Re: [Emc-users] Serial daisy chain protocol

[Nicklas SB Karlsson]

> > > > > > No SPI over long cable. For long cable RS422 will work, there
> > > > > > is a tradeoff between cable length and speed.
> > > > >
> > > > > Poor wording on my part. I had, once the terminations are
> > > > > sorted, visions of a 3 to 4 foot cable, defined as a long SPI
> > > > > cable.
> > > > >
> > > > > Right now, terms are not well characterized on either end and it
> > > > > takes some loading trickery to make a 6 inch cable work.  Bertho
> > > > > and I spent some time optimizing it to get 30.5 megabit one way,
> > > > > and 25 megabit the other direction.  And its now working well to
> > > > > a Mesa 7i90 with a 1" cable.
> > > >
> > > > Ethernet shielded twisted pair four pairs like this: sck+gnd,
> > > > miso+gnd, mosi+gnd, chip select + gnd. Or flat cable with a gnd
> > > > cable between each of them. Or of course Ethernet if available.
> > >
> > > No, the tcp is too slow, much latency.
> >
> > For TCP resend function to make sense there must be enough time to
> > resend packet.
> >
> > > However, the above scheme, made into ready-made hardware and a
> > > matching rj45 socket signal assignment on those mesa cards that can
> > > support spi, like the 7i90, ...
> >
> > Ethercat chips are available with SPI interface and I concluded then
> > distance is short enough and data is sent between it might be a good
> > choice to skip the Ethercat chips but I might be wrong.
> >
> I'd be for that skip myself, if we had real line drivers in the pi's 
> gpio.  The pi's pin drivers are just barely up to the task, and too 
> fragile electrically for true long term survival if the noise or edge 
> ringing exceeds 4 volts.  Ditto the i/o pins on the 7i90 and its ilk for 
> that matter.

If distance is long enough drivers are required it might be time to switch for 
Ethernet based or CAN.

Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Gene Heskett]

On Thursday 14 September 2017 11:47:22 Nicklas Karlsson wrote:

> On Thu, 14 Sep 2017 04:03:30 -0400
>
> Gene Heskett  wrote:
> > On Wednesday 13 September 2017 12:50:15 Nicklas Karlsson wrote:
> > > > > No SPI over long cable. For long cable RS422 will work, there
> > > > > is a tradeoff between cable length and speed.
> > > >
> > > > Poor wording on my part. I had, once the terminations are
> > > > sorted, visions of a 3 to 4 foot cable, defined as a long SPI
> > > > cable.
> > > >
> > > > Right now, terms are not well characterized on either end and it
> > > > takes some loading trickery to make a 6 inch cable work.  Bertho
> > > > and I spent some time optimizing it to get 30.5 megabit one way,
> > > > and 25 megabit the other direction.  And its now working well to
> > > > a Mesa 7i90 with a 1" cable.
> > >
> > > Ethernet shielded twisted pair four pairs like this: sck+gnd,
> > > miso+gnd, mosi+gnd, chip select + gnd. Or flat cable with a gnd
> > > cable between each of them. Or of course Ethernet if available.
> >
> > No, the tcp is too slow, much latency.
>
> For TCP resend function to make sense there must be enough time to
> resend packet.
>
> > However, the above scheme, made into ready-made hardware and a
> > matching rj45 socket signal assignment on those mesa cards that can
> > support spi, like the 7i90, ...
>
> Ethercat chips are available with SPI interface and I concluded then
> distance is short enough and data is sent between it might be a good
> choice to skip the Ethercat chips but I might be wrong.
>
I'd be for that skip myself, if we had real line drivers in the pi's 
gpio.  The pi's pin drivers are just barely up to the task, and too 
fragile electrically for true long term survival if the noise or edge 
ringing exceeds 4 volts.  Ditto the i/o pins on the 7i90 and its ilk for 
that matter.

