's all I've been trying to do. Someone else chime in if they want?

24-bit audio proposed for iTunes

Reply #108 –
Quote from: saratoga on
I'm not really sure how to interpret this response to my statement.  Are you disagreeing with me about nonlinear mixing, or did you just quote the wrong block of text?


My initial answer concerned DAC non-linearities. I generalized it somewhat shortly after and now just realized that the new version has no relation to the quoted text anymore.

24-bit audio proposed for iTunes

Reply #109 –
Quote from: saratoga on
Quote from: NwAvGuy on
It also can't measure the famous pitch error of the Sansa.


Thats a fair point, although this is a pretty rare problem.



My question is whether or not one can detect the approximate 0.25% pitch error from one of the existing test results that RMAA already provides.

For example, the RMAA THD plot has a strong carrier @ 1.000 KHz, which if memory serves from the tests I ran,  the Clip shifts to 1.0025 KHz.  Using the expanded frequency scale feature that RMAA provids, I think that this error can be properly detected and estimated.

A similar argument can be applied to jitter. While 1 KHz is not an ideal frequency for the purpose (10 or 11 KHz would be better), audible amounts of jitter can be detected and estimated in a similar fashion from the existing 1 KHz tests. Furthermore, RMAA allows you to move the existing test signals aroound in the configuration menu.

24-bit audio proposed for iTunes

Reply #110 –
Quote from: NwAvGuy on
Apparently hard facts, ohms law, etc. are not convincing enough for you? If frequency response swings of 8+ db are not "useful" enough, how about 13 dB more volume? Is that only academic to you as well? There are plenty of portable players, lots even, that can't drive plenty more low-sensitivity headphones to sufficient volumes. It's a very common complaint with some portable players. And 13 dB is huge.


You;'ve missed an improtant point. You don't need a $10K test set to show the effects of IEM load impedance on a digital player. All you need is RMAA,  a $5 Radio Shack 3.5 mm splitter, and a pair of headphones.

If you want to test the effects of standard resistive loads, buy a pack of 15 ohm and/or 33 ohm resistors for aboout a $buck each, and solder  them into some 3.5 mm male headphone plugs that you can get for I think less than $3 a pair at Radio Shack.

24-bit audio proposed for iTunes

Reply #111 –
Quote from: NwAvGuy on
Quote from: Arnold B. Krueger on
Quote from: NwAvGuy on
It also doesn't measure the DAC linearity (unless you count the -60 dB spectrum plot but virtually any DAC has near zero error there) or square wave response.


Sigma Delta DACs (the industry standard for about the past decade or more) can't possibly have linearity problems. That's one reason why virtually every DAC has low error at FS-60 dB .


I should probably give it up, but there are plenty of examples of Delta-Sigma DACs with linearity error.


I'm not at all convinced by unsubstantiated claims. I tested over 100 audio products with converters in them for my old PCAVTech web site (some were never posted), and I never saw a Sigma Delta converter that was appreciably more nonlinear than suggested by its noise floor. I tested some real garbage like the Avance semiconductor converters before Realtek bought them out and put them and us out of that misery.

Besides, there are virtually no products that drive their outputs with the true output of the DAC. They all add buffers or I/V converters in the same chip as the DAC or in another. There can be nonlinearity in them. But there is no doubt that RMAA tests for nonlinearity at both high and low levels. 

When people talk about DAC nonlinearity, they are usually talking about lack of monotonicity and/or missing codes. Sigma Delta DACs can't have those because strictly speaking the don't  decode the whole digital data word like a ladder DAC does.

I feel bad about beating up someone who is basically friendly to The Cause, but there is such a thing as going over the line and claiming omissions that aren't really omitted. At  worst, RMAA has a credible FFT analyser that you can break out and use all by itself. Some of your plots by other means are just FFT plots that you are analyzing in your comments.  You could do the same thing with RMAA.

24-bit audio proposed for iTunes

Reply #112 –
Quote from: Arnold B. Krueger on
Quote from: saratoga on
Quote from: NwAvGuy on
It also can't measure the famous pitch error of the Sansa.


Thats a fair point, although this is a pretty rare problem.



My question is whether or not one can detect the approximate 0.25% pitch error from one of the existing test results that RMAA already provides.

For example, the RMAA THD plot has a strong carrier @ 1.000 KHz, which if memory serves from the tests I ran,  the Clip shifts to 1.0025 KHz.  Using the expanded frequency scale feature that RMAA provides, I think that this error can be properly detected and estimated.


At the time I wrote this I was having problems finding the results of my RMAA tests on my OEM-firmware Clip+. Well I finally found them and checked out the detailed FFT analysis for the THD test.  The FFT plot very clearly shows that the 1 KHz tone is not centered on 1.000 KHz,. A quick visual estimate says that the 1 KHz tone was actually reproduced as 1.0025 KHz.

