Scientific reasons why cables matter for audio

[fpitas]

why e.g. a network switch or cheap power cable isn't a vector for noise.

There's a lot of digital processing on data from a network that tends to either correct errors or lose lock completely, if the noise is that bad. In the case of Ethernet, the common mode noise tends to get stripped away by the transformer (galvanic isolation). In every competent power supply, the designer has already accounted for line noise, which can be horrendous coming right from the outlet.

 

[kemmler3D]

In every competent power supply, the designer has already accounted for line noise, which can be horrendous coming right from the outlet.

Right, makes sense. I guess I'm not even clear how power noise would make it into the signal in the first place, but I guess it makes sense that if you're designing against a "SNR" of like, 10dB at the wall outlet... and delivering 80dB at the output, then reducing EMI getting in through cables is irrelevant in the way that removing a mosquito from the windscreen will not make a car go faster?

 

[fpitas]

Right, makes sense. I guess I'm not even clear how power noise would make it into the signal in the first place, but I guess it makes sense that if you're designing against a "SNR" of like, 10dB at the wall outlet... and delivering 80dB at the output, then reducing EMI getting in through cables is irrelevant in the way that removing a mosquito from the windscreen will not make a car go faster?

Yeah, anything the cord picks up is usually high frequency and easy to filter. The power line, on the other hand has some fascinating stuff on it, low and high frequencies.

 

[ErVikingo]

To me what matters when it comes to cables:

Gauge; wire needs to be thick enough to minimize resistance, especially for longer runs to your speakers

Shielding; decent shielding prevents the cable from electromagnetic interference or hum

Build quality: durable/reliable connector

Rest IMHO is jewelry effect.

Real exchange with a friend who is restoring a pair of speakers like mine. I suggested before he opened anything up to do so basic tests, including measuring ohms on the terminals of the woofer towers. Results left channel 3.3ohms, right 5.4ohms.

Went into the right one to inspect and found that internal wiring was cut and spliced with different gauge wire and with questionable welds on some and butt connectors in others. All wiring replaced with a similar gauge to OEM, now right channel measures 3.2Ohms! (in spec).

 

[Waveform Fidelity]

There's a lot of digital processing on data from a network that tends to either correct errors or lose lock completely, if the noise is that bad. In the case of Ethernet, the common mode noise tends to get stripped away by the transformer (galvanic isolation). In every competent power supply, the designer has already accounted for line noise, which can be horrendous coming right from the outlet.

Even incompetently designed power supplies, if linear, hits the transformer which acts as low pass filter, and if switch mode is drowned out by switching noise and filtering. If thats not enough the analog electronics PSRR stops noise getting on to the audio output. I am amazed at how many posts there are these topics and despite that how much ignorance still exists.

 

[DonH56]

For my own edification, I don't have any EE background... I think many arguments in favour of snake oil like network switches or fancy cables of all stripes rest on preventing noise from entering the audio signal.

I understand that audible noise doesn't actually get into the system in those ways (would have heard it if it did, obviously) but I'm not as clear on exactly why e.g. a network switch or cheap power cable isn't a vector for noise.

What everybody else already said.

Ethernet is galvanically(*) isolated per spec, virtually all high-speed data signals are differential so resist coupled and radiated noise, data is recovered in a mixed-signal circuit long before it is buffered and used, and much of the data (e.g. Ethernet, USB) is in the form of packets (groups) of data that must be processed before the DAC can do anything with them. There is a lot of isolation from data to audio, and that doesn't even account for the orders of magnitude difference in frequencies (so there has to be not only a sneak path but also a path that mixes the RF signal to audio). Most of the quasi-scientific focus on network switches IME is on ground isolation with the idea that ground noise creeps into the audio system from the switch. The argument against that (besides the basic galvanic isolation) is that any competent component isolates analog and digital ground, and chassis ground from signal ground, to minimize any such coupling.

The power supply provides additional very large isolation from the input line to output whether conventional or SMPS. The incoming signal is rectified and filtered before being applied to audio circuits, which in turn have their own power supply rejection ratio (PSRR), so 120+ dB isolation is not uncommon. That said, power supply spurs are significant problem for a lot of components, but that happens due to poor design and layout inside the box, not something a magic power cord will fix. But even that is not an issue for SMPS since the switching rate is well above the audio band. Except for a few bad designs the naysayers can point to as representative of "all" designs.

