dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
7155

CoxVegas
join:2011-07-25
Las Vegas, NV

CoxVegas to xymox1

Member

to xymox1

Re: Upgrade from dpc3000 to dpc3010 and slower upstream speeds

Starting to sound like some of the frankenmodems we have sitting in our modem lab here. (We take some apart, have cables hanging out of some for measurements, etc. etc.)

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

OOoOOo...

Well you have a HUGE advantage. The right test gear AND you can get the datasheets for a BCM3380. Its kinda crazy there is so much secrecy around those chips.

Because the chip is a all-in-one affair there is not much to do. Clean up power supplies is about it. Cant touch RF as thats way too complex and I am sure no QAM expert and dont have the gear to test it.

So all I can do is make the power supplies happier.

Of course you could tweak internal settings, adjust all sorts of stuff internally, do custom firmware... However I dont think there is anything that would really improve performance. The modem is capable of 300Mb/s and at most we will be getting 100Mb/s thru it.

I suppose, if a modem were made from scratch with a far more powerful CPU it might be possible to lower latency a bit because the modem could process packets faster. I wonder how fast it is currently doing small packets ? Say a 64 Byte Frame ? Doing 1518 Byte frames is one thing, 64 Byte Frames is another. Its now possible to buy a $99 router that can do 1 million packets per second at 64 byte frames. The BCM3380 sheet says it can do 500,000 pps but does not specify what size frames and more then likely its much larger frames. The router has dual cores and uses a MIPS64. »www.ubnt.com/edgemax#Edg ··· hardware

It important to remember that pps and frame size are the right way to evaluate a router and im sure that must extend to a cable modem. Just because it can do 300Mb/s at 1518 Byte frames does not mean it can do it at 64 byte frames.. Small frames are actually used quite a bit too, like DNS lookups.

Of course if the modem is being used as a gateway with NAT and a firewall all from the BCM3380, like a SB6580, then there is a bigger problem as the chip now has to also do all the networking as well. I believe in a separate modem and router.

On a whole different note.. I like using those little filters that the Cable Co has that filter out cable TV channels. I have 1 filtering out all that TV junk right before the modem. I feel its best to not feed the modem any more RF then it needs. Those little filters do a great job of removing all the stuff except the cable modem signals If you ask nicely the cable guy might give you one or 2. They are only a 1db insertion loss in the modem signal band but suck out like 50db of the entire specrum of TV channels and digital cable tv. Less stuff into the cable modem RF front end is always good.
xymox1

xymox1

Premium Member


Well after lots of playing around I am pretty solid with a DPC3010. I had some talks with some Cox people and they told me they see firmware updates from Cisco much faster and more often then from Motorola. Also the CMTS gear in my area is Cisco and I would think it might be better to use a cisco modem with a cisco CMTS.

So for those reasons I am going to stick with a DPC3010.

I have now played with 4 DPC3010's and 3 SB6180's. I was interested in what variations occurred between units. This actually did vary. I saw 3 db differences in levels in up and down stream.

In the mod department I decided to replace the 48Mhz crystal with a far better and tighter tollerance Xtal. The crystal in there now is a 100ppm drift part and I decided on using a 10ppm part. 10 times lower drift.

HOWEVER I was shocked when I measured the frequency of the current xtal. It measured 48.000000062 mhz and drifted around a center point that was almost the equal of my rubidium based freq counter ! I tried another dpc3010 and another. They were all very close to this ! This is a startling absolute precision for such a device. The only way this could really be is if there was some way to trim the freq in software in the bcm3380.. A software based calibration. That MUST be what occurs.

So that means during production a calibration must occur with some pretty serious test gear. That brings to mind what else gets calibrated ? Output levels, input levels, QAM settings.. Each modem must have its own unique calibration.

I should would like to be able to send off my modem for a careful calibration somewhere.

So the crystal upgrade is not possible as I cant software calibrate the BCM3380 to the new crystal

Soooooo.....

I think I have reached the end of my mods and my selection journey.

Now if only Cox would turn on 100Mb/s service and 8x4...

JohnDrenZ
Premium Member
join:2000-04-03
New River, AZ

JohnDrenZ

Premium Member

said by xymox1:

Now if only Cox would turn on 100Mb/s service and 8x4...

I couldn't agree anymore but don't see it happening anytime soon with the only competition in the valley being century link!

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ

xymox1

Premium Member

quote:
but don't see it happening anytime soon
Its at most a month or 2 away. Its supposed to be first part of January.

JohnDrenZ
Premium Member
join:2000-04-03
New River, AZ

JohnDrenZ

Premium Member

I didn't know but this would be sweet. Do you know what the upload is going to be bumped up to as well?

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

*Rumor* has it that it will be 100Mb/s down and 10Mb/s up. With speedboost and if you have the ultimate tier of service.

Also I would assume on a good day.

My 65Mb/s down and 5.5 up is rock solid now so I hope that 100/10 will be rock solid as well.

I test using speedtest.net and I use the San Jose server Smug Mug as its the fastest for some reason on the cox network from phoenix.

