Kitz Forum

Broadband Related => Broadband Technology => Topic started by: Weaver on January 23, 2018, 09:04:19 PM

Title: Virtual noise?
Post by: Weaver on January 23, 2018, 09:04:19 PM
Can anyone tell me what this term means?
Title: Re: Virtual noise?
Post by: burakkucat on January 23, 2018, 09:32:23 PM
Could you please show an example of its usage, please?

I suspect that without sight of the context in which the phrase has been used nobody will be able to offer an opinion.  :-\
Title: Re: Virtual noise?
Post by: Weaver on January 23, 2018, 09:42:34 PM
There was a very recent post - maybe today - that showed the status output from a VDSL2 modem, a show-state type command in a CLI. It was a label shown on one of the entries.
Title: Re: Virtual noise?
Post by: burakkucat on January 23, 2018, 11:13:59 PM
Ah, that narrows the field.  :) 

As I don't recognise the phrase as occurring in the CLI output of a Broadcom chipset equipped CPE, it must have been shown in the output of a Draytek device. I have to confess that I do not understand the quirky way that Draytek devices output the information . . . thus I am unable to help.  :no:  Sorry.
Title: Re: Virtual noise?
Post by: Weaver on January 24, 2018, 07:13:05 AM
I have heard the phrase mentioned in tech articles about dsl technology. A google search came up with a reference to an ?IEEE paper that I could not access.

I tried again with a google search for "virtual noise dsl" and had more luck
Title: Re: Virtual noise?
Post by: burakkucat on January 24, 2018, 05:37:22 PM
Using the Google search string "virtual noise dsl" the very first hit is to a Broadband Forum Technical Report, TR-197 (https://www.broadband-forum.org/technical/download/TR-197.pdf),  titled "DQS: DSL Quality Management Techniques and Nomenclature" dating from August 2012.
Title: Re: Virtual noise?
Post by: Westie on January 24, 2018, 06:07:30 PM
...a reference to an ?IEEE paper that I could not access.

Is this (http://kang.nt.e-technik.tu-darmstadt.de/nt/fileadmin/kt/Publikationen_PDFs/2012/ICC/JoinOpt_Sarhan.pdf) the IEEE paper you are referring to?
Title: Re: Virtual noise?
Post by: Black Sheep on January 24, 2018, 06:22:49 PM
Westie has provided a great link, and TBH I've never heard of virtual noise before today ??

But a quick search shows .... 'Compare with fixed noise margin for all the tones, The Virtual noise (VN) will improve the
downstream and upstream date rate , especially in short to medium loops. For the typical length in BT FTTC application, the VN does make sense.'

I'm sure Mr Cat or Kitz will be pouring over the Westie link, to gain a better idea of how it works !!  ;) ;D
Title: Re: Virtual noise?
Post by: Weaver on January 24, 2018, 07:28:07 PM
I was very confused by the language of TR-197 p 26

many thanks to the link to that paper from Universität Darmstadt - very good
Universität. It starts well

“The intuitive approach of assigning a higher SNR margin during initialization will lead to a tremendous loss in the overall data rate, since modems will use the same margin also when initialized during high noise times.”

Then I am immediately completely lost. But then I have just had a massive increase in pain medication, and everything is a bit of a blur.

Could anyone comment on that later paragraph?
Title: Re: Virtual noise?
Post by: Westie on January 24, 2018, 10:50:39 PM
I was pretty lost before the end of the first page, and skimmed from page 2 onwards.  :(

What I think the author was saying was that if the modem initialised using a higher SNR margin on a noisy line it would sync at a lower rate than by applying the same SNR margin to a theoretical (virtual) noise allowance, and when the noise abated the modem would be stuck on an artificially low rate.

But does this ignore DLM?

I look forward to hearing what the greater minds of B*kat & kitz make of it... ;D
Title: Re: Virtual noise?
Post by: Black Sheep on January 24, 2018, 11:02:01 PM
Ha ha .................. P1 = understandable'ish ............................ P2 = Martian language with equations.

 
Title: Re: Virtual noise?
Post by: Weaver on January 24, 2018, 11:27:43 PM
I'm with Black Sheep. I have a (very bad) first degree in Theoretical Physics, so the article ought to be aimed at my ilk, or vaguely similar, but I can't make head or tail of the English in the paragraph after that quote that I copied.

[Moderator edited to correct the output produced by predictive text . . . s/quite/quote/ ]
Title: Re: Virtual noise?
Post by: ejs on January 26, 2018, 06:38:35 PM
Does TR-197 provide an adequate explanation of what Virtual Noise does?

