What causes the 56kbps limit on dial-up internet connections?
gluc...@excite.com
What are the physical* reasons dial-up speeds cannot go above 56kbps?
One factor is that the phone line cuts of frequencies below 300 Hz and
above 3,000 Hz. What are the other physical causes of this limit?
Would this limit still exist if it weren't for the aforementioned
frequency-cutoffs?
bps = baud X number of bits per baud.
If only 1 baud is used, what is the maximum-bits-per-baud that can be
used on a phone line without the frequency of the analog electric
signal exceeding 3,000 Hz?
*NOTE: By "physical", I am referring to causes not associated with
legal regulations - such as limitations place by the FCC and
governments.
Thanks,
Radium
Jeff Liebermann
>What are the physical* reasons dial-up speeds cannot go above 56kbps?
>One factor is that the phone line cuts of frequencies below 300 Hz and
>above 3,000 Hz. What are the other physical causes of this limit?
>Would this limit still exist if it weren't for the aforementioned
>frequency-cutoffs?
The speed limit is the point where the error rate drops to the point
where communications is futile within the audio bandwidth. The V.90
spec allows 56Kbits/sec download, but only 33.6Kbits/sec upload.
(V.92 does 48Kbit/sec upload).
>bps = baud X number of bits per baud.
Yep. It would be 64Kbits/sec were it not for telco bit-robbing, where
every 6th frame is stolen for signaling. The audio channel can handle
the bandwidth, but when it hits the DS0 digital channels, bit robbing
limits the bandwidth to 56K. Therefore, you only get:
8Kbits/sec * 7 bits = 56 Kbits/sec.
What do you mean by a "band"? Are you perhaps thinking of bits per
baud?
>If only 1 baud is used, what is the maximum-bits-per-baud that can be
>used on a phone line without the frequency of the analog electric
>signal exceeding 3,000 Hz?
>
>*NOTE: By "physical", I am referring to causes not associated with
>legal regulations - such as limitations place by the FCC and
>governments.
You could go faster if you don't care about telco channel crosstalk or
hitting the line protectors thus creating distortion.
You may find the following of interest:
<http://www.LearnByDestroying.com/aty11/aty11.htm>
Hmmm... It's 7 years old. Maybe I should update it.
>Radium
Oh, it's you again.
--
Jeff Liebermann je...@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
gluc...@excite.com
> gluc...@excite.com hath wroth:
> >What are the physical* reasons dial-up speeds cannot go above 56kbps?
> >One factor is that the phone line cuts of frequencies below 300 Hz and
> >above 3,000 Hz. What are the other physical causes of this limit?
> >Would this limit still exist if it weren't for the aforementioned
> >frequency-cutoffs?
> The speed limit is the point where the error rate drops to the point
> where communications is futile within the audio bandwidth.
Usually, how much bps until this limit is reached?
> The V.90
> spec allows 56Kbits/sec download, but only 33.6Kbits/sec upload.
> (V.92 does 48Kbit/sec upload).
Yup.
> >bps = baud X number of bits per baud.
> Yep. It would be 64Kbits/sec were it not for telco bit-robbing, where
> every 6th frame is stolen for signaling. The audio channel can handle
> the bandwidth, but when it hits the DS0 digital channels, bit robbing
> limits the bandwidth to 56K. Therefore, you only get:
> 8Kbits/sec * 7 bits = 56 Kbits/sec.
I don't understand how you got that equation. Please clarify.
> What do you mean by a "band"?
I never used the word "band".
>Are you perhaps thinking of bits per
> baud?
Yes.
> >If only 1 baud is used, what is the maximum-bits-per-baud that can be
> >used on a phone line without the frequency of the analog electric
> >signal exceeding 3,000 Hz?
> >
> >*NOTE: By "physical", I am referring to causes not associated with
> >legal regulations - such as limitations place by the FCC and
> >governments.
> You could go faster if you don't care about telco channel crosstalk or
> hitting the line protectors thus creating distortion.
Okay, but if only 1 baud is used what is the maximum-bits-per-baud
that can be used on a phone line?
> You may find the following of interest:
> <http://www.LearnByDestroying.com/aty11/aty11.htm>
> Hmmm... It's 7 years old. Maybe I should update it.
Thanks for the link.
Phil Allison
"Jeff Liebermann"
>
>>What are the physical* reasons dial-up speeds cannot go above 56kbps?
>>One factor is that the phone line cuts of frequencies below 300 Hz and
>>above 3,000 Hz. What are the other physical causes of this limit?
>>Would this limit still exist if it weren't for the aforementioned
>>frequency-cutoffs?
** The limit is set by the restricted bandwidth AND the available to noise
ratio of a telephone network voice circuit.
A guy named Shannon work out the maths, loooong ago.
It kinda like the inverse of the Nyquist sampling theorem.
http://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem
The formula for when the s/n ratio >> unity is:
Bits/sec = 0.33 x bandwidth (Hz) x s/n (in dB)
For a voice circuit that becomes:
0.33 x 3000 x 45 ( typical good line)
= 45,000 b/s
The actual s/n ratio on a given voice circuit, which includes analogue
twisted wire lines and digital signals over co-ax or fibre varies - so
56K modems do an automatic line test to see just what capacity is available
and adjust themselves to that.
