About a month ago now, I interviewed at Google in Mountain View, California (I can’t elaborate at this time), and I had the privilege of flying with Southwest who offers Wifi on their long flights across the US.  I thought I’d append to my original technical comparison blog with some additional speed tests.  Southwest uses Hughesnet for their in-air internet service, so I expect that these results would be very similar to Hughesnet on the ground.

I’d first like to draw your attention to some simple pings (a test that tells us the time it takes for an internet data packet to travel to and from a computer on the internet / network).  The pings took place somewhere over Nebraska.  Matthouse.us is located in Atlanta, and int.matthouse.us is located in Chicago.  Both of these servers are serviced by a mixture of Level 3, Cogent, and other smaller providers (redundant links).

PING matthouse.us (98.142.216.195): 56 data bytes
64 bytes from 98.142.216.195: icmp_seq=0 ttl=44 time=835.283 ms
64 bytes from 98.142.216.195: icmp_seq=1 ttl=44 time=879.877 ms
64 bytes from 98.142.216.195: icmp_seq=2 ttl=44 time=924.125 ms
— matthouse.us ping statistics —
3 packets transmitted, 3 packets received, 0% packet loss
round-trip min/avg/max/stddev = 835.283/879.762/924.125/36.270 ms

PING int.matthouse.us (206.217.130.77): 56 data bytes
64 bytes from 206.217.130.77: icmp_seq=0 ttl=40 time=835.422 ms
Request timeout for icmp_seq 1
64 bytes from 206.217.130.77: icmp_seq=1 ttl=40 time=1153.039 ms
64 bytes from 206.217.130.77: icmp_seq=2 ttl=40 time=923.749 ms
64 bytes from 206.217.130.77: icmp_seq=3 ttl=40 time=758.116 ms
— int.matthouse.us ping statistics —
5 packets transmitted, 4 packets received, 20.0% packet loss
round-trip min/avg/max/stddev = 758.116/917.582/1153.039/148.035 ms

The speed test on Southwest around the time I took the above pings:

Overall, this isn’t bad, considering that I was going at 500mph or so at about 37,000 feet.

Finally, I thought it would be cool to share a speed test directly from Google’s wireless network in Mountain View!  I was impressed by their network, although I wasn’t surprised.  Google invests a lot into their network since they have a lot of traffic that they have to deal with!

Looking back at my history of internet usage, I can say that I have exhausted every possible connection that I can obtain.  I’m hoping that by writing this blog, I will save someone the agony that I have gone through in the past 7 years and provide a quicker route to getting usable internet (as I like to call it).  Before I even start, if you can get Cable, DSL, or FIOS, or another form of land based wired internet connectivity don’t even bother reading this blog since I don’t have them and can’t explain why they are better with proof (they just are).

I would like to start out with Dialup internet. Dialup is often the only form of affordable internet access for people in rural areas and this is why I have tried it and am blogging about it.  An example of a quality dialup provider is http://copper.net

First the disadvantages of dialup : Dialup has a major disadvantage of speed.  Dialup is very slow to connect (1 minute to dialup to the remote location) and is also very slow speedwise (56Kbps under ideal conditions).  In addition, Dialup uses your existing telephone line, therefore, while on dialup, your phone will be tied up and appear busy to outside callers.  In addition, Dialup requires that your phone line be fairly clear.  Unfortunately the phone line to our house is very old (ran circa 1930) and runs under a modern day wetland area (I would call it a swamp).  Obviously our phone line has a lot of noise in it which prevents dialup from connecting any faster than 24Kbps (in the city, I get 48Kbps).  Dialup has fairly good error control even though our phone line has a lot of noise and crackle in it.  Unfortunately, the noise in the line can be so excessive in spurts that it causes the connection to be lost.  Because of this, I was only able to get my own ping test, I had to borrow speedtests from elsewhere.

The upside to dialup is that it has a decent latency because it is ran via land.  Latency is the time it takes to send a signal to a remote computer and back over the connection, this is measured in milliseconds (ms) and is called a ping.  If you wish to try to ping google for instance, open up a command prompt and type “ping www.google.com”.  This will give you the latency of your test (it does 4 trials in Windows XP).  Latency on Dialup is effected by telephone line noise and connection distance to the local internet service provider (isp).

