Can enough memory fix digital audio?

[@ThinkingBetter]

Asynchronous buffering is an obvious need in digital audio processing. However, assuming the audio is buffered, the next topic is jitter in the master clock oscillator. The era of digital music started in the early 1980s and if oscillator jitter is such a big topic still today (which I will claim it is not with proper engineering), we should have some serious issues with recordings of the first CDs using digital sampling with 35 years ago oscillators, right? An audiophile person believing that he can hear jitter issues in modern equipment should easily be able to identify jitter issues in CDs from 35 years ago. Or what? :-) So which title of song from the early days in digital audio has audible jitter issues in the recording?

[@cebruthius]

I like the way you think :P

[@christianholmstedt8770]

Uh...ohh. Trying to use known science/engineering/knowledge is not going to convince PS Audio or its followers. I've tried. The discussion quickly devolves into some semi-religious new-wave style of thinking.

[@ThinkingBetter]

Christian Holmstedt Yes, there is too much snake oil talk in audio.

[@ayporos]

wait. I am confused here. A CD is a digital storage medium.... If the audio bitrate of a CD is known and standardized, then what's the point of buffering to fix/get rid of a clock signal that might be jittery?... If you read X amount of bits, then you know that's a second worth of audio... and then you read another X amount of bits, then you know that's another second worth of audio... Considering there are 2 parts of conversion there's 2 places for issues to occur: 1. recording - I.e. the clock of the recording device converting the analog source to a digital format to be stored on the CD 2. playback - I.e. the clock of the DAC converting the digital data stream from the CD to an analog signal What's all this talk about a clocksignal inherited from the CD-player?... are you telling me the DIGITAL output of a CD-player also has a DIGITAL clock output or it has an ANALOG clock signal?.. and why? Or is this clock simply linked to the speed at which the data is read from the disk? (which sounds even worse)... Of course you need a clock signal to convert digital data to an analog (audio) stream! Duh?!?! Am I correct to assume that for some strange reason people actually used the speed at which the digital data was streamed as some form of clock signal?... because that's one of the dumbest things I've ever heard. o.O

[@ThinkingBetter]

ayporos The DAC needs to clock each sample of audio out precisely. In a CD that happens 44,100 times per second with 32 bits of data, 16 bits for each channel left and right. The DAC must always have a sample of data to read from for every clock tick or you will hear an artifact. When data is read from a somewhat unreliable mechanical media like a CD or even streamed via the internet, the data is not flowing precisely with 44,100 samples per second and to ensure the DAC has data to work on always, some buffering is needed. It’s quite similar to how KZbin tries to make this video running smoothly by buffering even your internet is not always stable. The data flow is asynchronous from the DAC or your video decoder. Before the buffer runs empty, the buffering logic will pull more data to keep the DAC or your video decoder happy. Nobody is using the speed of which something is streamed directly as the effect would be terrible. Well, you can argue traditional radio and TV work in such way but they are based on broadcast signals.

[@JonAnderhub]

All CD and DVD players contain buffer memory between the EFM code demodulator (basically just a lookup table) and the CIRC decoder. It's this very buffer that is monitored to control the disc's rotation speed and ensure a constant linear velocity. This linear velocity is constantly being tweaked to ensure that the buffer never overflows or is ever starved for data. As far as jitter at the DAC stage is concerned, there isn't any, other than the jitter in the master clock. Each stage from the CIRC decoder to the DAC filter is slaved and synchronized to the chipset's master clock (which has an arithmetic relationship with Fs).

[@ThinkingBetter]

Avoiding timing jitter through buffering allowing an asynchronously flow of audio data and high precision DAC clock is decades old technology. First such buffering actually aimed at avoiding data interruptions on the Mini Disk (MD) format in 1992 (anti-skip). There really isn't any magic to this. What you want is just a very precise clock running the DAC and enough buffering so you never run out of data to feed into the DAC. Having buffering allows for some more sophisticated error corrections when the music originates from a CD. This includes re-reading the track. Essentially this is what anti-skip was born to do, but it's exactly the same mechanism. But for any digital audio playback it's a must-have approach to buffer the audio and ensure the DAC runs precisely. Rather than promoting this as something special, it would be dumb not to do this. But of course Paul has a point as some audio engineers are just poorly designing the clock architecture and end up with jitter even it's not actually hard to avoid.

