Ethernet bit rate

[radiolistener]

I think it's better to use the term "discrete signal" instead of "digital signal" if you want to emphasize that the signal represents discrete analog levels. Because term "digital" is used for binary discrete signals.

But the signal on ethernet pair is more complicated and cannot be treated as discrete, because it consists of analog combination between 5 levels DAC, analog gain of PGA which is used to tune signal power, reflected echo and the analog interference with a waveform of signal from remote endpoint. Also it has signal attenuation due to losses in the transmission line. And of course it has a sum of  environment noise and interference. So, this is very complex analog signal. And this is why I say that this is analog signal.

The same we cannot say that analog RF output of a cell phone is "discrete"or "digital". It is analog, despite the fact that a cell phone transfers digital data through that analog signal. It has analog RF frontend with analog mixer, analog amplifier, analog filters, analog switches, so it cannot be treated as discrete.

The same the signal on ethernet pair cannot be treated as discrete or digital because it is generated with DAC, amplified with PGA, and filtered with analog filters and passed through RF transformers inside Ethernet connector. This is all analog signal processing.

[wasedadoc]

It can be argued that until you get down to counting individual electrons and photons, all physical signals are analogue.  Those analogue signals may be conveying digitally encoded information.

[radiolistener]

No, there is no discrete level due to photon/electron quantization. The quantum of photon depends on the photon frequency, so you can divide that quantum by reduce frequency. And you can divide that quantum until zero value for DC frequency. 

I just wanted to say that the signal on ethernet pair is generated and processed with analog circuits and contains continuous frequency spectrum and infinite levels of amplitude, so it cannot be named as digital or discrete. This is classic analog signal. Yes, that analog signal carries digital data and depends on discrete transmitter state, but that doesn't means that it is digital or discrete.

If you pass the signal from ethernet pair through some discrete buffer with 5 levels to remove analog components it will break ethernet link. That break happens because the signal is analog and you destroy it by removing important analog components.

The cell phone also transfers digital data through electromagnetic waves, but this doesn't means that electromagnetic waves are digital or discrete.

[Siwastaja]

It's happens that the term "digital" is applied to a binary signal representation.

Of course it "happens", because binary signalling is, by definition, a subset of digital signal presentation, just like 3-level and 5-level are.

Did you ever seen DAC or ADC with a digital interface which is not binary and has multi-level amplitude on digital input/output?

No, but we are not discussing DAC or ADC interfaces, but Ethernet, which does have this multi-level signaling. Yet you are playing along like it does not exist anywhere.

What's with all this weird goal post shifting? Why it's so hard to admit you made a simple terminology mistake mixing up digital and binary (a very common mistake) and redkitedesign is right?

Regarding to the signal on ethernet pairs, it is 100% analog, because it may have infinite count of levels despite the fact that it is generated with DAC that has just 5 discrete levels.

But exactly the same is true for a binary signal, for example on SPI bus. "100% analog" means nothing and is useless distinction if you call every electronic signal "100% analog". In reality, the digital-ness comes from the fact that the signal can be interpreted in a way which produces a number, the same number used to generate the signal.

You can choose to call Ethernet analog signalling but others will disagree, because others consider analog something which does not have a concept of numbers at all, but instead deals with some continuous physical phenomena representing a value directly (analog clock (mechanical position), analog image (on a film), analog computer (opamps dealing with voltages)...)

[radiolistener]

But exactly the same is true for a binary signal, for example on SPI bus. "100% analog" means nothing and is useless distinction if you call every electronic signal "100% analog".

No, it is not the same. If you put digital (binary) buffer between SPI transmitter and SPI receiver it doesn't affect communication. It will still works. Because SPI communication uses digital (binary) signal.

But if you put the buffer which remove analog components from ethernet signal, it leads to a break of ethernet link, because ethernet PHY uses analog signal for communication with remote end-point.

Gigabit Ethenet PHY output is actually analog RF transceiver which has gain control, echo cancelation and analog signal substraction. It uses digital signal processing for analog signal, but this doesn't means that it uses digital signal for the link. Such signal cannot be digital.


Talking that ethernet pair has digital signal is the same as talking that mobile phone uses digital electromagnetic waves for communication. No, this is absurd. It uses classic analog radio communication for transfer digital data over electromagnetic waves and electromagnetic waves are not digital and not discrete, they are analog and continuous. It uses digitals signal processing and ADC/DAC for converting between digital and analog signal, but signal received on antenna and sent by antenna is analog, not digital.

The same with gigabit ethernet PHY, it uses digital processing for analog signal and ADC/DAC for converting digital domain signal into analog one and back, but signal on ethernet pairs is analog.

Just think, gigabit ethenet signal is a multi-level analog signal which has random and varying amplitude and added high power interference of known signal. If you put it on a digital logic element input which accept just 0 or 1 discrete value it just can see self transmitting signal and will be unable to listen for remote signal.

