Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weaver]

I realise that this is going back to when dinosaurs ruled the earth and will be a test of kitizens’ memory. Was the Broadcom PhyR error correction protocol available in Broadcom modems and DSLAMs if you were using G.992.1 ("ADSL1") rather than ADSL2 / ADSL2+ ?

I didn’t have the right modems when I was using G.992.1 + BT 20CN, and I’ve forgotten what the DSLAM was. My DSLAM would be mentioned in an old thread somewhere, not entirely sure what to search for. I had Netgear DG834v1/v2/v3 modem-routers back then and they were TI-based, not Broadcom. Is it at all possible that the pre-2015 (20CN) NSBFD DSLAM was also TI, and so ‘compatible’ if that’s even significant or meaningful in that case? I can’t remember what modems I used after that, but I had a period of using DLink DSL320B-Z1 modems from AA, which had a Mediatek chipset iirc, so no Broadcom yet either. Then at some point, can’t remember when, but a search of kitz.co.uk would tell me, I started using ZyXEL modems thanks to Burakkucat and hooray for Broadcom and the huge downstream reliability improvement that comes from PhyR, plus downstream 3dB SNRM and quite a bit more downstream speed.

[tubaman]

Is upstream PhyR enabled on your modems or is it like VDSL in that the exchange end decides which capabilities it will allow so it makes no difference if enabled or not?

[Weaver]

I can disable it at my end in the modem’s web config ui. The exchange has to be Broadcom and then I presume, but have no way of testing, that the exchange DSLAM always uses PhyR if you’ve got it and have enabled it. I’ve just had a look at the modem and I see that you can separately enable/disable PhyR in either direction, upstream or downstream. So that must mean that there is such a thing as PhyR upstream. I didn’t know that. I have both upstream and downstream ticked as well as a whole load of other wishes to Santa that are never going to happen, such as SRA, and PTM over ADSL2. I can long for PTM, I know that ADSL2 supports it, wish the DSLAM would, they’ve already written all the PTM code, for VDSL2. Following Burakkucat’s tip, I have selected a restriction to ADSL2 only; ie no ADSL2+.

[Actually, SRA would be a bit dodgy with bonded lines I think, because I have three IP-bonded links and the Firebrick needs to know what the upstream speed of each pipe is, in order to get the correct fraction of the upstream traffic sent to each of the links. Currently it gets told this info at PPP-connect time. If you had SRA, as in G.Fast, there would have to be some other protocol for dynamically updating the Firebrick’s egress speeds per link; maybe some PPP trickery if you’re using PPP, and if not, then who knows.]

No upstream PhyR unfortunately; perhaps that’s to do with Broadcom? Perhaps it’s to do with ADSL2 ? It would be incredibly valuable to me. I think perhaps Broadcom thought about downstream-only, because there was no market for upstream improvement this maybe being because they were thinking about the use case of IP TV which is a downstream-only thing. Broadcom mentions IP TV in the rationale section of a document of theirs about PhyR. They said they would need 9dB downstream SNRM routinely enforced to make IP TV work without L2 ReTX, as the link has to be completely reliable; in any case how would they and ISPs enforce such a margin? And what about SRA? Maybe making it work in two directions would have been a software nightmare? Does the G.INP protocol allow bidirectional ReTX? Thinking about whether ADSL2 might frustrate the possibility of bidirectional implementation, I think PhyR must be independent of ADSL2 because Broadcom surely just implement it over the top, as a higher layer (seeing as it was, I presume, developed independently), without needing to be constrained by anything particular in ADSL2? I certainly could do so, and have done the like, in the distant past.

I wonder why isn’t PhyR happening then on the upstream? Can anyone find out for me?

[burakkucat]

Does the G.INP protocol allow bidirectional ReTX?

Yes it does. G.Inp has been observed in an active state on the US direction for those VDSL2 circuits which are in "poor condition". It's a rare occurrence.

[tubaman]

Yes it does. G.Inp has been observed in an active state on the US direction for those VDSL2 circuits which are in "poor condition". It's a rare occurrence.

I've seen US G.Inp a couple of times on my VDSL line after bouts of high errors but it has never lasted more than a few days before being removed. It certainly appears that for VDSL the preferred state is 'off'.

[Weaver]

Why does G.INP get turned off at all? I haven’t read up enough so I don’t know what the byte bloat overheads are but I thought it wasn’t much ? I read something - maybe something that Kitz said - about 9% overhead with G.INP under certain special conditions- "G.INP High" it says in my notes, yet I have no idea what that means - some option / setting maybe? But I though that G.INP didn’t have a substantial byte bloat overhead normally, so why the need to turn it off at all? Sorry for my ignorance here.