...... Phil
Phil Allison
"Phil Allison"
** My previous post was to:
Whose post does not show on my server.
........ Phil
gluc...@excite.com
> ** The limit is set by the restricted bandwidth AND the available to noise
> ratio of a telephone network voice circuit.
>
> A guy named Shannon work out the maths, loooong ago.
>
> It kinda like the inverse of the Nyquist sampling theorem.
>
> http://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem
Okay.
> The formula for when the s/n ratio >> unity is:
>
> Bits/sec = 0.33 x bandwidth (Hz) x s/n (in dB)
What is the "0.33"?
> For a voice circuit that becomes:
>
> 0.33 x 3000 x 45 ( typical good line)
>
> = 45,000 b/s
>
>
> The actual s/n ratio on a given voice circuit, which includes analogue
> twisted wire lines and digital signals over co-ax or fibre varies - so
> 56K modems do an automatic line test to see just what capacity is available
> and adjust themselves to that.
If the max a voice circuit can take is 45 kbps, then how can dial-up
speeds get above that?
Also if frequencies below 300 Hz and above 3,000 Hz are cut-off,
wouldn't the bandwidth of a phone line be 2,700 Hz?
Eeyore
gluc...@excite.com wrote:
> Hi:
>
> What are the physical* reasons dial-up speeds cannot go above 56kbps?
> One factor is that the phone line cuts of frequencies below 300 Hz and
> above 3,000 Hz. What are the other physical causes of this limit?
Noise performance (signal to noise ratio).
> Would this limit still exist if it weren't for the aforementioned
> frequency-cutoffs?
Not the same limit for sure.
Graham
Eeyore
gluc...@excite.com wrote:
> Phil Allison wrote:
>
> > ** The limit is set by the restricted bandwidth AND the available to noise
> > ratio of a telephone network voice circuit.
> >
> > A guy named Shannon work out the maths, loooong ago.
> >
> > It kinda like the inverse of the Nyquist sampling theorem.
> >
> > http://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem
>
> Okay.
>
> > The formula for when the s/n ratio >> unity is:
> >
> > Bits/sec = 0.33 x bandwidth (Hz) x s/n (in dB)
>
> What is the "0.33"?
I dare say if you could be bothered to read the link supplied or google a bit
yourself for sampling theory you'd find a more complete answer.
Your desire for totally 'potted answers' that onvolve no effort on your poart is
one of the things that annoys so much. If you were genuinely interested you'd be
prepared to do some 'homework'
> > For a voice circuit that becomes:
> >
> > 0.33 x 3000 x 45 ( typical good line)
> >
> > = 45,000 b/s
> >
> >
> > The actual s/n ratio on a given voice circuit, which includes analogue
> > twisted wire lines and digital signals over co-ax or fibre varies - so
> > 56K modems do an automatic line test to see just what capacity is available
> > and adjust themselves to that.
>
> If the max a voice circuit can take is 45 kbps, then how can dial-up
> speeds get above that?
Are you really that STUPID ?
It's bloody onvious. The bandwidth is fixed and the 0.33 is fixed so what's the
remaining variable ? Hint - PA used the word *typical* to describe it.
I suggest you stick 56k into the equation and calculate the resulting answer.
Graham
Jeff Liebermann
>Jeff Liebermann wrote:
>
>> gluc...@excite.com hath wroth:
>
>> >What are the physical* reasons dial-up speeds cannot go above 56kbps?
>> >One factor is that the phone line cuts of frequencies below 300 Hz and
>> >above 3,000 Hz. What are the other physical causes of this limit?
>> >Would this limit still exist if it weren't for the aforementioned
>> >frequency-cutoffs?
>
>> The speed limit is the point where the error rate drops to the point
>> where communications is futile within the audio bandwidth.
>Usually, how much bps until this limit is reached?
Look at a communications channel this way. You can get *ANY* speed
through a bandwidth limited channel, up to a given error rate. If
your application requires a very low error rate to function, then you
have to have a good signal to noise ratio, and your thruput will be
fairly small. However, if you have a mess of forward error
correction, small packets, and a very tolerant application, you migtht
be able to squeeze some more thruput through the same channel.
I'm not going to expound on how V.90 works in detail. It gets messy
fast. There are modulation schemes for increasing the base 600 baud
modulation rate (bits per baud) to much higher bits/baud. Then add,
adaptive equalizers, echo cancellers, error detection, error
correction, data compression, etc. Anything to squeeze more thruput
into a rather ugly looking POTS line. However, that's just between
the user and the CO (central office). Once at the CO, everything gets
converted to digital and the rules change.
>> The V.90
>> spec allows 56Kbits/sec download, but only 33.6Kbits/sec upload.
>> (V.92 does 48Kbit/sec upload).
>
>Yup.
>
>> >bps = baud X number of bits per baud.
>
>> Yep. It would be 64Kbits/sec were it not for telco bit-robbing, where
>> every 6th frame is stolen for signaling. The audio channel can handle
>> the bandwidth, but when it hits the DS0 digital channels, bit robbing
>> limits the bandwidth to 56K. Therefore, you only get:
>> 8Kbits/sec * 7 bits = 56 Kbits/sec.