There is no "multi-level logic elements" inside gigabit PHY, it uses classic digitial circuit built with binary logic elements and ADC/DAC for converting between analog and digital domain. And pair between two PHY is analog domain....

You can choose to call Ethernet analog signalling but others will disagree, because others consider analog something which does not have a concept of numbers at all

There is no numbers on signal between two gigabit PHY, there is classic analog and continuous waves with continuous frequency spectrum on the pair between PHY.

And there is no way to process such analog signal with digital circuit without ADC/DAC converters to convert digital domain into analog and back. Because you cannot process analog signal with digital circuit, it needs proper conversion from analog domain to digital and from digital to analog.

Digital circuit consists of binary logic elements which can work with digital signal represented with binary level 1 and 0. And there is no way to process analog signal which consists of random and varying amplitude with digital circuit.

[radiolistener]

Here is waveform of 1000BASE-T signal, do you really think that this is a "digital signal" which can be processed with digital circuit by putting that signal on digital element input (which accepts just logic 1 or 0)? 

I even don't talk that amplitude of that signal can be changed on the fly by PHY during link negotiation.

[radiolistener]

Why it's so hard to admit you made a simple terminology mistake mixing up digital and binary (a very common mistake) and redkitedesign is right?

What I should admit? Digital signal can be processed just by putting it to a digital pin of FPGA. That's not the case for gigabit ethernet signal on a pair between two PHY. It needs at least ADC converter to convert such signal from analog domain to a digital domain.

I just don't see the way how such signal can be processed as digital without conversion from analog to digital. Do you see that way?

This is why I cannot accept your terminology, where you name analog signal (which requires significant dynamic range) as a digital.

[m k]

What about spectral lines.

[Siwastaja]

It can be argued that until you get down to counting individual electrons and photons, all physical signals are analogue.  Those analogue signals may be conveying digitally encoded information.

Exactly. And for convenience, the term "digital signal" means such signal that is successfully conveying digital data - numbers. If one does not agree with this, then the term "digital signal" does not exist at all (binary signalling would not be a special case, why would it - it still has noise etc.), which only leads to having to use more verbose language.

radiolistener's issue is the coexistence of two different logics that are contradictory without him noticing it and I am unable to explain this to him any better as he fails to see this simple and fundamental contradiction and instead is beating around bushes coming up more and more walls of text instead.

I accept disagreeing with terminology, even inventing their own no problem at all, but I would expect it to be internally coherent and logical, I just find it odd if one changes their mind randomly between bases b=2 and b=3. A scope trace of some I2C bus does not look any more "digital" than what was posted above, yet he would call that one digital.

[radiolistener]

And for convenience, the term "digital signal" means such signal that is successfully conveying digital data - numbers.

No. Digital signal is a signal that can be processed with digital circuit. Digital circuit is a circuit which consists of logic elements working with binary signal levels 0 and 1 and discrete in time according to the clock signal edges.

If you want to use digital signal processing for analog signal, you're needs to use mixed signal device, such as ADC (analog to digital converter) and DAC (digtal to analog converter).

[Siwastaja]

No. Digital signal is a signal that can be processed with digital circuit. Digital circuit is a circuit which consists of logic element working with binary signal levels 0 and 1.

You will have to explain this over and over again to everyone. Not best use of your time. Good luck.

[coppice]

DDR is either memory or a former communist country in Eastern Europe.

The term DDR is used a more broadly that just memory. It just means data exchanges on both edges of the clock, which is a widely used technique.

[radiolistener]

You will have to explain this over and over again to everyone.

In my opinion it's obvious, I don't understand why you didn't admit it.

Can you please show me example of digital circuit which can process 1000BAST-T signal without ADC converter for signal conversion from analog to digital?

I think it's pretty obvious that if you're needs ADC converter to convert signal from analog to digital, then it means that signal is analog, not digital. Because you cannot process it as digital. 

[wasedadoc]

Digital signal is a signal that can be processed with digital circuit. Digital circuit is a circuit which consists of logic elements working with binary signal levels 0 and 1 and discrete in time according to the clock signal edges.

For the final time. No.  "digital" is not synonymous with "binary".

Nor does there need to be a clock.  A 2-input gate is a digital circuit.  There is no clock input.

[radiolistener]

For the final time. No.  "digital" is not synonymous with "binary".

If you do not agree with what I said, please give an example of digital circuit which can accept signal which is not binary and can have many amplitude levels.

Note, that ADC and DAC is not a digital device, but a mixed signal device, because they are used on boundaries between digital and analog signal domain.

[coppice]

Nor does there need to be a clock.  A 2-input gate is a digital circuit.  There is no clock input.

In most cases there would be a clock, just not right at that gate. However, there are pure asynchronous logic designs that ripple through with no clock at all. The unpredictable nature of how long it takes for any one sample of the design to ripple has limited the use of pure asynchronous logic, but there have been a few examples of complex designs which claim to be completely clock free. I know of a few DSP designs of this type.