It has always flummoxed me a bit that somehow the cables and isolation techniques used to obtain 140~160 dB dynamic range in the RF world, and capture uV to fV signals over the air, somehow fall apart and are not good enough for audio and that somehow ears extract more information than any measurement can ever provide. Too many folk have no idea how the recording process works and the level of noise and distortion embedded in movies and music no matter the distribution medium. And marketeers want to keep it that way... Of course, if you cannot measure or hear it, then you ears, system, or measuring equipment is not good enough to hear what they hear, so the little mouse keeps racing in his little wheel.

Whew - Don

(*) Galvanically isolated is a fancy way of saying voltage and current paths are separated between the two sides, two components, or whatever. A very, very old term for electricity and electrical paths.

 

[Speedskater]

How many miles of copper did a phone call half way around the world go thru back in the day? But 3' of copper cable is supposed to be a problem?

Jim Brown retired Audio Engineering Society, Noise & Interference commentate co-chair has a paper on long audio lines.
Transmission Lines at Audio Frequencies, and a Bit of History

http://www.audiosystemsgroup.com/TransLines-LowFreq.pdf

 

[fpitas]

marketeers want to keep it that way

Yup!

 

[Waveform Fidelity]

What everybody else already said.

Ethernet is galvanically isolated per spec, virtually all high-speed data signals are differential so resist coupled and radiated noise, data is recovered in a mixed-signal circuit long before it is buffered and used, and much of the data (e.g. Ethernet, USB) is in the form of packets (groups) of data that must be processed before the DAC can do anything with them. There is a lot of isolation from data to audio, and that doesn't even account for the orders of magnitude difference in frequencies (so there has to be not only a sneak path but also a path that mixes the RF signal to audio). Most of the quasi-scientific focus on network switches IME is on ground isolation with the idea that ground noise creeps into the audio system from the switch. The argument against that (besides the basic galvanic isolation) is that any competent component isolates analog and digital ground, and chassis ground from signal ground, to minimize any such coupling.

The power supply provides additional very large isolation from the input line to output whether conventional or SMPS. The incoming signal is rectified and filtered before being applied to audio circuits, which in turn have their own power supply rejection ratio (PSRR), so 120+ dB isolation is not uncommon. That said, power supply spurs are significant problem for a lot of components, but that happens due to poor design and layout inside the box, not something a magic power cord will fix. But even that is not an issue for SMPS since the switching rate is well above the audio band. Except for a few bad designs the naysayers can point to as representative of "all" designs.

It has always flummoxed a bit that somehow the cables and isolation techniques used to obtain 140~160 dB dynamic range in the RF world, and capture uV to fV signals over the air, somehow fall apart and are not good enough for audio and that somehow ears extract more information than any measurement can ever provide. Too many folk have no idea how the recording process works and the level of noise and distortion embedded in movies and music no matter the distribution medium. And marketeers want to keep it that way... Of course, if you cannot measure or hear it, then you ears, system, or measuring equipment is not good enough to hear what they hear, so the little mouse keeps racing in his little wheel.

You must have been typing at the same time as me! I like your more details answer, I was trying to keep it simple

 

[DonH56]

You must have been typing at the same time as me! I like your more details answer, I was trying to keep it simple

Yah, sorry, I tend to get wordy. Simpler is usually better, but sometimes incomplete, and then I spend more time expanding my answer. Or get ignored by those who decide I don't know what I am talking about. There are lots of very sharp folk on ASR.

 

[fpitas]

Yah, sorry, I tend to get wordy. Simpler is usually better, but sometimes incomplete, and then I spend more time expanding my answer. Or get ignored by those who decide I don't know what I am talking about.

I liked your answer. I think it's a plus that different people explain things. Someone reading this is more likely to understand with different viewpoints presented.

 

[Salt]

To me what matters when it comes to cables:

Gauge; wire needs to be thick enough to minimize resistance, especially for longer runs to your speakers

Shielding; decent shielding prevents the cable from electromagnetic interference or hum

Build quality: durable/reliable connector

Rest IMHO is jewelry effect.

