Audiophile Ethernet exists: the AVB/TSN thread

[mppix]

Bruno Putzeys explained why this is a not such a great idea. If you are thinking of simple LP filtering the "digital" PWM pulses to generate the analog waveform, any imperfections in the pulses will manifest at the output as distortions after filtering. This is much much harder for the high voltage high current outputs needed to drive difficult loads such as loudspeakers, as opposed to the normal DAC that only need to drive an op-amp.

The Truth About Digital (Class D) Amplifiers

Firstly I'd like to point out that "digital amps" is a misnomer. There are two categories: Analog-controlled class D. Switching amplifiers…

www.audioholics.com

Quote:

​

One should ask the question: would any D/A converter designer in his right mind build a DAC using power components? Probably not. Then how about the old argument that digital-to-the-end is best? Well, I should think the D/A barrier is best put precisely where it allows the whole signal chain to perform at its best and why should we believe that this is necessarily right at the end? Quite obviously the concept of a digital class D amplifier was dreamt up by DSP folks who presumed that the signal should be kept out of the big bad analog world as long as possible, at the same time expecting the power stage, power supply and filter (all highly analog in nature) to perform flawlessly.​

Not sure if i follow: PWM is by definition sampled (-> digital). Low pass filtering of pulses is precisely what every class D amp does.

They also may use feedback correction but that can also be implemented in the digital or analog domain.

 

[MaxwellsEq]

Not sure if i follow: PWM is by definition sampled (-> digital). Low pass filtering of pulses is precisely what every class D amp does.

They also may use feedback correction but that can also be implemented in the digital or analog domain.

I recommend researching Bruno Putzeys experience and knowledge to date. Then ask yourself "I wonder if I know as much as this person and do I have as much real world experience in this field as he does?"

 

[voodooless]

I recommend researching Bruno Putzeys experience and knowledge to date. Then ask yourself "I wonder if I know as much as this person and do I have as much real world experience in this field as he does?"

To be fair, he has a point about feedback in the digital domain. Axign showed that one can make a digitally driven class D amp, although the complexity is quite high, and obviously there is an AD step needed to get the feedback signal to the digital domain again. Bruno also never said it would be impossible, just that’s it’s easier to do it in other ways.

This part though is wrong:

Not sure if i follow: PWM is by definition sampled (-> digital).

It is not by definition digital. You can have a non discrete PWM signal. It’s made fully with analog electronics and is the basis of basically almost all good performing Class D amps.

 

[MaxwellsEq]

To be fair, he has a point about feedback in the digital domain. Axign showed that one can make a digitally driven class D amp, although the complexity is quite high, and obviously there is an AD step needed to get the feedback signal to the digital domain again. Bruno also never said it would be impossible, just that’s it’s easier to it in other ways.

This part though is wrong:

It is not by definition digital. You can have a non discrete PWM signal. It’s made fully with analog electronics and is the basis of basically almost all week performing Class D amps.

I agree. I'd had a similar idea in the early days of domestic digital. Sony had released a successful PWM domestic amplifier and I started doodling an idea to take PCM bitstream and convert it into a direct PWM control signal. But I quickly realised what at first seems obvious is far from simple and (at the time) incredibly expensive to build for little benefit.

 

[Prana Ferox]

Perhaps everybody is waiting for Wireless Time-Sensitive Networking (WTSN)?

Personally I'm waiting on the Wireless Amplified Time-Sensitive Optical Networking variant, in this context I'd consider it elementary

 

[pierre]

The mechanisms to prioritize trafic are well known, understood and applied. As in anything, there are bad implementations, and you may find the odd GbE interface that can't sustain serious trafic... and there are, of course those that can and do it reliably. These are not terribly expensive either. We are at the point where 40 Gbit/s interfaces are commonplace, assuming the worst of these, would work at 25% of the rated throughput... you are left with 10 Gbit/s... I also fail to see the need for full duplex audio at 1 Gb/s , I could be wrong.
Nowadays it is commonplace for an enterprise network to carry enterprise data. voice, video, security (video and ACS) and Paging data, with no issue. This is most of the time transparent to the users. The IT people see to that and most of us (IT people), utilize for that, vanilla Ethernet...
I fail to see the need for a special flavor of Ethernet. I stand to be educated...

Peace

If I remember correctly AVB is standard ethernet. They need PTP, multicast, SRP and a a few other protocols which should be implemented in switches (and are in good implementations, name your favorite Cisco or similar). Ref: https://en.wikipedia.org/wiki/Audio_Video_Bridging#

AVB is not very useful at home where your typical 40$ switch is very far from being saturated (except when you torrent too much or when you backup something across the network maybe). For broadcast, that's a different story: think 100 cameras pushing raw footages at the real time, you need some QoS and possibly a complex priority queues system. At work, I have seen switches drop packets but we can push them a lot, normally it should not happen if you only fwd packets.

 

[pablolie]

The problem is, what happens if you have a busy network combining many workloads? You need the network to prioritise audio traffic properly.

In a big system with many channels and distribution over Ethernet you need clock synchronization.

And, last but not least, all Ethernet interfaces are not created equal. I have been testing Gigabit Ethernet adapters and I found out that the typical Realtek USB-C to GbE can´t sustain 1 Gbps in full duplex, consuming an insane amount of CPU in the process.

But that is completely irrelevant when it comes to any HD audio stream. Even a 24/192 connection will never go beyond 10Mbps. That is 1% of a 1Gbps Ethernet link... And if 1Gbps isn't enough, many would love to sell you 800Gbps interface switches...

 

[MaxwellsEq]

For broadcast, that's a different story: think 100 cameras pushing raw footages at the real time, you need some QoS and possibly a complex priority queues system. At work, I have seen switches drop packets but we can push them a lot, normally it should not happen if you only fwd packets.

This is absolutely the proper use case. If you also have multicamera and multimic shoots you need everything synchronised. If you are recording a large orchestra with multiple microphones, you need to prevent drift to avoid phasing effects.

Meanwhile, for domestic use, even something as simple as the Mikrotik hAP ax lite should be adequate running without queues or QoS.

 

[mppix]

I agree. I'd had a similar idea in the early days of domestic digital. Sony had released a successful PWM domestic amplifier and I started doodling an idea to take PCM bitstream and convert it into a direct PWM control signal. But I quickly realised what at first seems obvious is far from simple and (at the time) incredibly expensive to build for little benefit.

The beaglebone black (and later variants) has 2 realtime microcontrollers that can drive class D fets with sufficiently accurate pwm. It has also ADC capabilities that be used for feedback and it can also do TSN.
Not saying it is easy, just possible.