Thursday 4 September 2014

MEASUREMENTS: Digital Audio and the PlatterSpeed "Wow & Flutter" App

Shortly after my last post on the wow & flutter measurements with the Technics SL-1200 using the 7" test disk, Frans suggested in the comments a test using a CD to demonstrate what the results would look like using digital devices.

Great idea!

So, I whipped out good old Adobe Audition and generated a -10dBFS (same level as the Test LP signal) 3150Hz (for 33.3rpm) sine wave. I burned it on a standard Maxell CD-R - first time in a long time I've burned an audio CD!

To purposely not do anything special to reduce the "potential" for jitter (NB: I don't actually believe this anyway), the disk was burned at 48x on an older LiteOn iHAS DVD burner which I got either in late 2011 or early 2012. I did verify the burn to make sure the data was readable.  I used my electrically "noisy" quad-core home assembled i7 workstation to do the burn as well. Despite the strong temptation, I did not use any green "Stereophile recommended" StopLight pen to color the rim of the CD to improve the sound :-).

For the graphs, remember to look at the scale on the left to get a sense of the magnitude of frequency variation. I've aimed at measuring over the course of ~60 seconds.

To start, let me just show what the results look like with my network devices streaming off my server computer before I throw the CD into a few spinners:

Logitech Transporter - balanced analogue output, wired ethernet connection to music server (main soundroom system with Paradigm Signature S8 speakers):




Logitech Squeezebox 3 (Classic) - RCA analogue output, wired ethernet connection to music server (living room Tannoy MX2 bookshelf speakers):





Logitech Boom - wireless connection to music server at 30% signal strength 2 floors apart (built-in speakers):




Home Theatre PC to TEAC UD-501 DAC via USB using JRiver Media Center 19 to stream the shared file off my server machine in another room in the house (essentially the typical PC-DAC setup):




Alright, enough with the streaming devices... Time for some CD spinners now playing the burned Maxell CD-R...

Panasonic Blu-Ray DMP-BDT220 player (2012 model) in soundroom connected via HDMI to Onkyo TX-NR1009 receiver with Paradigm Signature S8 speakers:




Windows 8.1 Intel i7 computer with CD played off LiteOn iHAS DVD burner to ASUS Xonar Essence One DAC (USB) with desktop AudioEngine A2 speakers. To make this even more "challenging", I have 8 computing threads running doing WAV --> FLAC compression just to create 100% CPU load with heavy hard drive activity:




"Lo-Fi" Sony CFD-S05: Finally, my lowest fidelity device that plays a CD; an all-in-one cassette / CD / radio. About $70 and worth every penny just to play the occasional cassette tape :-).





So, what can we say looking at the above graphs and numerical summary? Exactly what the jitter measurements over the last year or so suggests. Timing anomalies as can be detected in frequency shifts are minuscule in the digital world compared to what is seen with a turntable set-up. Dunn J-Test utilizes detailed FFT analysis to look for tiny sideband distortions which typically show up below -100dB with most decent DACs these days. There is a world of difference between what's found with the J-Test compared to the timing errors with typical turntable measurements as reported by tests like PlatterSpeed.

Looking at the mean frequency (remember, target exactly 3150Hz), every one of the devices achieved essentially perfect scores except for the inexpensive "lo-fi" Sony all-in-one (average 3153.7 Hz). Even with the Sony, my guess would be that this is due to the low fidelity speakers or "defective by design" (remember the poor frequency response) rather than an underlying digital timing inaccuracy.

If we look at the "raw" frequency graphs, we see that the major variation is at the end when I press the STOP button and the iPad is picking up the "click". Otherwise, other than the Logitech Boom and the i7 computer connected to the ASUS Essence One with very low levels of frequency undulation (I've seen this with the Essence One before with the USB interface but not with SPDIF inputs), the rest of the plots are flat lines. I am impressed by the iPad and PlatterSpeed app for being able to pick up tiny <1Hz variations.

As for the numerical analysis... Frequency fluctuation results are essentially perfect across the board and better than any turntable set-up; this includes the inexpensive Sony all-in-one CD player. The small deviations are for the most part just the bits of ambient noise or that "click" at the end being detected.

Is any of this surprising? Of course not. It doesn't negate the fact that a good turntable system can sound very good, but is a reminder of the technical limitation of a physically spinning turntable and of human hearing (ie. a non-specialized "instrument" like the iPad can pick up imperfections easily, more so than the threshold of hearing that would interfere with enjoyment of music). In fact, we might even hypothesize that a small amount of wow and flutter using a turntable adds to the "texture" of the sound, making it less "digital" sounding perhaps.

No matter how one views it, these results are a reminder for those who aspire for "high fidelity" as synonymous with "high accuracy" that digital is clearly superior (at least in the time domain as per this discussion). I was frankly surprised how well the little Sony "lo-fi" fared, actually!

---------------------

Well, it was a blast zip-lining and white water rafting over the Labour Day long weekend! It's September so time to get back to work full swing... Hope everyone had great summer months in the northern hemisphere :-).

As the rainy and cool months approach here in Vancouver, I'm looking forward to some quality listening in the 'man cave'... Enjoy the tunes, everyone...