[radiolistener]

Nor does there need to be a clock.  A 2-input gate is a digital circuit.  There is no clock input.

yes, logic gate is not discrete in time, but it still has binary input and is used in clocked circuit.

If you don't use clock for your circuit with logic gates you will have metastability issues, in such case logic gate output may stay in undefined state (between 0 and 1), so your circuit cannot be named as clean digital. It's more mixed signal circuit, because it can get state which is not digital.

[m k]

Digital input includes an analog reference.

[wasedadoc]

For the final time. No.  "digital" is not synonymous with "binary".

If you do not agree with what I said, please give an example of digital circuit which can accept signal which is not binary and can have many amplitude levels.

Note, that ADC and DAC is not a digital device, but a mixed signal device, because they are used on boundaries between digital and analog signal domain.

1. By your reasoning even a simple inverter (eg 7404) is not a digital circuit.  It contains a 1 bit A-D converter.

2.  Although initially flash memory stored one bit per cell, many current high density ones store more than one bit per cell.  Their internal read sense amplifiers are not working with binary level inputs.

[radiolistener]

DDR is either memory or a former communist country in Eastern Europe.

DDR means Double Data Rate, it means that the data is transferred on both clock edges.
SDR means Single Data Rate, it means that the data is transferred on one clock edge (neg edge or pos edge).

There is also QDR which means Quad Data Rate, it means that the data is transferred on both clock edges and in the middle between clock edges.

DDR transfer is used in DDR memory, so this is why this term more familiar in terms of naming memory. But this does not mean that it is used only for memory

[wasedadoc]

Nor does there need to be a clock.  A 2-input gate is a digital circuit.  There is no clock input.

yes, logic gate is not discrete in time, but it still has binary input and is used in clocked circuit.

If you don't use clock for your circuit with logic gates you will have metastability issues, in such case logic gate output may stay in undefined state (between 0 and 1), so your circuit cannot be named as clean digital. It's more mixed signal circuit, because it can get state which is not digital.

That is more nonsense.

See my post at https://www.eevblog.com/forum/projects/building-a-philips-pm5530-sync-generator-equivalent/msg4656856/#msg4656856  About 20 digital ICs mostly TTL and about half of them are counters. No two of the ICs have a common clock.  No "metastability issues".

[radiolistener]

1. By your reasoning even a simple inverter (eg 7404) is not a digital circuit.  It contains a 1 bit A-D converter.

No, it doesn't have ADC. This is digital circuit because it is intended to work with binary signal which is represented with 0 or 1.

But if you use it not in digital circuit, it can produce undefined output which is not digital and in such case it will not be digital. But not because it is not digital, but because you're using it not in digital circuit

2.  Although initially flash memory stored one bit per cell, many current high density ones store more than one bit per cell.  Their internal read sense amplifiers are not working with binary level inputs.

I don't know about flash, but you can use digital signal processing of analog signal in digital circuit, but it requires signal conversion from analog to digital domain, this is what ADC is intended for.

About 20 digital ICs mostly TTL and about half of them are counters. No two of the ICs have a common clock.  No "metastability issues".

If they are used in such way that metastability cannot happens, then they always have binary signal state which is represented with 0 or 1. In such case they never enter into undefined state and can be declared as digital. 

But it's hard to implement asynchronous circuit on logic gates that doesn't suffer from meta-stability. If your circuit is async and don't use clock, there is a high risk that it has metastability issue, but it may happens rarely or you just don't notice it because it leaves undefined state too fast and such undefined state don't leads to noticeable change in circuit behavior.

[wasedadoc]

1. By your reasoning even a simple inverter (eg 7404) is not a digital circuit.  It contains a 1 bit A-D converter.

No, it doesn't have ADC. This is digital circuit because it is intended to work with binary signal which is represented with 0 or 1.

The input is an analogue signal.  Look at the data sheet. Any level between 0 and 0.8 Volts will be regarded as a Logical '0' for internal processing.  Any level between 2.4 and 5 Volts will be regarded as Logical '1' for internal processing.

[radiolistener]

The input is an analogue signal.

yes, you can put analog signal on digital circuit input. In such way digital circuit can get into undefined state and turns into not digital.

But the digital circuit means that all digital elements have digital signal on the input, not analog. In such way digital elements don't trap into undefined state and works as digital.

Also you can use digital elements to produce analog output, but it will not be a digital circuit.

[wasedadoc]

The input is an analogue signal.

yes, you can put analog signal on digital circuit input. In such way digital circuit can get into undefined state and turns into not digital.

But the digital circuit means that all digital elements have digital signal on the input, not analog. In such way digital elements don't trap into undefined state and works as digital.

Again, no.  Inputs between 0 and 0.8 or between 2.4 and 5 Volts do not result in any undefined state.  They are interpreted bythe chip as '0' and '1' inputs respectively.  How do you think it maps those ranges to the two logic values?