>
>I don't understand how you got that equation. Please clarify.
Nope. I don't want to get into how a DS0 (digital) line work. You
can get 64Kbits/sec out of a DS0 if you can use out of band
signalling. However, if you're using in band signalling, you're stuck
with 56Kbits/sec. Even if the analog part of the puzzle can go faster
than 56Kbits/sec, the digital part at the CO will limit the speed to
56Kbits/sec.
>> What do you mean by a "band"?
>I never used the word "band".
Sorry. There was a fly splattered on my monitor.
>>Are you perhaps thinking of bits per
>> baud?
>Yes.
>> >If only 1 baud is used, what is the maximum-bits-per-baud that can be
>> >used on a phone line without the frequency of the analog electric
>> >signal exceeding 3,000 Hz?
>> >
>> >*NOTE: By "physical", I am referring to causes not associated with
>> >legal regulations - such as limitations place by the FCC and
>> >governments.
>
>> You could go faster if you don't care about telco channel crosstalk or
>> hitting the line protectors thus creating distortion.
>Okay, but if only 1 baud is used what is the maximum-bits-per-baud
>that can be used on a phone line?
56Kbits/sec. The limit is NOT all from the analog part of the line.
The analog modem glop gets converted to digital at the CO and that's
limited to 56Kbits/sec. I could easily (well maybe not so easily) get
more than 56Kbit/sec thruput going between my house and the CO, but
the digital thruput at the switch will limit thruput to 56Kbits/sec.
The problem is worse when dealing with SLC (subscriber line
concentrators) where the analog to digital conversion is done outside
the CO, such as with Pair Gain. The best you can do with those is
perhaps 28.8Kbit/sec, mostly because the digital audio filter cuts off
at a much lower frequency than the filters at the CO. Since most of
the energy is in the higher frequency part of the audio spectrum, the
loss of the higher frequencies is fatal to higher speed modem
operation.
<http://en.wikipedia.org/wiki/Pair_gain>
>> You may find the following of interest:
>> <http://www.LearnByDestroying.com/aty11/aty11.htm>
>> Hmmm... It's 7 years old. Maybe I should update it.
>
>Thanks for the link.
You really should be asking in comp.dcom.modems.
John Larkin
The fact that all modern telephone exchanges digitize the voice signal
to 8 bits at a sample rate of 8 KHz, which means that the digital
paths within and between phone offices absolutely cannot transport
more than 64 KBPS. Some overheads, nonlinearities, and error
correction things resulted in the 56K download number. Upload is
burdened by some additional issues.
If the phone system were still truly analog, the number of bps you
could jam into 3 KHz of bandwidth would depend on the s/n ratio, per
Shannon, and could in some cases be a lot higher.
John
Jamie
You see, the internet is like a bunch of pipes with water running
through it, think of water as your information.
Plumbing is old, Pipes are small, You can only get some much through
those pipes!
It's true!
One of our Senators even said so! :)
--
"I'm never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5
Joerg
> On Sun, 09 Sep 2007 05:16:26 -0000, gluc...@excite.com wrote:
>
>
>>Hi:
>>
>>What are the physical* reasons dial-up speeds cannot go above 56kbps?
>>One factor is that the phone line cuts of frequencies below 300 Hz and
>>above 3,000 Hz. What are the other physical causes of this limit?
>>Would this limit still exist if it weren't for the aforementioned
>>frequency-cutoffs?
>>
>>bps = baud X number of bits per baud.
>>
>>If only 1 baud is used, what is the maximum-bits-per-baud that can be
>>used on a phone line without the frequency of the analog electric
>>signal exceeding 3,000 Hz?
>>
>>*NOTE: By "physical", I am referring to causes not associated with
>>legal regulations - such as limitations place by the FCC and
>>governments.
>>
>>
>>Thanks,
>>
>>Radium
>
>
> The fact that all modern telephone exchanges digitize the voice signal
> to 8 bits at a sample rate of 8 KHz, which means that the digital
> paths within and between phone offices absolutely cannot transport
> more than 64 KBPS. Some overheads, nonlinearities, and error
> correction things resulted in the 56K download number. Upload is
> burdened by some additional issues.
>
In reality I have rarely ever reached >30K in situations where I had to
use dial-up. Hotels and such.
> If the phone system were still truly analog, the number of bps you
> could jam into 3 KHz of bandwidth would depend on the s/n ratio, per
> Shannon, and could in some cases be a lot higher.
>
Lately our phone line here has a constant and very audible noise hash on
them. Maybe they don't want folks using dial-up and buy the DSL service ;-)
--
Regards, Joerg
Green Xenon [Radium]
http://groups.google.com/group/sci.electronics.design/msg/0247b0487037bc81?hl=en&
:
> gluceg...@excite.com hath wroth:
> >Usually, how much bps until this limit is reached?
> Look at a communications channel this way. You can get *ANY* speed
> through a bandwidth limited channel, up to a given error rate. If
> your application requires a very low error rate to function, then you
> have to have a good signal to noise ratio, and your thruput will be
> fairly small. However, if you have a mess of forward error
> correction, small packets, and a very tolerant application, you migtht
> be able to squeeze some more thruput through the same channel.