Now for some details on my testing for dialup:

This connection is connected to netzero (a free account that I had laying around).  It was connected at 24Kbps and I performed a latency test to google.  My average latency on dialup through my line was 258ms.  This isn’t bad for low latency applications such as SSH (remote linux terminal, completely text based) and VPN (virtual private networking) which both require very low latencies (below 300ms) to even connect.

Now comes the downfall.  I could not acquire a speedtest because I was disconnected from the internet (it took well over a half hour to complete which my connection couldn’t support).  I had to find speedtests which was a very hard task.  This is very close to what I would get if I let my speedtest complete without error.

This kind of speed is just useless for any kind of internet surfing.  Most webpages today are 100KB or more, and on this connection, 100KB will download in about 50 seconds to a minute.  This is the biggest downfall to dialup!

So now, lets move onto satellite internet.

I have wildblue satellite internet at home.  This is the connection that I have been using for the past few years.  It advertises 512Kbps down, 128Kbps up.

The advantages to Satellite Internet:

1. Stability – The connection rarely goes down.  Reasons for the connection going out is bad weather at my location or the WildBlue ground station.  Or if someone at WildBlue trips over the main internet line *ha ha*, but I bet it has happened a few times when our internet was out for no reason.
2. Always on and doesn’t tie up the phone line
3. It is fairly fast for web browsing (not considering latency just yet)

The disadvantages:

1. Latency will kill you – I will detail a bit about the high latency below.
2. Wildblue has several ground stations, but each ground station can’t take over another ground station’s frequency on the satellite, so if the weather there is bad, you won’t have a connection.  This happens frequently in the summer.
3. During peak usage times the connection is much worse than dialup (higher latency, slower speed).  I will detail this below also!
4. Your bandwidth (total data transfer allowance) is limited severely compared to land based solutions (7.5GB of downloads per month on the lowest plan or about 10 movie downloads at 800mb each)
5. Wildblue will cost an arm and a leg- the lowest priced package is 50 bucks a month

Hughesnet to Wildblue.  I’m using Wildblue’s value pack which offers 512Kbps down, 128Kbps with 7.5GB download and 2GB of uploads per month.  This is 50 dollars per month and if I go over the download / upload limit, I can be terminated by Wildblue and pay fees to them for this if I ever do it again (the first time is a warning). I’ve always been very careful of this contract limitation to prevent getting billed heavily.  Wildblue will also slow your speed down until you go back to 80% usage.  This means that if you downloaded 8GB the first day of service, you would be billed for the rest of that month and not be able to use your service (it is slowed down to dialup speeds with high latency), plus you can be terminated by Wildblue if it ever happens again.  Imagine 30 days waiting for that 8GB to disappear off your record before you get your speed back

On Hughesnet, their lowest plan will provide 1024Kbps down, 128Kbps up with a different limit system.  They limit you per day to 250MB of download and 50MB of upload (I think).  This also varys by plan.  If you go over this limit during the day, they will slow you down for a day then restore it with no contract violations or service outages.  Unlike Wildblue this is daily so you aren’t slowed down for possibly a full month.  In addition, Hughesnet allows you to download without this limit between 2am and 7am every morning when their system is underused by a lot. Also, hughesnet is more resistant to rain fade (weather effecting the signal strength and connectivity) and allows you to see how strong your signal is unlike wildblue.  I also found lower ping times on Hughesnet (800ms vs the wildblue average of 1300ms).

Unfortunately though, hughesnet and wildblue still suffer from peak time usage slowdowns every night and on weekends.

Lets show some details.  My plan is 512Kbps down, 128Kbps up remember (50 bucks per month).  First, I will show my latency tests.  The first series of pings was performed at 9:30pm at night, I left the computer on overnight with this window open to take the 2nd series of the ping test from non peak usage time (this was taken in the morning.  I used NASA.GOV because it is a government controlled server and *should* have really low ping response times for land based connections.

As you can see, during the morning the average ping is much lower than during the evening during peak usage hours.  now lets goto the speedtests:

This above one is for a peak usage time connection.  Notice I’m only getting 180Kbps down, 30Kbps up (compared to my advertised speed of 512Kbps down, 128Kbps up).  This connection is useless for even surfing because the latency is much higher with a slower speed (see above).

