Kitz Forum
Announcements => Site & Forum Discussion => Topic started by: ejs on November 15, 2018, 06:34:08 PM
-
https://kitz.co.uk/adsl/interleaving.htm
I think the table at the bottom of that page, titled "Relationship between Interleave Depth & Latency", is largely irrelevant to DSL. If you follow the link to the Juniper source, you'll find it's about DOCSIS (cable).
It is important to note that Interleave Depth & Interleave Delay are not necessarily the same thing as the additional amount of latency you will see when interleaving is switched on.
I'm pretty sure that at least for ADSL2/2+ and VDSL2, the interleave delay reported by the modem does indeed indicate the amount of additional latency due to the interleaving. And because it's a set amount of delay, the depth then tends to vary depending on the bandwidth of the line. More bandwidth, greater depth, same delay.
Another issue I don't like about that page is the notion that the modem will be doing any re-requesting of corrupt data. Your computer (TCP) might have to arrange some re-transmissions, but unless the modem is doing G.INP / PhyR, it won't be.
-
>>I think the table at the bottom of that page
Yes it is DOCSIS config, but as DSL also uses QAM I thought it was quite interesting. It's a table I happened to stumble across and why I linked back to it.
If it is of no relevance I can easily delete it.
>> I'm pretty sure that at least for ADSL2/2+ and VDSL2, the interleave delay reported by the modem does indeed indicate the amount of additional latency due to the interleaving.
What I was trying to explain was that some people would see say "Interleave depth = 8 " and assume that was the amount of delay. It was also written many moons ago when modems didn't give quite so much info in the stats and it wasn't always clear if when a modem said "Interleave" if it was delay, depth, max delay or whatever. Does interleave depth cause 1ms of delay.
If you can think of a way to better phrase it then I will happily change it. :)
>> that the modem will be doing any re-requesting of corrupt data
Agreed. My bad for over simplification. I've explained on another page about CRC that its actually done at a higher level, but agree it needs changing to PC.