Modifying a DOCSIS 3.0 cable modem for high performance

http://www.usbjtag.com/vbforum/showthread.php?t=8259

by whalesalad • 13 years ago

158 36 13 years ago

Looks really fun! Expensive, for sure. But I like that he was able to measure the effects in terms of noise reduction, signal power, and error correction. This is the sort of thing that is normally a "I have a buddy who swears swapping out this part for another makes it faster" kind of endeavor. Sure, in theory there should be an improvement, but you never get to actually detect it. It's awesome that he's able to measure the increases, though I'm sure the design team responsible for the devices felt those increases weren't substantial enough to warrant going all out on component costs (edit: though according to his buddy at Cox the increases were "dramatic").

whalesalad• 13 years ago

It's remarkable how sensitive these things are. I had to do some really gnarly stuff to deal with some voltage backfeeding into my modem from a PoE on a wireless AP. It helped a lot but ultimately I had to get a tech out (Charter in SoCal). The guy installed what he referred to as a spider, which was a hand-made mess of filters and grounding adapters on the coax.

Google "high upstream power level" and there are bazillions of people suffering from this problem with all kinds of different solutions.

I'm in VA now with Cox instead of Charter, but dealing with a very similar modem and now, very similar problems. I'm gonna order another one (fairly cheap online) and see if I can do some of these modifications to improve performance and upstream power levels.

That or i'll cave-in and switch to Fios.

fnordfnordfnord• 13 years ago

Here, take this: http://www.murata.com/products/catalog/pdf/c39e.pdf

mindslight• 13 years ago

You can probably achieve the same effect with more reasonable capacitors. I doubt those rails started off with one hundredth of the capacitance he added.

Low-inductance decoupling across every chip supply is the extremely important part, and for that you really want ceramic surface mount caps either replacing or stacked on top of existing ones. I'd recommend skipping tantalum caps altogether and filling that size/freq tier with larger ceramic multilayers since you're not designing for cost.

And of course low-esr electrolytics on the rails to reduce power supply ripple and all that. And some ceramic decoupling caps on both sides of switching regulators too. But definitely make sure the output of every voltage regulator (switching and linear) remains stable and hasn't started oscillating.

ksec2• 13 years ago

Still waiting for DOCSIS 3.1 to bring Cable up to speed.

sliverstorm2• 13 years ago

What do you mean, "to bring cable up to speed"? Commonly available DOSCIS 3 modems with 8 downstream channels can perform up to 300Mbps, and the manufacturer can add as many channels as needed.

That's not record-setting of course, but it more than covers the ISP speed offerings I'm aware of. Aside from fiber, which is not pre-laid for most last-mile applications, what would be faster?

ksec2• 13 years ago

I am pretty sure that is 300Mbps shared. And you rarely get that even in real world best case scenario. And much worst if you live in City and High Rise Buildings with High Density population.

DOCSIS 3.1 brings OFDM as well as up to 10 DL /1 UL Gbps. At that sort of speed, as well other added reliability features finally makes Cable a decent Alternative. Speaking from experience in EU and Asia Area. No idea if it is the same in US.

ParkerK• 13 years ago

It's not the DOCSIS standard, it's the providers that's holding it back.

guidedlight• 13 years ago

+1 In Australia, my DOCSIS 3.0 is 90Mbps down and 1.8Mbps up. And the government has announced they will rip up the cable and give everyone 25-50Mbps down and 4-5Mbps up.

Everybody here thinks it's a joke, unfortunately it's not. http://abbottsinternet.com.au/

For this I currently pay about $75 per month. Sigh.

staunch• 13 years ago

I'm an avid PC FPS gamer. I'd pay good money for a cable modem that brought my ping down a bit or even just reduced jitter. I'd kill for FIOS (which is apparently never expanding in coverage).

kodisha• 13 years ago

same here, someone should recognize this opportunity.

It was the same with PC monitors, for years LCD models had ENORMOUS input lag and crappy features, but now we see models with 120HZ and no input lag/passtrough mode appearing.

sounds• 13 years ago

Just went down due to being hit by so many requests. Here is a cached version:

http://webcache.googleusercontent.com/search?q=cache:iqkzgJs...

newman314• 13 years ago

I would pay to have someone do similar mods to my SB6120. I get terrible jitter, no thanks to Comcast...

voltagex_2• 13 years ago

I'm on ADSL2+ and am likely to be for a lot longer, thanks to the change in the Australian government. The phone line coming into my house seems to act as a big antenna and I get so much noise that I can't maintain a connection above 10Mbit down, 1Mbit up. I'd pay for mods to my modems too.

mctx2• 13 years ago

I'm also on ADSL2+ in Perth, my area isn't going to get NBN for the next 3 years. When I've finished moving house I'll open up my modem and have a see if there are similar power noise problems - I've got some decent electronics equipment (scope/SMD soldering station/dmms). If you're in Perth then you're welcome to join in?

NamTaf• 13 years ago

I get 3.4Mbit/0.65Mbit. 6km from the Brisbane CBD. I was slated to have the NBN start being built around my area within 12 months but god knows what'll happen now.

I've spent many hours in the past with Internode tech support trying to resolve some of my connection's issues to no avail. It does some really funky stuff too when I play with the settings, so I'm sure there's room to improve but I can't do it with just software.

klipt• 13 years ago

I don't understand why Comcast doesn't support the DPC3010; it seems to be one of the cheapest DOCSIS 3 modem available.

