Welcome to our new ADSL / VDSL Monitor site.

An invitation for Google to consider our community for their fiber optic network.
Supporting documents (under construction)

May 14, 2007 Breaking news!

Due to multiple problems (mainly recurring soft caps) we decided to cancel our AT&T ADSL service effective May, 15 2007. ADSL monitor will no longer be updated, bug fixes and support will still be honored on a best effort basis.

Back in 2000 and after many years of using GTE's ISDN we decided to switch to ADSL. At that time SBC's enhanced static ADSL plan (6016Kbps/384Kbps) was the best and fastest business/residential plan in the market. Unfortunately AT&T formerly SBC is no longer a leader and falling behind fast to other companies.

Here are some of the problems that made us consider another ISP. High error rates and suspiciously recurring soft caps from 6016Kbps down to 1300Kbps. The most troublesome problem was the soft cap. A soft cap is when your ADSLM modem syncs at your provisioned speed ie. 6Mbps, 3Mbps or 1.5Mbps but your download speed is capped at a much lower speed. This type of cap is usually implemented at the ISP's router by traffic shaping and policing which means that human intervention or bad configuration scripts are responsible for the cap. Our ADSL account was 'soft capped' three times during 2007 (Jan-May period). Perhaps some AT&T greedy gremlins are responsible for sabotaging our ADSL account?

Capped speed test results 1
Capped speed test results 2
High error rates

1. As for today it seems that AT&T's Uverse will not be coming to this area anytime soon (maybe 1-2 years.) UPDATE January 10, 2010 Four years and counting and uverse is still not here, take your time AT&T we're in no hurry.
2. We don't consider uverse to be a business oriented package so it is very unlikely that uverse will offer static IPs.
3. We will be forced to use 2Wire's odd and absurd gateway instead of our cisco equipment. Totally unacceptable for business oriented environments.
4. ADSL2+, VDSL or VDSL2+ look so feeble and uninspiring that we'd rather go to the dentist than decide for any one of these technologies.
5. AT&T FTTH/FTTP (all fiber system) is out of the question and even if AT&T wise up and correct their mistake by deploying fiber all the way to the customer it could take many years to reach our community.
6. U-verse is mainly a TV service with internet access and not the other way around. UPDATE January 10, 2010 We lost track of what uverse is. It is not for business nor it is for residential use it is only available in isolated spots covering less than 5% of any city's area. Low medium class and below will never see uverse deployed near them.
7. With AT&T dedicated prices falling fast we could have contracted a T1 or maybe two bonded T1s but for around $900/mo?.

1. The best TV service we can get. We have 3 HD TVs that we can use all at the same time.
2. VOD (video on demand) and DVR (digital video recorder).
3. Two way cable card equipment just around the corner.
4. A TV service that never goes down.
5. Two TW digital phone lines.
6. UPDATE January 10, 2010 Eagerly waiting for DOCSIS3.0 to hit our area.

We decided for the 10,000/1,000 Kbps plan at $179.17/mo plus a five static IP block since the price is similar to our old AT&T's static elite 6016/768 Kbps plan. We'll keep you informed with our installation experiences.

UPDATE May 15 2007
This morning without hesitation or trying to send us to AT&T's customer retention department AT&T's promptly diconnected our local phone and ADSL service. We never thought it was possible to be 100% AT&T free but here we are. Isn't competition great? A few years back we wouldn't have any other options, today we have TimeWarner's VoIP, Vonage, etc for voice and many other options for internet connectivity. We ordered Time Warner's business plan in the morning on Monday 14, 2007 and we were given an installation date for tomorrow Wednesday May 16 2007.

Some links, the online form, news and tools sections in this web site are not working due to the change of ISP. We hope we will be able to change the configuration in our cisco router in a couple of days. In the mean time we are using our backup residential (8000/384 kbps) roadrunner account for internet connectivity.

UPDATE May 16 2007
The TW sub-contractor arrived here at 8:30AM for the installation. The tech decided to add a second cable drop to connect our office, he finished the installation in 30 minutes and left. One hour later we received an email from TW's tier 3 tech-support with our new IP block and gateway information.

Unlike AT&T PPPoE static accounts TW handle static IP blocks as 'true' statics. All we did was to change the PPPoE configuration in our Cisco router to IRB (integrated routing and bridging) and viola true statics!

The following links show the speed test we ran just after getting our router working and some picures of the cable modem.