From section 6.3.9.1:
"[Virtual Noise] could be seen to have a similar effect to a frequency dependent Target Noise Margin."

The paper Westie found is about a method for optimizing DSL by a clever combination of Virtual Noise and the target SNRM.
Title: Re: Virtual noise?
Post by: kitz on January 26, 2018, 08:58:32 PM
I'm sure Mr Cat or Kitz will be pouring over the Westie link, to gain a better idea of how it works !!  ;) ;D

I'll pass.   Because if it is what I think it is,  then I think I already know from whats been said so far in this thread.

1) Its an invention that came about at about the same time as system level vectoring.
2) Its something Openreach hopefully won't have any need to implement, as there are better alternatives..  ie vectoring, g.fast, FTTP.  Openreach's long term plan appears to be to drive fibre to nearer the home, rather than look at something that is best used with SRA.
3) Openreach doesnt use Low power mode, thus its not as important.

Trying to think of a really easy way to explain it:-

Problem

Some lines become unstable due to them syncing before their crosstalker. Whilst the line may sync up at the set Target SNR, once the crosstalker comes online, then that 6.3dB could become 3dB and causing the line to incur far more errors at 3dB than it would have under normal circumstances at 6.3db.

If you are one of those countries where low power mode is used then this is going to become even more problematic each time the crosstalker enters low power mode.

Background

When looking at QLN graph, I or any other skilled linestats reader can easily spot signs of crosstalk. 
If fact I could spot it way before we got QLN just from the SNR per tone as thats the way I learnt to spot it on adsl2.  Only difference is QLN inverts the crosstalk as bumps rather than troughs.  Its why I will sometimes also ask for bitload graphs rather than just QLN.  We look for distinctive peaks and troughs whose shape is specifically indicative of crosstalk.
Again getting really simplistic, think how a child would draw a seagull, it doesnt always look like that but for the sake of envisioning this example think of seagull shape.

Solution.


What if a system could analyise bit load, SNR per tone and QLN and look for seagulls, deducing that this is crosstalk.

System then over time recognises which tones for that individual line are impacted by crosstalk and knows what is normal crosstalk behaviour for that line.  It is then also capable of recognising when the crosstalker is offline or in LP mode.

So to stop the line syncing at 6.3 dB and then suddenly be jolted down to 3dB when crosstalk comes on line..  what it does is create an artificial noise mask at the tones (and only at those specific frequencies) that the cross-talker impacts the line at. 

Result

Masks on the tones which are usually affected by crosstalk by applying artifical noise [margin] at those tones normally affected by x-talk so that those tones wont bit load more than usual and restricting the minimum SNR at those tones.
Stops line from getting an artificially inflated sync speed that will only drop when the crosstalker comes back online.
Ideal for use with SRA and countries which use LP mode.




----
Note...  above typed completely off top of head without checking any tech journals, thus no tech jargon or references or googlings... just pure blurb and so could be wrong.   
Hands extremely sore took waaaay longer than anticipated to type above so hitting send without checking for errors and signing back off for the night.


Title: Re: Virtual noise?
Post by: Westie on January 26, 2018, 09:26:29 PM
Trying to think of a really easy way to explain it:-

Thanks, kitz. That really helped. So it seems to me that "virtual noise" is a theoretical construct used 6 years ago to help solve a problem which is not relevant anymore due to developments in other areas.

And I love the picture of a crosstalking seagull on the top of a (SNR) bin.  :)
Title: Re: Virtual noise?
Post by: andyfitter on January 27, 2018, 01:14:33 AM
I agree, that’s a great explanation. It was kind of where I was in my head having muddled through the docs, but the way you laid it out was perfect.

Much appreciated.

Andy.
Title: Re: Virtual noise?
Post by: Weaver on January 27, 2018, 04:08:29 AM
Got it. So needs SRA to make sense.
Title: Re: Virtual noise?
Post by: kitz on January 27, 2018, 11:22:29 AM
Doesnt need SRA, just used that as an example where it may be useful. 

It's more useful with Low Power Standby Mode.   Refer to this post I made a few years ago about L2 power saving (https://www.facebook.com/Kitz.co.uk/posts/470860716385499:0). 

Quote
A White paper by Texas Instruments states that there are serious concerns about low power modes and how it can and does introduce Time Varying Crosstalk and for this reason all the major ISPs around the globe agreed that L2 mode has serious effects on FEXT. The ITU study group concluded that it should NOT be a VDSL2 specification.

