Merv's Politically Incorrect Audio Blog

Discussion in 'SBAF Blogs' started by purr1n, Dec 26, 2018.

  1. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Two reminders:
    Tutorial - Interpreting THD and THD+N specifications and graphs

    Tutorial - A tale of 6 amps THD+N and THD specifications

    Chasing numbers is a fool's errand. They are lossy compression equivalent for technical measurements with an enormous quantitative data reduction.

    Consider the following three amps THD+N, taken from their respective Left channel technical measurements posts:

    THD+N Left channel
    0.002% Liquid Platinum
    0.006% Lyr3
    0.009% DSHA-3F

    Those numbers are all respectably low and don't appear to differ by much.

    Now consider 50+7000Hz FFTs for the same three amps at approximately +3 dBu:

    50+7000 Hz Liquid Platinum
    04 20181110 LiqPlat 50+7000Hz 0dBu 300R 4XF.png

    50+7000 Hz Lyr3
    05 20180430 lyr3 50+7000Hz 0dBu 300R 4XF.png

    50+7000 Hz DSHA-3F
    06 20190221 DSHA-3F 50+7000Hz 0dBu 300R 4XF.png

    Those three amps have quite different FFT signatures.

    Measurements matter, when you know which ones are worth comparing. When chasing numbers the assumption is lower is better. I don't believe that is a good protocol. Looking at FFT signatures one might wish to find amps with signatures aligning with personal preferences. A superior protocol.
     
  2. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    CLASSIC SOUND DACS, MODERN SOUND DACS, AND DACS IN-BETWEEN

    This is a cut-and-paste from another thread with additional commentary. I felt this belongs here because a few people said it should be sticky and I've felt a need to refer to it in my DAC impressions. The answer goes beyond the original question below.

    Question: Do you think Schiit Multibit DACs are the only good DACs that exist?
    Answer:

    I'll answer this. Some history is required. This is the very long version which details my search for DACs many years ago: https://www.head-fi.org/threads/tho...-just-to-get-you-to-think-about-stuff.693798/

    Note that some people call me a Schiit shill, but this thread details my DAC journey starting in 2013, way before Yggdrasil even existed! The fact is, I stumbled (or re-stumbled) into the Theta DAC brands where Mike Moffat happened to be the designer. The entire Schiit mega-combo-burrito filter thing with its holographic imaging was already invented in the 1990s on the Theta Gen V. It should be noted that knew nothing about the Schiit filter until I sought out Mike Moffat @baldr - it was not the other way around - keep in mind that Changstar.com was avoided by manufacturers like the plague. I begged Mike to bring something like the Theta Gen V back. As an aside, if you use headphones, you are only getting 60% of what the Schiit Multibit DACs can do. Only speakers with good dispersion and off-axis characteristics will do justice to the mega-combo-burrito filter.

    Finally, I think a lot of people forget that Schiit started or was at least instrumental in popularizing or bringing back the R2R or R2R/2R sound. The Yggdrasil, Gungnir, and Bifrost are hybrid resistor ladder (R2R) and resistor string (2R) DACs. Resistor strings are better in some ways in that their architecture guarantees monotonicity. The downside is that 2R architectures require 2^n parts so it's infeasible to go beyond a handful of bits.

    --

    The Modi MB is the odd hand-me-down bargain product with specific traits quite different from the Bifrost, Gungnir, and Yggdrasil. This is 2019 and things could certainly change in the future. The Modi MB actually reminds me more of the vintage R2R DACs than what I would call the "Schiit Multibit" sound. To me, there are "Schiit Multibit DACs" and then there is the "Schiit Modi Multibit".

    During the time R2R DACs were being phased out in favor of sigma-delta "S-D" DACs in the early-mid 90s, it was obvious that most S-D DACs were becoming much more resolving than the R2R ladder DACs. In the 00s, holdouts such as AG-D circa 2010+, or Wadia a few years before, used multiple BB PCM1704 chips in parallel to wring out as much resolution as possible, but even then, their efforts fell slightly short. Note that the BB PCM1704, which had a loyal following, kind of sounded like ass. PCM1702 less so. PCM63 before that was the pinnacle of the R2R ladder chips.

