Prev: Energizer USB Duo battery charger hides a Trojan
Next: How to run FreeRtos on MSP430
From: Tim Wescott on 11 Mar 2010 14:15
RalfM wrote:
> Don Bruder wrote:
>> In article<hn9i18$m3h$1(a)speranza.aioe.org>, RalfM<rm(a)invalid.invalid>
>> wrote:
>>
>>> Such a bitwise transfer does need only a fast switching between 2
>>> voltage
>>> levels.
>>> Is it possible to have a transmitter/receiver that can use more than
>>> just 2 possible values per "bit period" (for example: a 12-bit ADC or
>>> DAC
>>> can detect 2^12=4096 different values (ie. voltages). Then why use
>>> only 2 values (0/1) in copper wired data transfers instead of using
>>> say 256 or 512 or 1024 or 2048 or 4096 bits etc.? (ie. make the cable
>>> a "8
>>> bit cable" :-) or more.
>>> IMHO one could dramatically increase the speed on wired copper medium
>>> (for example a 256-fold increase or even more would be possible).
>>> Are there such chips which can switch fast a voltage source say to 256
>>> different values? ie. fast DAC and ADCs, DSP maybe?
>>>
>>> Just some crazy thoughts of mine... :-)
>>
>> Congratulations - You've just re-invented analog transmission.
>
> Hmm. isn't analog transmission using AC?

No. "Analog" transmission (really multi-level transmission) just means
that you're using multiple levels. There's nothing to say it must be
AC, although it's often used over long wire runs, and it's often
convenient to AC couple such runs -- which means that it's often
convenient to design your signaling with no DC content.

I make the distinction because "real" analog transmission is where you
transmit a continuous voltage to mean a continuous quantity -- i.e. 3V =
3 gallons in a tank, or AM radio where the envelope represents the sound
you're trying to reproduce.

> I mean use DC like in Ethernet

Ethernet -- at least twisted-pair Ethernet -- is transformer coupled,
which means no DC can get through. The signaling is designed for zero
DC content, to make it all work.

For that matter, some Ethernet protocols use multilevel signaling. See
Wikipedia: http://en.wikipedia.org/wiki/10BASE-T.

> but instead of transmitting just 1 bit per cycle use instead say 8 bits
> per cycle,
> ie. 8 DC voltage levels much like done with a DAC and ADC.

That's been fiddled with. About the only place that it's really popular
is over really long stretches of wire. If the wire run is short (e.g.
USB, IEEE-1394, LVDS), or if it can be well controlled (e.g. lower speed
Ethernet) then the signaling is usually binary with some sort of BEC, or
FEC combined with BEC.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
From: RalfM on 11 Mar 2010 14:28
Mike Paff wrote:
> On Thu, 11 Mar 2010 02:44:52 +0100, RalfM<rm(a)invalid.invalid> wrote:
>
>> Such a bitwise transfer does need only a fast switching between 2 voltage levels.
>> Is it possible to have a transmitter/receiver that can use more than
>> just 2 possible values per "bit period" (for example: a 12-bit ADC or DAC
>> can detect 2^12=4096 different values (ie. voltages). Then why use
>> only 2 values (0/1) in copper wired data transfers instead of using
>> say 256 or 512 or 1024 or 2048 or 4096 bits etc.? (ie. make the cable a "8 bit cable" :-) or more.
>> IMHO one could dramatically increase the speed on wired copper medium
>> (for example a 256-fold increase or even more would be possible).
>> Are there such chips which can switch fast a voltage source say to 256
>> different values? ie. fast DAC and ADCs, DSP maybe?
>>
>
> Sounds good in theory, but is likely to fall apart once reality
> sets in.
>
> Think a little bit about how noise picked up by the transmission
> line will affect the reading at the receiver. Also consider the
> case where the transmitter and receiver have different ground
> references.

I think they can be solved easily by (a) using a twisted pair cable
per direction, much like in Ethernet, and (b) one of the wires would be
the common ground and the other obviously the DC level against the ground wire.
The only problem I see is to have a DAC and an ADC which can switch fast enough
the DC levels.
But I remember having seen in the specifications of these chips
that they can well do several Mega or even Giga samples per second,
so then it should suffice IMO.
From: RalfM on 11 Mar 2010 14:41
Tim Wescott wrote:
> RalfM wrote:
>
>> but instead of transmitting just 1 bit per cycle use instead say 8
>> bits per cycle,
>> ie. 8 DC voltage levels much like done with a DAC and ADC.
>
> That's been fiddled with. About the only place that it's really popular
> is over really long stretches of wire. If the wire run is short (e.g.
> USB, IEEE-1394, LVDS), or if it can be well controlled (e.g. lower speed
> Ethernet) then the signaling is usually binary with some sort of BEC, or
> FEC combined with BEC.