> Nicklas Karlsson
>
> --
> Check out the vibrant tech community on one of the world's
> most engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users


Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page 

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

On Thu, 14 Sep 2017 04:03:30 -0400
Gene Heskett  wrote:

> On Wednesday 13 September 2017 12:50:15 Nicklas Karlsson wrote:
> 
> > > > No SPI over long cable. For long cable RS422 will work, there is a
> > > > tradeoff between cable length and speed.
> > >
> > > Poor wording on my part. I had, once the terminations are sorted,
> > > visions of a 3 to 4 foot cable, defined as a long SPI cable.
> > >
> > > Right now, terms are not well characterized on either end and it
> > > takes some loading trickery to make a 6 inch cable work.  Bertho and
> > > I spent some time optimizing it to get 30.5 megabit one way, and 25
> > > megabit the other direction.  And its now working well to a Mesa
> > > 7i90 with a 1" cable.
> >
> > Ethernet shielded twisted pair four pairs like this: sck+gnd,
> > miso+gnd, mosi+gnd, chip select + gnd. Or flat cable with a gnd cable
> > between each of them. Or of course Ethernet if available.
> 
> No, the tcp is too slow, much latency.

For TCP resend function to make sense there must be enough time to resend 
packet.

> However, the above scheme, made into ready-made hardware and a matching 
> rj45 socket signal assignment on those mesa cards that can support spi, 
> like the 7i90, ...

Ethercat chips are available with SPI interface and I concluded then distance 
is short enough and data is sent between it might be a good choice to skip the 
Ethercat chips but I might be wrong.


Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

> > No SPI over long cable. For long cable RS422 will work, there is a
> > tradeoff between cable length and speed.
> >
> Poor wording on my part. I had, once the terminations are sorted, visions 
> of a 3 to 4 foot cable, defined as a long SPI cable.
> 
> Right now, terms are not well characterized on either end and it takes 
> some loading trickery to make a 6 inch cable work.  Bertho and I spent 
> some time optimizing it to get 30.5 megabit one way, and 25 megabit
> the other direction.  And its now working well to a Mesa 7i90 with a 1" 
> cable.

Ethernet shielded twisted pair four pairs like this: sck+gnd, miso+gnd, 
mosi+gnd, chip select + gnd. Or flat cable with a gnd cable between each of 
them. Or of course Ethernet if available.

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[Nicklas Karlsson]

> when I turned Lcnc & powerlink together I will post the results on the
> forum.

I think both are more than good enough but Powerlink will be cheaper because 
only one Ethernet port is needed.


> BUt I see that you try to implement canopen over rs422 mesa card? these is
> a nice project.

Well actually I use the CANopen message format just to get a standard format. I 
also use it over UDP/IP, the SDO communication is very useful for configuration.


--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[theman whosoldtheworld]

wikipedia explain better  the difference between field bus  ethercat is
better but need ET110 on slaves nodes + master & master in same deasy chain it
is not well digested  powerlink not have the same performance because
no particular chip is required on master or into slaves  most masters
are tolerated & GigaByte ethernet is implemented.

Says these, even these are things to consider  anyway my idea of
​​using cc-link ie is not bad either because the latest developments.

when I turned Lcnc & powerlink together I will post the results on the
forum.


BUt I see that you try to implement canopen over rs422 mesa card? these is
a nice project.


bkt

2017-09-12 20:33 GMT+02:00 Nicklas Karlsson :

> > for work buy device in china every mount from 3 year ... so I have my
> > little opinion ... You are in right about china ... but actually the
> > producer is focused on ethercat (due to state research grants and because
> > ethercat is other standards fieldbus). So that the future of powerlink is
> > much more connected to Europe and to USA than to china.
>
> As I understand Ethercat is daisy chain, like shift registers connected in
> series with hardware support to change bits while they are shifted thru and
> calcuta CRC. This make possible sending from several nodes in each message
> but there is a need for two Ethernet ports which add cost. There is an
> advantage with two rings in opposite direction for several reasons but it
> also increas price.
>
> As I understand Powerlink is the ordinary switch based approach, only one
> Ethernet port is needed but then sending from several devices more messages
> per second is required.
>
>
> Nicklas Karlsson
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[Nicklas Karlsson]

> for work buy device in china every mount from 3 year ... so I have my
> little opinion ... You are in right about china ... but actually the
> producer is focused on ethercat (due to state research grants and because
> ethercat is other standards fieldbus). So that the future of powerlink is
> much more connected to Europe and to USA than to china.

As I understand Ethercat is daisy chain, like shift registers connected in 
series with hardware support to change bits while they are shifted thru and 
calcuta CRC. This make possible sending from several nodes in each message but 
there is a need for two Ethernet ports which add cost. There is an advantage 
with two rings in opposite direction for several reasons but it also increas 
price.

As I understand Powerlink is the ordinary switch based approach, only one 
Ethernet port is needed but then sending from several devices more messages per 
second is required.


Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[theman whosoldtheworld]

>Chinas is very large and they export a lot of electronic products. It is
however a rather new phenomen, Shenzen was the first exprimental economic
zone opened in 1980, the area is a little bit larger than the "county" or
>similar there I grow up in countryside and in 1980 I think population was
similar around 30 000 but right now there is more than 10 million people.
The population growth in the pearl river delta have been insane and
>economy for Guangdong province have grown more than 100 times.

for work buy device in china every mount from 3 year ... so I have my
little opinion ... You are in right about china ... but actually the
producer is focused on ethercat (due to state research grants and because
ethercat is other standards fieldbus). So that the future of powerlink is
much more connected to Europe and to USA than to china.


>Requirement on performance is not very high. It is almost always need for
between three and six axis which will be around 6*32bits = 192 bits of
payload each period, IO may be slower. A more reasonable amount is
>8*64bits @ 1kHz = 512bit @ 1kHz = 512kbit/s of payload at 1kHz. All
counted full duplex.
>
>With control loop at linuxcnc computer there and one driver card for each
motor there would however be a need in the range of 2*3*1kHz to 2*8*1kHz
messages per second. Or (1+3)*1kHz to (1+8)*1kHz if multi casting is >used.


ok ... nice ... but I was referring to the fact that it needs an isolated
rt thread because powerlink works fine so that you will need to observe cpu
behavior first (my favorite is the I3 ... it is true that for a few moments
they are from 3.6Ghz.) to make a real powerlink-Lcnc robot.

regards
bkt

2017-09-12 18:31 GMT+02:00 Nicklas Karlsson :

> > Yes in china multiple manufactorer (drive) has powerlink ... but in
> Europe
> > the thing is different, ...
>
> Chinas is very large and they export a lot of electronic products. It is
> however a rather new phenomen, Shenzen was the first exprimental economic
> zone opened in 1980, the area is a little bit larger than the "county" or
> similar there I grow up in countryside and in 1980 I think population was
> similar around 30 000 but right now there is more than 10 million people.
> The population growth in the pearl river delta have been insane and economy
> for Guangdong province have grown more than 100 times.
>
> > said this last month I wanted to try powerlink and Lcnc together to be
> able
> > to observe the performance of a system with both running instances. ...
>
> Requirement on performance is not very high. It is almost always need for
> between three and six axis which will be around 6*32bits = 192 bits of
> payload each period, IO may be slower. A more reasonable amount is 8*64bits
> @ 1kHz = 512bit @ 1kHz = 512kbit/s of payload at 1kHz. All counted full
> duplex.
>
> With control loop at linuxcnc computer there and one driver card for each
> motor there would however be a need in the range of 2*3*1kHz to 2*8*1kHz
> messages per second. Or (1+3)*1kHz to (1+8)*1kHz if multi casting is used.
>
>
> Regards Nicklas Karlsson
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

> > No SPI over long cable. For long cable RS422 will work, there is a
> > tradeoff between cable length and speed.
> >
> Poor wording on my part. I had, once the terminations are sorted, visions 
> of a 3 to 4 foot cable, defined as a long SPI cable.

Yes it is probably a long SPI cable but it is possible to slow down speed. The 
Master In Slave Out MISO signal is a problem because there must be enough time 
for this signal to return before the next SCK clock flank.

Within circuit board SPI provide cheap and fast communication. I have need for 
more than one instance of the same circuit and planned to use several single 
instance cards, here it might also work.


> Right now, terms are not well characterized on either end and it takes 
> some loading trickery to make a 6 inch cable work.  Bertho and I spent 
> some time optimizing it to get 30.5 megabit one way, and 25 megabit
> the other direction.  And its now working well to a Mesa 7i90 with a 1" 
> cable.

If you have two SPI ports and do not use the MISO signal return path speed will 
increase. It might also be possible to skip the chip select signal and instead 
detect absence of clock signal as packet delimiter but it might be harder to 
implement than anticipated at least on some devices.

UART might be good if it is possible to get the correct bit rate at high speed, 
this might be a problem if there i n counts per bit. It is usually also more 
limited then it come to speed.


Regards Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[Nicklas Karlsson]

> Yes in china multiple manufactorer (drive) has powerlink ... but in Europe
> the thing is different, ...

