Is Digital Audio Transmission Really Analog?

0:01 hello everyone this is samir from audio science review i have a technical topic for you uh that's about how digital audio is transmitted and audiophile concerns around jitter and noise the motivation for the uh uh presentation came from people asking me about it but also this uh um uh video from darko audio a youtuber that covers audio and the meat of the topics actually after the introduction but the introduction starts with him holding up these two bottles of water saying that it's silly to ask for measurements of those uh when he could just open the bottle and and taste them and therefore what he's about to present

0:47 uh which i'll talk about in a second should not trigger anybody asking for measurements and uh he goes on and says you know it would be very silly for somebody to go and demand that measurements be provided for bottles of water and proceeds to taste them and i think he naturally says that fancier water is is better than the supermarket water it's a mistake uh in in what he's claiming in that uh we don't ask for measurements uh we'd love to see uh tasting uh and listening tests uh in his case all he's got to do is taste the two bottles of water blind not knowing which is which and then telling us which one's better

1:32 we'd love that kind of subjective control testing far more than measurements i don't even know what to do with measurements of water but i do know that a lot of research is done in in food and and tasting are done frequently blind or major uh cooking shows where the tasting is is done blind so i think his analogy is quite misplaced uh go ahead and have somebody else pour each bottle of water into two identical glasses and let's have him test that multiple times and then we can tell that his opinion is is not biased towards price and and looks and branding and so forth and uh the meat of this presentation as i said

2:18 comes later in that this is a review of of this streamer forget what the brand is at the bottom is a few thousand dollar digital audio streamer meaning that it it can book local content from your network or from the cloud and then feeds it to a dac and then you can listen to it and he's got a stack of them in here as you can see is the largest one that i think inuit was the one that it was testing down here i think this is uh uh not a sona so who's their competitor it'll come to me and he goes up this is zone is over here and then he's got uh an allo uh one over here and and then some other uh embedded one and uh sadly and naturally he says the the biggest and most expensive one sounds the best and then

3:04 you know progressively as the thing is smaller and cheaper the sound degrades um and then he confesses very quickly that you know this is incredulous you know to say that you know these digital sources have different sounds when the the source of the content is identical and he then goes on to explain that um what is actually transmitted as digital audio is actually analog and therefore he's got all the problems of analog and specifically has jitter in there so and he goes in and you know he's got a fancy production crew and they overlay graphics and say well this is what you think a digital audio looks like but in reality doesn't look like this and he's very correct about that i'm

3:50 going to talk about this in a second so is digital audio transmission analog the quick answer is absolutely yes and you know it doesn't have to be digital audio it can be digital anything inside your computer there are tons of data streams that are running around and the waveform can indeed be distorted but what distinguishes digital transmission from analog in is that we actually don't care about the uh the the small or even medium level changes in in the signal itself we set a threshold if look at the waveform on the right we set a threshold in here we said look anything above this threshold is a one anything below this threshold is a zero so i don't care that the waveform has been distorted over here i'm easily able to extract a one

4:38 and a zero and one's over here and indeed these signals can degrade very close to this line and all is well now once it gets really really bad and we flatline on the threshold then hell breaks loose but until that point you know the beauty of digital is that it can take an analog concept with variations and convert it into something that's got exceptional noise immunity and this is what makes it powerful and the fact that we can ignore noise and variation uh completely sort of counters to what what he's assuming and what he's talking about so just because something is analog doesn't mean that that is bad yes if we treat this as an analog voltage with all these variations that would be

5:24 bad that's why analog has noise in it but when we treat this digital that is not the case at all so on at a fundamental level there's a truism in what he's saying but he's completely bastardizing this into something and it's not now i could stop here and say this is completely wrong and i can get 100 digital engineers that sit next to me and just laugh at what is presented but i want to be fair there are issues in here that go beyond the superficial treatment and dismissal so let's uh go into that now this pretty graph that i showed you it only happens if i let's say i've got digital audio over very short distances instead of short distance let's say i have a broadcast stadium oops

