My MOCA experience setting up Netgear MOCA adapters with Comcast

[Gigahertz21]

I just setup my Netgear MOCA adapters yesterday. I couldn't get them to link to each other and thought the problem was most likely the signal booster (amplifier) in the garage. I couldn't open the Comcast cable box on the side of the house since they have it locked down with a special locking mechanism. So I called Comcast and they sent out a tech and he unlocked it and unhooked the booster from the system and then went and made sure the speeds of our cable modem and TV's was not affected. I also asked the tech if he could switch out the splitters in the cable box since they only went up to 1000MHz but that's all Comcast carries and they don't have any diplexers either.

I went and tried to link the MOCA adapters again but it still wouldn't work, I had one installed between my cable modem and router and the other one I was walking around the house with plugging into the coax outlet ports to see if the link light would come on but it wouldn't. The tech mentioned I should plug the adapter into a coax port that I know receives a cable connection. I figured all of the coax ports in the house were wired to receive a signal anyways but I followed his advice and unplugged a TV that was connected to a coax port on the wall and then plugged in the MOCA box and the link light finally came on.

So I think the whole time I was plugging my 2nd MOCA adapter into coax outlets in the house that were not wired to receive a signal. I probably didn't even need the signal amplifier to be removed, I talked to another tech and he mentioned the amplifiers they install are passive (not active) so MOCA adapters should still work with them. I'll never know now though since the amp is disconnected and I'm not sure how to reconnect it.

Anyways, I've only done one simple speed test. I used the upstairs computer wired to my router to transfer a file off an external USB hard drive that is attached to my downstairs WDTV Live media player which is connected to the network using a MOCA adapter. I was getting around 6MB/sec transfer speeds which should be more then enough to stream 720P/1080P 4-12GB h264 mkvs. But I know these MOCA adapters can get to around 10-12MB/sec and I'd like to get that speed. The Comcast cable guy left the cable box unlocked for me so I might change their 1000MHz splitter in there to one that goes to 2GHz (2000MHz) and see if that does anything.

The cables from the cable box on the side of the house go up into the attic so there is probably 1 or more splitters up there that only go to 1000MHz. I went up there to look around but the stuffing they put up in the attic for insulation is several feet thick and it would be a huge pain to try and follow those coax cables to see if they are attached to any lower frequency 1000MHz splitters.


For those that have an amplifier installed in their house by the cable company and it cannot be removed, I would suggest you go to this Amazon post and read it, the guy made a diagram on how he connected diplexers to route the MOCA signal around the amp. Also, read through the Amazon reviews, there are tons of explanations on how people got these adapters to work with their setup.

http://www.amazon.com/review/R3KCL6...e=UTF8&ASIN=B001N85NMI&nodeID=&tag=&linkCode=

 

[Bob K]

The cables from the cable box on the side of the house go up into the attic so there is probably 1 or more splitters up there that only go to 1000MHz. I went up there to look around but the stuffing they put up in the attic for insulation is several feet thick and it would be a huge pain to try and follow those coax cables to see if they are attached to any lower frequency 1000MHz splitters.

I tested my adapters and had no problem syncing through 5-1000MHz splitters while using the 1000MHz+ sync frequencies. It doesn't make much sense that I can but I've read that several others have experienced the same behavior with 5-1000MHz splitters. Weird.

If you think you are getting errors as the result of having 5-1000MHz splitters in your coax web, you could always go into the configuration on the MoCA and reduce the sync frequency to below 1000MHz. You'd lose functionality in the "coax out" port on the MoCA though. Of note, I noticed a consistent 2Mbps write speed drop at the 1.5GHz frequency (72Mbps) as compared to syncing the MoCAs at 875MHz (74 Mbps) even after I installed a good quality 2.4GHz splitter: My MCA1001 review on Amazon.

I'm interested to see what your experiences are when streaming high bitrate (16-35Mbps) videos. While many do it successfully, I seem to have a problem at the packet level that causes insufferable stuttering! MCA1001 TCP Window Troubleshooting

 

[stevech]

My MoCA maxes out at 9.3MB/s (72Mbps+) when doing a file transfer via windows 7 shares.

I've played HD1080i (not 1080p) videos across the MoCA bridge from a Synology NAS to an AMD E350 based HTPC - no problems. Using the WIn 7 IP stack RWIN size default which is 65KB.

 

[Bob K]

My MoCA maxes out at 9.3MB/s (72Mbps+) when doing a file transfer via windows 7 shares.

Using LAN Speed Test (www.totusoft.com), I get great speeds over MoCA (74Mbps write, 84Mbps read) when doing standard Win7 file transfers.

However, when I played around with Jperf, I noticed that a default TCP Window size of 8K would only give me 20Mbps throughput, As I increased the TCP Window size, the throughput steadily increased. A 128K TCP Window gave me over 90Mbps throughput.