Chinas is very large and they export a lot of electronic products. It is 
however a rather new phenomen, Shenzen was the first exprimental economic zone 
opened in 1980, the area is a little bit larger than the "county" or similar 
there I grow up in countryside and in 1980 I think population was similar 
around 30 000 but right now there is more than 10 million people. The 
population growth in the pearl river delta have been insane and economy for 
Guangdong province have grown more than 100 times.

> said this last month I wanted to try powerlink and Lcnc together to be able
> to observe the performance of a system with both running instances. ...

Requirement on performance is not very high. It is almost always need for 
between three and six axis which will be around 6*32bits = 192 bits of payload 
each period, IO may be slower. A more reasonable amount is 8*64bits @ 1kHz = 
512bit @ 1kHz = 512kbit/s of payload at 1kHz. All counted full duplex.

With control loop at linuxcnc computer there and one driver card for each motor 
there would however be a need in the range of 2*3*1kHz to 2*8*1kHz messages per 
second. Or (1+3)*1kHz to (1+8)*1kHz if multi casting is used.


Regards Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[theman whosoldtheworld]

@ Gene

your link but there are a lot of other ... I means that powerlink -> B
become part of ABB  so abb become manufactorer of plc and powerlink
device (abb have poverlink on yours drive  actully).
https://technology.ihs.com/591442/will-br-acquisition-give-abb-a-boost-in-industrial-automation

for 10Gb ethernet please see cc-link ie ..

bkt

2017-09-12 2:09 GMT+02:00 Gene Heskett :

> On Monday 11 September 2017 19:23:37 Chris Albertson wrote:
>
> > Why not SPI speeds over a long cable?   Gigabit Ether net is a good
> > example.   It is built into almost all computers and can be 1GHz over
> > 100 meters of cheap unshielded wire.
> >
> > They use only a few basic tricks, none of which add much to the cost
> >
> > 10/8 encoding to guarantee a minimum number of bit transitions even if
> > the data is a long string of zeroes.
> > Clock mixed with data so the receiver clock can be cheap non-precision
> > Differential to resister common mode noise
> > galvanic isolation to prevent ground loops
> >
> > There is also 10GHz Ethernet but it is not built into common consumer
> > equipment but even 10G data rates are easy to pushover 100 feet using
> > just RJ45 jacks
> >
> > The trick is to NOT think about how hard it is to push 10Gbps over a
> > wire. But to think how easy it is to build a microwave data link when
> > you can use wire in place of a clear air path.  No need for antennas
> > and you can use way ultra-low power.
> >
> > even for the case of SPI, think in terms is radio transmission lines
> > and the job becomes easier.
> >
> > On Mon, Sep 11, 2017 at 9:54 AM, Nicklas Karlsson <
> >
> > nicklas.karlsso...@gmail.com> wrote:
> > > > On Monday 11 September 2017 07:16:00 theman whosoldtheworld wrote:
> > > > > why not powerlink ie?
> > > >
> > > > Wikipedia description shows some promise. SPI speeds over a longer
> > > > cable, maybe.
> > >
> > > No SPI over long cable. For long cable RS422 will work, there is a
> > > tradeoff between cable length and speed.
> > >
> Or for mind boggling distances, how about a g-line? Virtually zero losses
> over a mile or 3 using common steel wire, single strand? No real reason
> it cannot be bidi by using 2 non-harmonically related carrier
> frequencies.
>
> 2 secrets to both low loss and low cost since the wire's ohmic losses
> aren't a major concern, its nothing but a waveguide with no external
> walls. The other is that when it goes past a supporting pole, it must be
> tight so the the included angle where its attached to the supporting
> insulator is less than 5 degrees, as a certain amount of the waveform
> energy will be radiated and lost as it continues in a straight line,
> leaving the confines of the field established by the central wire.
>
> This central wire exits from the center of a flat conductive panel that
> is 1 or more wavelengths in radius.  And re-enters the rx end in a
> similar structure.
>
> Or with small dishes and "gunn" diodes, which are innately bidirectional,
> the wire can be dispensed with. We have a remote broadcast linkage
> lashup running at 23 gigahertz, using 12" dishes on both ends, that has,
> when our satellite truck is on location up to several miles, I think 22
> is the record, which has saved the station its cost in satellite time
> not rented many times over.  If we can clear line of site see the
> studio's 255 foot STL towers top beacon we fire it up and cancel the
> satellite time.  Our unit isn't bidirectional, but thats a matter of
> changing out its video circuitry on both ends. Power output is 60
> milliwatts.
>
> If the sheckles can be found, distance vs bandwidth is a solveable
> problem. In this case the g-lines initial construction costs for
> planting all the support poles, and the cost of the gunn diode
> assemblies, both up front costs, are at best, still discouraging.
>
> Cheers, Gene Heskett
> --
> "There are four boxes to be used in defense of liberty:
>  soap, ballot, jury, and ammo. Please use in that order."
> -Ed Howdershelt (Author)
> Genes Web page 
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[theman whosoldtheworld]