Hmm. I don't understand why restrict yourself to use only binary signalling
when it can be done byte-wise (or even more) in the same time.
This could dramatically increase the speed, regardless of the distance.
From: Tim Wescott on 11 Mar 2010 15:05
RalfM wrote:
> Tim Wescott wrote:
>> RalfM wrote:
>>
>>> but instead of transmitting just 1 bit per cycle use instead say 8
>>> bits per cycle,
>>> ie. 8 DC voltage levels much like done with a DAC and ADC.
>>
>> That's been fiddled with. About the only place that it's really popular
>> is over really long stretches of wire. If the wire run is short (e.g.
>> USB, IEEE-1394, LVDS), or if it can be well controlled (e.g. lower speed
>> Ethernet) then the signaling is usually binary with some sort of BEC, or
>> FEC combined with BEC.
>
> Hmm. I don't understand why restrict yourself to use only binary signalling
> when it can be done byte-wise (or even more) in the same time.
> This could dramatically increase the speed, regardless of the distance.

For the same reason that most folks drive, ride or walk to work instead
of flying there in a helicopter. The helicopter dramatically increases
speed, regardless of distance, but it's expensive and takes constant
fiddling to keep it working.

Why don't you actually learn what Ethernet does? Then you would know
that even your examples don't always restrict themselves to binary
signaling, and you wouldn't keep repeating erroneous information.

While you're at it, learn what "baseband signaling" is, as opposed to
"DC", and ponder why the statement "Ethernet uses DC" is patently false.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
From: RalfM on 11 Mar 2010 15:06
Grant Edwards wrote:
> On 2010-03-11, RalfM<rm(a)invalid.invalid> wrote:
>> Mike Paff wrote:
>>> On Thu, 11 Mar 2010 02:44:52 +0100, RalfM<rm(a)invalid.invalid> wrote:
>>>
>>>> Such a bitwise transfer does need only a fast switching between 2 voltage levels.
>>>> Is it possible to have a transmitter/receiver that can use more than
>>>> just 2 possible values per "bit period" (for example: a 12-bit ADC or DAC
>>>> can detect 2^12=4096 different values (ie. voltages). Then why use
>>>> only 2 values (0/1) in copper wired data transfers instead of using
>>>> say 256 or 512 or 1024 or 2048 or 4096 bits etc.? (ie. make the cable a "8 bit cable" :-) or more.
>>>> IMHO one could dramatically increase the speed on wired copper medium
>>>> (for example a 256-fold increase or even more would be possible).
>>>> Are there such chips which can switch fast a voltage source say to 256
>>>> different values? ie. fast DAC and ADCs, DSP maybe?
>>>>
>>>
>>> Sounds good in theory, but is likely to fall apart once reality
>>> sets in.
>>>
>>> Think a little bit about how noise picked up by the transmission
>>> line will affect the reading at the receiver. Also consider the
>>> case where the transmitter and receiver have different ground
>>> references.
>>
>> I think they can be solved easily [...]
>
> Go ahead.
>
> I'm sure all those engineers at TI, Bell Labs, DEC, Intel, HP, etc.
> were all wrong.

I think they all thought of only bit-serial transmission, not going further in the dimension.

> You're going to have a lot of problems with waveforms changing too
> much. Have you ever looked at any eye-plots and and compared what
> comes out of the far end of a cable with what goes in?
>
>> by (a) using a twisted pair cable per direction, much like in
>> Ethernet, and (b) one of the wires would be the common ground and the
>> other obviously the DC level against the ground wire. The only
>> problem I see is to have a DAC and an ADC which can switch fast
>> enough the DC levels. But I remember having seen in the
>> specifications of these chips that they can well do several Mega or
>> even Giga samples per second, so then it should suffice IMO.
>
> Do you think that everybody goes to such measures to avoid DC
> signalling just for fun?

My understanding is that one can do it better, faster, and cheaper.
For example 1000BASE-T uses 4 pairs (!) of wires, IMHO a waste of wires and HW.
I think one can do it with only 2 pairs if one just uses "byte-serial" transmission
instead of bit-serial.
If byte-serial works then why not improve it further simply by using multiple bytes per cycle.
Ie. one could transmit even 32-bits in just 1 clock cycle; it all depends only on
the max range levels the DAC on the sending side and the ADC on the receiving side can distinguish.
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: Energizer USB Duo battery charger hides a Trojan
Next: How to run FreeRtos on MSP430