So if I don't care about the errors, I can get whatever speed I want?
What about getting 1 Gbps on dial-up if I have a baud of
1-symbol-per-second but 1-billion-bits-per-symbol? Is that possible?
> I'm not going to expound on how V.90 works in detail. It gets messy
> fast. There are modulation schemes for increasing the base 600 baud
> modulation rate (bits per baud) to much higher bits/baud. Then add,
> adaptive equalizers, echo cancellers, error detection, error
> correction, data compression, etc. Anything to squeeze more thruput
> into a rather ugly looking POTS line. However, that's just between
> the user and the CO (central office). Once at the CO, everything gets
> converted to digital and the rules change.
Okay
> >I don't understand how you got that equation. Please clarify.
> Nope. I don't want to get into how a DS0 (digital) line work. You
> can get 64Kbits/sec out of a DS0 if you can use out of band
> signalling. However, if you're using in band signalling, you're stuck
> with 56Kbits/sec. Even if the analog part of the puzzle can go faster
> than 56Kbits/sec, the digital part at the CO will limit the speed to
> 56Kbits/sec.
Why is the digital part limited to 56K?
> >Okay, but if only 1 baud is used what is the maximum-bits-per-baud
> >that can be used on a phone line?
> 56Kbits/sec. The limit is NOT all from the analog part of the line.
> The analog modem glop gets converted to digital at the CO and that's
> limited to 56Kbits/sec. I could easily (well maybe not so easily) get
> more than 56Kbit/sec thruput going between my house and the CO, but
> the digital thruput at the switch will limit thruput to 56Kbits/sec.
Can this be changed so that the digital throughput will go up to
1Gbit/sec instead of just 56Kbits/sec?
> The problem is worse when dealing with SLC (subscriber line
> concentrators) where the analog to digital conversion is done outside
> the CO, such as with Pair Gain. The best you can do with those is
> perhaps 28.8Kbit/sec, mostly because the digital audio filter cuts off
> at a much lower frequency than the filters at the CO. Since most of
> the energy is in the higher frequency part of the audio spectrum, the
> loss of the higher frequencies is fatal to higher speed modem
> operation.
> <http://en.wikipedia.org/wiki/Pair_gain>
Okay. Seems from the link that digital pair gain is more efficient than
the analog counterpart.
Robert Adsett
[Radium] says...
> On Sep 9, 9:13 am, Jeff Liebermann <je...@cruzio.com> wrote
> http://groups.google.com/group/sci.electronics.design/msg/0247b0487037bc81?hl=en&
> > through a bandwidth limited channel, up to a given error rate. If
> > your application requires a very low error rate to function, then you
> > have to have a good signal to noise ratio, and your thruput will be
> > fairly small. However, if you have a mess of forward error
> > correction, small packets, and a very tolerant application, you migtht
> > be able to squeeze some more thruput through the same channel.
>
> So if I don't care about the errors, I can get whatever speed I want?
> What about getting 1 Gbps on dial-up if I have a baud of
> 1-symbol-per-second but 1-billion-bits-per-symbol? Is that possible?
Sure, just don't expect them to be correct. And if you don't care about
errors why bother with the link at all? Just use random noise instead
and free up your phone line.
> > 56Kbits/sec. The limit is NOT all from the analog part of the line.
> > The analog modem glop gets converted to digital at the CO and that's
> > limited to 56Kbits/sec. I could easily (well maybe not so easily) get
> > more than 56Kbit/sec thruput going between my house and the CO, but
> > the digital thruput at the switch will limit thruput to 56Kbits/sec.
>
> Can this be changed so that the digital throughput will go up to
> 1Gbit/sec instead of just 56Kbits/sec?
There would be no point. It's meant to carry voice, conversations are
not going to significantly benefit from a significantly higher data
rate. If you simply want data you know where to get it. The current
most popular incarnation over phone lines is probably DSL.
Robert
--
Posted via a free Usenet account from http://www.teranews.com
colin
news:1189314986.2...@22g2000hsm.googlegroups.com...
The phone lines are digitised nowadays,
the rate at wich they digitise them is I gues about 56kbps
therefore no encoding scheme is going to get round this.
the modems carefully tune the signal levels for each
encoded state so that it matches the digitising of the phone line
to as close to 1:1 as possible.
Colin =^.^=
Don Bowey
"Green Xenon [Radium]" <gluc...@excite.com> wrote:
> On Sep 9, 9:13 am, Jeff Liebermann <je...@cruzio.com> wrote
> http://groups.google.com/group/sci.electronics.design/msg/0247b0487037bc81?hl=
> en&
> :
>
>> gluceg...@excite.com hath wroth:
>
>>> Usually, how much bps until this limit is reached?
>
>> Look at a communications channel this way. You can get *ANY* speed
>> through a bandwidth limited channel, up to a given error rate. If
>> your application requires a very low error rate to function, then you
>> have to have a good signal to noise ratio, and your thruput will be
>> fairly small. However, if you have a mess of forward error
>> correction, small packets, and a very tolerant application, you migtht
>> be able to squeeze some more thruput through the same channel.