Here is a speedtest above from the morning when there is little usage or normal usage.  Notice that I’m getting 530Kbps down, 60Kbps up.  Still not a good upload speed, but its better than 30Kbps which is comparable to dialup!  Also notice the pings, this isn’t quite true because these were taken around the same time of the day (different days though) when I did the latency ping test on the command prompt above.

Now lets move onto EVDO from Verizon (I know, this is a long blog)

Lets first start out with the advantages and disadvantages (for both rev 0 and rev a since they are both good)

Advantages:

1. Land based wireless communication, but its more land based and latency is much lower and comparable to DSL and Cable (especially under revision A, revision 0 matches Dialup latency more).
2. Speed is quite good.  It can exceed DSL speeds with up to 3.1Mbps download on Rev A, 1.8Mbps down on Rev 0 (1024 Kbps = 1Mbps).  Normally DSL gets about 1.5Mbps max download.  The upload on Rev a can reach 800kbps and on Rev 0, 128kbps.
3. Stability – if you get a good signal, this will rarely drop unless Verizon or the provider is doing tower repairs and such.
4. Mobile, this can move with you

Disadvantages:

1. Expensive.  EVDO can cost 60 bucks per month on Verizon.
2. Low Bandwidth allowance then excessive overage charges (5GB of downloading per month, then 5 to 10 cents per MegaByte (MB) over)
3. not the best replacement to DSL or Cable (but the only option in the country)
4. Works only if you get a cell signal on EV.  If you don’t get 3 bars, your connection will suck.  You need 4 bars for a decent connection.  There are cellphone boosters like the one I got that give excellent reception (see part 2 of my evdo quest).

So, I’m paying my heart out for an EVDO connection through Verizon and I love my connection speed thus far.  It has excellent latency and fair reliability.  The speed that comes with it is merely a bonus.

So here are some ping tests:

The above is from EVDO rev 0 with a perfect signal.  Notice the average latency beats dialup and satellite already at 236ms!  Lets do a speedtest:

Once again, much faster than Hughesnet, Wildblue, Dialup or any other connection that I have seen available to me!  And the really great part is, I haven’t even touched EVDO revision A yet.  Lets do that now, shall we?

Notice the average ping of 102ms.  DSL and Cable average at 30ms to 100ms, so I’d say this is a latent land connection.  EVDO rev a rocks, no?  Lets see a speedtest:

The speedtest above shows that rev a is a bit better than Rev 0!  This is now my primary connection method and I use satellite as a backup.  Maybe some day we will get DSL or something else land run, but until then, I’m quite happy with this setup.

Hopefully this will give you some insight into different connections that I use and sort of show why I like EvDo the best.  I’m sure something faster will come eventually, but until then, I’m now focused on stability of the connection to the Verizon tower, not so much the speed aspect.

Before I even start, I need to define a few terms that might confuse you.  I will use Kbps and Mbps quite often in this article.  These stand for Kilo/Mega bits per second.  These are used to measure connection speed.  DON’T CONFUSE THESE WITH KB/S AND MB/S which are your actual download speeds (KiloBytes/MegaBytes per second).

I live approximately 6 miles from the closest cellphone tower that I can get a half way decent link to.  There is another in the opposite direction about 5 miles away, but in our location, we could never get over the hill and trees to establish a connection of over 1 bar even with a specialized booster system.  I also should state that we live in a valley almost, although there is a passage through the hills that is relatively at the same level above sea level, this is the general direction the cell tower that we connect to is in.   If we were to connect to the other tower, we would have to bypass the highest hill in our entire county to get to the tower in that direction (Cattaraugus County, NY).

Before I get into any details, I connect to the Verizon EvDo network.  AT&T does not service our general area, Sprint roams in our area.  I’m not sure, but on TMobile, I get 1 bar in perfect conditions, so that may be an AT&T tower roaming.  I will leave that up to an expert to comment on if one wishes to do so.

Verizon EVDO is a specialized data network developed for cell phones to connect to the internet.  EVDO stands for Evolution-Data Optimized.  As you might guess, its made for high speed internet and is similar to the 3G technology offered by AT&T for the IPhone and other similar phones.

Verizon EvDo has a few different revisions, plus there is another 1x voice network that can handle internet data (although very slow, it does work).  Verizon’s 1x network that supports both voice and data is able to support a 128Kbps data speed (both uploading and downloading) and is suitable for voice.  The 1x network in many instances is stronger and able to maintain a connection in weaker signal areas, therefore in an area like mine, I normally connect to the 1x network without any kind of boost to the signal or locating my cellphone in good signal spots.