6:10 broadcast studio and and i've got signals have been transmitted between rooms and even buildings and now we're talking hundreds of feet and across hundreds of feet these voltages uh come down in value we get losses we get filtering these edges become less sharp and indeed damage can be done it's very seriously to to this way from to a point where it doesn't do it doesn't transmit digital data and good news is that when that happens that the impact is catastrophic in analog you know if you keep increasing noise it just gets noisier and noisier noisier digital doesn't work like that has this cliff where that the transmission is perfect to all of a sudden we can't capture something and it could be one bit or it could be a

6:56 thousand billion sequence that get corrupted and when that happens is very catastrophic in case of video all sudden you see pixels that are gone and and corrupted uh you probably have seen this especially if you have an antenna for digital uh video dtv standards and for audio uh most people assume they know what this look sounds like but they don't and i've produced some samples of data loss over uh uh usb for uh that's transmitting audio and i can assure you the sound degrades substantially there's glitches there's buzzing there's pops there's discontinuity it is not at all the subtle thing where micro details change and the timing has

7:42 changed bass is a little softer the data transmission has no notion of what that data looks like or what it is for it's just a bit and if you obliterate one or more bits the data samples can vary sharply from one another and that's what causes the ticks and pops and glitches there is no such thing as a subtle degradation now to make that data loss happen over usb i had to create a 30 foot cable with with 15 foot extensions a very very thin usb cable with a lousy transmitter and receiver to cause it to have data loss and in a good connection actually still couldn't get it to fail and indeed if you have a dac that's hooked up or you that's over your spitf coax or or

8:28 usb and it's any reasonably short length of cabling there is no data loss whatsoever just put that out of your vocabulary data loss can happen if you exceed the standards and that means going many many meters so if you're going from one room to another room or upstairs downstairs yes you can get data loss but in any kind of stereo setup where you got a stereo system sitting someplace and you got your streamer here and next to it there's a a dac that's five or six or eight or ten feet of cabling it absolutely will not have any data errors so assume 100 that there's no data errors and if there is no there's glitches and pops and it's not some subtle staging or micro

9:15 dynamics or any other silly subjectivist commentary now so far i've only covered one aspect of of digital audio which is what the data samples are he says this is the pcm value of this and the dac will take that pcm value and convert it to an analog value and it just follows whatever the digital stream uh is but we also need to tell the dac how fast the bits are coming what is basically the effective sample rate of the data is because we can have 44.1 48 96 from what have you and the dac needs to know that to be able to play it at the right timing if it plays at the wrong time that can generate distortion um the most classic links that we have is coax

10:03 and tossling both of them are called spdif and then there's a aes ebu professional standard that's a balance and whenever using these connections uh key thing to understand technically is that the host is a master so if it's a streamer or as a cd player or a computer that's playing the host is telling the dac i'm going to send you the split stream and you better follow the spit streams and only output your samples whenever i give them to you you might think well i'm playing 44.1 can i just tell the dac that it should play at 44.1 the answer is no that's a nominal rate for it to display on its display but the dac is not allowed onto on his own to assume there's a fixed data rate the host can actually change this by up to

10:49 five percent it can speed up and slow down even at displaying and uh the dac needs to stay in sync with it this is called a master slave kind of relationship and the dac is the slave this is a very poor architecture it was done for simplicity and the problem with now with this scheme is that the dac has now has to extract a timing of how fast the bits are coming at it and convert that to a sample rate internally and unfortunately with coax or any of these links there's only a two wires going there a shield and a center data pin there isn't a separate clock pin that says hey here's the sample rate that's coming at you and just take this the data itself has pulses and these

11:35 pulses need to be grabbed and the distance between the pulses is what tells the attack what the sample rate is and the second thing that can happen is that there can be noise on this line and that noise can actually be coupled from the host all the way out to the uh target toslink being an optical link helps a lot with this uh oftentimes people have ground loops and you go when they have a coax connection or usb connection and they switch to toslink and that goes away so this can be a problem and so the architecture here is limited and it can have problems um so let's look at this thing with with real uh detail and instrumentation i took my