Real exchange with a friend who is restoring a pair of speakers like mine. I suggested before he opened anything up to do so basic tests, including measuring ohms on the terminals of the woofer towers. Results left channel 3.3ohms, right 5.4ohms.

Went into the right one to inspect and found that internal wiring was cut and spliced with different gauge wire and with questionable welds on some and butt connectors in others. All wiring replaced with a similar gauge to OEM, now right channel measures 3.2Ohms! (in spec).View attachment 558758

At first glance I thought it was fine arising from that wire but the closer look confirmed it's damping material

 

[Cbdb2]

They had repeater amps.

Yes , but any distortion, "time smear", noise etc from the cable would be passed on and made worse by all the amps. Its amazing they made this work 70 years ago. TAT-1 was 4000km long . Thats 4,000,000 1m interconects.

 

[rcstevensonaz]

They had repeater amps.

And also substantial investment and engineering in echo cancelation.

 

[kemmler3D]

Very helpful, thanks @fpitas @DonH56 @Waveform Fidelity !

I definitely get why digital signals don't benefit from fancy equipment, but was less clear on specifically why electrical frippery is irrelevant.

To make sure I understand, in order to cause audible problems via power-related noise, you'd have to have shared grounds between different circuits in the device (DAC, streamer, amp?) or just poorly implemented filtering between the power supply and amp circuits?

To flip it around, if I wanted to create audible noise using a power cord, what would I have to do? Put it next to a large, unshielded motor, or something?

 

[fpitas]

Very helpful, thanks @fpitas @DonH56 @Waveform Fidelity !

I definitely get why digital signals don't benefit from fancy equipment, but was less clear on specifically why electrical frippery is irrelevant.

To make sure I understand, in order to cause audible problems via power-related noise, you'd have to have shared grounds between different circuits in the device (DAC, streamer, amp?) or just poorly implemented filtering between the power supply and amp circuits?

To flip it around, if I wanted to create audible noise using a power cord, what would I have to do? Put it next to a large, unshielded motor, or something?

It would be hard for the power cord to cause problems, other than the shared ground wire. And that's not a problem in the power supply. Hum is introduced into the amplifier input by ground rise. That's where the "ground" in one location has a voltage with respect to ground in another location.

 

[pablolie]

What everybody else already said.

Ethernet is galvanically(*) isolated per spec, virtually all high-speed data signals are differential so resist coupled and radiated noise, data is recovered in a mixed-signal circuit long before it is buffered and used, and much of the data (e.g. Ethernet, USB) is in the form of packets (groups) of data that must be processed before the DAC can do anything with them. There is a lot of isolation from data to audio ...

Indeed. The physical and the data layer are completely separate. Basically, Ethernet ALWAYS makes sure the zeros and ones in the packet payload arrive with the required integrity, no one needs perfectly shaped little squares. It's the transitions that trigger things, and a damn lot of noise is easily tolerated. Don't people get the fact that far more vital applications than our audio work with total integrity over the same infrastructure, and that the esoterica about oh-how-special-and particulary-demanding-audio-is are simply total ignorance? There is no such thing as "dirty" zeros and ones in Ethernet - it simply doesn't matter at all. Banking. Hospitals. Military. Don't you think they'd be worried if Ethernet was such a vulnerable transport medium? It's laughable. :-D

The situations where packet loss would happen in a home network are super trivial... either someone trips over a power or ethernet cable, or for some odd reason there's a sustained traffic peak that exceeds exisiting capacity. And that traffic peak would never be caused by audio in a Gigabit world, do the math. :-D

And every application can pick from a number of exisiting higher layer protocols to get the SLA (latency, jitter, packet loss recovery) it needs - if it doesn't, blame the developer's incompetence, not the network.

In your typical home network, there's absolutely nothing needed in a switch or router to make packet delivery "better" in any way, and it's downright stupid to think anything in those boxes can make audio sound better or worse. An idiotically overpriced "audio switch/router" is guaranteed to rely on the exact same internals as any halfway decent $19.99 Ethernet switch.

 

[DonH56]

Very helpful, thanks @fpitas @DonH56 @Waveform Fidelity !

I definitely get why digital signals don't benefit from fancy equipment, but was less clear on specifically why electrical frippery is irrelevant.

Had to look up "frippery".

