1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Moca works without POE filter - Do I still need to get one?

Discussion in 'TiVo Roamio DVRs' started by The TiVo Dude, Dec 25, 2013.

  1. HarperVision

    HarperVision TiVo's Italian Cuz!

    May 14, 2007
    Or complain to your cable co and they may give you one. That's what I did. It helps their system and behooves them to have one installed at your location.
  2. dianebrat

    dianebrat Uncontrolled Force

    Jul 6, 2002
    This issue is never that you need one for your Tivo to work, it's to prevent your data signals from backfilling out of your home and affecting your neighbors networks. If you're MoCA is active and you don't have a POE filter, your data could be available to your neighbors, and it can cause trouble for their installations. It's never about your Tivo not working.
  3. k2ue

    k2ue Retired RF Engineer

    May 9, 2002
    Victor, NY
    That's not going to be universally true: for MOCA to have a thruput of >100Mbs the path loss should not exceed 56dB, and it will choke about 6dB below that. Less than 50dB path loss will let it reach its max speed of about 200Mbs. So there is about a 12dB margin from hunky-dory to not working. TiVo is using MOCA at ~1100MHz, where cable loss is significant (~75 feet of (good) cable is equivalent to a 4-way split). High-grade splitters like cable companies use may have 30dB typical port isolation. So if you have a scenario of the TiVo on separate ports with 75 ft of line to each and a another 2-way split on the way to the TiVos, you would have 50dB path loss -- right at the limit for max thruput. Add any other liabilities (more splitters, old cable, poor connectors, RG-58, attenuators, etc) and you are going to lose thruput. Adding a POE filter would boost the MOCA level ~16dB if your first splitter is 4-way, or ~22dB if it is 2-way, so it can be helpful in those cases.
  4. RickBrant

    RickBrant New Member

    Sep 19, 2015
    Pardon me for resurrecting this thread, but I don't see how this can work.

    The cable is obviously far longer than a wavelength in the MoCA band (about 1 to 1.6 GHz; wavelength in the cable would be well under a foot). So the reflected signals are going to be arriving at the receiving modems in what can only be described as an unpredictable phase relationship with the direct.

    If you only have two MoCA modems this wouldn't be much of an issue (assuming the filter was installed very close to the "outward-facing" modem). But with three or more it's easy to imagine situations where signals sent by one will arrive at another by both "direct" and "reflected" paths, these differing hugely in transit time.

    Why wouldn't this corrupt the signals?

    Example: Suppose we pick 1200 MHz and assume the cable velocity factor is 0.7, then the wavelength is about 0.36 m. (The exact numbers don't matter; I'll show why in a moment.) Suppose we have one modem "B" that's 12m from the "head modem" ("A") and then another modem "C" that's 10m beyond B. B sends something. It's 10m to C by the direct path; it's 34m by the reflected path, a difference of 24m. 24m divided by 0.36 is 66.666... Which means the reflected signal will arrive at C 0.666 wavelengths, or 240 degrees out of phase, with the direct signal.

    That's a significant phase shift. If it was 180 out of phase it would be canceled if they were of equal amplitude.

    Of course the MoCA signal is very wideband (per Shannon-Hartley it has to be to get good data throughput), so we'll actually be getting reflections at all manner of phase shifts from the direct. Which is why the numbers above don't have to be exact. No matter how the stations are spaced there will be interfering reflections over a significant part of the MoCA band.) I just can't see why this can be ignored.

    I mean... I've never ever been in a reverberant room where the echoes improved intelligibility, let alone conveniently reinforced the sound that arrived by the direct path!
    Last edited: Sep 19, 2015
  5. RickBrant

    RickBrant New Member

    Sep 19, 2015
    Update: So, I got a MoCA filter and tried it. My configuration is similar to the one described in my post above: Three modems. The path from the first modem's coax "in" goes through at least 20m of coax to a splitter, modem "B" is about 2m from that splitter, then the other leg of the splitter goes via who-knows-how-much cable... but certainly at least 10m... to modem "C".