>
> So if I don't care about the errors, I can get whatever speed I want?
Of course not. You will have a 100% error rate at some point.
> What about getting 1 Gbps on dial-up if I have a baud of
> 1-symbol-per-second but 1-billion-bits-per-symbol? Is that possible?
>
>> I'm not going to expound on how V.90 works in detail. It gets messy
>> fast. There are modulation schemes for increasing the base 600 baud
>> modulation rate (bits per baud) to much higher bits/baud. Then add,
>> adaptive equalizers, echo cancellers, error detection, error
>> correction, data compression, etc. Anything to squeeze more thruput
>> into a rather ugly looking POTS line. However, that's just between
>> the user and the CO (central office). Once at the CO, everything gets
>> converted to digital and the rules change.
>
> Okay
>
>>> I don't understand how you got that equation. Please clarify.
>
>> Nope. I don't want to get into how a DS0 (digital) line work. You
>> can get 64Kbits/sec out of a DS0 if you can use out of band
>> signalling. However, if you're using in band signalling, you're stuck
>> with 56Kbits/sec. Even if the analog part of the puzzle can go faster
>> than 56Kbits/sec, the digital part at the CO will limit the speed to
>> 56Kbits/sec.
>
> Why is the digital part limited to 56K?
It isn't. The digital channel is 64 kbits. Robbed-bit signaling has
already been explained to you.
>
>>> Okay, but if only 1 baud is used what is the maximum-bits-per-baud
>>> that can be used on a phone line?
>
>> 56Kbits/sec. The limit is NOT all from the analog part of the line.
>> The analog modem glop gets converted to digital at the CO and that's
>> limited to 56Kbits/sec. I could easily (well maybe not so easily) get
>> more than 56Kbit/sec thruput going between my house and the CO, but
>> the digital thruput at the switch will limit thruput to 56Kbits/sec.
>
> Can this be changed so that the digital throughput will go up to
> 1Gbit/sec instead of just 56Kbits/sec?
Not for your dial-up service. Te reason has already been explained to you.
Phil Allison
"Jeff Liebermann"
> Look at a communications channel this way. You can get *ANY* speed
> through a bandwidth limited channel, up to a given error rate.
** Bollocks.
You need to read up on Shannon's theorem - fool.
> I'm not going to expound on how V.90 works in detail.
** Irrelevant to the inherent bit rate capacity of a voice communications
line.
> Sorry. There was a fly splattered on my monitor.
** Fly spots all over you as well.
> 56Kbits/sec. The limit is NOT all from the analog part of the line.
** In truth - it is.
From the user's position a dial up voice circuit IS analogue.
....... Phil
JosephKK
sci.electronics.design:
The number 1 reason is end equipment. If the end equipment does not
support any better than POTS that is all you get. See also
Shannon's theorem, which relates usable bandwidth, signal to noise
ratio, and datarate. If, as usual, the end equipment is the limiting
factor on a twisted pair line then that is the limit.
If you are looking for the limits on just a Cat 3 copper twisted pair,
then you are looking at something on the order of 30 MB/(second*mile)
based on the equipment that i have seen for sale.
Jeff Liebermann
>On Sep 9, 9:13 am, Jeff Liebermann <je...@cruzio.com> wrote
>http://groups.google.com/group/sci.electronics.design/msg/0247b0487037bc81?hl=en&
>:
>
> > gluceg...@excite.com hath wroth:
>
> > >Usually, how much bps until this limit is reached?
>
> > Look at a communications channel this way. You can get *ANY* speed
> > through a bandwidth limited channel, up to a given error rate. If
> > your application requires a very low error rate to function, then you
> > have to have a good signal to noise ratio, and your thruput will be
> > fairly small. However, if you have a mess of forward error
> > correction, small packets, and a very tolerant application, you migtht
> > be able to squeeze some more thruput through the same channel.
>So if I don't care about the errors, I can get whatever speed I want?
Yep. I've dealt with communications systems that initially send more
errors than usable data. Viturbi Decoders work that way. They
reconfigure themselves on the fly and optimize their decoding
characteristics based upon the channel characteristics of the moment.
Comm systems designed to operate in the presence of jamming and
interference also work that way. Take the bits you can get through
and retransmit the rest. For example, 802.11 was designed to
interleave with microwave oven 60Hz interference. Although thruput
drops drastically in the presense of microwave oven junk, quite a few
packets get through. Some of the original space communications
systems had more error correction code being transmitted, than data.
Calculating the optimium ECC code, retransmission rate, and optimum
baud rate is a major challenge.
>What about getting 1 Gbps on dial-up if I have a baud of
>1-symbol-per-second but 1-billion-bits-per-symbol? Is that possible?
Sure, no problem. 1024 QAM bursts with heavy ECC will do it just
fine. However, with a probable 99.9999% error rate (I can work out
the exact number later if anyone really wants it), requiring mutliple
retransmissions, you might be lucky and get perhaps 10Kbits/sec
thruput. As a general rule, if you can get about 75% of the packet
through without error, you have a workable system. Your 1 Gigabit
system doesn't even come close.