[@manardh7387]

I had one of those portable players for walking and listening. I always thought the NON-memory setting sounded better.

[@ThinkingBetter]

Manard H That might be due to the buffering going cheap by compressing the sound to save the RAM space. Some portable players did this.

[@TheTrueVoiceOfReason]

That's part of the reason why I bought my Lennoxsound Portable CD player back in the early 90's. It has a 1 minute buffer and plays CDs with perfection. While the headphone out is good for on the go, it's the line level output that sold me on this wonderful device. That and it came with the "optional" power supply. ;) ;)

[@ThinkingBetter]

World’s first digital audio recorder some 40 years ago actually had some awesome specifications. 32 channels, 50kHz sampling rate, 16 bits resolution, 20Hz-18kHz +/-0.3dB, 0.003% THD. It was made by 3M and called Digital Mastering System. Jitter was not specified but the sampling clock was by crystal oscillator. It even had CRC error check and correction. The recorder used 1800 Watts. You can check first generation digital audio jitter appearance e.g. listen to the very first digital recorded popular music album “Bob till you drop” by Ry Cooder using this recorder in 1979. Can you hear recording jitter on it? 😋

[@rich1051414]

So the 'digital lens', in other terms, is actually a multi-threaded DAC. As a programmer, that makes far more sense. You have one thread on the 'clock' that is only storing data into memory, then a separate thread, on it's very own clock, which is processing the data waiting on it, and the two threads have completely disconnected clocks, therefore there is no jitter to speak of.

[@cebruthius]

The digital lens is not a dac. SPDIF goes in, SPDIF goes out. It just "reconditions" the SPDIF signal by buffering it and sending it out on its own clock.

[@wilcalint]

FWIW: "Streaming" over the Internet is a misnomer. The data, video + audio, arrives in irregular bursts. If I look at the incoming data stream here for this KZbin video I see about one second long bursts sometimes reaching 1.5MiB/s in height every second or so. It is not a squarewave as at any given time in the burst the data rate varies widely. This is very common. So a large incoming buffer is to your advantage.

[@wilcalint]

One more thing. The bursting stops well before the video has completely played. Therefore the process is completely asynchronous.

[@cebruthius]

Audio over USB is packet based as well.

[@jonathansturm4163]

My friend from the mid-70s, Gary Bennett, lives in Stuttgart these days. He was a talented mathematician and physicist as well as a passable volleyball player in those days. But his ambition was to be a professional opera singer. His voice, though sweet, was deemed by the experts as too small. Fortunately, Gary ignored the experts and you can hear him singing one of my favourite arias, Puccini's Nessun Dorma, here: kzbin.info/www/bejne/gn2xgYqXm8iaask Gary is the very tall tenor on the left.

[@gerhardwestphalen]

You never really directly answered the question. Yes, this is what you guys are doing, but in reality does it make the source (and all cables leading up to the DAC) irrelevant? Or does it not completely eliminate that even though in theory it would?

[@Tom_Losh]

Yes, because it is transferring the digital bits, in their proper order, to a new media, then playing back those bits from that media using a precision clock. It really does eliminate the timing errors from reading the disc and transporting the bits to the DAC.

[@gerhardwestphalen]

By that logic, if you buy their DAC, then there's no point in buying their CD player as well and you can use a cheap CD player as long as it has a digital out. It would also eliminate all other "accessories" like isolation products and all cables on anything before the DAC. People would save thousands of dollars by getting their DAC as they could essentially sell everything else up the chain and replace with cheap components. Sounds too good to be true.

[@cebruthius]

> Sounds too good to be true. Well it is true. Digital is digital up to the point where you need a precision clock to turn it into the analog levels.

