Interesting SNR comparisons: Cisco 877, D-Link DSL-2500U & Telkom Marconi

P

Pada

Executive Member
Joined
Feb 18, 2009
Messages
8,192
Reaction score
6
Location
Stellenbosch
Here are the different SNR margin readings that me and my colleagues got at our office tonight,
with the routers synchronizing at 1024kbps/512kbps <-- down/up rates respectively.
Telkom says that we are about 2.9km from the exchange, and the line attenuation that the routers measured ranged from 45 - 51dB.
Cisco 877: 12.5 / 14.0dB
D-Link DSL-2500U: 21.8 / 14.0dB
Telkom Marconi: 29.9 / 14.0dB​
My guess would be that the Marconi would allow us to sync at 8Mbps/640kbps and it would be stable.

To give some background info:
We upgraded from 512kbps to 4Mbps in like March this year, and then about a month later we asked Telkom to change the sync profile to 2Mbps due to stability reasons.
Then about 1 week ago, we had to change it to 1Mbps due to the SNR margin on the Cisco 877 (supplied by MWeb) being too low for a stable 2Mbps connection even!

We've already contacted Telkom to have a look into the matter, but they were unable to do anything about it, which is why we now resorted to testing different ADSL modems.

I've always thought that the D-Link DSL-2500U with its Broadcom chipset was one of the best routers to use when you're in a situation where your ADSL is barely synchronizing. Clearly the really old Marconi's are still the best by far!

If you guys would want to see screenshots, I can easily provide screenshots of the D-Link & Marconi's web interfaces / MRTG graph of the Cisco's SNR :)
 
Last edited:
2 sides to this one. 1) The actual SNR is the same and will remain the same (provided all the routers use the same Tx power) for all the routers. The only difference is in how it is measured by each router. some routers (and I'm thinking marconi) are particularly rubbish and provide inaccurate values.

2) On the other hand, crappy electronics mean crappy transmit power, so could be that the ones showing a lower SNR actually do have a lower SNR due to their worse electronic design.

Obviously the best way to measure noise levels would be with a proper spectrum analyzer, but that's just too much $$$. There will also be variations with each attempt due to the channel (slightly) varying and the channel estimator (using the transmitted pilot symbols) calculating different values in order to equalize the channel (bit this is a minor variation since the variance of the channel in the phone line is almost minimal, unlike wireless).
 
Well the only real way to test this in your case is to ask telkom to bump your speeds back up again and see if the marconi or the dlink can keep a proper connection. ;)
 
D link is the best router out of those so trust that one.
 
Thanks for the input UnUnOctium & bullfrog.
I know that the SNR on the cable should stay the same, but like you said due to poor electronics, the transmit power as well as the ability to distinguish between the correct & incorrect signals would be affected, resulting in poor SNR margins.

We'll definitely try to negotiate with MWeb to supply us with a better router.
I'm 99% sure that both the D-Link and Marconi routers should be able to stay synchronized at 4Mbps without any problems, where as the Cisco one can't even keep it stable at 2Mbps!

* Update:
@Griz: I think that the Marconi is the best one in terms of SNR, where as the Cisco one is supposed to be the best router of them all.
The Marconi router is really quite nice router, since it has 4x 100Mbps Ethernet ports and it's capable of half-bridge mode too. There's not much more that you would want if you have a MikroTik router for the firewall and routing :)
I would really love to see if the Marconi can sync at 8Mbps and the D-Link at 7Mbps, because those are the rates that I would suspect they would be capable of due to their current SNR values that I read.
 
Last edited:
Yeah, I also wouldn't place too much significance on the reported values (variance). The sustained IP layer throughput attainable is what really counts. On many occasions I've seen an increase in sync rate result in a decease of actual IP throughput (with no upstream congestion/inefficiencies).

It all depends on the quality/condition of the copper loop (incl any external interference) and how the ADSL chipsets involved (both CPE & DSLAM) handle these conditions. Some combinations just handle adverse conditions better than others.
 
So far the SNR values that the D-Link DSL-2500U of mine reported were always spot on on whether the line would sync or not.

I also get that if you have a high sync rate, that you do stand a chance of getting a lower IP layer throughput. It would just be silly to set your sync profile so high that you have no SNR margin to play with.
Like currently we can set our sync profile to 2Mbps at our office and then our ADSL would work with an SNR margin of 8dB instead of the 12.5dB at 1Mbps, but the stability would be noticeably affected.

I would really love to ask Telkom to change the sync profiles and then to use a 10Mbps capable unshaped account to do some speed tests too, to make sure that the IP layer thoughput isn't affected negatively, like you said.
 
Roman4604 said:
Yeah, I also wouldn't place too much significance on the reported values (variance). The sustained IP layer throughput attainable is what really counts. On many occasions I've seen an increase in sync rate result in a decease of actual IP throughput (with no upstream congestion/inefficiencies).

It all depends on the quality/condition of the copper loop (incl any external interference) and how the ADSL chipsets involved (both CPE & DSLAM) handle these conditions. Some combinations just handle adverse conditions better than others.
Click to expand...