>Why is the digital part limited to 56K?
Because a DS0 channel is 64Kbits/sec by order of Ma Bell in her
manifestation as Bellcore and as inscribed in voluminous ANSI and ITU
specifications. So it is written, so it must be.
<http://en.wikipedia.org/wiki/DS0>
However, it's really a voice communications standard, which requires
some borrowed inband bandwidth for signalling. So, bit robbing
reduces the bandwidth to 56Kbit/sec.
This is the 2nd time I explained this. Is there a problem?
> > 56Kbits/sec. The limit is NOT all from the analog part of the line.
> > The analog modem glop gets converted to digital at the CO and that's
> > limited to 56Kbits/sec. I could easily (well maybe not so easily) get
> > more than 56Kbit/sec thruput going between my house and the CO, but
> > the digital thruput at the switch will limit thruput to 56Kbits/sec.
>
>Can this be changed so that the digital throughput will go up to
>1Gbit/sec instead of just 56Kbits/sec?
I suppose an act of God or miracle that modifies physics might change
it. Also, some rather bizarre quantum effects imply that it's
possible for the data to arrive before it's sent, thus increasing the
channel bandwidth. Perhaps a wormhole might help. Otherwise, I don't
think there's anything that you can do to inspire Ma Bell or have her
change the way the telephone network operates.
Think of it this way. Ma Bell operates a sewer system. It had big
pipes, medium size pipes, and small pipes. Amazingly, by the
judicious application of acronyms, it is possible for your home drain
pipe to shove more sewerage at the CO (central office) switch than it
can handle. It's pipe is slightly smaller than your home drain pipe.
It doesn't matter how much drek you shove down your sewer pipe, the
size of Ma Bell's pipe at the CO limits your capacity. You could
infinitely increase the size of your drain pipe, but Ma Bell will only
pass a limited amount. If your data doesn't pass the first time,
flush again later when the channel clears.
Jeff Liebermann
>>What about getting 1 Gbps on dial-up if I have a baud of
>>1-symbol-per-second but 1-billion-bits-per-symbol? Is that possible?
>Sure, no problem. 1024 QAM bursts with heavy ECC will do it just
>fine.
Oops. Using a 600 baud base modulation rate, 1 Gigabit/sec will
require:
1*10^9 bits / 600 baud = 1.66*10^6 bits/baud
which will require 21 bits of resolution. That would be:
2,097,152 QAM
which isn't going to happen over any kind of real audio channel.
We return you now to the Sci-Fi channel.
Green Xenon [Radium]
> "Green Xenon [Radium]" <gluc...@excite.com> hath wroth:
>>Why is the digital part limited to 56K?
> Because a DS0 channel is 64Kbits/sec by order of Ma Bell in her
> manifestation as Bellcore and as inscribed in voluminous ANSI and ITU
> specifications. So it is written, so it must be.
> <http://en.wikipedia.org/wiki/DS0>
> However, it's really a voice communications standard, which requires
> some borrowed inband bandwidth for signalling. So, bit robbing
> reduces the bandwidth to 56Kbit/sec.
>
> This is the 2nd time I explained this. Is there a problem?
Sorry I forget the part about bit-robbing.
My actual question was “why is the digital part limited to 64K”? Is this
restriction due to physical limits or legal limits?
Green Xenon [Radium]
> Oops. Using a 600 baud base modulation rate, 1 Gigabit/sec will
> require:
> 1*10^9 bits / 600 baud = 1.66*10^6 bits/baud
> which will require 21 bits of resolution. That would be:
> 2,097,152 QAM
> which isn't going to happen over any kind of real audio channel.
What if only 1 baud is used instead of 600 baud?
Jeff Liebermann
>My actual question was “why is the digital part limited to 64K”? Is this
>restriction due to physical limits or legal limits?
I think I also answered that. It's because Ma Bell, when they set the
specifications for the characteristics of MUX channels, groups,
super-groups, and digital technologies that they use, decided that
1/12th of a T1 is going to be a single DS0 voice channel with a
bandwidth of 64Kbits/sec. These standards are not arbitrary and were
(hopefully) carefully calculated to maximize the number of voice
channels in the available bandwidth. This was long before modems
became popular. Even today, Ma Bell will not guarantee anything
faster than v.32 (9600 baud) speeds. Some telcos still only guarantee
v.22 (1200 baud).
Jeff Liebermann
>Jeff Liebermann wrote:
Your modem will transmogrify into a quantum black hole and suck both
you and your ideas into another dimension.
This is a clone of the question you asked about 2 months ago. Same
answer as before. You cannot easily modulate a low frequency carrier
(1Hz), with a very high frequency modulation frequency (1GHz). Well,
actually you can do it, it's just that the results will be worthless
for doing anything useful.
For starters, the short term frequency stability (jitter) of the 1Hz
carrier will need to be at least half the modulation frequency in
accuracy. Otherwise, the modulation bits cannot maintain their
position in the constellation diagram. That's at least 1 part in
10^22 accuracy. A good cesium atomic clock might be good for 1 part
in 10^14, so this isn't going to happen:
<http://en.wikipedia.org/wiki/Atomic_clock> (see chart)
This might be of some interest:
"List of Device Bandwidths".