Yes in china multiple manufactorer (drive) has powerlink ... but in Europe
the thing is different, producers prefer to stay away from this protocol
they see as a competitor of current protocols too far out of the lines 
maybe now that it has become part of the galaxy ABB, things will start to
change ... we could only say it with fairs in May 2018 where we can count
the new powerlink implementations on existing devices from March 2017.

Any how I find that powerlink & opensafety is a good protocol for open
modern application ... Actuaqlly there are cc-link ie with 1Gb/s realtime
on normal gigabyte ethernet. in June or July not well remember, cc-link ie
and profinet have signed an agreement for convergence. cc-link ie has
recently signed up to OSDAL and has issued the codes is therefore an open
platform ... but has very few that implement it. Yet it remains the best
protocol candidate agaist fiber optics protocol who personally for the next
10 years at least I see very far from become open.

said this last month I wanted to try powerlink and Lcnc together to be able
to observe the performance of a system with both running instances. But I
did not have time. For now I just turned powerlink userspace on a pc 
it works fine. maybe I'll have time this month even if I'm already devoting
myself to an ethercat project.
That being said, the biggest problem for me is not the drives ... but the
plc ... I did not find European or American builders who use the protocol,
unless we're b & r.

if others want to try you can start from this great site ... many have
already applied applications with small PCs that go out of fashion now ...
https://sourceforge.net/p/openpowerlink/discussion/


bkt

2017-09-11 17:55 GMT+02:00 Nicklas Karlsson :

> On Mon, 11 Sep 2017 13:16:00 +0200
> theman whosoldtheworld  wrote:
>
> > why not powerlink ie? no need special chip, no need special hardware ...
> > but it run on every device ... so you can must to limit it to a single
> core
> > realtime working ... so other core is free for Lcnc ... not tested yet
> with
> > Lcnc istances so I'm not able to report latency performance with Lcnc
> > istance + powerlink ie istances ... its only limit is that there are few
> > device manufacturers who use it, ...
>
> It might actually be a good choice and as POWERLINK Ethernet protocol
> comply with authorized standardization level for communication technologies
> in China plenty of devices could be expected.
>
> https://www.ceasiamag.com/2012/05/powerlink-awarded-
> national-standard-in-china/
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Gene Heskett]

On Monday 11 September 2017 19:23:37 Chris Albertson wrote:

> Why not SPI speeds over a long cable?   Gigabit Ether net is a good
> example.   It is built into almost all computers and can be 1GHz over
> 100 meters of cheap unshielded wire.
>
> They use only a few basic tricks, none of which add much to the cost
>
> 10/8 encoding to guarantee a minimum number of bit transitions even if
> the data is a long string of zeroes.
> Clock mixed with data so the receiver clock can be cheap non-precision
> Differential to resister common mode noise
> galvanic isolation to prevent ground loops
>
> There is also 10GHz Ethernet but it is not built into common consumer
> equipment but even 10G data rates are easy to pushover 100 feet using
> just RJ45 jacks
>
> The trick is to NOT think about how hard it is to push 10Gbps over a
> wire. But to think how easy it is to build a microwave data link when
> you can use wire in place of a clear air path.  No need for antennas
> and you can use way ultra-low power.
>
> even for the case of SPI, think in terms is radio transmission lines
> and the job becomes easier.
>
> On Mon, Sep 11, 2017 at 9:54 AM, Nicklas Karlsson <
>
> nicklas.karlsso...@gmail.com> wrote:
> > > On Monday 11 September 2017 07:16:00 theman whosoldtheworld wrote:
> > > > why not powerlink ie?
> > >
> > > Wikipedia description shows some promise. SPI speeds over a longer
> > > cable, maybe.
> >
> > No SPI over long cable. For long cable RS422 will work, there is a
> > tradeoff between cable length and speed.
> >
Or for mind boggling distances, how about a g-line? Virtually zero losses 
over a mile or 3 using common steel wire, single strand? No real reason 
it cannot be bidi by using 2 non-harmonically related carrier 
frequencies.