[@Synthematix]

Well to begin with digital was invented to replace all the ageing analog technology, when people say "digital is crap" this really annoys me, eer, no, cassette tapes were crap, not even a nak dragon can match the cheapest of cd players ;)

[@manardh7387]

This gets very strange since I understand the USB cables still can have an affect on sound quality. To many people the transport quality still makes a sometimes BIG difference on the quality. In other words, if we rip CD's and then play them on some memory storage (buffered) DAC's, we should get equal sound quality to playing straight off a transport into the DAC. I haven't read anything from reviewers on this, at least as I can recall.

[@WillyJunior]

USB cables absolutely do not have an effect on sound quality.

[@christianholmstedt8770]

The statement above is 100% true and correct.

[@CRAZYCR1T1C]

Reviewers are not engineers and use magic and voodoo to justify their jobs.

[@indranilsen51]

I don't think that Paul addressed the 2nd part of the question which is about a cheap CD player feeding a very well designed DAC. If the DAC is really well designed with buffering capability as Paul talked about would a cheap player feeding it produce the same sound as a relatively expensive transport like Direct Stream Memory play? Paul could you please comment?

[@cebruthius]

Yes, that is correct. However since the communication for SPDIF is unidirectional, you might get buffer underruns if the CD is too slow or buffer overruns if it's too fast. However that concern is mostly theoretical.

[@isettech]

Paul mentioned lightly the buffer with 2 clocks. What is important is type, not the size of the buffer. The buffer is a memory which can be written to and read at the same time. CD players have this from the very first manufacture. The clock for data from the disk has jitter, and wow. The clock is retrieved from the data from the optical disk. The disk motor speed is controlled only by the buffer fill percentage. The buffer runs about 1/2 full. When the buffer is less than 1/2 full, the motor is sped up to make the buffer 1/2 full. When over half full, the motor is slowed until the buffer is 1/2 full. A jitter free clock pulls data from the buffer at a constant rate and clocks the DAC with the same clock. This data stream is jitter free. Why Paul is renaming this a lens for a CD player I don't understand. My credentials is as an ISCET certified electronics technician. During the economic down turn in the 1980's I worked as a consumer electronics repairman and took the Phillips repair class on these when they first came out. At that time I learned about the 8 to 14 modulation used on the disk, the error correction, the pickup, motor control, tracking, etc. I have since progressed into engineering as the compensation and rewards are better and consumer electronics became to cheap to repair for a profit. For more info on this Google International Society of Certified Electronics Technicians, Solomon and Reed error correction, Eight to Fourteen Modulation, Compact Disc spindle servo, and static dual port RAM. Not sure what this lens fixes other than sources other than a CD such as Bluetooth, and internet streaming audio. Buffer size should be big enough to never overfill and never go complete empty, and should be true DUAL PORT Static RAM. Adding more computer system memory does not meet the critical real time clock requirements for a jitter free clock as this has to work around read and write cycles and dynamic RAM refresh cycles.

[@rich1051414]

So the 'digital lens', in other terms, is actually a multi-threaded DAC from what I can tell. As a programmer, that makes far more sense. You have one thread on the 'clock' that is only storing data into memory, then a separate thread, on it's very own clock, which is processing the data waiting on it, and the two threads have completely disconnected clocks, therefore there is no jitter to speak of.

[@Peter_S_]

On your last point, you're talking about RAM from the 1980s. It's not the 1980s any more. My credentials include having designed and built DSP based audio effects processors since 1989. DDR4 is clocked at 200-400MHz and a CD delivers 1.4112 megabits per second. Given a 200MHz clock that means you need to fill a word wide audio data output register once every 2267 memory clocks. Clearly this is a vanishingly simple task with today's DRAM and refresh is transparent to it. Any bargain basement DRAM combined with a double buffered output register in front of the shift register is up to the task. This has been the case for a couple decades now.