Therefore, the statement that RMAA "...can't measure the famous pitch error of the Sansa." is simply false.

24-bit audio proposed for iTunes

Reply #113 –
For 99.95% of consumers this technology will be nothing more than a marketing gimmick.

And I really doubt it will take off in the nearest 10 years (or ever).

24-bit audio proposed for iTunes

Reply #114 –
Quote from: Arnold B. Krueger on
I feel bad about beating up someone who is basically friendly to The Cause, but there is such a thing as going over the line and claiming omissions that aren't really omitted.

We are on the same side of the cause and I'm sorry if I crossed some line. But I'm not sure what line that is? You put the above sentence in the DAC linearity reply. So I'll start there.

Quote from: Arnold B. Krueger on
I'm not at all convinced by unsubstantiated claims. I tested over 100 audio products with converters in them for my old PCAVTech web site (some were never posted), and I never saw a Sigma Delta converter that was appreciably more nonlinear than suggested by its noise floor.

Fo DAC non linearity I was talking about the linearity of the entire device. Obviously, I'm not opening up the players and probing the raw DAC output which generally isn't available anyway with a "SOC" (System On a Chip) design. I have seen considerable errors, and I've also seen published Audio Precision reviews that document significant non-linearity errors. Here's an example of a -6 dB error at - 90 dBFS--the value I run my tests at:



You can find the entire test here:

hrt-music-streamer-usb-da-converter-measurements

I don't know if that helps "substantiate my claim" for you, or not, but the above is the sort of error I was talking about. I consider a -6 dB error at -90 dBFS to be a significant, and measurable, linearity error that at least some would find interesting. Apparently John Atkinson at Stereophile agrees it's worthwhile to measure linearity as well because he does it regularly. If all Delta Sigma DAC devices, as you claim, have not had any linearity problems in the last 10 years, why bother to make the measurement? If you want to get into the semantics of terminology, and accuse me of incorrectly wording exactly what sort of error I'm talking about, fine. But I think we both know what sort of real world measurement I was talking about (i.e. the one above). What line did I cross here?

Quote from: Arnold B. Krueger on
At the time I wrote this I was having problems finding the results of my RMAA tests on my OEM-firmware Clip+. Well I finally found them and checked out the detailed FFT analysis for the THD test.  The FFT plot very clearly shows that the 1 KHz tone is not centered on 1.000 KHz,. A quick visual estimate says that the 1 KHz tone was actually reproduced as 1.0025 KHz.

Therefore, the statement that RMAA "...can't measure the famous pitch error of the Sansa." is simply false.

I guess the key phrase is "visual estimate". I haven't tried zooming in to the max amount on RMAA. But my earlier comment was that you won't see those results directly in a table, etc. But I agree with your earlier statement that it's a rare error and not generally something of interest.

Quote from: Arnold B. Krueger on
You;'ve missed an improtant point. You don't need a $10K test set to show the effects of IEM load impedance on a digital player. All you need is RMAA, a $5 Radio Shack 3.5 mm splitter, and a pair of headphones.

I didn't miss it at all. In the RMAA blog article I talk about the pros and cons of using headphones vs resistors as a load and recommend people use a load for testing with RMAA. I completely agree you can go buy resistors and I advise doing just that in my RMAA article. My debate with Saratoga was over his suggestions that (paraphrasing):

A - Output impedance doesn't matter to anyone so it's not worth measuring

B - All 16 ohm headphones are essentially similar enough to be used as a load for a product review that can be fairly compared to other product reviews

As for "needing a $10K test set" I've never suggested anyone who wants to publish a review run out and buy an expensive audio analyzer. I have said many times here, and on my blog, RMAA is a great tool when it's used properly. And I've tried to provide some useful advice to help people do just that.

My big issue with RMAA is it's often viewed as this sort of "Gold Standard" on many of the forums, etc. Many people just accept virtually anyone's published RMAA results as being correct. That's far from being true. As I said earlier, you're not a typical RMAA user. You know how to use it correctly, most don't.

So the point of the "$10K test set" was to point out such analyzers are the "Gold Standard" not RMAA. If you really want to have complete, definitive, verifiable and repeatable measurements that conform established ndustry standards, RMAA is not the correct tool for the job. Yes, with enough knowledge, other hardware, other software, etc. you can kludge your way around some of it, but that's also not very practical for a lot of people.