JohnDrenZ
Premium Member
join:2000-04-03
New River, AZ

JohnDrenZ

Premium Member

I get similar results from San Jose or Scottsdale which is funny.

Lets hope the rumors are true but I would like a slightly higher up load






xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

The San Jose Smug Mug is more consistent if you do it a lot. Which i sure have..

The 10Mb/s is *supposedly* a cap and the only way to get more is to get a business acct where you will *supposedly* be able to get 20Mb/s..

Thats rumor and Cox will not confirm or deny. However that seems to be what other people are reporting in other Cox cites when they switch on 100Mb/s..

JohnDrenZ
Premium Member
join:2000-04-03
New River, AZ

JohnDrenZ

Premium Member

I do have a business account as well so we shall see what Cox plans to do.
bchandler02
join:2011-07-08
Oklahoma City, OK

bchandler02 to xymox1

Member

to xymox1
This was a very interesting post, thanks for sharingso much details!

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

well im not actually done yet. I got a very recent production SB6180 last week. It had some production changes and had some caps right where i decided to put them ! Only 2 tho where I have now put 17 all over in key spots.

But this caused me to look over and test the sb6180 all over again.

In fact I dont think its a equal to the dpc3010. While the RF section *might* be better the poor design of the power supplies to the BCM3380 is a big oversight. The fact they had to add caps later in production - tacked down with silicone - is clearly a rather emergency measure to fix a known issue hurridly rather then do a board redesign.

The complete lack of electrolytic caps on the power supply rails to the chips is almost bizarre. Measuring the noise on these rails reveals what seems clearly like too much noise for proper operation.

I would think this would result in strange behavior, more then normal reboots and hangs.

I also now see another big difference bewteen the dpc3010 and sb6180. Under the bcm3380 chip are small caps on the power leading into the chip. These are standard stuff on any CPU. They keep noise out of the power going to the CPU. On the SB6180 the values of these caps is 10 times smaller then on the dpc3010. In fact they are so small it seems really weird. When you add on to that the lack of electrolytic caps after the regulators then this design seems dangerously noisy.

So my opnion now is that the DPC3010 out of the box is a superior device from a engineering stand point.

Also the DPC3010 is much easier to mod and add further tweaks to to make it even better.

Now BOTH the SB6180 ( and 6580 ) and the dpc3010 ( dpc3825 ) have SERIOUS thermal management issues and run WAY TOO HOT. Ive now clearly seen that older units rather quickly loose performance as the heat bakes the bcm3380 and degrades its RF performance. All the used older modems I have seen all have degraded RF performance and this must be related to the bcm3380 and I feel heat. So the addition of a fan to blow the heat outta these boxes will keep performance up over the life of the unit. Not to mention the huge increase in lifetime cooling the BCM3380 down will provide. The complete stupidity in allowing these super high temps on this chip are just stunning. These are all going to fail.

Overall I dont like the design and layout.
xymox1

xymox1

Premium Member

HOWEVER there is one very important thing that might make the SB6180 a better choice, a better heat sink. The sb6180 is far more likely to survive longer from the excessive heat then the 3010. The thermal management on the 6180/6580 is better and this alone might be a reason to buy it over the 3010...

Adding a small fan to these tho then makes the 3010 the better choice as it has better electrical design and better firmware support.

The above is my final conclusion.

dvd536
as Mr. Pink as they come
Premium Member
join:2001-04-27
Phoenix, AZ

dvd536 to xymox1

Premium Member

to xymox1
current firmware is may 16 2011. your firmware is old.

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ

xymox1

Premium Member

Yes, as is covered in the thread many posts ago my firmware was old. Important to read the whole thread before posting. In fact there is a *** UPDARE *** i put on the very first post that says that.
thedunlap
join:2005-01-31
Naples, FL

1 edit

thedunlap to xymox1

Member

to xymox1
I acquired a DPC3010 when Cox first started offering them for free and my experience was poor.

First, I discovered that the Cox caps were much stricter on the DPC3010 than on the SB6120. In fact, for about a year, I experienced a 20-35Mbit powerboost advantage over the 55mbit limit on the DPC3010. However, that advantage was eventually and rightfully taken away by an update.

Secondly, and most importantly, my DPC3010 was always flakey. Which I found very odd, especially b/c the DPC3010 always had 3-4db advantages (higher SNR 4+db) (lower upstream -4db) compared to my SB6120. I think you hit the nail on the head with your analysis of the the deficient heatsink in the DPC3010. At the time, it was not unheard of to be 1-2+Tbytes/month, and my guess is it overworked the broadcom chip and things got too hot in the DPC3010, which caused it to flake out.

Switching back to the SB6120 always resulted in a rock solid and stable connection in comparison.

floridaguy
@cox.net

floridaguy to xymox1

Anon

to xymox1
At first, I was like this guys a little crazy. But, then I realized he's just my kind of crazy. Awesome thread. Xymox1, if you ever get into router modding, would be an insanely interesting thread. Only thing I wish you would have done differently was figure out a more aesthetically pleasing design scheme for your frankenmodem fan mod.

Btw, xymox1, how often are you calling in to get your account's MAC switched?