    The Schiit Multibit DACs (again, excluding the Modi Multibit) have their own sound. They do not totally sound like the R2R DACs of decades ago. And whether you believe in INL/DNL (think of this as amplitude jitter) over strict linearity measurements, the mil-spec tolerances of the chips used by Schiit in their DACs did contribute to their DACs edging out current S-D designs in terms of resolution. The latest S-D chips from AKM seem to have the best potential, but all other S-D chips from ESS (including the latest 9038), TI/Burr-Brown, Analog Devices, Cirrus, Wolfson, all seem to be at least a notch or two behind. Typically, the older the chip, the less resolving, all things being equal.

    During the time when Yggdrasil first came out, and this was years ago, almost every S-D DAC had fricking or odd nasty timbre to varying extents. It's too easy to forget this. Bottom line: Schiit Multibit DACs are extremely resolving of low-level information, more than even the latest S-D DACs, but they do this with a hint of that vintage R2R sound timbre.

    Basically, it comes down to four categories (NOS is an extra qualifier that needs to be discussed elsewhere) when it comes down to tone and timbre:

    Classic R2R
    UltraAnalog, PCM1704, PCM1702, discrete R2R ladders. I tend to detest everything here except for the UltraAnalog or Theta stuff, or the modern $5000+ Metrum and MSB models. Metrum, Wadia, Sonic Frontiers, Linn (their old stuff), CAL, A-GD, Airist (RDAC), Theta, Modi-Multibit and MOS16 (to a large extent, but not totally). Also Resolution Audio Cantata (which was not resolving).

    Reborn "R2R"
    Rockna, Schiit, Holo (Spring 3), Soekris (these three are also very different from each other, but as a group is much more different than the above). Some discrete R2R ladder designs here, but they sound more modern and less syrupy or dark compared to the above.

    Bad Delta-Sigma
    ESS before their latest Sabre chips, badly implemented new ESS, badly implemented Wolfson, Cirrus (boring), BB (dull). This reflects the sad state of audio engineering. There's not enough money in it for the chip companies to care. Just think 5-10 years ago when most DACs sounded like ass: Mytek, Benchmark, etc. Others like Cambridge attempted to make things listenable by smearing everything with poo.

    Modern "good" Delta-Sigma
    AKM and Analog Devices, maybe ESS Sabre 9038, but need more time to tell. Newer implementations sound better. There seems to be more awareness on the part of designers of how S-D designs of the past exhibited unlistenable "digital" highs (rasp, glare, hardness, edginess, grain, etc.)

    Now coming back to the answer to the original question, here is the answer in bullet points:
    • Some people will want more of the classic R2R sound, that dense timbre, thicker bass, liquid mids, rolled highs, slower more romantic presentation, which the Schiit Multibit DACs (again, Modi Multibit excluded) only have to some extent. They will go elsewhere.
    • Some people dislike the closer placement of vocals or instruments mixed to be more upfront or the generally closer soundstage (row 10 seating) of Schiit DACs. They will go elsewhere.
    • Some people will complain that the Schiit Multibit DACs are too incisive and not relaxed enough. They will go elsewhere.
    • Some people will feel the Schiit Multibit DACs are not incisive enough. They will go elsewhere.
    • Some people will bitch that the Schiit Multibit DACs don't have a sufficiently "black" background. They will go elsewhere.
    • Some people will disregard the Schiit Multibit DACs because don't cost enough. They will go elsewhere.
    • Some people will dislike the irreverent "I don't give a f**k about what you think" attitude of Schiit. They will go elsewhere.
    • Some people will hate Schiit because they are popular. They will hate on them, bork measurements of their DACs, or present only information that suits their narratives.
    • Some people will move on because it's human nature to get bored.
     
