aussievintage Posted December 15, 2019 Posted December 15, 2019 (edited) 13 minutes ago, rmpfyf said: Furthermore - if you believe Ethenet packets so wildly aperiodic, I challenge you to remove any oscillators from your routers and switches and see how you go. (Oh that's right - all digital processes are clocked out, and have inherent periodicity). That's not what it means to say the packets are aperiodic. Within the packets, data may be clocked at a set rate, but each packet gets sent independently, may even take a different path through the network, then the train of packets gets received aperiodically at the other end. 13 minutes ago, rmpfyf said: Unless you buffer and clock out independently at the very end, all jitter on shared interrupt devices matters. That's pretty much exactly what happens when the data is sent over USB to a DAC. Edited December 15, 2019 by aussievintage
Stereophilus Posted December 15, 2019 Posted December 15, 2019 16 minutes ago, aussievintage said: This would explain maybe why some people don't hear an improvement with better ethernet cables - because their downstream equipment is better (properly) designed and it copes with the problem reducing it to inaudibility. Perhaps so. 18 minutes ago, aussievintage said: Unless the noise causes data errors, it will not affect the digital stuff downstream, and proper design of power supplies and shielding should be protecting that same downstream equipment analogue sections from noise coming in via digital cable (and generated internally). My understanding is that noise entering via data ports (as opposed to power supplies) can propogate widely affecting clocking circuits and DAC chips / circuits directly. These circuits are very sensitive to noise. This would cause jitter, which is widely accepted as being audible. Circuits designs can vary widely in their ability to detect and minimise this noise.
Guest rmpfyf Posted December 15, 2019 Posted December 15, 2019 6 minutes ago, aussievintage said: That's not what it means to say the packets are aperiodic. Within the packets, data may be clocked at a set rate, but each packet gets sent independently, may even take a different path through the network, then the train of packets gets received aperiodically at the other end. You don't say. Though this is significantly different from true randomness. Particularly when your intended output depends on periodicity. 7 minutes ago, aussievintage said: That's pretty much exactly what happens when the data is sent over USB to a DAC. Nope. Go read USB audio class async mode. There's no requirement for independent clocking out of the receive buffer. None. In many devices the same device dealing with USB packets is responsible for clocking out. In many more it's a synthetic signal. Where it's not - and it's additional hardware to do so - then yes, there's true ibdeoendence, and the only jitter is in clock and local circuitry It's not enough to point at a buffer, state it's isolated.
aussievintage Posted December 15, 2019 Posted December 15, 2019 8 minutes ago, rmpfyf said: Where it's not - and it's additional hardware to do so - then yes, there's true ibdeoendence, and the only jitter is in clock and local circuitry That's what I saw when looking at a block diagram of a fairly typical USB DAC chipset. (didn't I post it somewhere in a thread a little while back?) I suspect that, now it's implemented by chipset manufacturers, that a lot of hardware these days performs better than your fears predict.
aussievintage Posted December 15, 2019 Posted December 15, 2019 19 minutes ago, Stereophilus said: My understanding is that noise entering via data ports (as opposed to power supplies) can propogate widely affecting clocking circuits and DAC chips / circuits directly. These circuits are very sensitive to noise. This would cause jitter, which is widely accepted as being audible. Circuits designs can vary widely in their ability to detect and minimise this noise. When you look at the very low noise figures of even relatively modestly priced DACs these days, you can see it isn't that hard to keep the noise out of even the sensitive analogue circuitry. I cannot see how it would be a problem for digital circuits much less prone to noise problems. Yes circuits vary, but I am pleasantly surprised at how good things have become, and I think it's partly due to modern multi-layer circuit boards with small SMD components, and well designed multi-function chipsets that come with design layouts that give the OEM a very solid well-performing starting point.
