|
Posted by Ed Nielsen on December 26, 2006, 10:38 pm
If you were Registered and logged in, you could reply and use other advanced thread options
That is an excellent splitter, although even excellent ones can have an
occasional bad apple in the lot. Other than the suggestion to try
grounding that splitter, all I can probably say is good luck pinpointing
the cause if the issue. I read your other post about what you are
seeing, and it really sounds like a network issue rather than something
local.
CIAO!
Ed N.
Timothy Daniels wrote:
> Now the splitter seems to make no difference in packet loss (as measured
> by "ping -t"), except to drop the downstream signal level to the modem
> to 6-7 dBmV and to cause the modem to raise its signal level to 50 dBmV.
> The splitter is a TVC unit that's rated for use to 1 GHz, and it's
> solder sealed.
> On pings to yahoo, google, and mit, packet loss is reported as 2-3%
> (although
> I observe twice that) and to rr.com it's reported around 10%. There seems
> to be no pattern, now.
>
> *TimDaniels*
>
> "Ed Nielsen" wrote:
>> Sorry -- I had misread something you had written. Warren had given a
>> good analogy of that, though.
>>
>> Technically, a splitter is a power divider. In fact, as you probably
>> already know, in the commercial world that is what they are called.
>> When the voltage is cut in half (half of the voltage goes down each
>> leg of a 2-way splitter), there is a 3dB reduction in signal strength.
>> Actually, it's 3.01. Throw in a little more for imperfect materials
>> used in the manufacturing process and we have ~3.5. That number
>> varies slightly with different manufacturers and it also varies a
>> little with frequency. As the frequency increases, so does the loss
>> through the splitter. The insertion loss through a 2-way splitter is
>> typically from ~3.3dB (5MHz) to ~4.2dB (1GHz). A 4-way loses ~6.7dB
>> (5MHz) to ~8dB (1GHz).
>>
>> It is possible that a splitter could cause disruption amongst the
>> packets. Make sure that the one you have in there has a high return
>> loss on all ports. Also, of course, solder-backed (no epoxy).
>>
>> HAVE A VERY MERRY CHRISTMAS!
>>
>>
>> CIAO!
>>
>> Ed N.
>>
>> Timothy Daniels wrote:
>>> I never posed that it was packet loss that caused the modem
>>> to increase its upstream signal level.
>>>
>>> You say that it is the CMTS at the headend that tells the modem
>>> to turn its signal level up or down. And when the splitter is
>>> substituted
>>> for the barrel connetor, the output level of the modem is increased.
>>> That suggests that the splitter itself is attenuating the upstream
>>> signal
>>> from the modem to the CMTS, as nothing else has changed. It may
>>> also be the splitter which is mangling the packets. I'll search around
>>> for another splitter to see if that helps.
>>>
>>> *TimDaniels*
>>>
>>>
>>> "Ed Nielsen" wrote:
>>>> It isn't packet loss that causes the modem to pump out more -- nor
>>>> is it because the downstream has changed. [You could use a couple
>>>> of diplex filters (not the satellite ones but similar in purpose)
>>>> that separate the forward path from the return path, pad the
>>>> downstream down as much as you want (or even amplify it) and you
>>>> would see virtually no change in the transmit level of the cable
>>>> modem.] It is merely that because of the reduced attenuation on the
>>>> drop, the cable modem is hitting the CMTS too hot, so it is told to
>>>> turn down.
>>>>
>>>> That suggests to me that there might be something (noise) getting
>>>> into the drop between the cable modem and the splitter/barrel
>>>> location. When you insert the splitter, you lower the signal level
>>>> hitting the CMTS, so it tells the cable modem to increase its
>>>> transmit level. The signal level from the splice location back to
>>>> the lockbox remains pretty much the same whether the splitter is in
>>>> there or not -- it is only in the portion of the drop from that
>>>> location to the outlet that sees the increase/decrease in signal
>>>> level as the modem's transmit level changes. Could that piece run
>>>> parallel to some A.C. lines for awhile? There could be something
>>>> that is induced onto and is riding on the shield until it finds a
>>>> spot where it can leak onto the center conductor. Have you tried
>>>> grounding the splitter? If there is something riding on the shield,
>>>> that could give it somewhere to go where it will do no harm.
>>>>
>>>>
>>>> CIAO!
>>>>
>>>> Ed N.
>>>>
>>>> Timothy Daniels wrote:
>>>>> Yup, that all sounds right. :-) And it sounds like the cable modem
>>>>> is receiving and transmitting the right signal levels. But why would
>>>>> the packet loss increase when the barrel connector is substituted
>>>>> for the splitter, thus causing the upstream signal level to rise from
>>>>> +47dBmV to +49 dBmV as a result of the downstream signal level
>>>>> dropping from +10 dBmv to +7 dBmV? I'd think that the louder
>>>>> outgoing signal would better the signal-to-noise ratio. as seen by
>>>>> the network.