I've played around with TCP Optimizer but no combination of settings gets rid of the stuttering. More details here: MCA1001 Packet Troubleshooting

I've played HD1080i (not 1080p) videos across the MoCA bridge from a Synology NAS to an AMD E350 based HTPC - no problems. Using the WIn 7 IP stack RWIN size default which is 65KB.

I'm hoping the Wireshark guys in the link above will be able to help me out with my capture files. Despite disabling the automatic TCP Window scaling in Win7 and setting a manual TCP Window size, I haven't found the problem. Something significant is going on, but I just don't have the skill yet to identify the problem in my Wireshark capture files.

 

[stevech]

something wrong that Windows PC default window size is 8KB.
Which direction? There's a window for each.

Some cheap non-PC devices (media player?) have limited buffer space and might assert a smaller window size in the receive direction.

 

[Bob K]

something wrong that Windows PC default window size is 8KB. Which direction? There's a window for each.

Sorry, 8K was the default TCP Window size used in Jperf, not the TCP Window size used during actual Win7 transmissions.

Some cheap non-PC devices (media player?) have limited buffer space and might assert a smaller window size in the receive direction.

The high bitrate video files play flawlessly in any configuration where the MCA1001 is not in the transmission path. That indicates the LG BD670 (and LG BD570) bluray players I use are quite capable of handling the streaming. Only when the MoCA adapters are introduced do I experience video stutter.

If you have some time, I'd be grateful if you'd take a look at the information in the MCA1001 Packet Troubleshooting link I posted earlier. It answers a lot of the troubleshooting questions you've asked. If you have Wireshark, would you be willing to download the capture files to see if you can detect a problem? I've spent about 10 hours so far analyzing the captures. Everything looks normal in both the "stutter" and "non-stutter" captures; however, I fully admit that I'm a junior in the TCP world and am unable to detect all potential errors.

 

[stevech]

I'll try to take a look at the big capture files. Meanwhile...

I doubt that your video streaming is using TCP for the video data. It might have a second flow that uses TCP. There's a lack of good standards for transporting compressed video; the compressors are standardized (CODECs) but the transports aren't.

The MoCA device are layer 2 bridges- this means that they see IP packets and, like simple ethernet switches, they don't process they layer 3 protocols like TCP. I think most video "bearer traffic" is unreliable datagrams like UDP/RTP.

Did I understand you to say that if you run less than 1080p, say, 1080i or less, there is no quality problem?

if there's any way to beg/borrow a test setup for streaming (I don't think jperf does that).. it would be helpful.
And if jperf has a rate-driven test using UDP or RTP, where you set the source at, say, 20Mbps constant and see if the receiving end gets overrun.

This is a hard one.

 

[Bob K]

I doubt t that your video streaming is using TCP for the video data. It might have a second flow that uses TCP. There's a lack of good standards for transporting compressed video; the compressors are standardized (CODECs) but the transports aren't.

I can only base it off of what I see in Wireshark, which lists "TCP" in the Protocol column for the tens of thousands of packets transmitted to and acknowledged by the bluray player. If there's a masked underlying UDP layer in there somewhere, I don't have the smarts to see it right now.

The MoCA device are layer 2 bridges- this means that they see IP packets and, like simple ethernet switches, they don't process they layer 3 protocols like TCP. I think most video "bearer traffic" is unreliable datagrams like UDP/RTP.

That's why this is so baffling to me. If the MoCA is a layer 2 bridge and only routes untouched packets, why the stuttering only on bitrates above about 20Mbps? There is no packet loss, and one of the guys at Wireshark thinks it might be related to packet Round Trip Time (RTT): Maybe the MoCA is somehow delaying communication just enough to foul something up.

Did I understand you to say that if you run less than 1080p, say, 1080i or less, there is no quality problem?

Yes, the topology will flawlessly play any compatible video that stays under 20Mbps or so.

if there's any way to beg/borrow a test setup for streaming (I don't think jperf does that).. it would be helpful. And if jperf has a rate-driven test using UDP or RTP, where you set the source at, say, 20Mbps constant and see if the receiving end gets overrun.

I'm going to do more testing with Jperf. Specifically, I'm going to connect my two PCs together to see if the throughput with the default 8K TCP Window gets stuck at 20Mbps (as it does withthe MoCAs in the topology) or if the PCs automatically scale to allow the highest throughput. I think that'll give me another datapoint to determine if Jperf's results can be trusted.

 

[stevech]

TCP may be used for signalling and flow control whereas RTP or a form of UDP might be carrying the compressed video such as MPEG.
I suppose that if RTP/UDP packet arrival rate at the bridge exceeds the speed on the other side (MoCA RF), which it does, it's up to the TCP protocol session, if present, to do flow control. Maybe the bridge implementation is wrong and they don't buffer properly at high rates.