PS: Back to School question of the week - in 500 words or less, why would anyone want to buy the 24/96 version of Maroon 5's "V" on HDTracks for $22.98? Please provide plenty of subjective descriptions :-).

5 comments:

  1. This was a fun test... unfortunately it is worthless because the StopLight pen wasn't used. A small omission

    A few interesting things can still be seen though.
    Those looking at the plots may notice the freq variations at the beginning and end of the plot (which skews the generated numbers significantly by the way) and the smaller variations on some of the plots during the 60 seconds.
    This is clearly the result of inferior USB cords used for this test showing jitter..
    Ooohh.. sorry my bad, I mean clearly caused by the measurement device which was probably moved (ever so slightly) during the test.
    The pulses at the beginning and end are (I assume) caused by operating buttons on the screen of the used phone starting and stopping the test ?
    The (very small) variations come from moving a tiny bit towards the sound source while measuring (frequency increases) and moving away (frequency decreases).
    Think sirens of emergency vehicles passing-by or horns of trains at railroad crossings seemingly altering frequency while they don't in reality.
    Makes you wonder how much moving your head while listening to speakers will affect the perceived pitch.
    Headbanging seems not to be recommended for serious listeners.

    Speed deviations (very small only 0,003% off in some cases) may be caused by the used x'tal oscillator frequency(ies).
    This can be the case for gear that uses one X'tal for both 44.1 and 48kHz bitrates.
    For both data-rates the pitch will be ever so slightly 'off' as a division with one X'tal it isn't practically possible to get perfect pitch for both data-rates in this case.
    The ones with perfect pitch probably have 2 X'tals (and switches between them depending on data rate) or can only do 44.1 based data-rates exactly and are somewhat 'off' on 48kHz based rates.
    The Sony CFD-S05 probably has a X'tal (or a simple and cheap ceramic resonator) that is slightly too high in frequency as it plays 0.11% too fast.

    The Logitech Boom 'hunts' a bit, not that dissimilar from the SL1200, but more smoothly, slightly slower and much less in amplitude. Probably a PLL thing.
    A PLL adjusts its output frequency (which is used to clock the DAC in many cases) depending on a difference in feedback signal coming from the platter (TT) or derived from (USB) data-rate and is compared to the (multiple times divided) reference X'tal frequency.
    When these two input frequencies differ slightly in phase (hence Phase Locked Loop) the output frequency of the PLL is adjusted and the speed of the motor (or DAC chip) is also slightly altered.
    The reaction time of that PLL control circuit is determined by a low pass filter which causes the (very low frequency) 'hunting' as it always is a bit 'too late' and then slowly corrects itself in the other direction agian.
    The internal clock is probably slightly slower than the data-stream it receives and it adjusts it's PLL so there are no buffer over- or under-runs possible.
    Still the 'wow' it has (slow speed variations) are 75x better, in the order of 0.002%, than those of the SL1200 !

    The WIN8.1 PC (that was very busy doing other chores as well) and Xonar Essence one combo show a well known phenomenon where the Xonar shifts its frequency slightly (0.1Hz = 0,003%) periodically, about every 2 seconds in this case, to prevent buffer under-/over-run as the clock of the PC (which determines the USB rate).
    Highly unlikely this would be audible.
    Another question that pops up is how the plot would look when the PC isn't doing its extra chores (probably exactly the same ?)

    The other devices all perform at least 1500x better than the SL1200 (in maintaining a constant speed/pitch) which is a good thing to know.

    Thanks for doing this enlightening test which shows the dramatic difference in pitch stability between vinyl and digital.

    ReplyDelete
  2. Thanks for the observations Frans. Very helpful in understanding the results and the demonstration of just how many *times* less accurate the turntable is in comparison!

    Yeah, the Xonar looks like that whether the computer is doing the extra tasks or not.

    However, I'm sure if I burned the CD to a black disk at 1X speed, used the StopLight pen, cryo treated, paid at least 5 figures for new power, speaker, USB & interconnect cables, used cable risers, phoned up Machina Dynamica to do the Teleportation tweak, bought the Dark Matter CD coating, wore the Clever Little Watch, installed a Shakti Hallograph, and gave Ted Denny a bunch of bucks for a Tranquility Base, HFT/FEQ/XOT/ECT/WTF... The Xonar's graph would be flat as the horizon off Hawaii on a calm evening with not a jitter in sight. :-)

    ReplyDelete
  3. Some information about the deviation by eccentricity of vinyl records at http://www.thesoliduniverse.com/2014/10/vinyl-audio-eccentric-records.html

    ReplyDelete
  4. "So, I whipped out good old Adobe Audition and generated a -10dBFS (same level as the Test LP signal) 3150Hz (for 33.3rpm) sine wave." Love your sense of humor. It is like frequency response rating of headphones. 20Hz-20kHz Period. Is it 20Hz -3dB, -0.5dB or -100dB? 20/20. :o)
    "Even with the Sony, my guess would be that this is due to the low fidelity speakers or "defective by design" (remember the poor frequency response) rather than an underlying digital timing inaccuracy." I don't understand your logic here though. How is non-linearity of frequency response you've measured could affect jitter of sinusoidal signal (3150Hz)?

    ReplyDelete