>>>>>
>>>>> *TimDaniels*
>>>>>
>>>>>
>>>>> "Ed Nielsen" wrote:
>>>>>> That "reference level" (as you call it), like most numbers that
>>>>>> have a (+) or a (-) before them is 0. DOCSIS says that cable
>>>>>> modems have an input level window of -15 - +15dBmV, which simply
>>>>>> means that as long as the input signal level is within that
>>>>>> window, it will work. The output level window, as dictated by
>>>>>> DOCSIS, is +8 - +58dBmV. Again, as long as the output level of
>>>>>> the cable modem is within that window, it will work.
>>>>>>
>>>>>> The input level, of course, is determined by the plant/drop
>>>>>> system. For example: There is a line extender (amplifier) at the
>>>>>> pedestal that feeds your house. [Single family home with a white
>>>>>> picket fence and a couple of elm trees in the front yard out in
>>>>>> the suburbs. Can't forget the 2 dogs and a cat, a swing on the
>>>>>> front porch, a vegetable garden, and a few fruit trees out back,
>>>>>> etc.] The tap on the output of that L.E. is a 26 value tap, and
>>>>>> the length of your drop is 100', and the DOCSIS frequency is
>>>>>> 600MHz, and it is a QAM256 system. The L.E is running out at
>>>>>> 44dBmV (no slope, the output is flat), which equates to 18dBmV
>>>>>> coming out of the tap. 100 feet of RG 6 loses ~5dB@600MHz, so you
>>>>>> a level of 13dBmV at the groundblock. In a QAM256 system, the
>>>>>> data carrier runs at 6dB below the nearest analog channel, so that
>>>>>> 13dBmV becomes 7dBmV. Place a cable modem right at the
>>>>>> groundblock, and it will see an input level of +7dBmV. Say you
>>>>>> have a TV set in the family room and your computer is in a room
>>>>>> 25' from the groundblock. A 2-way splitter loses ~3.9dB at 600MHz
>>>>>> and the 25' drop to the cable modem loses 1.25dB. That 7dBmV at
>>>>>> the groundblock then becomes (+7-3.9-1.25=) 1.85dBmV at the input
>>>>>> of the cable modem. Suppose you have 4 TV sets. The input level
>>>>>> at your cable modem would then be (7-7.9-1.25=) -2.15dBmV. Change
>>>>>> the attenuation and the input level changes by the same amount. A
>>>>>> 30dB wide input level, cable modems have.
>>>>>>
>>>>>> On the output side in the same scenario, the return amp in the
>>>>>> L.E. wants to see ~10dBmV. The pathway from cable modem is the
>>>>>> drop and the tap. Let's say that the upstream frequency is
>>>>>> 22MHz. RG 6 loses ~1.25dB/100' at 22MHz. The pathway to that amp
>>>>>> has the signal losing 1.25dB (drop) and 26dB (tap). 10 (required
>>>>>> input level to the amp) + 26 (tap) + 1.25 (drop) would have the
>>>>>> cable modem pumping out 37.25dBmV. One TV (2-way splitter) and the
>>>>>> cable modem 25' from the groundblock would have it pumping out
>>>>>> (10+26+1.253.6+.31=) 41.16. Four TVs would necessitate that the
>>>>>> modem transmit at 44.56dBmV.
>>>>>>
>>>>>> The CMTS (Cable Modem Termination System (in the headend)) wants a
>>>>>> signal level of 0dBmV at its input. No wide window like cable
>>>>>> modems have -- it wants 0dBmV. If the attenuation at a cable
>>>>>> modem changes, the CMTS tells the modem to increase or decrease
>>>>>> its output level accordingly. If a cable modem runs so as to hit
>>>>>> the CMTS with too hot of a signal, the CMTS tells it to turn
>>>>>> itself down. Too little and it tells the modem to turn itself
>>>>>> up. When you remove the splitter from your drop, you reduce the
>>>>>> attenuation of your drop by 3.5dB, which means that there is
>>>>>> 3.5dBmV more signal hitting the CMTS than previously, so the CMTS
>>>>>> tells the modem to turn itself down. Put it back in and the modem
>>>>>> is told to crank up the power to compensate for the increase in
>>>>>> attenuation. That is why the transmit levels vary when you
>>>>>> remove/insert the splitter.
>>>>>>
>>>>>> If noise gets into the line, whether it be in the drop or
>>>>>> somewhere in the plant, the CMTS can't hear the cable modem as
>>>>>> well as it needs to, so it tells the modem to "speak louder," same
>>>>>> thing as you do when conversing with someone on the sidewalk and a
>>>>>> loud car drives by. Noise can be ingress from a local broadcaster
>>>>>> or impulse noise such as from a blow dryer or vacuum cleaner.
>>>>>>
>>>>>>
>>>>>> CIAO!