12:22 audio precision analyzer which has both the spdif output and input and i uh had it generate this bit of signal and then i took that and hooked it up to my uh oscilloscope tektronix oscilloscope and had it gen display what is called an eye pattern so if you follow my cursor so every bit sort of looks like this is nothing is basically negative meaning it's a zero and then it gradually goes up and then stays constant here then it comes down so you're seeing one bit what is overlaid on it is another bit that has arrived and it's out of cycle with it uh this is a digital scope so it's able to catch these things and if you let it run often enough where a lot of bits get

13:07 overlaid on top of each other you get what is called an eye pattern it's called an eye pattern because these things look like two eyes if you look at the left and right eyes and engineers talk about the clear eye in that if this area in the middle is very clear and very wide the receiver has no trouble extracting the data from this link and indeed that's the case in here even though i have a cable is three to five feet between the audio precision and tektronix scope we can see the eye patterns is extremely clear but we also see that as darko has said is that there's no this is not digital uh the waveform the waveform actually has a slope to it and gradually goes up to one and the uh dac actually grabs this uh

13:54 zero crossing over here as the timing point of view so it says okay start a clock here okay the next one came over here and these two points are now the the distance between them and measured in time is now your sample or bit rate that's coming at the dac if you take this waveform now and uh jiggle it left and right what we call jitter or make it noisy where the edges are fluttering over here this timing measurement can get uh changed as well you can have random errors in here or it can have um a deterministic so let's say the waveform jumps once a second left and right while they'll also transmit to the deck

14:40 now good news is that the decks have been designed to actually work in a far more hostile environment all the way up to what the spec requires so you you know you could have 1500 feet of coil cable so they have an internal flywheel mechanism if you know how turntables use the momentum of the platter to get a steady speed they have their own waveform called pll or phase lock loop and the phase lock loop uh it basically measures these this timing and generates an internal clock that's immune to high frequency changes low frequency can still get through but high frequency one it just filters them out and doesn't care so if there's very high frequency noise in here let's say from your cell phone or wi-fi or you know am station fm

15:28 station everything it it happily filters all that and gets a clean one and we can see in here the waveform is very clean anyway this is an actual waveform this is not a made up thing and the scope is receiving and displaying it and we can see that we have no trouble over here and these edges in here are quite precise so we don't need that but if we do need it um the mechanism is inside your dac and that's why when you go change one digital cable to another digital cable you know you don't notice anything because while the filtering is there and the fidelity of the signal is already quite high usb audio started off using a similar kind of uh mechanism the people designed the usb audio initially and the audio devices worked the same way in that the uh the

16:13 pc was the master and the dak was a slave that actually caused worse jitter than than than spdif because you know it's a computer and the computer just you know doesn't care if that timing so precise as first usb pack is coming in but after a few years of usb audio being out there um some clever folks came up with a mechanism to conv use a asynchronous mode of usb where the usb is still actually trying to transmit real-time audio stream but the the dac picks its own clock so this time he says hey i'm playing 44.1 and i'm going to tell the host if it's five percent faster i'm going to say hey slow down slow and slow down until it slows down the four four to forty four point

16:59 one and if if it's going too slow i can actually tell to speed up so you know it's like you as being in charge of a um uh you know autopilot you know you're basically accelerating decelerating uh cruise control in your car and that's what this is a mechanism where because usb products are smart they have memory and they can have a cushion against the host not being in sync with it is able to basically eliminate any dependency whatsoever on a digital clock in a in a usb so usb is also serial link it does have a separate clock but its clock could also have all the noise issues that i've talked about but that clock on usb is only used to extract the data uh once the data is extracted it's

17:46 never used for timing the timing the the dac is the master and you can have a very high precision high quality clock that's right next to the dac chip doesn't care at all about like speed of trying to measure those corners and and arrive at good timing okay so in case of usb we don't worry about data errors and we don't worry about jitter over usb bus it simply is not something that's in the vocabulary that we have so do we say that the bits are bits is a common objectivist thing to say and laugh at anybody who says you know digital audio can degrade performance the answer is no you can't say that because i've talked already about noise and jitter and let's talk a little bit about more about noise a usb cable