    • Like Like x 77
    • Epic Epic x 4
    • List
    Last edited: Nov 14, 2019
  3. rlow

    rlow A happy woofer

    Pyrate Contributor
    Joined:
    Jul 18, 2017
    Likes Received:
    7,738
    Trophy Points:
    113
    Location:
    Canada
    ^ Holy cow that’s great insight. Forget “sticky” this should just be posted on the homepage! ;)

    On a side note, I think that people moving on purely due to better measuring sigma-delta DACs should be called out seperately (with a note that says, show me what DAC measurements will actually tell you anything about how a modern DAC will actually sound, other than a totally garbage one). I realize the 2nd last bullet sort of speaks to this but not explicitly wrt S-D vs R2R measurements.
     
  4. geniekid

    geniekid Facebook Friend

    Contributor
    Joined:
    Mar 13, 2018
    Likes Received:
    137
    Trophy Points:
    33
    Location:
    Chicago, IL
    I feel like there should be another bullet point - the reason I went elsewhere is because I don't have the ears/perceptiveness to hear the sonic differences mentioned previously and therefore prioritized on other features.
     
  5. Ainsworth

    Ainsworth New

    Joined:
    Jun 19, 2019
    Likes Received:
    11
    Trophy Points:
    13
    Location:
    Allen Park, MI
    Greetings, @purr1n. I notice that you keep referring to the Modi Multi-bit, but it is not mentioned in the referenced Head-fi post. A quick search for Modi Multi-bit returns only the above quoted post. What I was wondering: Do you consider this DAC to be good Schiit or just Schiit ?

    You may recall me from the ZMF Verite / Aeolus thread. I pulled the trigger on the Sennheiser/MassDrop HD6XX, and I am thinking if I like those well enough, I might go a different direction and upgrade my DAC instead. I am OK with my current DAC (internal to my Cambridge CXN network player) but It does tend to be a little shrill with some material.

    I know that is not a lot of info, but I have not done the legwork yet to figure out what I like and don't like about my current DAC. Plus I don't have a lot experience with them. My only other DAC (other than those in my phones) was an ifi DSD nano which bit the dust after about 18 months of use.

    So I am going to stop short of asking for a recommendation, though a suggestion or two would be great. Or should I just got to the Schiit site and buy the highest end DAC I can afford from there. I like Schiit.

    Your reference to row 10 seating above is interesting. In Opera (no, I am not a fan of the genre), Row 9 is considered the row to be in. (Thus the term Dressed to the 9's)
     
    Last edited: Jun 21, 2019
  6. skem

    skem Friend

    Pyrate
    Joined:
    Nov 1, 2017
    Likes Received:
    1,911
    Trophy Points:
    93
    Location:
    Charles River
    But if you have $$...
     
  7. iFi audio

    iFi audio MOT iFi Audio

    Joined:
    Oct 3, 2017
    Likes Received:
    1,284
    Trophy Points:
    113
    Location:
    Many places.
    Home Page:
    Very enjoyable and quite accurate take on DACs in general. Nice one.
     
  8. tomchr

    tomchr MOT - Neurochrome

    Joined:
    Oct 28, 2016
    Likes Received:
    513
    Trophy Points:
    93
    Location:
    Calgary, Canada
    Home Page:
    I agree that focusing exclusively on THD/THD+N/SINAD is a bit narrow-minded. It just so happens that THD is one of the most well-established measurements in the industry, hence, known by most. There are other types of measurements out there, however.

    Those of us who use measurements to characterize electronic circuits and wish to get out of the "one number says it all" mindset, have started using multi-tone IMD as a way to assess the performance of a piece of equipment under conditions that resemble music reproduction. I use a 32-tone signal in my measurements, for example. The test signal sounds a bit like an out-of-tune pipe organ.
    A measurement of the noise floor is also highly relevant to a headphone amp. You can't have a low noise floor without also having good THD+N, so THD+N is certainly a relevant measurement for a headphone amp.

