Re: subscription :Megaco query
2004-04-01 09:13:05 GMT
hi Kevin ,
What will be the state of the terminations which are present at
the time of MGW get fail ?
What mechanism provided by the megaco to maintain the same state at
the both end MGC and MGW ?
regards,
prateek
Kevin Boyle wrote:
-- PRATEEK RAJ SINGH PAWAR Research Engineer Mobile Solutions C-DOT Ph(O):24677525 ext. 324 Room No 824,9th Floor Cell : +919810652039 Akber Bhavan ,Chankya mail :pprateekraj <at> yahoo Puri , New Delhi-21 pratraj <at> rediff ----------------------------------------------------------------------------Procedures for MGC control of MGs are described in the spec. Essentially, the MGC would have to wait for the MG to return to service.
Kevin
-----Original Message-----
From: Prateek Raj Singh Pawar [mailto:prateek <at> cdotd.ernet.in]
Sent: Wednesday, March 24, 2004 11:38 PM
To: Megaco <at> ietf.org
Cc: Prateek Raj Singh
Subject: [Megaco] subscription :Megaco queryHi ,
This is parteek , working in C-DOT Indian goverment telecom research center . we are currently working on the 3G MSC Server .
Pls add me in the mailing list of megaco .
I h've some queries related with the megaco .
What are the control procedure if MGC control only one MGW configuration and MGW will fail .
with regards,
PRATEEK RAJ SINGH PAWAR
Research Engineer
3G Mobile Solutions
C-DOT Ph(O):24677525 ext. 324
Room No 824,9th Floor Cell : +919810652039 Akber Bhavan ,Chankya mail :pprateekraj <at> yahoo
Puri , New Delhi-21 pratraj <at> rediff
----------------------------------------------------------------------------
_______________________________________________
Megaco mailing list
Megaco <at> ietf.org
https://www1.ietf.org/mailman/listinfo/megaco
> Do you agree with "near end", "half way" definitions?
> Is my view inline with yours?
>
> Perhaps one issue is the use of term "near end" in NGN environments,
> whereas G.168 precisely talks about "cancelled end" since the 2002 version.
>
> Any G.168 experts view? Please correct me if I'm wrong.
>
> Best regards,
> Albrecht
>
>
>
>
> (See attached file: NGN_EchoCanceller_Principle_r02.PPT)(See attached file:
> NGN_EchoCanceller_Principle_r02.pdf)
>
>
>
>
> "Soncodi, Adrian"
>
> <Adrian.Soncodi <at> t To: Albrecht
> SCHWARZ/DE/ALCATEL <at> ALCATEL
>
> ekelec.com> cc: <megaco <at> ietf.org>
>
> Subject: RE: [Megaco]
> Question on echo control direction
>
> 19.03.2004 18:46
>
>
>
>
>
>
>
> Albrecht,
>
> Let's say both switches perform near-end, half way EC as you suggest. Then
> the question is: what do you mean by "near-end half way EC"? Is it:
> a) Cancelling echo towards the near-end user, so that he does not hear
> echo? or
> b) Cancelling echo generated by the near end, so that the far end user does
> not hear echo?
>
> In other words, which one happens when the MGC instructs the MG to turn EC
> on? I'm not saying it should be one or another, I'm saying that the H.248
> standard should specify it, because it matters.
>
> Kevin says it should be a). I just point out that most of the MGs today do
> b). And I found one MG that does a). So if MGCs follow the ISUP procedures
> to turn EC on at both ends using b), they may have the surprise that some
> EC devices are "flipped over", and thus they become ineffective as they
> work over the long path. Or, instead of being complementary, both EC
> devices cancel in the same direction ... you see the problem.
>
> I don't quite understand the rest of you e-mail. What do you mean by "Any
> EC requirement in addition ... must be signalled correspondingly." Signaled
> where? in H.248 you only have one ec property.
>
> Are you suggesting that the MGC just signals "ec=on", and then the exact
> behavior of the MG (i.e. the direction of the half-way EC device etc.)
> depends on the local MG provisioning? This could be a solution, but even
> so, the H.248 standard should make it clear. And the MGC would still have
> to know what is the case in order to properly implement two-way echo
> control procedures.
>
> To conclude, I'm still trying to find out what the MG's expected behavior
> is. See my long e-mail for pros and cons of a) and b). But I believe if we
> don't clarify it, there will be interop problems.
>
> Adrian
>
>
>
>
> -----Original Message-----
> From: Albrecht.Schwarz <at> alcatel.de [mailto:Albrecht.Schwarz <at> alcatel.de]
> Sent: Friday, March 19, 2004 3:06 AM
> To: Soncodi, Adrian
> Cc: Kevin Boyle; megaco <at> ietf.org; megaco-admin <at> ietf.org
> Subject: RE: [Megaco] Question on echo control direction
>
>
>
> Adrian,
> I couldn't find any contradiction between Kevin's and your statements.
>
> My understanding is that the "basic" EC settings for circuit-to-packet MGs,
> where the POTS/ISDN terminal is accessed via a circuit-switched network,
> are:
> - near-end EC, and
> - half way only ("I don't know the exact G.168 terminology").
>
> I think that this common understanding may be derived from legacy PSTN/ISDN
> networks.
> Any EC requirement in addition, e.g., like
> - far-end EC,
> - 2 near-end "H.248 TDM Terminations" (in a TDM-to-TDM call), and/or
> - full way
> must be signalled correspondingly.
>
> My impression is that the "basic settings" are sufficient for almost all
> cases. The additional EC capabilities are valid network scenarios (e.g.,
> inter-operater interworking, replacement of an international exchange,
> tandem echo cancellation, two satellite links in the TDM-to-TDM call
> scenario, missing near-end EC capability in peer MG, etc), but rather
> exeptional cases (like in pre-NGN TDM networks).