18:32 has data and glock but it also has power and it also has a ground some dags run on that usb power i had to shut modi ii that any noise on a five volt supply would translate into hideous jitter on its analog output and you could take that back and plug it into five different computers you get five different measurements and fortunately that's an example of something that just shouldn't have ever gotten out and even at that time there were a ton of other decks that didn't have this problem and today the quality of usb receivers or anything i test and recommend is just superb on this front although occasionally i still run into this in some avr's i see this thing some you know smart amplifiers i've found

19:17 that their usb receivers are not robust and that can pick up noise through either uh ground or or voltage um the uh excuse me one of the common symptoms a lot of people report for example is that with an internal dac to a computer or even a crappy dac externally when their computer gets busy they actually they can actually hear the activities especially when the gpu the graphics process and inside the computer starts to get going they can hear you know different kinds of noises all the way through their audio device so that proves that bits are not bits because remember the data hasn't changed so that from that point of view is the same but because we have an electrical link that electrical link ultimately can

20:03 transmit stuff there's a you can have a galvanic isolator that you can buy either separately or one that's in in some dax that really provides complete isolation between the two uh fortunately if you get a good deck 99.99 of time it's not a problem and if he is balanced audio output as i always recommend that actually becomes quite immune to this problem it's only mainly an issue when you have rca connections on your deck okay now does this mean we don't get jitter noise actually we get plenty of it and i measure it all the time this legit the main source of jitter and noise isn't the lane the link or the host behind it it's actually internally generated a dac is a mixed signal device it has both digital and analog subsystems and a

20:49 lot of times the noise inside its own digital section can bleed into the analog or you could have lousy power supply that has noise on it all of that gets translated to an analog output of the dac reason is that audio dacs have extremely high resolution the bit depth even a crappy one say 16 bit cd audio we can go the way up 24 bits so even slightly you know movement in those samples generates error so from that point of view it requires a very very clean implementation fortunately we get that in a lot of devices now what i've highlighted in here is that to know that if we have a problem here we measure we can measure jitter we can measure noise all the things that darko says well you know we can't measure we should measure it's like no

21:35 if you're going to talk about something technical being a problem and you name it as jitter well we have tools that let us measure jitter so here's an example of a topping dm7 and a channel deck that i measured and this graph is just from the review itself and i feed a special signal to it it's called the j test and uh we can see that the noise floor down here this in the analysis i can bring this noise floor down and able to even resolve these spikes in here there this is a valid set of bits that's a single bit toggling at the limits of 24 bits so we're able to actually correctly see the sequence of bits in here being resolved and if you look at the spectrum

22:21 in here just absolutely clean all we see just randomness of noise at extremely low levels so it's an example of a case where this is running from my computer which is a noisy source and a noisy streamer and it's wailing with 16 cores in it and gpu and everything else yet this stack absolutely does not care and it's producing exceptionally clean signal resolving it down to 24 bits of resolution but it's not always that way as i mentioned this you know these things can degrade and here's an on qr tx rz50 that i tested i think a year ago and uh we can actually a few months ago uh 212. and we see that not only does it have all these spikes this is identical measurement so it should be just like

23:06 the other one with noise down here and just a few bits down here but instead we get this regular pattern this regular pattern is actually part of the j test signal and is bleeding from the digital section that's receiving that data into the analog section so remember all these measurements are actually measuring what comes out of the deck i'm not actually measuring this clock to see how accurate it is i don't care it's a black box i just say give me your waveform and the waveform should only have this one spike in it at 12 kilos anything else is unacceptable so all of a sudden we see an excessive uh pattern over here over coax over hdmi uh is much cleaner but it's noise floor still way up here much worse than what i just showed you