2 secrets to both low loss and low cost since the wire's ohmic losses 
aren't a major concern, its nothing but a waveguide with no external 
walls. The other is that when it goes past a supporting pole, it must be 
tight so the the included angle where its attached to the supporting 
insulator is less than 5 degrees, as a certain amount of the waveform 
energy will be radiated and lost as it continues in a straight line, 
leaving the confines of the field established by the central wire.

This central wire exits from the center of a flat conductive panel that 
is 1 or more wavelengths in radius.  And re-enters the rx end in a 
similar structure.

Or with small dishes and "gunn" diodes, which are innately bidirectional, 
the wire can be dispensed with. We have a remote broadcast linkage 
lashup running at 23 gigahertz, using 12" dishes on both ends, that has, 
when our satellite truck is on location up to several miles, I think 22 
is the record, which has saved the station its cost in satellite time 
not rented many times over.  If we can clear line of site see the 
studio's 255 foot STL towers top beacon we fire it up and cancel the 
satellite time.  Our unit isn't bidirectional, but thats a matter of 
changing out its video circuitry on both ends. Power output is 60 
milliwatts.

If the sheckles can be found, distance vs bandwidth is a solveable 
problem. In this case the g-lines initial construction costs for 
planting all the support poles, and the cost of the gunn diode 
assemblies, both up front costs, are at best, still discouraging.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page 

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Chris Albertson]

Why not SPI speeds over a long cable?   Gigabit Ether net is a good
example.   It is built into almost all computers and can be 1GHz over 100
meters of cheap unshielded wire.

They use only a few basic tricks, none of which add much to the cost

10/8 encoding to guarantee a minimum number of bit transitions even if the
data is a long string of zeroes.
Clock mixed with data so the receiver clock can be cheap non-precision
Differential to resister common mode noise
galvanic isolation to prevent ground loops

There is also 10GHz Ethernet but it is not built into common consumer
equipment but even 10G data rates are easy to pushover 100 feet using just
RJ45 jacks

The trick is to NOT think about how hard it is to push 10Gbps over a wire.
  But to think how easy it is to build a microwave data link when you can
use wire in place of a clear air path.  No need for antennas and you can
use way ultra-low power.

even for the case of SPI, think in terms is radio transmission lines and
the job becomes easier.

On Mon, Sep 11, 2017 at 9:54 AM, Nicklas Karlsson <
nicklas.karlsso...@gmail.com> wrote:

> > On Monday 11 September 2017 07:16:00 theman whosoldtheworld wrote:
> >
> > > why not powerlink ie?
> >
> > Wikipedia description shows some promise. SPI speeds over a longer cable,
> > maybe.
>
> No SPI over long cable. For long cable RS422 will work, there is a
> tradeoff between cable length and speed.
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>



-- 

Chris Albertson
Redondo Beach, California
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Gene Heskett]

On Monday 11 September 2017 12:54:46 Nicklas Karlsson wrote:

> > On Monday 11 September 2017 07:16:00 theman whosoldtheworld wrote:
> > > why not powerlink ie?
> >
> > Wikipedia description shows some promise. SPI speeds over a longer
> > cable, maybe.
>
> No SPI over long cable. For long cable RS422 will work, there is a
> tradeoff between cable length and speed.
>
Poor wording on my part. I had, once the terminations are sorted, visions 
of a 3 to 4 foot cable, defined as a long SPI cable.

Right now, terms are not well characterized on either end and it takes 
some loading trickery to make a 6 inch cable work.  Bertho and I spent 
some time optimizing it to get 30.5 megabit one way, and 25 megabit
the other direction.  And its now working well to a Mesa 7i90 with a 1" 
cable.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page 

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

> On Monday 11 September 2017 07:16:00 theman whosoldtheworld wrote:
> 
> > why not powerlink ie?
> 
> Wikipedia description shows some promise. SPI speeds over a longer cable, 
> maybe.

No SPI over long cable. For long cable RS422 will work, there is a tradeoff 
between cable length and speed.

