How many bits in a megabit?

Trying to get my ISP to tune a very long line with poor and fluctuating SNR
margin.

Line length according to BT technicians who attended the site last Thursday
is 8km

Router is Vigor V2600

Initially:
up speed = 448000
down speed = 544000 but varied
SNR fluctuated from 3.0 dB to 7.0 dB
Line attentuation 73.5 dB

Discussed capping the line speed to make best use of available SNR margin.
ISP suggests capping at 500kbits/sec.

Router now shows:
up speed = 288000
down speed = 576000
SNR fluctuates from 1.0 dB to 4.0 dB
Line attentuation 73.5 dB

Not surprisingly SNR margin is lower and reliability is not so good.

ISP agrees that capping at 500kbits/sec actually means capping at 512
kbits/sec. So why is down speed apparently now capped at 576kbits/sec?

Assuming 1 kbit is actually 1024 bits, and we multiply by 512 we should see
524288 bits/sec

Is my maths wrong, or does the reported speed include some overhead?

--
Graham J

On 01/12/2007 17:38, Graham J wrote:

> Assuming 1 kbit is actually 1024 bits, and we multiply by 512 we should see
> 524288 bits/sec

When used for comms the kilo/mega/giga prefixes are always used to refer
to powers of 10 not powers of 2, so just stick with 1000 in this case.

> Is my maths wrong, or does the reported speed include some overhead?

The 288000/576000 numbers you see is the raw ATM rate your line is
synched at, there is then overhead in using PPPoA, IP and TCP/UDP on top
of that.

> ...ISP agrees that capping at 500kbits/sec actually means capping at 512
> kbits/sec. So why is down speed apparently now capped at 576kbits/sec?
>
> Assuming 1 kbit is actually 1024 bits, and we multiply by 512 we should
> see 524288 bits/sec
>
> Is my maths wrong, or does the reported speed include some overhead?

Data rates are usually quoted in "real" units, so there should be 1,000,000
bits every second in a 1 Mbit/s stream. The problem is, memory sizes are
quoted to the nearest power of 2 for historic reasons, which means that file
sizes are sometimes quoted similarly (but shouldn't be, IMHO), which means
that you can get a misleading result if dividing a file size by time to
determine a data rate.

There is some overhead within the ISP's network, but they should allow for
that when capping. The bottom line is, if they cap you at 500 kbit/s you
should really expect to see at least 500,000 bits.

Having said that, congestion anywhere in the network could reduce that at
any time, so try measuring it at off-peak times and average several
measurements to get any sort of realistic result.

"Phil McKerracher" <(E-Mail Removed)> wrote in message
news:5ih4j.56187$(E-Mail Removed) .uk...
>> ...ISP agrees that capping at 500kbits/sec actually means capping at 512
>> kbits/sec. So why is down speed apparently now capped at 576kbits/sec?
>>
>> Assuming 1 kbit is actually 1024 bits, and we multiply by 512 we should
>> see 524288 bits/sec
>>
>> Is my maths wrong, or does the reported speed include some overhead?
>
> Data rates are usually quoted in "real" units, so there should be
> 1,000,000 bits every second in a 1 Mbit/s stream. The problem is, memory
> sizes are quoted to the nearest power of 2 for historic reasons, which
> means that file sizes are sometimes quoted similarly (but shouldn't be,
> IMHO), which means that you can get a misleading result if dividing a file
> size by time to determine a data rate.
>
> There is some overhead within the ISP's network, but they should allow for
> that when capping. The bottom line is, if they cap you at 500 kbit/s you
> should really expect to see at least 500,000 bits.
>
> Having said that, congestion anywhere in the network could reduce that at
> any time, so try measuring it at off-peak times and average several
> measurements to get any sort of realistic result.


You've missed the point: if the ISP says they capped at 500,000bits why does
the router report MORE than that ????

--
Graham J

Literally one million.

Historically in the IT business it is 2**20 = 1,048,576.

oddly in telecoms a 2 Meg PCM trunk (E1) is 2,048,000 bits per second.

whereas on Ethernet a 10Meg network runs at 10,000,000Hz

By the time you get to a terabyte drive the difference between 10**12 and
2**40 is nearly 10%.



"Graham J" <(E-Mail Removed)> wrote in message
news:47519c29$0$13926$(E-Mail Removed)...
> Trying to get my ISP to tune a very long line with poor and fluctuating
> SNR margin.
>
> Line length according to BT technicians who attended the site last
> Thursday is 8km
>
> Router is Vigor V2600
>
> Initially:
> up speed = 448000
> down speed = 544000 but varied
> SNR fluctuated from 3.0 dB to 7.0 dB
> Line attentuation 73.5 dB
>
> Discussed capping the line speed to make best use of available SNR margin.
> ISP suggests capping at 500kbits/sec.
>
> Router now shows:
> up speed = 288000
> down speed = 576000
> SNR fluctuates from 1.0 dB to 4.0 dB
> Line attentuation 73.5 dB
>
> Not surprisingly SNR margin is lower and reliability is not so good.
>
> ISP agrees that capping at 500kbits/sec actually means capping at 512
> kbits/sec. So why is down speed apparently now capped at 576kbits/sec?
>
> Assuming 1 kbit is actually 1024 bits, and we multiply by 512 we should
> see 524288 bits/sec
>
> Is my maths wrong, or does the reported speed include some overhead?
>
> --
> Graham J
>
>
>
>

Graham J wrote:
> You've missed the point: if the ISP says they capped at 500,000bits why does
> the router report MORE than that ????

You've been regraded from ADSL MAX to "Home 500". This gives a maximum
downstream rate fo 512kbps and upstream of 256kbps. The ATM rate for
that same service is 576kbps down and 288kbps up. The difference is the
ATM overhead.

Andy Burns wrote:
> On 01/12/2007 17:38, Graham J wrote:
>
>> Assuming 1 kbit is actually 1024 bits, and we multiply by 512 we
>> should see 524288 bits/sec
>
> When used for comms the kilo/mega/giga prefixes are always used to refer
> to powers of 10 not powers of 2, so just stick with 1000 in this case.
>
>> Is my maths wrong, or does the reported speed include some overhead?
>
> The 288000/576000 numbers you see is the raw ATM rate

Raw ADSL rate..

your line is
> synched at, there is then overhead in using PPPoA,

ATM, the BRAS throttling on the ATM, ..(if on ADSLMax)


>IP and TCP/UDP on top
> of that.
>
E.G. connected at 576kbps, you could expect around 50-52Kbytes/sec.

On 01/12/2007 18:48, The Natural Philosopher wrote:

> Andy Burns wrote:
>> The 288000/576000 numbers you see is the raw ATM rate
>
> Raw ADSL rate..

difference being?

Graham J wrote:

> You've missed the point: if the ISP says they capped at 500,000bits why does
> the router report MORE than that ????

I think you've missed the point too - the difference is the unusable
protocol overhead.

And if you think that's bad, check out the protocol overheads in 802.11
- and in that case, the headline rate *includes* rather than excludes it...

On Sat, 1 Dec 2007 18:20:11 -0000, "R. Mark Clayton"
<(E-Mail Removed)> wrote:

>Literally one million.
>
>Historically in the IT business it is 2**20 = 1,048,576.
>
>oddly in telecoms a 2 Meg PCM trunk (E1) is 2,048,000 bits per second.

Isn't the difference accounted for by time Slot 16 not being used? or
has my memory gone tits up?
--
Cheers

Peter