<http://en.wikipedia.org/wiki/List_of_device_bandwidths>
Green Xenon [Radium]
> "Green Xenon [Radium]" <gluc...@excite.com> hath wroth:
>>What if only 1 baud is used instead of 600 baud?
> You cannot easily modulate a low frequency carrier
> (1Hz), with a very high frequency modulation frequency (1GHz). Well,
> actually you can do it, it's just that the results will be worthless
> for doing anything useful.
Okay, but you put 600 baud in the equation*. 600 Hz -- just like 1 Hz --
is much lower than 1 GHz, so how can a 600 Hz carrier be modulated by 1
GHz signal?
*You said “Using a 600 baud base modulation rate, 1 Gigabit/sec will
require: 1*109 bits/ 600 baud = 1.66*106 bits/baud”
BTW, does 1 baud equate to a 1 Hz carrier?
> For starters, the short term frequency stability (jitter) of the 1Hz
> carrier will need to be at least half the modulation frequency in
> accuracy. Otherwise, the modulation bits cannot maintain their
> position in the constellation diagram. That's at least 1 part in
> 10^22 accuracy.
Okay.
>A good cesium atomic clock might be good for 1 part
> in 10^14, so this isn't going to happen:
> <http://en.wikipedia.org/wiki/Atomic_clock> (see chart)
Thanks.
> This might be of some interest:
> "List of Device Bandwidths".
> <http://en.wikipedia.org/wiki/List_of_device_bandwidths>
Already seen this list many times before but thanks anyway.
Gary Tait
>>1Gbit/sec instead of just 56Kbits/sec?
Yes and no.
Yes, they could change out all the telephone infrastructure to support such
data rates over what is now a DS0.
The fact is they just won't. There are other more practical and cost
effective means to get high-speed data over twisted pair than "opening up"
the voice channel bandwidth.
Joel Kolstad
news:7RZEi.2710$3Y1...@newssvr17.news.prodigy.net...
> In reality I have rarely ever reached >30K in situations where I had to use
> dial-up. Hotels and such.
There are a lot of old PBXes out there that really mess things up.
On good "direct line to the CO" connections, back when I used dial-up (>5
years ago now) I would routinely get ~45-48kbps. The only time I got the
fabled 53kbps was with an internal (modern :-) ) PBX connection to an inside
modem pool .
45-48kbps is really quite good given the transmission medium used, I think.
> Lately our phone line here has a constant and very audible noise hash on
> them. Maybe they don't want folks using dial-up and buy the DSL service ;-)
A lot of phone companies officially only support something like 14.4kbps and
will do absolutely nothing to improve line quality if they can demonstrate a
successful 14.4kbps connection. I can't imagine that part of that policy
wasn't largely influenced by marketing... (Even so, of course, most lines
will do much better.)
On the other hand, I'm not sure there ever was a means of determining that
voice quality on a POTS line was "inadequate," was there?
I read an article the other day about some ham who was tracking down huge HF
band interference at night. He quickly figured out that it was a bad ballast
on a street lamp, but the power company made him go around and note all the
bad lights he could find in something like a 2-3 mile radius. Supposedly they
were "willing" to do this themselves, but the emissions/interference guys only
worked regular 9-5 hours and were "not allowed" to work overtime... such as at
night, when the street lamps would actually be *on*. What nonsense that is...
I have to imagine there's a union in there somewhere... what hiring manager
would ever hire someone to mitigate interference from streetlamps with the
understanding that they'd never have to work outside of 9-5 unless overtime
was paid?
---Joel
Joel Kolstad
news:C309D711.768DC%dbo...@comcast.net...
>> So if I don't care about the errors, I can get whatever speed I want?
> Of course not. You will have a 100% error rate at some point.
Actully, being able to achieve a *bit* error rate of 100%would be wonderful --
to fix it you just flip the bit!
:-)
Bit Error rates of 50% are essentially the same as "random noise," of course.
Robert Baer
> gluc...@excite.com hath wroth:
>
>
>>What are the physical* reasons dial-up speeds cannot go above 56kbps?
>>One factor is that the phone line cuts of frequencies below 300 Hz and
>>above 3,000 Hz. What are the other physical causes of this limit?
>>Would this limit still exist if it weren't for the aforementioned
>>frequency-cutoffs?
>
>
> The speed limit is the point where the error rate drops to the point
> where communications is futile within the audio bandwidth. The V.90
> spec allows 56Kbits/sec download, but only 33.6Kbits/sec upload.
> (V.92 does 48Kbit/sec upload).
>
>
>>bps = baud X number of bits per baud.
>
>
> Yep. It would be 64Kbits/sec were it not for telco bit-robbing, where
> every 6th frame is stolen for signaling. The audio channel can handle
> the bandwidth, but when it hits the DS0 digital channels, bit robbing
> limits the bandwidth to 56K. Therefore, you only get:
> 8Kbits/sec * 7 bits = 56 Kbits/sec.
>
> What do you mean by a "band"? Are you perhaps thinking of bits per
> baud?