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol --> Powerlink

[Nicklas Karlsson]

On Mon, 11 Sep 2017 13:16:00 +0200
theman whosoldtheworld  wrote:

> why not powerlink ie? no need special chip, no need special hardware ...
> but it run on every device ... so you can must to limit it to a single core
> realtime working ... so other core is free for Lcnc ... not tested yet with
> Lcnc istances so I'm not able to report latency performance with Lcnc
> istance + powerlink ie istances ... its only limit is that there are few
> device manufacturers who use it, ...

It might actually be a good choice and as POWERLINK Ethernet protocol comply 
with authorized standardization level for communication technologies in China 
plenty of devices could be expected.

https://www.ceasiamag.com/2012/05/powerlink-awarded-national-standard-in-china/

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Gene Heskett]

On Monday 11 September 2017 07:16:00 theman whosoldtheworld wrote:

> why not powerlink ie?

Wikipedia description shows some promise. SPI speeds over a longer cable, 
maybe.

> no need special chip, no need special hardware 
> ... but it run on every device ... so you can must to limit it to a
> single core realtime working ... so other core is free for Lcnc ...
> not tested yet with Lcnc istances so I'm not able to report latency
> performance with Lcnc istance + powerlink ie istances ... its only
> limit is that there are few device manufacturers who use it, we see it
> will happen next year as it has now become an ABB protocol.

Unfortunately, wikipedia never heard of the ABB protocol.  Other links 
and cites please?

Thanks.

> rebards
> bkt
>
> 2017-09-10 11:10 GMT+02:00 Nicklas Karlsson 
:
> > > In practice if you have ever had to set up a daisy chain topology
> > > network you will not like it.  One problem is that one connector
> > > or device
> >
> > failure
> >
> > >  takes down the entire network.   The new switch based Ethernet is
> > > much more reliable especially now with self test and green LED
> > > built into
> >
> > every
> >
> > > end or every cable.
> >
> > Plan was to connect like shift register which should be very similar
> > to Ethercat. What's worse is micro controller have FIFO so it is not
> > possible to feed thru with good performance since virtual shift
> > register length will be long in each device. Green LED should be
> > possible also with daisy chain but there is a drawback all devices
> > must work.
> >
> > Single chip will of course always work. For buses Ethercat for high
> > bandwidth and CANopen for low bandwidth seems to be the choice. I
> > think there is nothing wrong with Sercos but it seems to be more
> > expensive for less performance.
> >
> >
> > Regards Nicklas Karlsson
> >
> > 
> > --
> > Check out the vibrant tech community on one of the world's most
> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> > ___
> > Emc-users mailing list
> > Emc-users@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/emc-users
>
> --
> Check out the vibrant tech community on one of the world's
> most engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users


Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page 

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[theman whosoldtheworld]

why not powerlink ie? no need special chip, no need special hardware ...
but it run on every device ... so you can must to limit it to a single core
realtime working ... so other core is free for Lcnc ... not tested yet with
Lcnc istances so I'm not able to report latency performance with Lcnc
istance + powerlink ie istances ... its only limit is that there are few
device manufacturers who use it, we see it will happen next year as it has
now become an ABB protocol.

rebards
bkt

2017-09-10 11:10 GMT+02:00 Nicklas Karlsson :

> > In practice if you have ever had to set up a daisy chain topology network
> > you will not like it.  One problem is that one connector or device
> failure
> >  takes down the entire network.   The new switch based Ethernet is much
> > more reliable especially now with self test and green LED built into
> every
> > end or every cable.
>
> Plan was to connect like shift register which should be very similar to
> Ethercat. What's worse is micro controller have FIFO so it is not possible
> to feed thru with good performance since virtual shift register length will
> be long in each device. Green LED should be possible also with daisy chain
> but there is a drawback all devices must work.
>
> Single chip will of course always work. For buses Ethercat for high
> bandwidth and CANopen for low bandwidth seems to be the choice. I think
> there is nothing wrong with Sercos but it seems to be more expensive for
> less performance.
>
>
> Regards Nicklas Karlsson
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

> In practice if you have ever had to set up a daisy chain topology network
> you will not like it.  One problem is that one connector or device failure
>  takes down the entire network.   The new switch based Ethernet is much
> more reliable especially now with self test and green LED built into every
> end or every cable.

Plan was to connect like shift register which should be very similar to 
Ethercat. What's worse is micro controller have FIFO so it is not possible to 
feed thru with good performance since virtual shift register length will be 
long in each device. Green LED should be possible also with daisy chain but 
there is a drawback all devices must work.