    The top speed of a car is a rather poor analogy for THD. After all, THD is a measure of linearity whereas top speed is not. A better analogy would be a measurement of how close to a straight line the car follows when you hold the steering wheel completely still. We've probably all had the experience of driving a car where you constantly have to fight to keep the car driving straight. It's a rather different experience than driving a car that can follow a straight line without much driver input. I much prefer the latter. Others are free to disagree.

    Tom
     
  9. frenchbat

    frenchbat Almost "Made"

    Joined:
    Sep 26, 2015
    Likes Received:
    1,369
    Trophy Points:
    93
    Do you publish the results of this 32 tone IMD test somewhere ? Your website maybe ?
     
  10. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Yeah except you can have similar THD profiles, or two profiles below what would be considered audible, yet get wildly different sounding DACs, amps, etc.

    A properly aligned car with a centered steering wheel will feel like just that. Won’t matter the make or model. It’s a known objective goal with known results, whereas THD and human audibility and perception is still debated.

    I don’t hear anyone debating if there’s more pleasantness if your car pulls to the left or right.

    Also while alignments are done on nice machines, steering wheel centering, AFAIK, is done based on a human’s best effort. (Either that or machines can’t center a steering wheel for shit.)

    And they will align you to be “close enough within a narrow spec,” and yet some drivers can tell when something is slightly off. It’s frustrating when you feel it’s not quite right and get told to pound sand by the tire shop!

    In that regard, maybe it is more akin to static audio measurements than I suggested. We can go off hard numbers all we want, but only hope engineers and designers listen when someone says a piece of equipment sounds like ass regardless of how it measures.
     
  11. tomchr

    tomchr MOT - Neurochrome

    Joined:
    Oct 28, 2016
    Likes Received:
    513
    Trophy Points:
    93
    Location:
    Calgary, Canada
    Home Page:
  12. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Top speed for a car is similar to SINAD for an audio reproduction device in that it's just as useless as figure.

    Top speed tells us almost nothing about a car so long as it isn't incompetent like that if a Metro Geo with max speed of 80mph. Same goes for SINAD or THD+N. As long as SINAD is better than around 80db, it's good enough. Most recordings have noise floors around that level or higher. It they don't, it's because an audio engineer ran the tracks through a noise filter in Adobe Audition or Pro Tools.

    Note that Car and Driver doesn't splash their homepage with a ranking of cars based on top speed. Anyone who splashes their homepage with a ranking of SINAD of DACs is way way oversimplifying things.

    The inclusion of top speed with SINAD with my plots was never intended to be an exact analogy. It was intended to ridicule the ridiculousness of any simplified objective approach. To point out that there is no one ring (one measurement) that rules them all.

    As far as measured performance, it's easy to get good numbers. Just use chips and massive amounts of negative feedback. Heck, just follow the implementation guidance from the chipmakers' data sheets.
     
    Last edited: Jul 6, 2019
  13. allegro

    allegro Friend

    Pyrate
    Joined:
    Nov 4, 2017
    Likes Received:
    637
    Trophy Points:
    93
    Location:
    Florida
    The objectivist extremists at ASR insist that gear that measures better has to sound better full stop. They are not interested in dialogue. Any argument to the contrary is usually first met with the challenge that "you did not match levels" during comparisons.
     
  14. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    It's not science when we don't answer how THD, SINAD, the SMPTE IMD, and AP's 12-tone IMD correlate to what we perceive. It's just measurement for the sake of measurements. We need studies on psycho-acoustics for it to be science because ultimately we are talking about the science of perception.

    At least top speed and 1/4 mile times tell us a lot more about how powerful a car will feel. There is a huge correlation with 1/4 miles that correlate with "butt-dyno".

    The entire discussion seems to have been railroaded into another direction. My point is that a decent THD figures of 0.05% (specifically at 0dbFS for 1kHz) are "good enough", and that's all we can really say about THD. After this certain point, I don't think THD says any more about how something sounds.

