We currently have 42 nodes broken down into 5 upstream (return) node groups. On the downstream (forward) side of the plant, we have a single node group. We would like to break the downstream into 4 node groups. The way the math works, we will have 10 downstream channels per node group and we will have 2 downstream channels that cover all 4 node groups. We will be using an ATX Q-Series amplifier to provide channel isolation between the four node groups. Will it be possible to have an uneven number of downstream and upstream node groups or will we need to have a 1:1 ratio between the two?
Your modems will only bond to 8 channels at a time, why would you send 10 to each node group? You could make a 5th node group with the extra downstreams. But that depends on what kind of cards you have and how your configured, yay variables. Not sure what to say on the 2 extra channels, probably not a good idea to spread them over all of the groups, I'd put them in with your heaviest node group.
As Capm says, it's hard to give advice on such topics without a lot more info
Your terminology isnt super familar to me. I assume on the downstream your "node group" is what I would call a "mac domain" or a "service area" (eg int cable 5/0/0, 5/0/1 etc).
Your design will come down to the combination of headend hardware you have :
* what linecard
* what forward combiner
* how many TX lasers, and how many nodes optically split off each TX laser
* what return combiner
Plus it depends on what types of modems you have in the field : D1, D2, D3 4x4, D3 8x4. Oh and dont forget 16X and 24X modems are just around the corner!
You will have to decide how many US do you want to bond for D3 (1, 2, 3 or 4)
Then you also have to factor in stuff like how many modems are there per node (is it pretty even or vary widely?)
When it comes to spanning some DS across mutiple mac domain, there are some benefits to this. For example say you had a Cisco MC20x20V linecard. This gives you 20 DS to play with. You might have lots of 8x4 D3 modems on the network. You want to provide 8X DS bonding but if you made 2 x mac domains with 8 DS each, you are then left with 4 DS "stranded". Instead you could make 4 mac domains with 4 DS each, and then "span" the remaining 4 DS across all 4 mac domain. This way each mac domain can provide 8x bonding. You wont get full 8x performance if traffic levels are heavy, but generally traffic bursts will come and go and this design allows those 4 spanned DS to provide extra capacity to which every mac domains are busiest at the time. Spanning some DS is also handy trick to distrubute multicast IPTV to multiple mac-domains (as a "secondary bonding group")
The way I design the wiring is to use a spreadsheet. Left to right across the page it is showing all the mappings between :
CMTS DS ports -> TX combiner inputs -> TX combiner output -> node -> RX combiner input -> RX combiner output -> CMTS US ports
You can then plan out which nodes get connected where. To evenly distribute traffic you can pair up nodes that are big/small, busy/quite, CBD(daytime)/residential(nighttime).
Typically I would be sending 8 DS to each mac domain, and then having 2 upstreams per 1, 2, 3 or 4 nodes (depending how big they are). 4 to 8 upstreams per mac-domain.
If you spread an upstream over too many nodes, it leaves you too exposed for noise. Should there be noise in node 1 it could have knock on effects for all the other nodes it is shared with.
In my network I have it easy, in that we got rid of all the old modems and now only have D2 and D3. This means you dont have to stuff around with tdma special upstreams to service ancient modems. Instead we can just run straight atdma to get access to the best features. D3 modems can bond, and D2 modems can be load balanced.
I have a big collection of PDF and PPT docs I have collected from the internet over the last few years since design and deploy of our D3 network. Some of them have some really good wiring/design examples in them. I guess I could also publish an example of my wiring spreadsheet if people thought that might be useful.