Forums

Universal Parallel Bus -- why not?

Started by GreenXenon March 20, 2010
On Thu, 25 Mar 2010 03:44:36 -0700, "JosephKK"<quiettechblue@yahoo.com> wrote:

>On Sun, 21 Mar 2010 18:14:01 -0500, "krw@att.bizzzzzzzzzzzz" <krw@att.bizzzzzzzzzzzz> wrote: > >>On Sun, 21 Mar 2010 17:04:49 -0500, "Tim Williams" <tmoranwms@charter.net> >>wrote: >> >>><krw@att.bizzzzzzzzzzzz> wrote in message >>>news:3clcq552ic5jonnrut4n2g7ejp6hn8lg4u@4ax.com... >>>>>> Nonsense. You could attach anything you wanted to them. The parallel >>>>>> port on >>>>>> the PC has always been bidirectional. >>>>> >>>>>Nope. >>>>>http://en.wikipedia.org/wiki/Parallel_port >>>> >>>> Hoisted by your own petard: >>>> >>>> In early parallel ports the data lines were unidirectional (data out >>>> only) >>>> so it was not easily possible to feed data in to the computer. >>> >>>Duh: the data lines were not bidirectional. >> >>Bullshit. >> >>The Monochrome and printer adapter pinout: >> >>Pin No Signal Direction Register-bit >>DB25 Name >>1 nStrobe Out Control-0 >>2 Data0 In/Out Data-0 >>3 Data1 In/Out Data-1 >>4 Data2 In/Out Data-2 >>5 Data3 In/Out Data-3 >>6 Data4 In/Out Data-4 >>7 Data5 In/Out Data-5 >>8 Data6 In/Out Data-6 >>9 Data7 In/Out Data-7 >>10 nAck In Status-6 >>11 Busy In Status-7 >>12 Paper-Out In Status-5 >>13 Select In Status-4 >>14 Linefeed Out Control-1 >>15 nError In Status-3 >>16 nInitialize Out Control-2 >>17 nSelect Out Control-3 >>18-25 Ground >> >>>I do not know of any other definition of "bidirectional" which allows >>>different unidirectional wires to count as bidirectional. Bidirectional >>>means two directions on the same wire (and in terms of implementation, >>>usually two directions on the same port address, which certainly isn't true >>>of the four status bits on the parallel port). >>> >>>> The 8255 *was* bidirectional. >>> >>>Still is. But they never used it. They used discrete latches, hence >>>unidirectional. > >Not 8255. The 8 databus lines of the printer port (Centronics) were >implemented with a 74LS374 in XT and earlier. Bidirectional was later.
Yes, but the ORIGINAL PC's printer port *WAS* bidirectional, even though it was TTL. The above pinout is from the 5150's Monochrome and Printer Adapter.
On Thu, 25 Mar 2010 04:00:24 -0700, "JosephKK"<quiettechblue@yahoo.com> wrote:

>On Sun, 21 Mar 2010 11:11:40 -0500, Spehro Pefhany <speffSNIP@interlogDOTyou.knowwhat> wrote: > >>On Sun, 21 Mar 2010 13:07:01 -0800, the renowned Robert Baer >><robertbaer@localnet.com> wrote: >> >>>whit3rd wrote: >>>> On Mar 20, 7:33 am, GreenXenon <glucege...@gmail.com> wrote: >>>> >>>>> I keep hearing about Universal Serial Bus [USB]. Why hasn't a >>>>> Universal Parallel Bus [UPB] been implemented yet? >>>> >>>> It has been done, twice. IEEE-488 for instruments, and SCSI >>>> for small computers. >>>> >>>> Expensive cables made IEEE-488 a boutique item, and SCSI >>>> (which got up to 320 MBytes/sec) was usually kinda high-end >>>> Most Macintosh computers from 1986 to 1998 used SCSI. >>>> >>>> Parallel ports DO NOT COUNT, because they aren't a bus; >>>> the early ones weren't even bidirectional, and there was >>>> never any good support for more than two connections. >>>> Bus, from 'omnibus' meaning 'for everyone' implies that all >>>> bus signals are served by all devices, not just two. >>> Well, it turns out that the ORIGINAL PC/XT in 1980 had all of the >>>hardware on board to do full 8-bit I/O. >>> Are you perhaps alluding to an earlier version?? >> >>You sure about that? I recall that inputs had to be done 4 bits at a >>time (nibble mode) to assure compatibility. I have a chunk of code >>around somewhere that defines all the printer control signals in a >>sensible way for general purpose control (some are inverted IIRC...) >> >>Checking Jan Axelson's seminal publication "Parallel Port Complete", I >>see she says:- >> >>The original PC's parallel port had eight outputs, five inputs, and >>four bidirectional lines.... On many newer PCs, the eight oututs can >>also serve as inputs..." >> >>Now, the really crotchety old farts will remember that there was a >>hardware modification you could do to the original parallel port cards >>(wot used LS TTL parts) to cut the /ENABLE pin of the output latch so >>it wasn't permanently grounded, and jumper it to allow it to be >>controlled for birectional I/O, but I wouldn't claim that this "had >>all the hardware on board" when soldering irons and dremels are >>involved... >> >>Of course all this stuff has been in a corner of an LSI chip for >>decades now, so you're stuck with whatever IP is used for the chip, >>but you used to be able to do it. >> >> >> >>Best regards, >>Spehro Pefhany > >Yep, and any "0" being output by the ls374 tended to override any "1" >from any other device. Wired and, which ever device said "0" (low) won. >With the PC or XT ls374 wired on, it did not really make that good of >bidirectional.
It is still bidirectional, even if you have to write a '1' to the port to receive. That's not a whole lot different than programming a UC GPIO port to IN/OUT.
On Thu, 25 Mar 2010 03:51:18 -0700, "JosephKK"<quiettechblue@yahoo.com> wrote:

>On Sun, 21 Mar 2010 13:07:01 -0800, Robert Baer <robertbaer@localnet.com> wrote: > >>whit3rd wrote: >>> On Mar 20, 7:33 am, GreenXenon <glucege...@gmail.com> wrote: >>> >>>> I keep hearing about Universal Serial Bus [USB]. Why hasn't a >>>> Universal Parallel Bus [UPB] been implemented yet? >>> >>> It has been done, twice. IEEE-488 for instruments, and SCSI >>> for small computers. >>> >>> Expensive cables made IEEE-488 a boutique item, and SCSI >>> (which got up to 320 MBytes/sec) was usually kinda high-end >>> Most Macintosh computers from 1986 to 1998 used SCSI. >>> >>> Parallel ports DO NOT COUNT, because they aren't a bus; >>> the early ones weren't even bidirectional, and there was >>> never any good support for more than two connections. >>> Bus, from 'omnibus' meaning 'for everyone' implies that all >>> bus signals are served by all devices, not just two. >> Well, it turns out that the ORIGINAL PC/XT in 1980 had all of the >>hardware on board to do full 8-bit I/O. >> Are you perhaps alluding to an earlier version?? > >You are just referring to the wrong connector, those were ISA card slots. >The 74LS374 running the printer port data lines is not a bidirectional device.
...but the '244 around it was the other half.
>Get a schematic if you do not believe me.
On Thu, 25 Mar 2010 17:32:06 -0500, "krw@att.bizzzzzzzzzzzz" =
<krw@att.bizzzzzzzzzzzz> wrote:

>On Wed, 24 Mar 2010 22:58:37 -0700, "JosephKK"<quiettechblue@yahoo.com> =
wrote:
> >>On Sat, 20 Mar 2010 09:01:16 -0700, Charlie E. <edmondson@ieee.org> =
wrote:
>> >>>On Sat, 20 Mar 2010 10:38:44 -0500, Jamie >>><jamie_ka1lpa_not_valid_after_ka1lpa_@charter.net> wrote: >>> >>>>GreenXenon wrote: >>>> >>>>> Hi: >>>>>=20 >>>>> I keep hearing about Universal Serial Bus [USB]. Why hasn't a >>>>> Universal Parallel Bus [UPB] been implemented yet? >>>>>=20 >>>>> Wouldn't a UPB be faster than a USB at the same clock rate? If so, >>>>> this would mean the same speed with less energy comsumption. Right? >>>>>=20 >>>>>=20 >>>>> Thanks, >>>>>=20 >>>>> Green Xenon >>>>Yes it would be how ever, that cost more.. >>>> >>>Against my better judgemnet, since I know GX is just a troll, I will >>>answer his question... >>> >>>When you have one signal line, and you send a bit, then once you are >>>able to determine the status of that bit, you are ready for the next >>>one. You can do this very quickly. >>> >>>However, when you have two lines, then you have to be able to know, >>>not just the status of one of those lines, but that of both of them, >>>at the same time. This takes a little longer to be certain, and >>>therefore, it goes slower. One line or the other may be a little >>>longer, or have a slightly different impedance, or any number of other >>>reasons that makes it different. Yes, this is ofen more than twice as >>>long as for a single line. >>> >>>As you add signal lines, the problem grows. You have to wait till you >>>can be sure that ALL the lines have reached the correct state, before >>>you can move on to the next. Now, you may think that this doesn't >>>grow too fast, and you are right. But, at the same time, your >>>hardware has also increased. Your clock speed has gone down, and your >>>hardware cost has gone up. This makes a parallel signal not very >>>attractive=20 >>> >>>So, how do you get faster throughput? You use multiple serial lines, >>>like LVDS, each taking a part of the flow, and add them all togther at >>>the the other end. >>> >>>Charlie >> >>Yep, that is just what PCIExpress does, up to 16 lanes (serial line=20 >>group?) does. > >Does PCI-E dynamically deskew each lane?
Not really, it aggregates packets over the lanes, more like store and = forward. Nor do the packets on any lane have to be related to the packets of any = other lane.
On Thu, 25 Mar 2010 17:34:12 -0500, "krw@att.bizzzzzzzzzzzz" =
<krw@att.bizzzzzzzzzzzz> wrote:

>On Thu, 25 Mar 2010 03:44:36 -0700, "JosephKK"<quiettechblue@yahoo.com> =
wrote:
> >>On Sun, 21 Mar 2010 18:14:01 -0500, "krw@att.bizzzzzzzzzzzz" =
<krw@att.bizzzzzzzzzzzz> wrote:
>> >>>On Sun, 21 Mar 2010 17:04:49 -0500, "Tim Williams" =
<tmoranwms@charter.net>
>>>wrote: >>> >>>><krw@att.bizzzzzzzzzzzz> wrote in message=20 >>>>news:3clcq552ic5jonnrut4n2g7ejp6hn8lg4u@4ax.com... >>>>>>> Nonsense. You could attach anything you wanted to them. The =
parallel=20
>>>>>>> port on >>>>>>> the PC has always been bidirectional. >>>>>> >>>>>>Nope. >>>>>>http://en.wikipedia.org/wiki/Parallel_port >>>>> >>>>> Hoisted by your own petard: >>>>> >>>>> In early parallel ports the data lines were unidirectional (data =
out=20
>>>>> only) >>>>> so it was not easily possible to feed data in to the computer. >>>> >>>>Duh: the data lines were not bidirectional. >>> >>>Bullshit. >>> >>>The Monochrome and printer adapter pinout: >>> >>>Pin No Signal Direction Register-bit >>>DB25 Name >>>1 nStrobe Out Control-0 =09 >>>2 Data0 In/Out Data-0 =09 >>>3 Data1 In/Out Data-1 =09 >>>4 Data2 In/Out Data-2 =09 >>>5 Data3 In/Out Data-3 =09 >>>6 Data4 In/Out Data-4 =09 >>>7 Data5 In/Out Data-5 =09 >>>8 Data6 In/Out Data-6 =09 >>>9 Data7 In/Out Data-7 =09 >>>10 nAck In Status-6=20 >>>11 Busy In Status-7=20 >>>12 Paper-Out In Status-5=20 >>>13 Select In Status-4=20 >>>14 Linefeed Out Control-1 >>>15 nError In Status-3=20 >>>16 nInitialize Out Control-2 >>>17 nSelect Out Control-3 >>>18-25 Ground =20 >>> >>>>I do not know of any other definition of "bidirectional" which allows=
=20
>>>>different unidirectional wires to count as bidirectional. =
Bidirectional=20
>>>>means two directions on the same wire (and in terms of =
implementation,=20
>>>>usually two directions on the same port address, which certainly =
isn't true=20
>>>>of the four status bits on the parallel port). >>>> >>>>> The 8255 *was* bidirectional. >>>> >>>>Still is. But they never used it. They used discrete latches, hence=
=20
>>>>unidirectional. >> >>Not 8255. The 8 databus lines of the printer port (Centronics) were=20 >>implemented with a 74LS374 in XT and earlier. Bidirectional was later. > >Yes, but the ORIGINAL PC's printer port *WAS* bidirectional, even though=
it
>was TTL. The above pinout is from the 5150's Monochrome and Printer =
Adapter.=20 What documentation i can find of the original IBM hardware shows this to = be true. However many "clone" cards omitted the input possibility thus saving a = part or two. Nor was the bidirectionality all that good, to get input on those pins = writing "FF"=20 before hand was necessary (the ls374 output enable pin was hardwired).
How about one fiboroptic bus with photonic logic processor made out of  etched glass? I like the UPB idea,
but I think the future of computing will be computers that contain optics and LED's.  
ljnormanhot@gmail.com wrote:
> How about one fiboroptic bus with photonic logic processor made out of etched glass? I like the UPB idea, > but I think the future of computing will be computers that contain optics and LED's.
Optics, yes. Sure and hell NOT parallel bus unless one manages to wavelength multiplex 8, 16 or 32 channels. LEDs? are they not a bit slow?
On Fri, 02 Dec 2016 21:49:24 -0800, Robert Baer
<robertbaer@localnet.com> wrote:

>ljnormanhot@gmail.com wrote: >> How about one fiboroptic bus with photonic logic processor made out of etched glass? I like the UPB idea, >> but I think the future of computing will be computers that contain optics and LED's.
One problem with high speed parallel connections is the need to keep the path lengths exactly the same for each connection to be able to direct sample full words. At high speeds, you need to use some self clocking on each wire and after clock skew elimination to build a full word. The device complexity is similar as multiple serial connections bundled together. You could as well send serially all bits of a word on one wire, the nest word on the next wire etc.
> Optics, yes. Sure and hell NOT parallel bus unless one manages to >wavelength multiplex 8, 16 or 32 channels.
CWDM would be attractive, but would need lasers. What is the propagation spread on different wavelengths for different fiber types ? Would it be possible to sample each wavelength in parallel or would some self clocking signal be needed for each lamda ?
> LEDs? are they not a bit slow?
The spectral purity would be an issue even for 2-4 lamda CWDM.
upsidedown@downunder.com wrote:
> On Fri, 02 Dec 2016 21:49:24 -0800, Robert Baer > <robertbaer@localnet.com> wrote: > >> ljnormanhot@gmail.com wrote: >>> How about one fiboroptic bus with photonic logic processor made >>> out of etched glass? I like the UPB idea, but I think the future >>> of computing will be computers that contain optics and LED's. > > One problem with high speed parallel connections is the need to keep > the path lengths exactly the same for each connection to be able to > direct sample full words. At high speeds, you need to use some self > clocking on each wire and after clock skew elimination to build a > full word. > > The device complexity is similar as multiple serial connections > bundled together. You could as well send serially all bits of a word > on one wire, the nest word on the next wire etc. > >> Optics, yes. Sure and hell NOT parallel bus unless one manages to >> wavelength multiplex 8, 16 or 32 channels. > > CWDM would be attractive, but would need lasers.
* Which are a bit faster than LEDs and so a better match to this concept.
> > What is the propagation spread on different wavelengths for > different fiber types ?
* a definite problem; a dumb solution is to have each channel "clocked" at slow rate commensurate to guarantee clean recovery in all channels.
> Would it be possible to sample each wavelength in parallel or would > some self clocking signal be needed for each lamda ?
* Self-clocking is possible for each channel, but the slowest one will limit the overall data rate.
> >> LEDs? are they not a bit slow? > > The spectral purity would be an issue even for 2-4 lamda CWDM.
* Well so far, this is just an idea, sort-of sci-fi. Does not mean not possible or not practical.
>
On Saturday, 3 December 2016 10:47:53 UTC, upsid...@downunder.com  wrote:

> One problem with high speed parallel connections is the need to keep > the path lengths exactly the same for each connection to be able to > direct sample full words. At high speeds, you need to use some self > clocking on each wire and after clock skew elimination to build a full > word.
Even if the path lengths are exactly the same there is dispersion to deal with. The propagation velocity is frequency dependent, so data- dependent timing errors creep in. Long ago I was involved in transmitting digital TV signals around a building. To cut costs the designer tried using twisted pair telephone wire with one bit per pair plus a clock pair. It was impossible to get reliable operation at 50m range because of dispersion no matter how much tweaking of timing delays was done. John