23:52 but you still notice this comment in here as bad as this is with the eye being insulted that there's all this in here compared to the other that i look at these measurements and i say look the threshold of hearing is around -115 so almost all of these things are below threshold of hearing the few of them over here above threshold of hearing but they're right next to a very strong 12 kilohertz tone and therefore masking will make sure that you don't hear that and on top of that real content doesn't have full ample to 12 kilohertz anyway usually so as bad as this is you're not gonna hear jitter from this uh avr you know it technically is poorly engineered you shouldn't have this but the fact that it does is still not a bothersome thing now

24:40 there's an automatic assumption that the better the story and higher the price and the fancier the brand the cleaner these things must be uh it's not so i've tested a lot of products indeed unfortunately many high-end products have inverted performance where the price goes up but performance also goes down and how do we know that we measure it so this is a twenty five hundred dollar deck um this is an audio gd master at seven uh china chinese dac that is revered used to be so it's a massive box with tons of over design circuits in there but look at this jitter performance usb's down here is still not as clean as what i showed you in topping it does have some jitter components but again they're at minus 130 db or so when i switch to coax and

25:27 toslink all of a sudden the problem gets far worse remember now it has to go extract the clock itself and when he's trying to do that it just has a is doing a very poor job in here and is it going to be audible no this threshold of hearing is way up here and just shows poor engineering it's like if you can have clean jitter and noise why not get that and and put up with this thing and now you wouldn't pay twenty five hundred dollars and still got boars performance um this is a recent example that i tested this name unity atom uh this is coax input and look at the noise floor up here is way elevator in here which means got random noise that can cause jitter it's one of the things but one of the other thing it does is that it masks

26:12 all the jitter components below this so these you know if the noise floor was down here we would see a very horrible looking picture with all these spikes being jitter components over here and right now they look like they're small but in reality they're way above a good noise floor of another system clearly this is not something that you want to be proud of you paid a lot of money for this dac i forget how much it is it's thousands of dollars and and you know look at the noise floor that you get so more money didn't get you less noise and less jitter another example 7 500 uh sacd player a cdsa cd player from rans sa-10 and look at these pronounced

26:57 spikes in here these are all digital activities inside the uh the unit they're bleeding into the clock circuits or our reference voltage of the dac resulting in this deterministic spikes coming out here again i'm always fair i do a psychoacoustics analysis real quick and i say well threshold of hearing is at -115 so no this is not going to be audible but by you know you pay 7 500 big heavy cd blur but you know you're getting far worse performance than my computer this is my computer for driving a dac that's just a few hundred dollars and it's a channel that i can get this clean and then you go get a

27:43 you know 7 500 transport thinking oh my god it's going to be so much better digital stream that it outputs it's like no that it's not very clean um let's talk about uh streamer back to the you know what darker talked about i've measured the digital stream coming out of a lot of streamers that have spdif output and i'll tell you the it's just gonna become boring because the amount of jitter they have on those links on a short length of cable it is so small as to not matter and uh it did any little bit that it does have the dac will filter it anyway i can show jitter down to one or two picoseconds and i've shown that but it's like one or two picoseconds i mean it's just not

28:28 something that's in our vocabulary as far as impact audibly now let's look at an example of a streamer that was way on top that he was showing and this streamer not only is it a streamer it's built on raspberry pi a single board computer but also the dac built in and the whole thing cost 210 209 look at its performance over here look how clean it is it's not as perfect as what i just showed you in the topping stand alone deck we have you know few spikes in here it could be jitter asparagus things but their levels are minus 140 db in here so it's absolutely utterly inaudible and not a concern at all and if a 200 dollar deck and streamer combo can produce this why would they worry would to you know say

29:14 well is there something that's cleaner than this um i i've already been explaining this to you um every time i make this jitter measurement i do as uh psychoacoustic analysis as i've been showing you and you know maybe once or twice i've seen something that made me think oh wait you know that could be audible in general just shows you know poor engineering and i said look you could have designed this product without this and costs nothing more i mean basically it's just poor engineering design skills hygiene that causes those that lack of isolation and poor clock extraction there's no reason to have jitter this has been you know people have had 30 40 years of experience dealing with a digital audio jitter um on this thing so