    I cringe when noobs rag on gear with 94db SINAD as measuring "bad" just because a ChiFi designer (competent, good at layout, but not terribly creative) who lifted the designs from a chip datasheet made a DAC with 114db SINAD. A SINAD of 94db is fricking amazing.

    Again, what's the SINAD (just the +N portion) of your listening environment? The ambient noise of your house? What's the SINAD of your 6.5" woofer at 100db playing 37Hz?
     
    • Like Like x 4
    • Epic Epic x 1
    • List
    Last edited: Jul 6, 2019
  15. Senorx12562

    Senorx12562 Case of the mondays

    Pyrate
    Joined:
    Oct 26, 2015
    Likes Received:
    3,274
    Trophy Points:
    113
    Location:
    Bird-watcher's paradise
    I understand that the use of auto performance was really not an attempt to analogize to sound. That said, in terms of analogy, I think the handling of a car (and even moreso a motorcycle) is closer to analogous to sound in terms of the relationship between the measureables and the end goal than anything related to how fast it is in a straight line. Measurements, whether physical, like wheelbase, track, weight, caster, camber, etc., or measured performance, such as gforce around a skid pad or speed through a set of cones, don't tell one much about how fun it is to drive (or ride), or in fact, to listen to.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Exactly, it's still a subjective hobby, much like cars.

    However, when it comes to measurables, people shouting "oh horrible, 93db SINAD, bad engineering", that's like saying "horrible, only 325hp", just because it's 2019 and 450hp+ can be easily obtained for under $45k. It's all about context.
     
  17. tomchr

    tomchr MOT - Neurochrome

    Joined:
    Oct 28, 2016
    Likes Received:
    513
    Trophy Points:
    93
    Location:
    Calgary, Canada
    Home Page:
    I guess I'm an objectivist extremist then. After all, I participate on ASR... :) We don't all fit the same mold, ya know.

    There's a lot of peer-reviewed science around measurements (plural!) vs listening experience. Harman Kardon used to have the results of Sean Olive and Floyd Toole's research published on their website. Sadly, they've taking it down, so I guess you need to dig out the original AES papers. They conducted their experiments in a controlled setting and applied proper scientific method (good sample size, double blind, proper statistical analysis of results, etc.) They concluded that equipment that measured better was rated as better sounding by the listeners. They tested this with a wide array of listeners: Students, Harman Kardon employees, audiophiles, studio engineers, professional musicians, HiFi reviewers, etc. and found that all the groups of listeners rated gear that measured well as better sounding than equipment that measured worse.

    Olive & Toole also performed a series of sighted trials. They concluded that better looking equipment is rated as sounding better. They also found that if the price of the equipment was known, the higher priced equipment tended to be rated as better sounding in sighted trials.
    They also found that Harman Kardon rated Harman Kardon gear as better sounding than competitors' gear in sighted trials. Imagine that. :)

    Some of the experiments are described in Floyd Toole's book on sound reproduction. It's a very worthwhile read.

    I'm not aware of any hobbyist that went to these lengths in their "experiments".

    Humans are not rational. Not only are we dealing with psychoacoustics. We're also dealing with a range of cognitive biases.

    Asking about level matching is perfectly valid, by the way. One of the oldest sales tricks in the book is to nudge the volume control up a little when you introduce the gear you'd like to sell. Louder -> better sound. That effect is established in science as well.

    There are tons of studies out there. Renew your AES membership and have a look...

    Yes and no. You can glean some from a THD measurement. You can get even more information from a measurement of the harmonic spectrum. If you like tube amps, you'll likely find an amp with ~1% THD and the harmonic spectrum dominated by the 2nd and 3rd order harmonics to be good sounding. If you like a clean sound, the 0.05% THD is likely a better target.

    On the other hand, if the designer took a 114 dB SINAD chip and turned it into a 94 dB SINAD product, they degraded the performance by 20 dB. While both numbers may be good or acceptable, it comes across as putting a Geo Metro suspension in a Ferrari. It may drive fine across town, but I kinda wanted the Ferrari.