On the other hand, maybe it's only cheap because Comcast doesn't support it?

wmf• 13 years ago

Comcast has strict certification requirements, so it may have something to do with that. http://www.publicknowledge.org/blog/zoom-shows-how-comcast-a...

I've noticed that Time Warner rents really cheap modems but for BYO they only support the most expensive models. I don't know why that is (one might speculate that they want to keep you renting since it's more profitable for them).

thrownaway2424• 13 years ago

I'm skeptical that the shown modifications lead to the claimed performance differences. Those leaded capacitors have no effect above 1MHz because their performance is dominated by the inductance of their package. Cable modems operate between 5MHz and 200MHz, where all power supply transients will be served by those little ceramic surface-mount caps the author bypassed with leaded tantalums. Above about 1GHz even those are useless and the power supply decoupling is achieved by the capacitance between the copper planes of the circuit board.

So it kinda looks like racing stripes on a Geo to me. The claimed improvements could be due to almost anything including unplugging the coax and plugging it back in differently, atmospheric conditions, time of day, etc.

tomstokes• 13 years ago

A fun endeavor for sure and I like that he measured before & after results for comparison, but I'm not convinced that he made a real improvement. He notes that he saw "a 1db increase in downstream levels occurred along with a full 1db increase in S/N" but 1dB is well within the margin of error and day-to-day variations, especially for the imprecise internal power measurement circuits of the modem. In fact, simply adjusting and re-tightening your coaxial cable can easily result in more than a 1dB change. Also, I'm not quite sure where he came up with the 1000:1 reduction in power rail noise, because his waveforms show significantly less (although still significant) noise reduction. Finally, having a 10-ping average of 7ms vs 6ms is very much within the margin of error one would expect to see.

If he really wanted to improve the RF performance of his modem, he should have used small, low-value, surface-mount capacitors instead of the large tantalum capacitors with relatively long leads and large electrolytic capacitors. Cable modems and other RF equipment operate at very high frequencies by definition, and tantalum & electrolytic capacitors with long leads are ineffective at these high frequencies. The long leads of the capacitors act as inductors, which will prevent the high frequency fluctuations from ever even reaching the capacitor.

If you look closely, you can see that he added his tantalum capacitors to the already-present small ceramic decoupling capacitors. If he really wanted to improve the RF performance of the circuit, he would have been better off adding more low-value, surface-mount ceramic capacitors in parallel with the pre-existing ones to help with the high-frequency decoupling, where it counts in an RF circuit. If he was really adventurous, he could splice some ferrite beads in series with the supply lines to form an LC filter for even better noise rejection.

Furthermore, he could have taken measures to improve the S/N ratio of the circuit by improving the shielding around the sensitive RF circuits in the middle of the board. You can see where the designer originally made room for a shield "can" to be soldered over the sensitive components, indicated by the exposed copper rectangle around the heatsink area. However, you may get unlucky if your shield can volume has resonant frequencies in the operating range of the circuit which will trigger feedback and render the circuit inoperable. A safer option would be to add some RF-absorbing foam over the sensitive area to absorb the noise.

jonknee• 13 years ago

Finally, having a 10-ping average of 7ms vs 6ms is very much within the margin of error one would expect to see.

It's not like he did a google.com ping on each and called it a day... From the post:

I used Multiping set to do 10 pings a second and chart it. My target was the CMTS IP. I did weeks of charting and then did mods and did more charting. The differences were quite interesting.

Weeks of charting should more than correct for other factors.

alanctgardner2• 13 years ago

But we don't get to see the 'interesting' charts. You're just kind of taking it on faith that he noticed monotonically increasing performance as he soldered more and more expensive caps to his power supply.

raverbashing• 13 years ago

Maybe I understood wrong, but the capacitors he added are to the DC circuit, not the RF circuit

By reducing power noise in that area (including noise created by the consumption variation due to the processor/circuitry itself) you end up reducing some noise in the RF section as well

tonyb• 13 years ago

I concur. He also says " The upstream power decreased by 1.5db indicating that the transmitted signal was being seen more clearly now" but that is inaccurate.

A cable modem's upstream power level has nothing to do with noise. The CMTS will always power adjust the modem so that it hits the CMTS at 0db, it doesn't power adjust based on noise. A 1.5 db change in Upstream SNR would be a different story, but that is a reading that can only be obtained from the CMTS so there isn't any data to backup that claim.

I also would dispute the clam that a modem vendor or MSO is trying to only minimize the cost of the cable modem. If minor internal changes reduced the number of service calls and truck rolls a small increase in upfront cost would quickly be recovered.

X-Istence• 13 years ago

I don't see any waveforms for the noise reduction in the power circuit in the post.

Also, he reduced the noise on the DC part of the circuit, from looking the pictures posted it doesn't look like he touched the RF part at all.

callesgg• 13 years ago

I improved my cable modems signal with 7db by soldering the cable directly to the circuit-bord. Thats a lot more than 1db and the price was essentially zero.

taralx• 13 years ago

This. So much this. Cheap connectors are a huge source of loss in these systems, and eliminating them can be a much bigger win than expensive capacitors.

kabouseng• 13 years ago

If the Power supply he added is switch mode, I hope he checked that the modems' nominal current draw is enough for the switch modes minimum load... Although if it already ran 3 months then it probably is...

Modifying a DOCSIS 3.0 cable modem for high performance

Related Stories