Time Warner speed test.
Ambit U10C020 cable modem view 1 view 2 view 3
Ambit web site

So far we are extremely satisfied with Time Warner's business class. The days of thunderstorm, HAM radio, AM station interference and marginal SNR are over. Our cable modem RF statistics are rock solid. The good thing is that upgrading to a higher speed package will be painless since cable is not distance limited like ADSL. AT&T customers with T1s or frame relay should find TW very interesting since the price for the top tier speed package 15,000/2,000 is a lot less than the price of a T1 1,500/1,500.

We are still working in updating our DNS, mail and web configuration so some of the links in this web site will remain 'broken.'

UPDATE May 17 2007
Our web, mail and DNS servers are back online. The online support form, news and tools sections are 100% operational. The migration from AT&T's ADSL static package to Time Warner's business class is now complete.

UPDATE May 20 2007
One very interesting thing is that our new TW business class connection is very quiet by this I mean that hacker attacks and requests are very sporadic. Before AT&T closed the following ports in their static accounts we had hundreds of connection attempts per minute to TCP ports 135, 137, 139, 445 and others in addition to the messenger and VNC attacks. With our new TW business class we are lucky if we see ten in a 1 minute period. Probably this is due to the virgin and pristine condition of TW IP blocks.

UPDATE May 24 2007
One small inconvenience that we've found with Time Warner business class is the way RFC-2317 is implemented. RFC-2317 is a method of delegation for non octet boundary classless in-addr.arpa address spaces. Time Warner will not delegate DNS requests to your own DNS servers, instead you have to call and give them the reverse PTR name entries of your IP block. Basically you have no control over the reverse DNS requests however you have full control of the forward DNS requests.

So far so good, speed tests are consistently 9,887 Kbps down and 982 Kbps up. RF signal statistics are rock solid even during thunderstorm activity. Modem SNR down is in the 38-40 db range with small variations. Latency is low, immunity to noise and interference is superb this is expected as the downstream frequency is so high and much less subject to interference present in lower frequencies.

UPDATE June 4 2007
Time Warner have a nice help and support page explaining the options and features for their business class service: Time Warner business class help and knowledge base

Apparently there was a misundertanding and TW tech support gave us full access to the DNS management page. Instead of having to call tier 3 technical support to make changes to our DNS records now we can login into the above link and manage our domain records.

UPDATE June 13 2007
We have received a couple of emails from curious readers asking us why Time Warner's business and residential class is less prone to interference. In order to answer this question we decided to run some tests in our cable system and compare the results to last year results of our now defunct ADSL line.

Before we get into details let's review the characteristics of some ISP services.

Note 1. The full VDSL allocated spectrum is not used by each subscriber. Instead different sub-bands ( profiles ) are used depending on geographical locations, countries and application ( via central office CO or cabinet RT ).

Note 2. In the US cable use 6 MHz and in Europe 8 MHz wide channels for downstream. Upstream channel bandwidth varies from 0.200 up to 6.4 MHz depending on the upload speed.

For example VDSL plan 998 used by North America, Asia and various European countries assign 1.1 - 3.75 MHz and 5.2 - 8.5 MHz for downstream 1 and 2. Upstream 1 and 2 use 3.75 - 5.2 MHz and 8.5 - 12 MHz.

Image 1. Frequency allocation for xDSL services up to VDSL2+ (lower limit in KHz and upper limit in MHz).

Discussion:

By nature frequencies used in xDSL technologies in the the LF, MF and HF bands are very noisy, by this we mean that in addition to man made radio frequency (RF) broadcasts, natural phenomena also manifests and interferes in these bands. All this and the vulnerability of twisted pair cable for ingress interference make xDSL technologies much more vulnerable compared to cable or fiber.

Table 3 shows a few images of our now defunct ADSL line showing the effects of RFI and electrical noise. During some times of the day the noise surpassed the ADSL signal! Dr. Samuel Sheng in this pdf report talks about this possibility. At the end of page 28 he states "RFI can be significantly larger than the received signal!" we have confirmed this in various ADSL installations. VDSLx should be even more vulnerable at similar distances.

Image 1 of table 3 shows a relatively clean spectrum, we had to hunt for this condition very early in the mornings when nearby air conditioners and electrical motors were not running. We found this unusual calm to be in the order of less than one percent of the day! In this image ADSL occupies one fifth of the first graticule division ~1 MHz bandwidth. The whole VDSL2+ spectrum (profile 30a) occupies the first six graticule divisions which represent the 0 - 30 MHz segment. In reality AT&T's implementation of VDSL2+ occupies only the spectrum from 0 - 8.5MHz.