I believe there are some significant things RMAA doesn't directly measure or provide and/or you rarely (if ever) see in published RMAA measurements. They include:
  • Test Levels For Repeatability and Comparability (a meter solves this, but again, it's hardly ever published with RMAA results)
  • Maximum Output Power (requires a good meter and lots of trial and error with RMAA or other software and is rarely done)
  • Output Impedance Of Headphone Outputs (can also be calculated if you have a known load, but few ever do)
  • Linearity Error (requires other perhaps paid software and a quiet PC sound interface)
  • Square Wave Response (requires other software)
  • 11025 hz Standard Jitter Test (could be done with other software but I haven't seen any free analyzers with sufficient resolution)

My goal was to point out the common real-world problems with most published RMAA results. If you've used the software much, I'm sure you've seen it provide invalid results--there are several on my blog such as:



The above is the same PC headphone output, with 3 different loads, all calibrated correctly using RMAA's calibration dialog. RMAA did not detect any clipping or report any errors. Look at the THD and IMD for the SuperFi headphones. RMAA reports 100+ times more distortion just from using different headphones. Why? It likely clipped due to the impedance swing of the headphones. But would the average person using RMAA know that? Here's the frequency response graph from RMAA for the same test run:



The SuperFi headphones (see the impedance plot I posted earlier in this thread) caused this wild frequency response which swings from +6 dB to nearly -9 dB. This is also another good argument for knowing the output impedance. I don't think anyone would argue the above frequency response change is inaudible or doesn't matter. RMAA got the graph reasonably correct but really messed up the distortion measurements. This kind of problem would be obvious on an analyzer, but it's rather hidden in RMAA.

And below is another actual run with RMAA. The 3rd column was the music player on my cell phone. It really doesn't have 237% IMD, honest. But RMAA thinks it does. There were no error messages from RMAA. It got the noise and dynamic range close to correct but fell on its face for all the other measurements:



RMAA is basically a "magic black box" and you have almost no control over, or sometimes even knowledge of, what it does internally in the software. You just set the levels and click the GO button. It either works or it doesn't. And sometimes it doesn't, and sometimes it only half works. You can't change the test signals, or the measurement bandwidth of the distortion analysis, or anything else besides the calibration level. It's a rather limited tool in many ways.

For example, when I run RMAA with a PC interface that supports 24/192, does RMAA cut off the THD measurements at the audible limit of 20 Khz? Does it measure out to 96 Khz? Is it something in between? I certainly can't control it, and I don't even know what it is. Perhaps there are posts in a forum somewhere that answer these questions, but are they correct or someone's guess?

Are the noise measurements unweighted? Weighed? Bandwidth limited? Are they calculated out to the measurement limit? What if I want an A Weighted reading to compare to the dBA spec many manufactures use?

The program hasn't been updated in a few years, and the developer apparently has lost interest in it. It has some blatant bugs that many have complained about. I'm not trying to say it's useless, only point out its limitations and how it's very commonly used incorrectly or in ways that are misleading.

It's so popular because it's free and seems easy to use. But, in reality, it's also really easy to get bad numbers with RMAA. And to use it correctly, and overcome all the limitations of PC sound interfaces, not having absolute levels, gain matching, external divider networks, grounding, supplementing it with other software, etc., RMAA ends up being far more difficult (and error prone) than using a genuine analyzer.

24-bit audio proposed for iTunes

Reply #115 –
And, back on topic, does anyone know if the newly announced iPad 2 has the much talked about 24 bit Wolfson DAC Apple is supposed to be switching to? I've tried my luck on Google and come up empty. Both are related to recent announcements, so you get plenty of hits, but nothing I've found so far mentions the iPad 2 itself having 24 bit support.

The sole audio performance spec on the Apple iPad 2 Tech Specs page is a useless: "Frequency response: 20Hz to 20,000Hz". And under supported formats, there's no mention of 24 bit:

Apple iPad 2 Specs

24-bit audio proposed for iTunes

Reply #116 –
Quote from: NwAvGuy on
Quote from: Arnold B. Krueger on
I feel bad about beating up someone who is basically friendly to The Cause, but there is such a thing as going over the line and claiming omissions that aren't really omitted.

We are on the same side of the cause and I'm sorry if I crossed some line. But I'm not sure what line that is? You put the above sentence in the DAC linearity reply. So I'll start there.

Quote from: Arnold B. Krueger on
I'm not at all convinced by unsubstantiated claims. I tested over 100 audio products with converters in them for my old PCAVTech web site (some were never posted), and I never saw a Sigma Delta converter that was appreciably more nonlinear than suggested by its noise floor.

Fo DAC non linearity I was talking about the linearity of the entire device. Obviously, I'm not opening up the players and probing the raw DAC output which generally isn't available anyway with a "SOC" (System On a Chip) design. I have seen considerable errors, and I've also seen published Audio Precision reviews that document significant non-linearity errors. Here's an example of a -6 dB error at - 90 dBFS--the value I run my tests at:



You can find the entire test here:

hrt-music-streamer-usb-da-converter-measurements

I don't know if that helps "substantiate my claim" for you, or not, but the above is the sort of error I was talking about. I consider a -6 dB error at -90 dBFS to be a significant, and measurable, linearity error that at least some would find interesting. Apparently John Atkinson at Stereophile agrees it's worthwhile to measure linearity as well because he does it regularly. If all Delta Sigma DAC devices, as you claim, have not had any linearity problems in the last 10 years, why bother to make the measurement? If you want to get into the semantics of terminology, and accuse me of incorrectly wording exactly what sort of error I'm talking about, fine. But I think we both know what sort of real world measurement I was talking about (i.e. the one above). What line did I cross here?