Stereophilus Posted December 15, 2019 Posted December 15, 2019 11 minutes ago, aussievintage said: When you look at the very low noise figures of even relatively modestly priced DACs these days, you can see it isn't that hard to keep the noise out of even the sensitive analogue circuitry. I cannot see how it would be a problem for digital circuits much less prone to noise problems. Noise measurements at the analogue output of a DAC are completely unrelated to how noise at an ethernet (or USB) input can cause jitter. 1
davewantsmoore Posted December 15, 2019 Posted December 15, 2019 40 minutes ago, rmpfyf said: In many devices the same device dealing with USB packets is responsible for clocking out. In many more it's a synthetic signal. ... whether this is a (big) issue or not, depends on the DA converter.
davewantsmoore Posted December 15, 2019 Posted December 15, 2019 8 minutes ago, Stereophilus said: Noise measurements at the analogue output of a DAC are completely unrelated to how noise at an ethernet (or USB) input can cause jitter. 0_o
davewantsmoore Posted December 15, 2019 Posted December 15, 2019 2 hours ago, Ittaku said: I beg to differ. Ethernet packet timing and sizes are so wildly aperiodic and asynchronous from the audio stream that's constructed from the data that it should have no effect no matter how many extra interrupts are generated, nor when they're generated, nor what timing and periodicity they're generated with, and resends are actually unbelievably rare as well. The good thing about this is it very simple to test in a controlled way..... By blasting network traffic at the victim computer/server/player..... anything from benign to super-extreme amounts can be used. This could be a listening test, or a "measure the analogue output of the DAC" test.
Ittaku Posted December 15, 2019 Posted December 15, 2019 1 hour ago, rmpfyf said: Depends what's at the other end. We differ on that point too. Only some of the first USB implementations in early DACs that used isochronous transfer could be affected. The vast majority are asynchronous.
BAM Posted December 15, 2019 Posted December 15, 2019 (edited) On 13/12/2019 at 8:09 PM, Colourless- said: Hi guys, I have recently upgraded to a digital streamer. I was using USB input for my old CAS setup and I can hear the difference between USB cables/USB purifiers. Since streaming uses TCP protocol with error correction and data is not affected from jitter, I would like to ask that does ethernet cables make any sound difference just like USB cable? I am curious. Are you wanting to hear from people who have tried different Ethernet cables with results either way or hearing technical theories as to why Ethernet cables should or shouldn't sound different or both.? Edited December 15, 2019 by BAM 2
aussievintage Posted December 15, 2019 Posted December 15, 2019 1 hour ago, Stereophilus said: Noise measurements at the analogue output of a DAC are completely unrelated to how noise at an ethernet (or USB) input can cause jitter. Yes but my point was simply that if they can keep the noise out the most sensitive parts of the circuit, then they can keep it out of those parts that cause jitter as well.
Neo Posted December 15, 2019 Posted December 15, 2019 22 hours ago, Muon N' said: People use to say similar about media playback programs, and then USB cables, that's It's all 1's and 0's so can't sound different. Me, I don't know and am happy with no definitive answer, so I'll wait for the truth to come out one way or the other as it usually does eventually. Well said From my experience there is a difference in Ethernet cables Neo 1
Guest rmpfyf Posted December 15, 2019 Posted December 15, 2019 (edited) 3 hours ago, aussievintage said: That's what I saw when looking at a block diagram of a fairly typical USB DAC chipset. (didn't I post it somewhere in a thread a little while back?) I suspect that, now it's implemented by chipset manufacturers, that a lot of hardware these days performs better than your fears predict. No fears just circuit diags and data. Read the async spec carefully. It's not a set rate. It simply states what generates the clock. Don't get me wrong here, it is possible to make a system completely independent of what's upstream. Building one presently and many here have such rigs. Not everyone does however. Presently here... Cables make a difference. Rather wish they didn't. Engineering accordingly. Edited December 15, 2019 by rmpfyf
Guest rmpfyf Posted December 15, 2019 Posted December 15, 2019 3 hours ago, davewantsmoore said: ... whether this is a (big) issue or not, depends on the DA converter. Yup. And what's just before it.
Guest rmpfyf Posted December 15, 2019 Posted December 15, 2019 2 hours ago, Ittaku said: We differ on that point too. Only some of the first USB implementations in early DACs that used isochronous transfer could be affected. The vast majority are asynchronous. Async does not mean true independence, though it can form the basis of a good implementation.
