We are looking mainly at the power levels on each channel, too low a power means that the modem can't properly 'hear' what data is being sent, counter-intuitively, too high a power can have a similar effect - think of it as someone whispering very quietly or someone yelling at full volume directly into your ear. In both cases it is hard to understand the message they are trying to convey.
The official DOCSIS (ie the underlaying cable internet technology) specifies a power range between +15 dBmV and -15 dBmV per channel, in case you are wondering how a power level can be negative, in this case in can be, just trust me on this one! In practice, though, you don't want to be anywhere near those extremes and each cable operator knows the reasonable operational limits of their own equipment. For VM a general 'rule of thumb' is a value between +10 and -6 dBmV, but because of the variable tolerances of electronic components, it doesn't necessarily follow that if the levels fall slightly out of this range then the connection will immediately fail, neither does it mean that a value of 0 is 'better' than a value of -3. But the closer it is to 0, the more leeway there is to cope with changes.
And things do change, as the weather gets warmer, the cables expand in the heat, their impedance goes up and so the power reaching the hub goes down. It is quite possible for there to be a 3dB swing between a cold winter's day and a very hot summer one. So a value of -4, say, on a bitterly cold day in January is fine but might give a cause for concern come August!
Signal to noise ratio S/N should be as high as possible, it is a measure of how 'clear' the signal is or how much extraneous 'noise' has gotten in. Anything above the low to mid 30dB should be OK but again there really isn't any hard and fast rule. I generally think that seeing values above 38dB is a good indicator.
Now the upstream is a bit more complex, the hub knows the power of the downstream channels it is receiving, what it doesn't know is how well the messages it is sending out are being received. All it can say is that it is broadcasting using x power levels and it is getting a response back so this is probably OK. Similarly, it has no idea of what S/N ratio the far end of the cable is getting so there is less useful information from the hub end.
Now there is a maximum value of power that the hub is capable of pushing out, if the values for output power look too high then it indicates that the hub is HAVING TO SHOUT to get its message through, which is often an indication of a cabling fault, bad connection, cracked co-ax cable etc. Acceptable values would be around 35 to 50 dBmV. Don't forget the change that happens as the cable expands in the summer - somewhat conversely, as the cable expands and the impedance rises, then the hub has to increase power to compensate, so you have the reverse situation from the downstream. Again, you really don't want to be looking at an upstream power of 50dBmV in the middle of winter, because it could well struggle in the summer.
Something else we often look at is the upstream channel modulation, not without going into the maths and drawing pretty constellation diagrams (actually Wikipedia has some nice animated diagrams showing how this works) QAM is a method of encoding binary data onto two analogue signals. The higher the QAM level, the more data can be 'packed' in but the better the connection has to be and the lower the noise level to be able to extract the data. VM's upstream runs at 64 QAM but it self-regulates, if it detects that a particular channel is a bit 'noisy' and can't reliably sustain this modulation, then it will drop to 32 or even 16 QAM because that is better than losing the channel entirely or corrupting data which would need to be resent.
This is not necessarily something to worry about, yes it is an indication that something isn't quite right but it might not be as important as it looks;
If you look at the upstream values you might see something similar to this
1 25800273 38.1 5120 64 qam
The first number is simply the channel ID, the second is the channel frequency (now the 25.8 MHz frequency is known to be a bit noisy, there are many sources of external noise which ideally shouldn't get into the cable but if there any imperfections in the connection that this is likely to be the channel most affected).
The third number is the outgoing power level, that's fine, the fourth mysterious number is the symbol rate in thousands of symbols per second. As I mentioned earlier QAM works by packing bits of digital data into 'symbols' and it is these that are actually transmitted to the hub. And lastly we have the modulation which tells us how many bits of useful information can be packed into each symbol.
So you might think that if the modulation falls from 64 to 32 then you have halved the data rate? Not so, because of the way QAM works it is a binary representation, 64 QAM is six bits per symbol (2 raised to the power of 6 is 64), and similarly 32 QAM is five bits per symbol, so you haven't halved the rate just reduced it by about 17%. If you want to know what that means in real terms well we can do (roughly) some maths.
A single channel at 64 QAM and a symbol rate of 5120 ks/s is - and not making allowance for overheads because I'm only interested in a comparison;
5120 x 6 = 30720 kbits per second or 31457280 bits per second or about 30 Mb/s (slightly more, I rounded the numbers and if we do allow for the DOCSIS overheads it's closer to 27 Mb/s anyway) - and each modem bonds four channels so a maximum theoretical upstream speed of c. 120 Mb/s
Now in reality the cable operator doesn't offer speeds that high because congestion and the way DOCSIS upstream works wouldn't let you get anywhere near the theoretical maximum reliably. VM's fastest upstream speed is 52 Mb/s, comfortably within the limit.
Now suppose one channel drops to 32 QAM, same symbol rate but it is now 5120 x 5 (rather than 6) or 25600 kb/s or 25 Mb/s - if the other three channels are still on 64 QAM or even if they have all dropped, the theoretical maximum upstream speed has dropped from 120 to 115 or 100 Mb/s (roughly), so it still shouldn't affect your actual connection rate in practice.
Should they all be at 64 QAM? Well yes. Is it an indication that the sky is falling in? Absolutely not!
Now lastly, when looking at the logs themselves, you can get a sort of measure of how important each message is by looking at the priority which on the Hub 3 and earlier mentions Critical, Warning, Notice and Error and on the later hubs these are replaced by numbers where 6 is the lowest importance and 1 is the highest. The categories however don’t always mean what you might think. Notice just means something has happened which you might be interested in, ie someone has logged in successfully or failed to log in because the password was wrong. ‘Error’ is far less worrying than it sounds - the hub has received some information which it doesn’t understand or has no idea what to do with and so has decided to just ignore it and carry on. The DHCP messages are like this and can be safely disregarded.
Now ‘Warning’, ie the RCS Partial Service messages, imply that something has happened but the hub can work around it, basically indefinitely, without noticeably impacting your service. Yes it’s not ideal and the hub will act to fix it at some point but don’t lose too much sleep over it.
Lastly ‘Critical’ now these are things that the hub can’t work around and will, if left, impact your service, now whether or not you notice this is another matter. The infamous No Ranging Response is typical of this, one of the upstream channels is not getting the response back that it expects, and if it doesn’t recover in a fairly short time, the hub will need to drop and renegotiate that connection.
Here endeth the lesson........
John