Single chip will of course always work. For buses Ethercat for high bandwidth 
and CANopen for low bandwidth seems to be the choice. I think there is nothing 
wrong with Sercos but it seems to be more expensive for less performance.


Regards Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

> In practice if you have ever had to set up a daisy chain topology network
> you will not like it.  One problem is that one connector or device failure
>  takes down the entire network.   The new switch based Ethernet is much
> more reliable especially now with self test and green LED built into every
> end or every cable.

For between machine communication and user interface I will use ordinary 
Ethernet and maybe WiFi. For realtime I am pretty suer Ethercat should be a 
better choice for a similar price but I have not yet read all details about the 
protocol.


Regards Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

> In practice if you have ever had to set up a daisy chain topology network
> you will not like it.  One problem is that one connector or device failure
>  takes down the entire network.   The new switch based Ethernet is much
> more reliable especially now with self test and green LED built into every
> end or every cable.

I give up on ordinary Ethernet, Sercos might be good but no one seems to care 
so Ethercat seems to be the choice.


> Talk about cables.  I just got email from Leadshine and Stepper-on-line
> that they are promoting a new line of stepper motor with integrated
> driver.  The driver is permanently fixed to the motor housing.  This means
> two things (1) the driver is optimized for that exact motor and (2) there
> are no high current/voltage cables needed you run DC power and digital
> step/direction all the way to the motor.

Might be good for them but where do I earn my money? To use my chain saw and 
sell trees is an option but I avoid it if possible, machines earn more money 
per hour if there is enough trees.

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Re: [Emc-users] Serial daisy chain protocol

[Chris Albertson]

In practice if you have ever had to set up a daisy chain topology network
you will not like it.  One problem is that one connector or device failure
 takes down the entire network.   The new switch based Ethernet is much
more reliable especially now with self test and green LED built into every
end or every cable.

Talk about cables.  I just got email from Leadshine and Stepper-on-line
that they are promoting a new line of stepper motor with integrated
driver.  The driver is permanently fixed to the motor housing.  This means
two things (1) the driver is optimized for that exact motor and (2) there
are no high current/voltage cables needed you run DC power and digital
step/direction all the way to the motor.

They also have new series of closed loop steppers but information and price
was not available yet

On Mon, Sep 4, 2017 at 8:56 AM, Nicklas Karlsson <
nicklas.karlsso...@gmail.com> wrote:

> To daisy chain serial communication in a ring is well known old technology
> which may have good real time performance. It is sometimes used for SPI and
> JTAG devices. It is also used in the DMX protocol. Not the least it is used
> on the new state of the of the art Ethercat protocol. Do anybody know about
> a serial daisy chain protocol?
>
> I figured to send CANopen packages would work rather well but a standard
> protocol is better. I have a choice to send exactly one message to each
> node every time or allow to insert a few extra which may be used for
> CANopen SDO messages but add complexity. Low overhead is good but if more
> than half of bandwidth may be used for payload there is not to much to gain.
>
> Even though Ethercat is a good protocol I still think there is a place for
> UART and SPI not the least for communication within circuit board and a
> daisy protocol would work around the problem there is not enough available
> ports available then only a fraction of the bandwidth is used.
>
>
> Regards Nicklas Karlsson
>
> 
> --
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> ___
> Emc-users mailing list
> Emc-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-users
>



-- 

Chris Albertson
Redondo Beach, California
--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

[Emc-users] Serial daisy chain protocol

[Nicklas Karlsson]

To daisy chain serial communication in a ring is well known old technology 
which may have good real time performance. It is sometimes used for SPI and 
JTAG devices. It is also used in the DMX protocol. Not the least it is used on 
the new state of the of the art Ethercat protocol. Do anybody know about a 
serial daisy chain protocol?

I figured to send CANopen packages would work rather well but a standard 
protocol is better. I have a choice to send exactly one message to each node 
every time or allow to insert a few extra which may be used for CANopen SDO 
messages but add complexity. Low overhead is good but if more than half of 
bandwidth may be used for payload there is not to much to gain.

Even though Ethercat is a good protocol I still think there is a place for UART 
and SPI not the least for communication within circuit board and a daisy 
protocol would work around the problem there is not enough available ports 
available then only a fraction of the bandwidth is used.


Regards Nicklas Karlsson

--
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
___
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users