30:02 now even if we do have jitter then it means that saying that jittery cause bad sound is not valid because you know across almost 400 devices that i've tested maybe one or two of them have had bad jetter the other ones no matter how bad they are they haven't risen up to the level of audibility let alone be something that's that's nasty but we only know that because we measure so if you go by random five thousand dollar streamer just like that moran's transport it may actually be worse anyway so this lay assumption that i paid more and it's more expensive and it's a bigger box therefore must be cleaner is wrong and certainly wrong to just run around and say oh our differences in sound on this

30:48 thing okay so translating this into streamers basically if the streamer is outputting a digital stream and it's got a usb connection or even speed of connection and you're getting a good deck that i recommend you know which goes down to 100 bucks it's just a non-issue at all and and performance is the same regardless of the noise of the source a lot of people say oh the computer's noisy therefore dedicated streaming must sound better no you know every measurement that i make i actually use my computer as a streamer when i test streamers and there are streamers that measure that sign it's 120 db there's as squeaky clean as low noise as you can possibly

31:34 get so the it is true that the sources can be noisier than others but it is not true that you can just you know wet them in the air and say oh this is noisy and the other one therefore must be worse or better as dark was doing there's zero legitimate reasoning behind that we can measure it and across all these devices i've measured what he's doing is just totally totally incorrect okay so where are we it's just not a concern somebody i get this all the time what streamers should i buy i said buy the one has the features and the price and support that you want that's what i do you know it just makes no difference my main system i built a silent windows pc and that runs the room server for me and that is my streamer device would i do it

32:20 again i don't know i don't like the fact that windows update constantly reboots it and sometimes roon doesn't start and i gotta go manually restart it and uh so from that point it is not good know that every other streamer you buy that is not a windows or mac pc is gonna run an operating system called linux because of all the networking other features you need you need a full blown os therefore it's a computer usually it's a multi-core cpu yes it's not as as high performance as a computer is but it's not a noise-free little you know custom circuit it's it's a full-blown computer that's in there and you have challenges in that you may not get updates if that company goes out of business how you're going to get it to work and support new features in the

33:06 future uh nice thing about the pc or mac is that you know it's an open platform and forever you're in charge of that the downside is the manageability of it so you got to decide you know what works for you but that is the uh important thing the next important thing is to get a dac that i recommend and i've recommended i don't know probably 100 dacs and out of them 50 of them are just amazing performance state-of-the-art um for good measure use a balanced xlr connection if you can build a dac and a preamp and amplifier that all balanced you save yourself a lot of grief with ground noise and ground loops on that and you do that you'll be one in a million that you you're still gonna get noise and what have you jitter is just not a concern in any of those

33:52 now i've talked about a lot of measurements but i started the whole presentation on on subjective testing and i absolutely absolutely love to see subjective testing from any reviewer but subjective testing needs to be blind it can't be based on price and size and who gave you a freebie unit the next day the previous day and what have you if you think the sound is so easily distinguishable from one to the other just do the listening test blind the digital levels likely are the same so just have someone do an a b test for you to repeat it a dozen times and let's see if you get eight or ten times uh out of 12 that you guessed correctly which one is which then it's fantastic we'll chuck all these measurements aside and we'll

34:38 go by that shoot a video of that document the protocol how you did that that's great but if you're going to do a sided test as dark goes down it's just as worthless as a fortune cookie at the chinese restaurant you know it just means nothing he's so biased by by classic folklore out there and i can't blame him personally because so many people are biased to think if you pay more you get more that's how real life works you know you buy a hundred thousand dollar car it's comfortable and faster than than a used five thousand dollar car but just know that when it comes to audio unfortunately that equation has been thrown out long time ago companies are happy to sell you stuff that's super expensive because so much of it went into the the channel that distributes it the machinery the you know all this other

35:25 stuff that is not performance related hopefully you learned something out of this uh presentation and you're more educated both when you're talking to objectivists and subjectivists because there are people mistaken about these concepts on both sides of the equation okay see in a future video bye-bye