Guest rmpfyf Posted December 15, 2019 Posted December 15, 2019 (edited) 3 hours ago, davewantsmoore said: The good thing about this is it very simple to test in a controlled way..... By blasting network traffic at the victim computer/server/player..... anything from benign to super-extreme amounts can be used. This could be a listening test, or a "measure the analogue output of the DAC" test. Easier way - just pull the cable and listen. Edited December 15, 2019 by rmpfyf
Ittaku Posted December 15, 2019 Posted December 15, 2019 (edited) 2 minutes ago, rmpfyf said: Async does not mean true independence, though it can form the basis of a good implementation. And this is ultimately where we differ where all the previous disagreements come to a head. I say all jitter and timing short of a complete buffer underrun is totally irrelevant even in the face of the piddliest smallest async USB buffer. USB packets are every 1ms. The timing of that is so coarse it can't be used to influence anything downstream. Edited December 15, 2019 by Ittaku
aussievintage Posted December 15, 2019 Posted December 15, 2019 17 minutes ago, rmpfyf said: Not everyone does however. I think many more products are on the market, than you currently suspect, that have eliminated any problems in this area. Even quite modest products are benefiting from advancements that have migrated into the building blocks now routinely used by many designers.
davewantsmoore Posted December 15, 2019 Posted December 15, 2019 18 minutes ago, rmpfyf said: Easier way - just pull the cable and listen. Harder to test this with a good test (A/B switching of cables from the listening chair with a button) .... without the right network switch, and some gymnastics at the client end.
davewantsmoore Posted December 15, 2019 Posted December 15, 2019 3 hours ago, Ittaku said: The vast majority are asynchronous. ... but this doesn't mean "totally immune from any effect". 25 minutes ago, Ittaku said: And this is ultimately where we differ where all the previous disagreements come to a head. I say all jitter and timing short of a complete buffer underrun is totally irrelevant even in the face of the piddliest smallest async USB buffer. Measurements on the other side of the 'buffer' says otherwise.... but it's very hard to generalise the effect it will have. Perhaps none, but.
Ittaku Posted December 15, 2019 Posted December 15, 2019 4 minutes ago, davewantsmoore said: Measurements on the other side of the 'buffer' says otherwise.... but it's very hard to generalise the effect it will have. Perhaps none, but. What "measurements" are these?
davewantsmoore Posted December 15, 2019 Posted December 15, 2019 Just now, Ittaku said: What "measurements" are these? Noise. ie. jitter
Ittaku Posted December 15, 2019 Posted December 15, 2019 Just now, davewantsmoore said: Noise. ie. jitter Yeah, I'm wondering where you've seen said measurements I mean.
bob_m_54 Posted December 15, 2019 Posted December 15, 2019 5 hours ago, Stereophilus said: Agreed. I think we just went over this a few posts back... Ethernet cables can conduct noise to downstream components. That conducted noise can (potentially) affect sensitive downstream electronic thereby affecting sound quality. The discussion about data integrity is irrelevent in relation to the conduction of noise on ethernet. OK, the discussion seems to be divulging from whether or not a particular brand (price) of Ethernet cable will sound better than another, to whether or not noise on your network will cause audible degradation. For the original question I'll say that the brand of cable, as long as it conforms to the IEEE standard for that particular class of Ethernet cable, will not make an audible difference to your system. Simply put, conforming cables will be equal in data/audio performance. Now if you are talking about whether or not noise on your network cabling will audibly affect your system, it may, but that is a characteristic of the upstream and downstream hardware, and not something the audiophile priced cable, in comparison to the complying, non audiophile priced cable, can fix. The audiophile priced cable will transmit that noise just as faithfully as the non audiophile priced cable of the same category. If you are alluding to the noise being induced into the cable in question, from a noisy piece of hardware, then again, providing the cables are of the same category (and shielded/unshielded etc), they will perform the same and transmit or reject said noise the same. An audiophile priced Ethernet cable will not give you a wider sound stage, better instrument separation, blacker blacks, more openness and airiness and not even a tighter base. Transmitted data doesn't even understand these concepts. In the olden days, people used to play chess by mail. This worked well up to the point when rain entered your mailbox and left the letter unreadable, or someone nicked your mail, and the worst scenario when there was a mail strike, and nothing got through. 1
Recommended Posts