<- The ultimate goal, best way to test which router is the best. This all depends on the BER (bit error rate). Routers with bad RSSI measuring circuitry, if it is bad enough, will load the sub-channels incorrectly (i.e. make the constellation too big). This will give you a higher constellation (hence higher sync rate) but the BER will become really high and you'll be getting bit errors left, right and centre.

Quick graph: http://www.mathworks.se/matlabcentral/fx_files/19717/1/ExactBER.jpg <- notice how QPSK always has a lower BER than QAM64 (4 vs 64 constellation size). On the flipside (as is always in engineering), QPSK has a much lower data rate [hence sync rate].
 
A MWeb rep contacted me on myBB and told me that we can use another ADSL router in bridge mode and still let the Cisco router dial the PPPoE and L2TP VPN, which is exactly what we're doing now with our D-Link DSL-2500U + Cisco 877.

So far the line is stable at 4Mbps, using the D-Link DSL-2500U, instead of the Cisco 877 as an ADSL modem.

Here are the line statistics reported by the D-Link:
Code: 
                          DS        US
SNR Margin (dB):          12.5      14.0 
Attenuation (dB):         44.0      25.0 
Output Power (dBm):       9.1       13.0 
Attainable Rate (Kbps):   7808      920 
Rate (Kbps):              4096      512

Here are a few speed test results:
http://www.speedtest.net/
Download Speed: 3.29Mbps
Upload Speed: 0.27 Mbps
Latency: 70 ms

http://speedtest.neotel.co.za/
Download Speed: 3389 kbps (423.6 KB/sec transfer rate)
Upload Speed: 406 kbps (50.8 KB/sec transfer rate)

http://speedtest-rba.isdsl.net/
IS ADSL Local Speed Test to Internet Solutions Rosebank Data Centre
Start Time: 18:14:05
End time: 18:14:06
Download Size: 409600 bytes
Time Delta: 1.0428249835968 seconds
Download Speed: 384 KBytes/s
Line Speed: 3072 kbits/s

http://www.my-speedtest.com/speedtest.htm
Download Speed: 3327 kbps (415.9 KB/sec transfer rate)
Upload Speed: 321 kbps (40.1 KB/sec transfer rate)
Latency: 295 ms
​

I will keep you guys posted on the line stability.
 
This is really interesting. I was just wondering if changing my modem would yield a higher more stable sync. Currently I have an old Dlink 604GT (dont recall exact model) and syncing at 5Mpbs. I am 3.2Km from exchange and have about the same attenuation as you Pada. I did try a netgear834 - but couldnt go beyond a stable 5Mpbs. My 2500u broke, but considering getting another one. I know a number people using this modem, it is pretty solid, especially for around R280!
 
Well, when those modems are made, they all have minor varation in the electronics (ICs etc.). The major ones come in the design, some modems are just better designed than others. The variations will be minor and 99% of the time won't warrant buying a new modem, but if you have some crappy one then it will make a big difference.
 
...apparently (as per Laurie of Cybersmart) the current 10 MB Telkom ADSL sync profile is best utilised by a trend chipset router.
 
who.is.michael said:
...apparently (as per Laurie of Cybersmart) the current 10 MB Telkom ADSL sync profile is best utilised by a trend chipset router.
Click to expand...
Well that is dependent on what make/model of Telkom DSLAM its mated to.

There are differences in performance between the various CPE chipsets and the Ti, Infideon, Centillium etc. chipsets used by Telkom's DSLAMs. When the copper loop is in good condition & distances are shortish, this will be minor, but as conditions degrade the differences become quite noticable.
 
who.is.michael said:
...apparently (as per Laurie of Cybersmart) the current 10 MB Telkom ADSL sync profile is best utilised by a trend chipset router.
Click to expand...

Laurie doesn't know anything. Trend is a developer of security software, not xDSL ICs (or electronics for that matter). A very good way to tell how good your router will be is to google up what IC it uses for the digital signal processing. Google that IC make & model, then compare receiver sensitivity. The lower, the better (i.e. -103 dBm is better than -95 dBm). Ofcourse, this can easily be nullified by a crappy front-end design.
 
Last edited:
UnUnOctium said:
Trend is a developer of security software, not xDSL ICs (or electronics for that matter).
Click to expand...
Think he meant TrendChip (one word), a Taiwanese producer of xDSL chipsets. Most of the lower-end Billion routers use them.
 
Roman4604 said:
Think he meant TrendChip (one word), a Taiwanese producer of xDSL chipsets. Most of the lower-end Billion routers use them.
Click to expand...

Ah, you mean Ralink ;) They're ok but they have nothing on higher end Broacoms in my opinion.
 
Last edited:
Netgear (TexasInstruments) DG834; Linksys AM300; Siemens Gigaset 762SX... best 3 modems I've used. If you're comparing stuff to the Telkom Marconi router, then chances are all those routers being compared are crappy, as the marconi doesn't compare to any of the ones I suggested =)
 
I somehow find it unlikely that the 45 - 51dB range is attributable to purely differences in electronics quality. dB is on a log scale so thats a massive range. Me thinks the routers are gaming the measurements.

So to accurately test them we'd need a line that isn't capable of 10mbps & then push the sync rate up in slow increments until one of the routers bails.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X