    File transfers from a machine attached to modem "B" to one attached to modem "C" were completely unaffected. Not better, not worse. One might expect that if there was an improvement in signal strength there would be an improvement in throughput. However the throughput I'm seeing was already at the limit of the MoCA modems' Ethernet interfaces (100 Mbit/s) so improvement really shouldn't have been possible. On the other hand it didn't get worse either, so clearly my concerns about destructive interference were not realized here.

    There's no harm in leaving the filter in, and it might avoid complaints from the cable company, so I'll leave it in, but as far as I am concerned the claims of improved performance remain unproven.

    Sure wish I hadn't sold my spectrum analyzer - I'd like to have seen what's happening on the cable with and without the filter! (But that would have been its first use in many years, so...)
  6. snerd

    snerd New Member

    Jun 6, 2008
    From a purely theoretical perspective, if you could build a system with perfect splitters, then with no PoE filter in place the MoCA devices would not be able to communicate at all. I'll explain why using a simplified system using a single "perfect" splitter with say 20 meters of coax between the splitter and each of the (A,B) MoCA adapters. We'll assume everything is properly terminated so that there are no reflections anywhere in the system. Now when A transmits, when the RF signals reach the splitter, they go through the splitter and out toward the cable company, never to return. Since we've assumed the splitters are perfect, the "downstream" ports of the splitter have perfect isolation, which means that none of the upstream signal from MoCA A ever reaches MoCA B. Adding the PoE filter causes the signal to be reflected back down through all of the splitters so that the MoCA signals are kept in the house rather than leaking out to the neighbors.

    Real splitters aren't perfect, having port isolation of maybe 25 dB to 30 dB, so when an upstream signal hits the splitter, a tiny fraction of the RF power from MoCA A will be coupled to the other port to be received by MoCA B. A splitter with 30 dB of port isolation will allow 0.1% of the upstream RF power to couple to the other port, while about 45% of the upstream RF power leaves the house, and the rest is absorbed (and wasted) in the splitter. The 45% figure is based on a 3.5 dB insertion loss for a 2-port splitter, since the insertion loss is the same for downstream and upstream signals. Adding a PoE filters mean that the upstream signal will lose 3.5 dB going upstream through the splitter, then a small loss (may 1 dB) as it is reflected by the PoE filter, then another 3.5 dB loss going back down through the splitter for a net loss of 8 dB plus loss in the coax. So overall the signal is 22 dB stronger than if no PoE filter is used.

    There is still an issue with interference between the part of the signal that is due to imperfect port isolation having a different phase than the part of the signal that is reflected by the PoE filter, and I don't really know how the MoCA system resolves that problem.

    The TiVO menus include Network diagnostics which show the PHY data rates and some numbers about signals levels. You might try looking at those numbers with and without the PoE filter to see how much difference it makes, if any. The difficulty is the MoCA is designed to operate essentially at full speed as long as the total loss is below about 50dB, so you might need several layers of splitter and some long cables before you see much degradation in performance.
  7. RickBrant

    RickBrant New Member

    Sep 19, 2015
    snerd: Thank you for your well-thought-out and -expressed answer.

    I think you've answered it! If the reflected singal is 22 dB stronger than what leaks through the splitter, then the leaked signal is pretty far down in the noise. 22 dB is a LOT of difference with modern (DSP) techniques, with a modulation as well-characterized as QAM.

    Practically speaking, we have to conclude that obviously the MoCA demod can cope with it. If it couldn't, my throughput would have been much poorer without the POE filter.

    Ah, well, I... (looks down, draws circles with toe) I don't actually have any TiVO hardware. I'm just using NIM100 modems. I was directed to this thread for the tech details on why the POE filter is supposed to improve MoCA, independent of whether it keeps the cable company happy. I have some info on getting into the management interface of the NIM100 but I don't know if they have something that reads out SNR. I'll probably look at it someday but there's no pressing need atm.

    Thanks again for the answer, this clears up a lot.

Share This Page