Discussion in 'Measurement Techniques Discussion' started by atomicbob, Feb 11, 2018.
Yes, this is a blatant example of masking. Yes, lossy compression schemes also exploit the masking principle of the basilar membrane. No this example isn't mp3. I'll post yet another auditory example.
Attached are three files. One is a pure sine. The other two have have 0.1% THD+N added. Can you hear which is which?
WARNING - these files are -1 dBFS and LOUD. Turn volume down and bring up to a comfortable level.
Please listen before going to the DAW and viewing the FFT of the files. For those of you with foobar2000 an ABX wavefile tester is available. Adds to the educational opportunities.
Again, these are examples to help put perspective on audible vs. measurements. This in no way is an attempt to invalidate either listening or measurements but is meant as a self educational tool. Both listening evaluation and measurements are valuable tools for audio device assessments. These simple examples do not begin to cover all the possibilities.
At a later time I will post the same files with audible cues added which make it easier to hear the distortion. Also will post some wave files demonstrating how incredibly sensitive the ear can be, detecting 1 to 5 samples out of place in a context allowing such sensitivity.
Spoiler: FFT displays of the three files - not labeled
Just guessing here:
A is pure sine
B is with low frequency
C is with high frequency
I'm pretty sure of C.
I still could be wrong. All too close to call.
Just analyzed the files myself....I was wrong....WAY wrong.
I have both a fully original-spec Yggdrasil (USB Gen 3 and original DAC/analog boards) and a unit with pre-release USB Gen 5 and Analog 2 boards and the DSP update. I have comprehensive comparisons written up for that comparison, done back-to-back.
However, I would want to validate that the shipping versions of those boards is the same as the pre-release versions I have before making any formal comments on them.
I will say that the pre-release stuff I’ve heard is, as @Marvey alluded, in the “Well, Schiit ... this changes everything ...” domain.
More when I am back from New Orleans ... (assuming I don’t die of debris and beignet overdose ...).
Be cool guys.
<Looks at thread title related to content of Torq's post>
Let's not mention it to him as long as he keeps posting.
Yeah, please try not to die of excess quite yet. Look forward to reading it- hopefully this renders Gumbreya-style hacks somewhat unnecessary for most people.
Too bad I lost touch with the massive beardy jazz geetarist I used to know who had a town house in the old quarter and his main home on stilts in a 'gator swamp, or I'd want to try to cross the streams. (Motherfucker played with Jacho and jammed with Herbie Hancock.. such tales. I got him into Aphex Twin- was that wrong?)
Excellent work AB, as usual.
And among other significant actions taken by "a, the professional", I thought that as a professional he would accord those who are professionals, a professional courtesy and at least ask before using your work (or anyone's really).
Not to mention, that when obtaining results that are at such variance with other results, to pause and look into them a bit more deeply BEFORE publishing them. And to at least contact those (like the manufacturer) who know what the measurements should be.
I'm reminded of how Tyll handled the differences between his measurements vs those that Audeze had generated.
When discrepancies arise, working to resolve them vs. defending them, usually results in a win-win-win scenario.
Just a few thoughts.
Jonathon Novick is very diplomatic but does offer an RTFM comment where he suggests something along the lines of, if you look in the lower bottom corner of the graphical interface, the software gives you some clues as to what its doing. That small piece is rather revealing of test equipment user competence. Certainly doesn't inspire confidence in the numbers and analysis that the bride-from-the-land-of-the-rising-sun dude is offering up out there.
Thanks @bobsherman for pointing to the youtube vid. Most educational.
Most audio analyzers have status lines. These are complex instruments meant to be used by Scientists and Engineers, not Appliance Operators. The dScope I use has a similar status line in the lower right corner as can be seen in my Technical Measurements screen shots. Nothing is hidden from view. The only graphs in my measurement suites without this information are those produced by the dScope print function. However there are associated screenshots made with the same settings where status line is available.
Reading the dScope status bar (as can be viewed in my Technical Measurements suites)
Look at the area outlined in blue, lower right corner.
1 - Analogue Outputs Range tile
This tile indicates the signal status of the Analogue Outputs; possible indications are:
(grey background) Indicates that the Analogue Outputs are operating normally.
(red background) Indicates that the Analogue Outputs are muted because the selected amplitude or frequency of the Signal Generator exceeds their range.
2 - Digital Outputs Range tile
This tile indicates the signal status of the Digital Outputs; possible indications are:
(grey background) Indicates that the Digital Outputs are operating normally.