>
> Regards,
> Albrecht
>
>
>
>
>
>
> "Soncodi, Adrian"
>
> <Adrian.Soncodi <at> t To: "Kevin Boyle"
> <kboyle <at> nortelnetworks.com>, <megaco <at> ietf.org>
>
> ekelec.com> cc:
>
> Sent by: Subject: RE: [Megaco]
> Question on echo control direction
>
> megaco-admin <at> ietf
>
> .org
>
>
>
> 18.03.2004 18:13
>
>
>
>
>
>
>
> Kevin,
>
> In your message you say:
>
>
> "I cannot find anything that specifies this, but I don't know why someone
> would use an echo canceller to cancel echo inside the context... Why not
> just fix the DSP to not introduce the echo in the first place? You don't
> generally find echo on the transition from TDM<->packet."
>
>
> Actually, it's not the DSP that introduces echo. It's the 4-wire to 2-wire
> conversion beyond the TDM termination. So for a TDM-packet-TDM call, we
> have the following:
>
> - Typically, echo is cancelled at both ends of the call. Originating and
> terminating switch each cancel in different directions.
>
> - For an echo device to cancel in one direction, it has to take the other
> path as reference, wait for the signal to turn around from the user, and
> then compare the two and clean the echo. This involves a delay buffer (echo
> tail), typically up to 64 ms, sometimes up to 128 ms (I don't know of MGs
> that can do more).
>
> - As a consequence, each switch typically cancels echo over the short path.
> For a TDM-packet-TDM call, the short path is usually the TDM, while the
> long path is the packet network plus the other TDM. If you wait for the
> signal to turn around on the packet side, many times the delay is bigger
> than the buffer size and the EC device becomes ineffective.
>
> - Consider two MG-s interconnected by a packet network. If these were
> traditional TDM switches, each one would clean the stream going towards the
> other (for example, the echo indications in ISUP IAM/ACM are designed to
> achieve this result). This corresponds to the direction into the packet
> network.
>
> - Another reason to clean the stream into the packet network is that if the
> stream has echo, then the encoding/decoding performed by some compression
> algorithms results in lower quality voice (i.e. some algorithms are
> optimized for "clean" voice).
>
> - A solution for this would be to turn echo control on the packet-side
> terminations. Then if these terminations work like you say they should, the
> desired result would be achieved. Unfortunately, all the MGs I know cannot
> do this. They can only apply a half EC device on the TDM termination
> (probably because the package name is TDM)
>
> - Another aspect: For a conference call with N TDM terminations in a
> context, in order cancel the echo generated by one TDM circuit it is
> simpler to clean its stream going into the context than the other N-1 going
> out of the context.
>
> - Until now, I thought I knew the answer to this one. Although this does
> not seem to be clearly specified either in MGCP or in MEGACO, for all the
> MGs I've seen, if you turn on echo on a TDM termination they clear the
> stream going into the packet network. They reference the backwards stream
> to TDM, which is turned around over the short path. So it looks like most
> of the existing MGs do not work like you recommended -- people, correct me
> if I'm wrong here.
>
> - However, recently I stumbled upon an MG that behaves exactly the
> opposite. So now I really have a problem, because there is no way for an
> MGC to know whether the ec property results in the insertion of a full EC
> device or a half device, and which direction.
>
> - By analogy with signals (although ec is a propery not a signal), this
> should work as you said, i.e. towards the outside of the context. But if
> you would like to standardize this behavior, already a lot of MGs would be
> non-compliant (again, people?).
>
> - There is also no way for an MGC to audit for this behavior. But if MGs
> were allowed to behave both ways, then you could have different MGs at the
> two ends and then of a TDM-packet-TDM call, and it becomes impossible to
> provide two-way echo cancellation.
>
> So how do you propose we clarify all this?
>
> Thanks,
>
> Adrian
>
>
> -----Original Message-----
> From: Kevin Boyle [mailto:kboyle <at> nortelnetworks.com]
> Sent: Wednesday, March 17, 2004 6:12 PM
> To: Soncodi, Adrian; megaco <at> ietf.org
> Subject: RE: [Megaco] Question on echo control direction
>
>
>
> Ecan generally applies to the network transition point to avoid
> reflections off the transition. Therefore, it would make sense that
> the echo canceller on the TDM termination would cancel echo outwards
> toward the TDM facility.
>
>
> I cannot find anything that specifies this, but I don't know why
> someone would use an echo canceller to cancel echo inside the
> context... Why not just fix the DSP to not introduce the echo in the
> first place? You don't generally find echo on the transition from
> TDM<->packet.
>
>
> Kevin
>
>
> -----Original Message-----
> From: Soncodi, Adrian [mailto:Adrian.Soncodi <at> tekelec.com]
> Sent: Wednesday, March 17, 2004 6:51 PM
> To: megaco <at> ietf.org
> Subject: [Megaco] Question on echo control direction
>
>
> Hi,
>
>
> When an MCG turns on echo control for a TDM termination on the MG,
> which one of the following audio streams becomes "clean" of echo:
>
>
> a) The stream from the termination outwards to the TDM facility, or
> b) The stream from the termination inwards into the context
> (presumably going out on another facility associated with another
> termination in the context) (I am assuming that only a half-echo
> canceller is inserted, because most of the MGs can only do that. The
> other half would be on the other termination.)
>
>
> Where is this specified? the TDM package in H.248 annex E does not
> say.
>
>
> Thanks,
>
>
> Adrian
>
>
>
>
>
>
> _______________________________________________
> Megaco mailing list
> Megaco <at> ietf.org
> 
RSS Feed