I don't think that the above shows "DAC error" as such.

First a little history, which you may or may not be aware of.

When digital audio first came generally available the complaint "measures good but sounds bad" was often heard. In retrospect, it is not clear that all of the complaints were based on reliable listening.

At any rate, there was supposedly bad-sounding digital equipment that had great frequency response, low noise and low nonlinear distortion that "didn't sound right". One actual technical difficulty was that the resistor ladder DACs of the day were prone to errors of a kind that had no direct equivalent in the analog domain.  Therefore, means were devised to test for them. One of these "new forms of distortion" were due to errors in the resistor ladders that lead to incorrect step sizes when an ascending or descending sequence of digital values were applied to the DAC. Isolated steps could be so wrong that the desired staircase was had steps that either didn't exist or even went in the wrong direction. This is known as lack of monotonicity. 

While Stereophile has tested for this kind of error for decades, AFAIK no actual instances equipment with this fault it have been published. 

My review of the Stereophile archives suggests to me that by the time they had the resources to adequately test digital gear, the new digital gear had improved to the point where monotonicity failure was rare or non-existent in high end audio gear. DACs based on resistor ladders were passing from the marketplace, and/or such resistor ladder based equipment as was being built in the high end price category was too highly perfected for gross errors of this kind to appear.

The review you cited does not appear to show  difficulty with classic "DAC Error". Instead, it shows a product with relatively high nonlinear distortion of a fairly ordinary kind. If there were actual monotonicity errors, then there would be large steps  and discontinuities in Figure 4 of the review you cited. Instead I see the effects of a gently curved transfer function which is confirmed by figures 6 and 7.

If there was classic "DAC error", the nonlinearities shown in figure 4 would have some dramatic breaks or steps. Instead it is just curved. It would not be as nonlinear in terms of traditional harmonic generation and IM as is shown in figures 6 and 7.  The source of the nonlinearity could be the I/V converter or any buffer stage that followed it.

I believe that the far less costly Behringer UCA 202 would generally outperform the HRT Streamer. I suspect that one possible purpose of the Stereophile review was to show the inadvisability of buying low cost equipment. It might be might that they could have gotten something better if they spent far less! ;-)

24-bit audio proposed for iTunes

Reply #117 –
There are renewed rumours that Apple will start selling 24-bit, 96 kHz or 192 kHz audio. Robert Hutton claims it will launch at the beginning of June with the release of three Led Zeppelin remasters. Pricing is supposed to be one dollar above the current iTunes Plus.

Not sure where he got his info from. He somewhat supports his claim by noting that certain labels have stopped catering to HDtracks. He also argues that Apple has already got the largest HD catalog of HD audio, as they've requested labels to upload HD content for a number of years.

EDIT: German tech site heise.de sees other signs of a major revamp of the iTunes music store which could include HD audio:

1.) Apple is increasingly concerned about losing sales to Pandora, Spotify, and friends. And it doesn't look as if iTunes Radio is gaining any traction. (At the very least, it doesn't generate significant *sales* for the music store.)

2.) Neil Young's crowdfunding campain was a success. So maybe the momentum for HD audio is there.

3.) A launch date in early June would fit well in Apple's calendar. WWDC runs June 2 through 6.

4.) Apple is hiring in several areas for the iTunes music store.
 

24-bit audio proposed for iTunes

Reply #118 –
I'm not sure if that's accurate or not.  At this point, anything not officially announced by Apple has to be viewed as rumor or conjecture.  It would be nice if Apple started selling lossless content (I could care less about placebophile 24-bit, 96KHz/192KHz content) as I could then stop buying CDs all together.  However, if the rumor is true, I'm still confused on their pricing.  Is it $1.00 extra per track or $1.00 extra per album?  Hutton's blog makes it appear that it's $1 per track meaning a 12 track album won't cost $9.99 but rather $21.99.  Other albums, like Led Zeppelin IV, would cost $17.99 and albums like 1 from The Beatles would cost $39.99.  Those prices seem a bit excessive even for the iTunes Store which normally sells albums at the going rate ($9.99) or $1-$3 more compared to Google Play and the Amazon mp3 Store.  If Apple did launch such a service and the price rumor was accurate, they would be pricing themselves out of the market.  An extra $1 per album would make more sense.

Then again, this has to be taken with a big grain of salt seeing as how the whole "lossless on the iTunes Store" rumor has been circulating for the past 4 years (if not more).