Notice the strong envelope interference in image 2 of table 3. Peaks are -60db and max at -50db. In communications a strong signal is considered to be -70db ~ -80db! The left series of peaks and valleys in image 1 and 2 of table 3 corresponds to the local AM radio stations. The AM spectrum segment falls right in the top two thirds of the ADSL allocated bandwidth!

Image 3 of table 3 shows the devastating effects of crosstalk in twisted pair cable. Since May 15, 2007 we no longer have adsl at our office but we can still see the foot prints of other adsl modems in our disconnected phone line. This effect is called crosstalk (near and far end). The first graticule square is the adsl modem upstream, the trasmit signal is in the order of -55 dBm. The next series of strong spikes are AM radio stations that completely drown parts of the adsl downstream signal. Interfering local AM radio stations are located at 590, 710, 780, 810, 910, 940, and 970 KHz. This DMT allocation spreadsheet of our defunct adsl service match perfectly the same interference pattern found in our neighbor's adsl modem. The 'DMT Matrix' column represent the number of allocated bits to each tone in a cisco WIC1-ADSL adsl modem, tones above 255 were not used. Tones with severe interference have less or none allocated bits.

The HF band is full of RF interference and noise (International broadcasts, HAM radio operations, etc) but this is not all, twisted pair cable is not very good for high frequency transmission. As the length of the local loop increases so the attenuation at higher frequencies. This is the reason why some uverse subscribers are capable of syncing higher than others. The closer to the distribution device (VRAD) the higher they could sync. This is a technology of haves and have-not's the closer you are to the distribution device the better off your signal will be. Being close to a VRAD or RT or central office does not guarantee immunity to noise. By reading some reports from uverse subscribers we can conclude that loops longer than 1,500 ft will fall into the have-not's category.

Clean 0-50 MHz spectrum	Noisy 0-50 Mhz spectrum	Crosstalk

In contrast to xDSL technologies cable HFC systems are more robust in many aspects. First the downstream frequencies are much more higher in fact in the VHF range making it less vulnerable to RFI present in the lower bands (LF, MF, HF.) Second the high quality of cable installations in the form of a continuously shielded coaxial cable and equipment make cable systems much less vulnerable to ingress RFI. See the first 4 pictures below and compare them to what the telco has to offer you in picture 5. Perhaps your local telco have the phone lines buried in you area and you can't see them but the quality should be the same... from bad to worse. We don't know why there are so many wires coming out from the telco panel in picture 5, we only have three homes in the area. Welcome to bridge taps!

Trunk Bi-directional Amp	Trunk Bi-directional Amp	Trunk Bi-directional Amp	Tap box	Telco twisted pair

Below image 1 of table 4 shows the absence of noise in our assigned 3.2 MHz wide upstream channel. The TDMA signal present in the picture is our cable modem uploading data to a server. Something very important to note is that the noise present in figure 2 of table 3 when the spectrum analyzer was connected to our end of AT&T's twisted pair cable is not showing up when the spectrum analyzer is connected to our end of the coaxial cable (figure 1 of table 4). This prooves the superiority of coaxial cable over twisted pair immunity to ingress RFI.

In image 2 of table 4 we measured the external spectrum using an external antenna, notice how clean is the the 6 MHz wide slot allocated to our modem downstream channel. You can see two signals to the left probably from mobile transmitters or harmonics. To the right is a 6 MHz wide video channel used by our local UHF TV channel 31 (notice power level at -80db).

Image 3 of table 4 shows the spectrum at our end of the coaxial cable, our modem downstream channel frequency is at 567 MHz, notice the power level at -67db (a very strong signal). Cable users using the next channel located at ~ 573 MHz should have no problems in the improbable event of external RFI ingress since the modem signal completely over-powers the external signal as shown in image 2 of table 4.

Cable companies have so much bandwidth available (+800 MHz and in some areas +1GHz) that they can evaluate which channels could be problematic (overlapping with external sources) and completely avoid them.

Clean upstream 34 MHz	At antenna 567 Mhz	At coaxial 567 Mhz

Perhaps the most definitive proof of cable's superiority over xDSL to ingress RFI is the image in table 5. The two spectrum analyzer readings were taken at the same time.

Our spectrum analyzer was set to envelope (to store the maximum reading) 100 KHz resolution bandwidth, 5 MHz per division (50 MHz total), 10 dbM per vertical division. Notice the two large spikes to the left of the top trace they are about 8 and 13 MHz. In addition to the ubiquitous interference in the xDSL assigned spectrum spikes are the type of interference that will make VDSL very problematic for the majority of uverse customers. In the top xDSL trace notice the large spike to the left of the first two graticule lines. Any ADSL modem would experienece high CRC error count rates and possibly periodic loss of sync.