It's not like vectoring which cancels out effects of x-talk.
Virtual Noise cancels out the 3dB <-> 6dB swings when crosstalk disturbers go on and off line.   
It does this by artificially recreating the crosstalk environment even when the disturber is offline, by mimicking effects FeXT affected tones and ensuring those tones dont ever drop below 'x' dB compared to other tones.

Its not a cure for FEXT, in fact it does the opposite by ensuring the line constantly has the affects of FEXT regardless if the disturber is on or off line.
Title: Re: Virtual noise?
Post by: Black Sheep on January 27, 2018, 12:04:27 PM
Lovely piece of feedback, ...... many thanks for the layman's interpretation.  :)
Title: Re: Virtual noise?
Post by: Weaver on January 27, 2018, 06:31:05 PM
I'm with you about low power mode. That's a very good point, that goodness someone realised that. Road to hell and something about good intentions and all so on.

I was referring to the part in your helpful explanation where you talked about increasing speeds on the basis of a false impression gained when disturbers had temporarily gone away. So that would require Sra so I assumed, otherwise it would not be possible to increase speeds unless i have misunderstood. (Well there is the scenario of resynching getting an increased speed because of that only because disturbers have temporarily gone away and then your system keeps going round a long-term loop like that again and again.)

It does make me wonder about my own situation with the Dlinks where the initial reported downstream SNRM does not match the supposed target SNRM for downstream (but this does not apply to upstream).
Title: Re: Virtual noise?
Post by: Weaver on January 27, 2018, 06:33:55 PM
btw i didn't understand the term ‘noise mask’. This is some kind of frequency-domain function applied to the actual observed noise PSD?
Title: Re: Virtual noise?
Post by: ejs on January 27, 2018, 07:31:19 PM
I think what happens is that the noise level for a particular tone, used to determine how many bits to allocate to it, is the maximum of either the actual noise, or the virtual noise configured for that tone.
Title: Re: Virtual noise?
Post by: Weaver on January 28, 2018, 02:28:03 AM
Ejs, surely that has to be correct.
Title: Re: Virtual noise?
Post by: konrado5 on January 29, 2018, 11:01:38 PM
I've came across the "virtual noise" term a few years earlier and read about it in some technical documents. I've understood that virtual noise is fictional predefined at DSLAM noise causing the lower SNR per tones despite shortage of real noise. For example it can be predefined that virtual noise is 120 dBm/Hz per all tones. :)
Title: Re: Virtual noise?
Post by: konrado5 on January 29, 2018, 11:13:41 PM
I've looked at G992.3 specifiaction and found.
Quote from: G992.3
Configuration parameter TXREFVN defines the transmitter-referred virtual noise PSD to be used in
determining the SNR margin.
(p. 72)
I've also found.
Quote from: G992.3
The
received virtual noise PSD as defined in SNRM_MODE = 2 shall not be taken into account in
SNR(f).
(p. 108)

So the virtual noise artifically lower SNR margin but it doesn't lower SNR per tones.
http://www.itu.int/rec/T-REC-G.992.3/en
Title: Re: Virtual noise?
Post by: Weaver on April 08, 2023, 10:47:02 AM
This is the best bit of that paper, TR-197. Even I can actually understand the following:

“6.3.9 Virtual Noise (VN)

6.3.9.1 Objective – basic variant

The Virtual Noise function defines a Noise Mask (either transmitter or receiver referred), frequency shaped, which at the receiver is converted into a shape that represents the expected noise on the line. This could be seen to have a similar effect to a frequency dependent Target Noise Margin.
Normally the crosstalk noise changes with frequency and VN can attempt to model such change. If the Target Noise Margin is configured only on the peak value it could be too limiting in case of high values of crosstalk noise.
This way the xDSL line is configured to withstand a noise level equal to the configured Virtual Noise Mask, hence decreasing the bitrate with the expectation of increased stability.”

It then goes on to say more (I’ll need to read the whole thing) about how this is mentioned in the standards.

Is a “noise mask” then just a function whose output is either an SNR allowance or an actual current measured SNR, not sure which, and whose argument is the tone/bin number/bin?

I have seen virtual noise mentioned in G.994.1 where it very briefly mentions a flag in the modem capabilities and settings data set. This is declared as on/off as to whether or not a modem can employ virtual noise mathematics or is going to do so currently. I can’t remember which it is, but it’s telling the other end about whether VN is going to be/might be used by the this-end modem.



Also, I found this Broadcom patent relating to virtual noise (https://patents.justia.com/patent/7362828). I admit that I haven’t read it yet, but it looks interesting.