I have a 56k modem that usually connects at 49.2Kbps. Divide 49,200 bits by 8 and my maximum download speed should be 6.15KBps.
However, when I start a download, it’s not uncommon for the windows download dialoge box to initially report between 9 and 13KBps, then throttle down to about 5 or so after 15 or 20 seconds. This is even more common in Win2K than it was in Win98.
So what’s the deal? Is the data actually travelling that fast or is it a flaw in how windows is determining the throughput?
Someone once told me that there were “FCC packets” that somehow limit the throughput on telephone lines so that data packets do not take up all the bandwidth and crowd out voice packets, and it takes a few seconds for them to “kick in” so to speak.
Is there any truth to this? If so, it still doesn’t explain how the download could exceed the connection speed to my ISP. Can anyone explain?
On a 56K modem you should generally see an average speed of around 4-4.5K/second. Connection speeds should range roughly between 45K-50K/second.
This speed can sometimes be ‘increased’ if compression is happening so you get more data in a smaller size. The data is expanded once it reaches your computer.
This is rare however since most stuff on the internet is generally as compressed as it’ll ever be so there’s no room for more compression on the way to your computer.
The next thing that happens is there are frequently errors in transmission. This is quite common and your modem is built to take care of it. Basically it sends a message back to the sender saying I didn’t quite get that last bit would you mind resending it.
In the end your max speed hovers around 4 - 4.5 as mentioned above. This is the average speed of download and generally you’ll find that Windows reports pretty close to what you’re getting. In my experience I have found Netscape to do a better job than Internet Explorer in telling you what you’re getting.
As for the FCC thing if you look at the box of any 56K modem somewhere in small print you’ll see a notice that the FCC limits the speed of modems to 52K/second. The FCC won’t allow devices with higher power (it’s not a ‘packet’ thing) to connect to standard phone lines. Presumably your modem is capable of up to 56K but you will never see it.
At a guess I’d say there was some compression happening. As I said it doesn’t happen often but sometimes it can. It also usually requires similar (read same manufacturer if not actually the same model) modems on each end.
If you go into your modem properties you’ll probably see your modem set to 115,000K. By doing this your are telling your computer to go ahead and use compression if it can even though it ‘knows’ it only has a 56K modem.
Er, typical modem connections add one start bit and one stop bit for each byte. Each byte requires 10 bits of data. A packet based protocol, which you use to access the Internet, adds a little more overhead but it’s difficult to calculate how much. So, divide your connection speed by 10 and round down.
Nope, I just picked that site at random, and started the download to make a screen shot. This occurs on any fast site. Some sites start off as fast as 20K or so.
Jon,
Ok, I see your point. But not adding the overhead just makes it more of a mystery.
w/r/t the unusually high kbps number ramping down, consider the following:
The number is not calculated directly from monitoring the line, but is calculated based on total # of bytes transfered (sp?) divided by elapsed time.
The download starts before the window in your screenshot is drawn.
The timer doesn’t start until after the window comes up (I THINK. Anyway, the timer definitely starts AFTER the download has already begun).
Therefore,
If you get a fast initial throughput, when the first kbps value is calculated, you get an artificially high number. For long downloads, the number eventually asymptotically approaches reality as the offset between download start and timer start becomes a smaller and smaller chuck of the total time. For short downloads, it probably is never accurate.
also
You may notice that, during a long download, there might be a temporary pause in the data stream coming in. Suddenly the kbps counter begins to fall rapidly, because of 1).
Actually, divide 49,200 by 8 and yes, you get 6,150 Bps (Bytes per second). However, to get to KB per second you need to divide again, not by 1,000 but by 1,024 (1KB = 1,024 B in bin-speak). Therefore, your top rate should only be 6.00 KBps.
Lots of windows programs are notorious for giving a really inflated download speed at first, it has to do with working with a small sample of data divided by time. Or the timer is slightly off. It also might have to do with lag of windows deciding where to put the file, checking for associations, drawing the box etc. Maybe that’s whats throwing off the timer. Your real download time will be at what you’re getting towards the end, its impossible to get 12kB/sec on a dialup downloading zipped files.
As to why you’re got getting 6kB/sec downloads, you’re wasting lots of data in the form of (nice diagram here)packet headers. Windows is showing you the data its getting from the TCP/IP stack, not how many bits are flowing down the line. Maybe you could get 6kB/sec if you dialed up your ISP not using PPP or SLIP, but with a serial terminal like Hyperterminal and downloading using a different file transfer protocal like xmodem, kermit, zmodem or something with little or no headers. Unless you’re on a college system you’ll probably be hard pressed to find an ISP that’ll offer this old BBS style dial-up.
As pointed by other posters your mis-leadingly high transfer rate is a windows artifact, not the promised land of throughput receding away. Windows will via various methods read ahead and cache portions of header data prior to download so it is common to get an interesting but misleading speed burst at the start of Dl’ing.
Ok then, why do I see the transfer rate drop from 9K to 7K then back up to 9K? If the number was artificially high it would never exceed its last highest point.
I really don’t know. It does usually behave as you mention
with some minor up/down fluctuations as the transfer rate descends to some baseline point. I believe that the up/down jumping around is related to the real variations in DL speed that occurs while the calculation of the transfer rate for the immediately available portion of the data that is/was cached is blended into the descending transfer rate.
I don’t know how the transfer rate calculation code operates so this is my best guess regarding the jumping. To give the perception of improved speed Windows/IE 5 caches many more things than most people realize. I would be glad to stand corrected if there is a more correct/plausible explanation for the jumping.
The 115K setting has nothing to do with compression. That is the speed of the port, i.e. how fast the computer talks to the modem. The modem speed of 56k is how fast it can talk to other modems (on the phone line). Setting the port to 115K will not have any affect on your download speed, as long as it is kept faster than or equal to your modem speed.
I have been wondering something similar. According to FCC regulations, modems can not provide speeds faster than 56Kbps. However, I am not in the US, and don’t think I need to worry about FCC rules. My modem driver has only one option: the country. It says:
“Selecting a country other than the one in which you are currently located may cause your modem to be configured in a way that violates the telecommunication regulations/laws of that country.”
Well, my country is not in their list. So which country has the most liberal regulations in regard to modem speed?