Verizon’s Evdo has 2 current Revisions that I know of.  The Revision 0 standard can handle a 128Kbps upload speed and 1.8Mbps download speed in optimal signal conditions (about 16KB/s up, 150KB/s down).   This is slightly older though, the current standard is EvDo Rev A.  This standard can handle 2.1Mbps upload, 3.1Mbps download under optimal conditions and with appropriate equipment (or about 268KB/s up, 400KB/s down).  In most conditions, you won’t ever achieve these speeds unless you are right next to the tower, but in most cities, you should achieve a very fast connection because you need an ideal signal.

For the past year, I have been using a tethering plan to get data on the Verizon network with my EvDo Rev 0 phone.  I noticed at the end of last summer, I was always getting on the 1x network, but I could always go outside and fix that problem.  This year, I know I have all summer, and I want decent internet access for voice over ip, etc.

Knowing that boosters existed, I had my father pick one up (on my tab).  I had a Zboost yx500 cel (800mhz only) booster here and we put the external antenna outside at the highest point.  Like the product said, it did give me 4 bars of signal inside, but it was only within an inch of the unit.  Data also wasn’t much better than what I could get when taking my phone outside with a computer.  I was let down, and had the zboost returned a day later.

Now comes the amazing part of using the dish: I was looking online at more powerful boosters and discovered that many people use parabolic antennas for reception.  We have a huge 10′ cband satellite dish in the back yard that has been unhooked for at least 5 years now.   We were actually considering scrapping it for junk money!  I’m glad we didn’t do that now.  Anyways, I wondered for a few days if I could use that to get a decent signal. I was bored about a week ago so I thought, why not take my vx6800 evdo rev a phone out to the dish and put it in front of it and try to focus a signal to it.  Surprisingly although the 6800 has a horrible antenna (no signal most places here), I got 1 bar with the dish pointed way off, so I took some tools out and pointed the dish with my father the other day.  To our pleasant surprise, I started getting 4 bars and an excellent signal when we got the dish aligned appropriately and used its focal length to focus a signal onto the phone. To put the signal in perspective (using ##debug on the phone pad to get the field test screen), I went from a -110db signal to a -70db signal.  Thats about a 40db gain, stronger than most amplifiers will guarantee.

To point the satellite dish, I sort of cheated using GPS.  I know where both towers are exactly, so I got the altitude and coordinates of the tower.  I then used the GPS to point the satellite dish correctly onto the tower.  It took a little time (vs a guess and check approach), but it provided a really good signal on our first try.

Knowing that the zboost would bring the signal inside (and boost it slightly), I had it brought back and we mounted the antenna correctly onto the dish.  Since the old dish had RG6 satellite cable already ran into the house, we used the existing low loss cable to connect the booster to the antenna (instead of the RG6 cable provided with it).

As the box says, the signal did come into the house as 4 bars within 4 feet of the base unit.  I found that 2 bars were still visible within 25 feet of the unit (before I go behind a lot of walls).   Unlike Jordan (In his blog here) and using the yx500 single band unit, my signal did go further.  I know this is because I have a very strong signal outside for it unlike Jordan had. I would still say the unit is a bit far fetched on its estimates of signal coverage area, but its more than fine for me (I don’t even use a cell phone for voice in my room, although I can now).  So I recommend if you get a zboost system, make sure you have a very strong signal outside.

Now for the boost that the zboost provided.  Inside within an inch of the unit, my previous -115db signal is now -35 to -40db.  This is better than it would be in the city (I get about -70 on that 6800).  I’m very happy with this setup, although I may consider finding ways of pushing the signal further into the house so I can talk in another room without problems if I ever would want to.   I also can use 2 data connections at maximum speed at the same time.  I will compare my exact speed results, but for now, I will say that I am averaging about 2Mbps down, .4Mbps up on my evdo rev a connection.

So now for photos of my setup:

This is by far the best connection to the internet I’ve ever had and my next blog in the next day or two will explain exactly why I say this!  So the next time you have the signal blues, try using a big dish to focus directly on your signal, you might be surprised, I know I was!

Links to other blogs that also may interest you:

• 10 Ways to hide your Satellite Dish