I have a question about how download DSL speed is measured.
Previously I had 1.5Mb DSL service. When I went to a site such as
SpeakEasy to measure my d/l
speed, it would show approximately 1.1 - 1.3 Mb which is what I would
expect. However when I would use a .bat script in a WinXP console
window to measure the speed I would only calculate about half that
The .bat script works as follows:
1) turn on a timer
2) use ftp to download a 1.5MB file from the server of my ISP.
3) turn off timer & save elapsed time
4) calculate d/l speed from the known file size and d/l time
I've recently upgraded to 3Mb service and I have seen my d/l speeds
double. However I still see the same relative disparity between the
two methods of d/l speed measurement as described above; i.e. the
speed measured at SpeakEasy is about 2.9Mb/sec but the speed
calculated by the script is about 1/2 that.
Can anyone offer any insight as to why the large disparity?
You're including all the Windows and program overhead in your timing.
It takes time for Windows to load and close your FTP program, and a
significant amount of time for your FTP program to look up and connect
to the remote server, negotiate with the remote server to open the file
and then save the downloaded file on your PC at the end.
I suppose you could try to swamp the overhead by downloading a very
large file, but the overhead will still be there.
Testing can be useful if you have reason to think that something funny
is going on upstream.
My router usually trains up at 6592/896 Kbps and I usually get 5Mbps
downloads from close-in, capable sites.
When things seemed to go sour a few months ago, I was able to use test
results to first convince myself that something was going on and to get
The Phone Company and my ISP to look into things.
Unbelievable as it seems, The Phone Company actually admitted that
they'd over-committed the DSLAM at the local central office, and my ISP
did some load balancing on their circuits and things have been fine
Unfortunately, you are chasing shadows there.....;-)
There are so many variables that are added in there it almost seems
pointless to do any "testing" because all testing is moot. If you
REALLY want to know what your Hardware NEGOTIATED Line rate is, then
let your Hardware tell you what speed it obtained. For a DSL service,
most DSL H/W provides a status display similar to this one (please
excuse the formatting) -
ATU-R (DS) ATU-C (US)
Modem Status: Showtime (DMTDSL_SHOWTIME)
DSL Mode: ITU G.992.1 (G.DMT)
ITU STD NUM: 0x01 0x01
Vendor ID: 'ALCB' 'BDCM'
Vendor Specific: 0x0000 0x9191
Vendor Country: 0x00 0xB5
Capacity Used: 99% 39%
Noise Margin: 12.0 dB 29.0 dB
Output Power: 16.5 dBm 11.5 dBm
Attenuation: 57.5 dB 31.5 dB
Defect Status: None None
Last Fail Code: None
Selftest Result: 0x49
Interrupts: 52250 (78 spurious)
SW Version: 3.8129
FW Version: 0x1A04
Interleave Fast Interleave
Speed (kbps): 1344 0 160
Reed-Solomon EC: 499 0 0
CRC Errors: 7 0 0
Header Errors: 10 0 0
Bit Errors: 0 0
BER Valid sec: 0 0
BER Invalid sec: 0 0
The line to note is this one -
Speed (kbps): 1344 0 160
Also useful is tihs one -
Noise Margin: 12.0 dB
The SPEED tells me that this ADSL line has negotiated a DS
(DOWNSTREAM) connection speed of 1.344 Mb, and an US (UPSTREAM)
connection speed of 160Kbps, and the NOISE MARGIN tells me that the
Router NEGOTIATE a rate that is likely to be LOWER than the rate that
is possible ON THIS LINE. Ideally, the NOISE MARGIN should be
somewhere around 4 - 8db to be closer to optimal (note there is no 1
guaranteed value that is truly optimal) This can be for a number of
reasons, some inside your (or your ISP) control, some outside you
control. EG Currently its extremely wet outside and such weather can
play havoc with the performance of the Copper cable in our area, so
the displayed rate is likely to be lower what than I am normally able
to get (I normally get about 2.0Mb Down, 160 Up), which is close to my
contracted service, I get less due to my large distance from the
That is the most accurate form of line speed between YOUR house and
your CONNECTION point and it is EXACTLY correct! Nothing else is in
the equation. Thre are 2 things to note -
1. This is an ATM (DSL) line rate so the framing on that line will
be in ATM cell (block) size of 53 Bytes, using a 48 Byte payload,
meaning effectively the USER visible operating speed is 1.334Mb LESS
about 5/53 or something like 9-10%.
2. Usually, DSl H/W continually evaluates its connection parameters
and adjusts the line signaling to obtain values that usually work
best. It is CRITICAL to understand that copper in the ground is NOT a
static environment, there are MANY variables that mean it's
performance can and will vary, sometimes by wide margins. ALL of this
is usually OUTSIDE your control, all a user can do is ensure they
operate their hardware within its working parameters, and these days
there is very little required from the user in this respect.
For really accurate readings you are adding in too many uncontrollable
variables into your figures to worry about trivia like this. Please
dont waste your time...
HOWEVER, having said all that the REAL issue for users is the
bandwidth between YOUR machine and the destination Server for EACH
site you are trying to communicate with, with whatever SERVICE you are
trying to reach. All thee factors will affect performance and are near
impossible to calculate on a public network such as the internet.
The reality is that it all comes down to the question "Is your ISP
delivering the service they promised?",and that can be a really tough
question to PROVE, let alone answer.
Simple, the methodology used for calculating these figures is flawed,
and its like a cat chasing its own tail. This does not stop cats
desiring to catch their own tail.....;-)
I hope this helps.........................pk.