    Heh. Yeah... :)

    It can be an objective hobby too if you want it to be. I think both approaches are valid.

    Tom
     
  18. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Wasn't Sean Olive one of the doofuses whose research gave us a headphone measurement comp curve with a metric fuckton of extra bass added because the listeners preferred it?

    I'm all for a bit of color to suit tastes or environmental situations, but ain't no way that should be suggested as a reference for testing purposes.

    I'd like to see studies on why some can successfully and blindly pick out differences in equipment that should sound identical based on supposed audibility thresholds in a wide variety of measurements.

    Because that kind of throws a wrench in all this and suggests we haven't really figured it out. Not to say we shouldn't use existing measurement techniques, but they shouldn't be the final say unless results are just awful.

    Or maybe do long term, in house studies to see what people end up gravitating towards. I've certainly had times where a "clean" sound is preferential in short bursts, or that I can hear might be more textbook accurate, but over a long period of time it becomes annoying, unfulfilling, or grating.

    So, we still don't know how to measure headphones for shit. DAC/amp measurements seem lacking beyond small use cases and limited studies.

    Unrelated to sound, but there's a lot of relatively new research on how light, depending on spectrum and intensity (or stability, ie flicker), affect us through our eyes, even at specific angles and diffusion levels, through our skin, etc. And it seems like every day there's something new about how CFLs and LEDs are kinda f'ing us over, most notably with screen time.

    Similar to short burst audio listening, a bright 6500K light may seem invigorating at first, but it's hard to keep up for long indoors. (Though improvements in CRI, flicker, where the blue light peak lays on the spectrum if using LED, etc may mitigate that as we learn more about how lights affect us.)

    Point being: My money is on us just only having scratched the surface when it comes to audio perception in relation to objective tests. It's good to use what we have today while we keep digging with open minds! But I think doubling down on existing measurements today as the final word, ala ASR style, is just myopic and self serving for those that can't or don't want to consider the possibilities beyond.
     
  19. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    It's 2019 and people are still having the same arguments that I was seeing in 2012 on Head-Fi Sound Science forum.

    Is anyone actually changing their mind? Is the discussion actually moving forward? Anyone operating rationally has checked out of this arena.
     
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I remember the test that Sean Olive did in regards to transducers (headphones). However, transducers have very high levels of distortion. We are really talking about different things. Gross levels of distortion in headphones and minute levels, so minute in DACs, even the worst measuring.

    So let me ask you, did Sean Olive and Floyd Tool correlate non-linear distortion from -95db to -100db to -115db to better sound quality? Or this is one of those appeals to authority arguments?

    In any event, I don't have confidence in Sean Olive's work considering that his latest headphone target advocates increasing the bass on the HD650 below 180Hz by 100Hz it is +3db and 50Hz +6db, increasing the upper-mids +6db, and mid-treble +9db.

    I would argue that the loudness trick only works with an inexperienced listener. This common refrain is a crutch that "disbelievers" oftenrely on. I should point out that I am all for level matched blind tests: https://www.superbestaudiofriends.org/index.php?threads/dac-blind-test-series.7518/. I believe a few others have also performed blind tests. I would also argue, based on my own experience, that level matching actually makes it easier for ears to tell the difference.

    As far as cognitive bias, this also rings true for people who believe that measurements matter. "No difference" is often heard.

    Uh no. Maybe if the tube amps were broken. 1% is too much. Or maybe I would say that I don't like tube amps that measure like that.

    I do think that if distortion is on the cusp on audibility, say 0.1% or 0.05%, then this pattern of 2nd order followed by successively lower levels with each harmonic is found to be more acceptable or normal compared to say just an obvious 3rd order sticking out.

    Bad analogy. The problem is that 94db SINAD is still extremely good. I'd venture to say better than any recording can exceed.
     
    Last edited: Jul 7, 2019

Share This Page