>>>>>>
>>>>>> Ed N.
>>>>>>
>>>>>>
>>>>>>
>>>>>> Timothy Daniels wrote:
>>>>>>> I think you got that backwards. First, the modem appears to be
>>>>>>> reporting a positive offset - a level *above* some reference level,
>>>>>>> not a negative offset from some reference. Thus, with the splitter
>>>>>>> in the line, the downstream level drops in magnitude from 10 dBmV
>>>>>>> to 9 dBmV (this morning, and during the day yesterday, the downs-
>>>>>>> stream level with the splitter in the line is/was 7 dBmV). And the
>>>>>>> upstream level rose to 49 dBmV from 47 dBmV with the splitter
>>>>>>> (today it's at 50 dBmV with the splitter.)
>>>>>>>
>>>>>>> That's agrees with the model that the Transmit upstream level varies
>>>>>>> counter to the downstream Received level, and it also suggests that
>>>>>>> the upstream level is a positive value, not a value below a
>>>>>>> reference
>>>>>>> level. Given that, it looks as if the high downstream level was set
>>>>>>> too high for there not to be a splitter since the packet loss
>>>>>>> *increases*
>>>>>>> when I substitute a barrel connector for the splitter.
>>>>>>>
>>>>>>> This would make sense if the cable company always assumes that
>>>>>>> there would be one or two TVs hung off the same line - a likely
>>>>>>> scenario for the average unit in a condo building, but just not the
>>>>>>> scenario in my condo unit.
>>>>>>>
>>>>>>> *TimDaniels*
>>>>>>>
>>>>>>>
>>>>>>> "Ed Nielsen" wrote:
>>>>>>>> Lower packet loss with the modem pumping out more suggests that
>>>>>>>> there may be some noise (impulse, ingress) hopping onto your
>>>>>>>> drop. Where did they pad your cable? Right at the modem? If
>>>>>>>> you can get to the main splitter, try padding there. Don't
>>>>>>>> worry about the input -- you have plenty of room there to do
>>>>>>>> what you want. You can go down to ~-12dBmV before there would
>>>>>>>> start to be any cause of concern. Heck, I used to run at
>>>>>>>> ~-12/52 for about 4 years until some plant issues caused me to
>>>>>>>> alter that. Now, I'm at +7/49, and I hate it. One of these
>>>>>>>> days soon, I'll get around to swapping out my DC to fix it.
>>>>>>>>
>>>>>>>>
>>>>>>>> CIAO!
>>>>>>>>
>>>>>>>> Ed N.
>>>>>>>>
>>>>>>>> Timothy Daniels wrote:
>>>>>>>>> Without the 2:1 splitter, the downstream signal level rises to
>>>>>>>>> 10 dBmv
>>>>>>>>> (from 9 dBmV) and the upstream signal level drops to 47 dBmV (from
>>>>>>>>> 49 dBmV). The tracerts show a quadrupling (or more) of
>>>>>>>>> timeouts with
>>>>>>>>> the splitter removed. It seems that the lower transmit level
>>>>>>>>> hinders the
>>>>>>>>> packets' survival rate, and that the splitter, in reducing the
>>>>>>>>> receive level,
>>>>>>>>> causes the modem to drop its transmit level and thus increases the
>>>>>>>>> timeout incidence. Is that right?
>>>>>>>>>
>>>>>>>>> BTW, *I* put the splitter between the modem and the RR network
>>>>>>>>> for analog cable TV reception of local stations. Whenever the
>>>>>>>>> techs
>>>>>>>>> were here, the splitter had been replaced by a barrel connector
>>>>>>>>> (don't ask). As far as TW is concerned, my only service from them
>>>>>>>>> is RoadRunner. But no instruments measured levels within my condo
>>>>>>>>> unit. The only measurements that they took directly were on
>>>>>>>>> the level
>>>>>>>>> from the amp in the vault at the curb and on the output from
>>>>>>>>> the building's
>>>>>>>>> amp that resulted from their padding down the signal coming
>>>>>>>>> from the
>>>>>>>>> curb amp. The only measurements that they took specific to my
>>>>>>>>> modem
>>>>>>>>> were the remote readings reported to them from by cell phone
>>>>>>>>> from the
>>>>>>>>> central office. They said that it gave them more information
>>>>>>>>> than what
>>>>>>>>> I got from the modem's webpage and, presumably, what they could
>>>>>>>>> learn
>>>>>>>>> by tapping into the line between the wall outlet and my modem.
>>>>>>>>> So it's
>>>>>>>>> strange that the timeouts are so much less frequent when I put
>>>>>>>>> my 2:1
>>>>>>>>> splitter in the line - it is not what they thought the usual
>>>>>>>>> configuration was.
>>>>>>>>>
>>>>>>>>> *TimDaniels*
| 
Yahoo! 
Google 
Windows Live 
del.icio.us 
digg 
Netscape 
XML site map