(red background) Indicates that the Digital Outputs are muted because the selected amplitude or frequency of the Signal Generator exceeds their range.
3 – Digital Output Sample Rate and Bit Depth
4 - Digital Outputs Source tile
This tile indicates the selected source for the Digital Outputs; possible indications are:
Indicates that the Digital Outputs are sourced normally from the Signal Generator.
Indicates that the Digital Outputs are 'looped through' from the Digital Inputs, for in-line testing.
Indicates that the Digital Outputs contain the Channel Check sequence, operating at a wordlength of 24 bits.
Indicates that the Digital Outputs contain the Channel Check sequence, operating at a wordlength of 20 bits.
Indicates that the Digital Outputs contain the Channel Check sequence, operating at a wordlength of 16 bits.
5 - Analogue Inputs Source tile
This tile indicates the selected source for the Signal Analyzer's Analogue Inputs; possible indications are:
Indicates that the Analogue Inputs are sourced normally from the front-panel connectors.
Indicates that the Analogue Inputs are sourced from the Digital Input jitter demodulator, operating in 'fs jitter' mode.
Indicates that the Analogue Inputs are sourced from the Digital Input jitter demodulator, operating in 'data jitter' mode.
Indicates that the Analogue Inputs are sourced directly from the Analogue Outputs.
Indicates that the Analogue Inputs are connected 'pre and post' the EUT's A–channel. Channel A is sourced from the Channel A front-panel connectors, while Channel B is sourced from the dScope's Channel A Analogue Output.
Indicates that the Analogue Inputs are connected 'pre and post' the EUT's B–channel. Channel B is sourced from the Channel B front-panel connectors, while Channel A is sourced from the dScope's Channel B Analogue Output.
Note that the tile's background is grey to indicate normal operation of the Analogue Inputs, but is red if the selected terminating impedance has been over-ridden to 100kR because excessive amplitude has been detected which would risk damage to the terminating resistor.
6 - Digital Inputs Source tile
This tile indicates the selected source for the Signal Analyzer's Digital Inputs; possible indications are:
Indicates that the Digital Inputs are sourced normally from the front-panel XLR connector.
Indicates that the Digital Inputs are sourced normally from the front-panel BNC connector.
Indicates that the Digital Inputs are sourced normally from the front-panel TOSLINK connector.
Indicates that the Digital Inputs are sourced directly from the XLR Digital Output.
Indicates that the Digital Inputs are sourced directly from the XLR Digital Output.
Note that the tile's background is grey to indicate normal operation of the Digital Inputs, or red if no digital input carrier is detected on the selected source ('input unlocked'). A yellow background indicates that one or more of a number of warning conditions of the Digital Input have been detected. These are 'biphase violation', 'block-length error', 'eye-narrowing near-fail', 'asynchronous wrt generator' or an error in Channel Check operation if selected.
7 – Digital Input Sample Rate and Bit Depth
8 - Digital Inputs Channel Status tile
This tile contains the legend "CS", normally with a grey background. The background is red if either the Digital Input is unlocked or if 'inconsistencies' in the incoming Channel Status are detected (as defined in the 'highlight inconsistencies' function of the Input Channel Status dialogue box).
9 - FFT Progress tile
This tile shows a red progress bar for acquisition and calculation of the FFT buffer. If the FFT Analyzer is operating in two-channel mode, two bars are shown – the upper bar indicates the A–channel and the lower the B–channel. If FFT averaging is enabled, the tile also contains an indication of the number of averages completed.
Updated post #2 replacing KMixer model diagrams with Audio Engine model for Windows Audio Architecture. KMixer was replaced with Audio Engine in MS Windows Vista and later.
I tried this one couple of hours before and I was able to hear the square-wave only, but not the sine-wave; instead, it was like a pause between the main 440Hz sound and the square-wave. Now, at 12:32 AM, fridge is not working, cat + wife + kid got to sleep (I estimate less than 25dB background noise), I am happy to say that I could hear the sine-wave in the middle too.
I was using FOSTEX T50RP-mk3 (slightly modded) and BURSON CV2+, also...I'm sure with my other 2 closed-back cans I should be able to hear the sinewave too (Solo2 and K550). Also, DAC and headamp not sure they matter much for this test...I can do it again with ODAC and O2 if someone asks.
L.E.: I've done it again with "Beats Solo 2" closed cans connected to MacBook Pro (mid-2014), MacOS Sierra and VOX 3.0 as player. I really think it's more about how quiet is the surrounding background, but using closed cans as well.
Separate names with a comma.