>
>
>>If only 1 baud is used, what is the maximum-bits-per-baud that can be
>>used on a phone line without the frequency of the analog electric
>>signal exceeding 3,000 Hz?
>>
>>*NOTE: By "physical", I am referring to causes not associated with
>>legal regulations - such as limitations place by the FCC and
>>governments.
>
>
> You could go faster if you don't care about telco channel crosstalk or
> hitting the line protectors thus creating distortion.
>
> You may find the following of interest:
> <http://www.LearnByDestroying.com/aty11/aty11.htm>
> Hmmm... It's 7 years old. Maybe I should update it.
>
>
>>Radium
>
>
> Oh, it's you again.
>
Theoretical as well as calculated limits may be nice to dream about,
but the fact is that almost every telco throttles the speed to about
48K, making them liars when they say that they support 56K.
Don Bowey
Baer" <rober...@earthlink.net> wrote:
What telco says they support 56 kbit/s? I think you may be the liar.
Jeff Liebermann
> Theoretical as well as calculated limits may be nice to dream about,
>but the fact is that almost every telco throttles the speed to about
>48K, making them liars when they say that they support 56K.
Everyone lies, but that's ok because nobody listens.
I routinely get 49.3Kbits/sec on my dialup connections. Sometimes, it
goes to 50.3Kbits/sec. I guess the mythical throttle doesn't apply to
me.
The actual maximum speed is about 53Kbits/sec. The FCC limits the
line levels on dialup lines to prevent crosstalk. Without sufficient
line level, the S/N ratio drops, which causes the error rate to climb.
The modem compensates by reducing its speed.
Incidentally, the local telco only "supports" 1200 or 9600 (v.32)
baud, depending on whom I ask. Just try calling 611 with a modem
problem and see what happens.
Eeyore
Jeff Liebermann wrote:
> Robert Baer <rober...@earthlink.net> hath wroth:
>
> > Theoretical as well as calculated limits may be nice to dream about,
> >but the fact is that almost every telco throttles the speed to about
> >48K, making them liars when they say that they support 56K.
>
> Everyone lies, but that's ok because nobody listens.
>
> I routinely get 49.3Kbits/sec on my dialup connections. Sometimes, it
> goes to 50.3Kbits/sec. I guess the mythical throttle doesn't apply to
> me.
>
> The actual maximum speed is about 53Kbits/sec. The FCC limits the
> line levels on dialup lines to prevent crosstalk. Without sufficient
> line level, the S/N ratio drops, which causes the error rate to climb.
> The modem compensates by reducing its speed.
When I still used dialup, my reported connection speed was typically ~ 52kbps.
This is on a line that was 'new infrastructure' when it was installed.
Graham
Joerg
> "Joerg" <notthis...@removethispacbell.net> wrote in message
> news:7RZEi.2710$3Y1...@newssvr17.news.prodigy.net...
>
>>In reality I have rarely ever reached >30K in situations where I had to use
>>dial-up. Hotels and such.
>
>
> There are a lot of old PBXes out there that really mess things up.
>
> On good "direct line to the CO" connections, back when I used dial-up (>5
> years ago now) I would routinely get ~45-48kbps. The only time I got the
> fabled 53kbps was with an internal (modern :-) ) PBX connection to an inside
> modem pool .
>
> 45-48kbps is really quite good given the transmission medium used, I think.
>
>
>>Lately our phone line here has a constant and very audible noise hash on
>>them. Maybe they don't want folks using dial-up and buy the DSL service ;-)
>
>
> A lot of phone companies officially only support something like 14.4kbps and
> will do absolutely nothing to improve line quality if they can demonstrate a
> successful 14.4kbps connection. I can't imagine that part of that policy
> wasn't largely influenced by marketing... (Even so, of course, most lines
> will do much better.)
>
After all, nearly all of them also offer DSL over that same line and
they would really, really like that dial-up customer to fork over those
additional monthly fees.
> On the other hand, I'm not sure there ever was a means of determining that
> voice quality on a POTS line was "inadequate," was there?
>
> I read an article the other day about some ham who was tracking down huge HF
> band interference at night. He quickly figured out that it was a bad ballast
> on a street lamp, but the power company made him go around and note all the
> bad lights he could find in something like a 2-3 mile radius. Supposedly they
> were "willing" to do this themselves, but the emissions/interference guys only
> worked regular 9-5 hours and were "not allowed" to work overtime... such as at
> night, when the street lamps would actually be *on*. What nonsense that is...
> I have to imagine there's a union in there somewhere...
You can almost bet on that.
> ... what hiring manager
> would ever hire someone to mitigate interference from streetlamps with the
> understanding that they'd never have to work outside of 9-5 unless overtime
> was paid?
>
One reason why I have never worked in any larger company. Of course that
didn't help us when we got that dreaded new "anything over 8hrs/day must
be paid as overtime" rule in CA. Before, we could do 4/10 shifts which
really helped families with sick relatives or lots of kids plus it
helped us maintain a 4hr PM window. This stupidity has cost jobs. It
also destroyed home-care for some families. I had people in my office in
tears. I guess the unions thought they won but the reality is that the
people they were supposed to protect lost. Much of this work is now
outsourced. What choice do you have if they press you out of the
profitability margins?