[@isettech]

This is inexpensive and quite good. DSP's do indeed have registers for just in time delivery to a DAC. Technology has marched on with vast improvements. My point is there never was an issue with clock jitter from the buffer to the DAC from day one of the CD player. It is most often a crystal with a base accuracy of 0.005% at worst. A little component noise may introduce infinitesimal noise jitter. The simple task of moving a word to a register every 2267 clock cycles is completely valid, however most interrupt routines are many clock cycles long. Some interrupts are higher priority than others. Loading a register within a specific time window is simple. If the same system is responsible for moving it from the register to the DAC, there lies the problem, unless the DAC is running on a dedicated clock asynchronous and jitter free from the processor. With the move to high clock speed DRAM and without a dual port RAM or DSP or other correction, there is an introduction of clock jitter for the reasons mentioned in PC based solutions. Audiophiles are claiming they can hear this, so it needs corrected. Stand alone CD players never had the jitter. They came with dual port RAM buffer for the spindle speed control. That was why I mentioned there is a solution already for moving between a variable time data buss and packet info to a clocked memory to DAC jitter free solution as used in the earliest CD players. Audio based DSP projects depending on design sometimes have jitter as they run on a single crystal that is not directly divisible to the DAC clock which introduces jitter. For example a 200 megahertz clock will not divide evenly to the CD clock rate. There is a remainder. A perfectionist will want a jitter free clock. One dedicated to the data to the DAC transfer. Speaking of spindle speed. Paul mentioned about 600 RPM on a CD. LOL. CD is a constant velocity medium, not a constant angular velocity medium. Speed in RPM varies considerably from edge to center.

[@shintsu01]

is there way to disable this large bufferr in the dac to a small one? like maybe lets say 20-40ms so that it can used as a sound card?

[@TheMB2333]

Based on your answer, I wish I knew why a company invests so much in their source's clocks. Source devices like Rasberry Pi or SOtM Ultra. These devices connect via Asynchronous USB and it's always been my understanding that the clock is controlled by the DAC for all asynchronous connections. That being the case, why is everyone raving, folks like Hans Beekhuyzen, about the performance of the source to a modern-day high-end DAC? Yes, I realize these sources are cleaning up the signal in other respects, but the CLOCK is always mentioned as important.

[@Crokto]

one possibility is that this method requires the injection of lag into the system, and there are use cases where that's unacceptable

[@dandonna3904]

how does work on iPhone when I have 2,500 songs on the iphone

[@4G12]

If maximum sound quality for sound playback was the only goal that's a great idea, but this won't do for real time applications where latency is unacceptable though.

[@ThinkingBetter]

Buffering audio is critical for latency in the case of video playback where you want to avoid lip-synch issues. Lip-synch means you can see the lips of people being in synch with what they say. Often you actually need to buffer audio to delay the sound to match the video as modern video processing includes video frame rate conversion and video noise filtering algorithms that work over multiple video frames creating delays. You might have heard of gamers complaining about using a TV for gaming as the video delay is too long for fast action. So what Paul is talking about in this video is actually done in pretty much every TV you can buy but for a different purpose of ensuring video and audio is in synch. However, since the video and audio coming into a TV is PUSHED by the source, the actual clock is mastered from the source and PLLed in your TV to minimize or avoid audible jitter.

[@dfelo93]

I truly wish I could work for Paul, I would learn so much.

[@Mark-lq3sb]

Absolutely amazing. Anybody and everybody who knows this, that and everything else has to post and let you know about this, that and everything else. To anybody and everybody: Please start your own KZbin channel.

[@WillyJunior]

No you.

[@Mark-lq3sb]

I don't qualify. Didn't you read what I posted? People, my post went up and over Boris's head.

[@WillyJunior]

@@Mark-lq3sb You really need to grow up.

[@Mark-lq3sb]

Hey Boris, ...and your post saying "No you", exactly how a child would respond. Remember Boris, no one forced you to reply to my post. Yet, you did.

[@WillyJunior]

@@Mark-lq3sb It's a meme, pal.

[@chrisvinicombe9947]

this letter was picked well to sell your cd player 😄😄😄