The bottom coaxial cable trace shows no sign of interference except for the first two sub divisions in the first graticule square. These first two sub divisions correspond to the AM broadcasting band. ADSL and ADSL2+ are totally dependant on the AM broadcasting band spectrum segment. Coaxial cable severely attenuates the very strong AM broadcasts (20 ~ 30 dBm). Unlike xDSL, HFC cable systems do not use the AM spectrum segment at all. HFC allocated upstream band starts at 5 MHz (the second graticule square) and channels are usually assigned above 30 MHz. The noise floor is about -100db.

Noise envelope in ADSL line (top) Cable's coaxial (bottom)

UPDATE June 19 2007
A very interesting lightreading article or here talks about the deployment of fiber to the home FTTH/FTTP in cable systems. One comment that caught my attention was "In most cable systems today, the 5MHz to 42MHz portion of the spectrum is dedicated to the upstream, but just more than half of that is considered usable. A mid-split would allow operators to get more upstream by using the cleaner range of 42MHz to 108MHz." I wonder if the 'usable' part corresponds to the upper half (25 ~ 42 MHz) of the allocated spectrum?

Considering that our modem's upstream spectrum is so clean despite external interference what in the world is the author of the lightreading article talking about? Where does all this leave VDSL technologies that depend on this spectrum segment for upstream and downstream? It seems that the future looks bright (at least in non-Verizon Fios territory) for cable companies using today's proven HFC (fiber to the node and last mile via coaxial cable) and preparing to deploy tomorrow's FTTH/FTTP (all fiber) technology as this physorg article or here reports.

UPDATE August 7 2007
In the past couple of weeks we've noticed that our TWBC connection feels faster uploading and downloading files. We did a speed test and we got the following result:

Apparently TWBC made modifications to their business and residential plans and now offer 10 Mbps/512 Kbps as the bare minimum and 15/2 Mbps for their top tier plan. New Time Warner business class tiers TWBC gave us a pleasant surprise with these new speeds. These changes are very interesting from the technological and commercial point of view. What is going to be next? Powerboost technologies? What is going to happen when DOCSIS 3.0 is introduced at the end of 2007 or first half of 2008?

Another speed test.

Thanks to AT&T for letting us use their server.

Speakeasy speed test server.

TIPS and TRICKS June 28 2007
If in addition to your cable modem you use a router like a cisco or xyzel or any other router capable of supporting multiple public IP addresses you can configure your router to use all your public IP addresses (including your network and broadcast IPs) with the exception of the gateway address. For example if TW assigned you 5 static IPs your usable IP block and gateway will look something similar to this: 68.x.x.178 to 68.x.x.182 net mask 255.255.255.248 and gateway 68.x.x.177. The following configuration is for a cisco 2611XM with two fast ethernet ports (inside and outside interfaces). Other routers should have a similar configuration. This can be accomplished if and only if the cable modem routes the broadcast and network addresses to the inside. We discovered this behavior by chance while programming our router.

Under the default configuration the outside interface must be assigned one IP address from our public IP block and the provided netmask. In this case the router will only be able to use 5 public IPs (with one public IP used by the outside interface) but if you change the net mask of the outside interface to a class C network the router then will be able to use all the public IPs with the exception of the gateway address. Under this configuration the broadcast and network addresses will change to 68.x.x.255 and 68.x.x.0. Notice that with the default configuration you can't assign 68.x.x.176 to the outside interface because it conflicts with the netmask (68.x.x.176 is the network address under the original configuration the same applies to the broadcast address which is 68.x.x.183).

Now there is the issue of routing the rest of the class C addresses that doesn't belong to us (68.x.x.0 ~ 68.x.x.175 and 68.x.x.184 ~ 68.x.x.255) Our cisco router assumes that the whole class C network is directly connected to the outside interface, this assumption is not correct. The addresses that doesn't belong to us are outside our cable modem. To correct his we need cisco IOS 'ip route' command to route these addresses trough our gateway. We subdivided the address range in 5 different groups.

By using this scheme we are able to use our originally assigned network and broadcast IP address as regular IPs. We assigned the first IP address (68.x.x.176) to the cisco's outside interface so any hosts without a static nat translation will use this IP. We are still evaluating the implications of this configuration.

Please use our online form to report any errors or problems with this page.