To make sure I understand, in order to cause audible problems via power-related noise, you'd have to have shared grounds between different circuits in the device (DAC, streamer, amp?) or just poorly implemented filtering between the power supply and amp circuits?

Or decoupling to an inappropriate place, like a filter (RFI suppression) capacitor connected from the input side to the output side of the power supply or data interface. Basically poor design or layout practice. For a "linear" supply, the transformer isolates the input from the the output windings, unless a poor design connects them inappropriately (that is simplified; there are all sorts of caveats and details beyond the scope of discussion here).

To flip it around, if I wanted to create audible noise using a power cord, what would I have to do? Put it next to a large, unshielded motor, or something?

Again, power filtering happens inside the component, so if putting the power cord next to a large noise source causes audible problems then there is a problem with the component's power input design. I imagine the most common sources of powerline noise in a home are motors and generators from the HVAC and refrigeration systems, light dimmers, and various other power supplies connected to the power line (including LED lighting). Any decent power supply and grounding scheme should reject, suppress, and/or isolate that noise to inaudibility at the audio output. There are also safety issues that drive isolation; for instance, computer cases are typically isolated from the neutral line connection in the USA, using the safety ground to prevent accidental shocks from equipment failure.

I do not recall the powerline standards (I have a copy on another computer) but I think 5 or 10 percent distortion is the max allowed (hopefully a power engineer can verify or correct me). That is fairly easy to reject to complete inaudibility by the time the incoming voltage is converted to DC and filtered. Feedback in the component will also help suppress noise from the power supplies.

Note that is not true of audio signal cables since the signal level is much lower and not directly isolated from the audio signal (since a signal cable is carrying the audio signal). Then cable routing and shielding becomes more important.

There are always examples of components with inadequate isolation or RFI suppression but those are in the minority IME/IMO. As are claims of massive noise on the wall outlet or RFI from external sources. Those are both present, but again filtered by reasonable components, or Amir's measurements would indicate problems. Most of us do not live next door to a megawatt AM transmitter or have massive nose sources inside our homes. Like many things audio, the reality is much less than the marketing hype.

IME/IMO - Don

 

[Ruediger]

In your typical home network, there's absolutely nothing needed in a switch or router to make packet delivery "better" in any way, and it's downright stupid to think anything in those boxes can make audio sound better or worse. An idiotically overpriced "audio switch/router" is guaranteed to rely on the exact same internals as any halfway decent $19.99 Ethernet switch.

What they do in these audiofool switches is basically: copper in -> convert to optical -> convert to copper -> copper out

So it's not nothing, the coming and going copper cables are isolated from each other. Paired with a good build quality, a nice internal power supply.

The thing is that it is an answer to a question no one asked. Problems from the RJ45 copper cable are extremely rare, so no need to isolate. And when looking at the tech used the price of such devices is at a high multiplier from it's actual value.

Anyone can do this at home - copper/fiber/copper - with some basic switches and SFP modules. But there is no need to.

The costs of higher rank switches are from processing power and software. A switch with no layer management at all is the lowest rank. I don't know if these snakeoil switches even have any features that would be of use as an actual good switch. I suspect not as the production numbers vs. a software would be a challenge. Or they would have to use something off the shelf and re-brand it.

Bottom line: complete nonsense products.

 

[arqwfxdv]

I know their tactics is to drive people mad. But this is epic:

@Bergen-e4s
To the people summoning Amir, Galen is more ASR than ASR.

To the people bringing up the cable blind test paper, Galen's approach is more scientific than using a 14-bit player.

@arqwfxdv
To the people summoning Galen, Amir is more iconoclastic than Iconoclast.

To the people bringing up the White Paper thingies, Amir's approach is more scientific than using tons of irrelevant calculations, pictures, and poor technobabble.

And what does a 14-bit player have to do with an ABX listening test?

@arqwfxdv
Oh, and you still haven't answered my question as to why me using the McGurk effect as an illustration of psychoacoustics is “false equivalence”. Three days in.

@Rower30-ye6qn
@arqwfxdv Using physics is irrelavant? I hope you don't drive over any bridges that look good in pictures but have no design calculations. - Galen

@arqwfxdv
@Rower30-ye6qn Are you capable of saying something that makes sense? Or are you just running around moving goalposts?

Where are your measurements?

LINK