Multiple signals on single Serial TX line

[Syd Rumpo]

That was my first intention, but it takes a new microcontroller and the code for the serial protocol has already been written and this would not be welcome by the sw guys

I could use a small microcontroller to act as the glitch detection, but that would add another microcontroller to be programmed which adds production costs

Cheers

Klaus

This should be transparent to any existing protocol, except in that it
would add a small delay. I'm adding a pair of uC's across the barrier.
All they're doing is converting and passing on the existing serial
comms at a higher speed so that the extra digital line can be added.

Unless I've misunderstood the problem, which is quite likely, this
solution is easy. I would say 'trivial' but every time I use that word,
extra work seems to appear from nowhere.

AIUI now we have...
115k tx--->barrier--->115k rx

But we could have...
115k tx--->Syd's uC1--->barrier--->Syd's uC2-->115k rx

So uC1 receives the existing 115k serial data and sends it at a higher
speed across the barrier. UC2 receives this higher speed serial signal
and sends it on at 115k.

uC1 also has a digital input for the extra signal. uC2 has a
corresponding digital output.

Say, for example, the new barrier rate is 500kbaud, then we could send
alternate bytes across the barrier using a 9-bit protocol. With bit 9
set these could be re-transmitted transparently by uC2 at 115k, with bit
9 clear, these could be your digital data.

Cheers

 

[Klaus Kragelund]

So do the ADI iCouplers. Quads, even. RS-232 and RS-488, too. ;-)

The ADUM2201 is considered, has lower current than the Si8421

/Klaus

 

[[email protected]]

On a sunny day (Tue, 15 Jan 2013 22:37:40 +0200) it happened
[email protected] wrote in
<[email protected]>:

Yea, we had to replace a kilometer of screened RS232 in a factory
with glass fiber as the max usable baudrate dropped to 75 Bd
due to cable capacitance, I have measured it.

You made the same mistake as those building the first trans-Atlantic
telegraph cable. They tried to increase the signaling rate by
increasing the voltage levels, finally destroying the cable.

The RS-232 model is a high voltage, high impedance voltage source
driving a capacitive load. Ultimately, this will kill the performance
quite rapidly.

Low voltage signaling, such as RS-422/485 use much lower signaling
voltages and hence much less current is needed to charge and discharge
the line capacitance. The low +/-200 mV receiver threshold would allow
much higher line speeds even when looking at the situation as a simple
voltage source and capacitive load case.

Of course, running 115k2 at 1 km relies on the transmission line
approach.

IIRC ADSL does equalizing, and probably is corectly terminated.

If you do a frequency scan for some residential telephone pairs, there
are going to be some deep notches at certain frequencies due to 1/4
wave stubs in the wiring system. The ADSL system will ignore these
subcarriers as well as those subcarriers suffering from heavy external
interference, such as local medium wave radio broadcasts. Thus the
real ADSL throughput is less than advertized.

RS232 is NOT terminated, so the cable is used pure capacitive and charged
by an approx. 10mA current source.
Now you can calculate the rise time.

My suggestion to my customers is to use RS-232 only within the same
equipment rack, partly due to the speed issues, but mainly due to
ground potential issues.

 

[[email protected]]

I did not make that mistake.
I was called as consultant to help PLC guys that had a problem
with communication in some factory.
Took a day of travel, and 5 minutes with a scope to find the problem.
Reported it, made a screenshot, showed and explained it to big boss,
made some recommendations,
and then they decided to replace it with glassfiber.

Sorry for my bad English, by "you" (plural) I was referring to the
original situation.

Interference is an other thing, especially in an industrial environment
where large machines are switched on and of, and you get magnetic induction,
ground currents too.
AFAIK from that moment on they used glassfiber for all communications.

When working with power plant people, they always demand fibers, even
if 1 kV galvanic optoisolation would be quite sufficient.

 

[Jasen Betts]

Hi

We are using an Icoupler to transfer TX serial data across a barrier with 115kBaud

I need to transfer a digital signal with low bandwidth across the same barrier

115200 NRZ has a null in the spectrum at 115200Hz, so you can put yout sidechannel
signal on that frequency.

if you dothe 115K2 BPS in software and add glitches into the
stream when you need to generate the sidechannel signal.

eg if every bit is 16 "uart clocks" long
during a 0 in the main signal send a 1 during the fourth clock and
during a 1 send a 0 half a bit-time later - in the twelveth clock

that'll show up as a 115200Hz tone which is right in the first trough
of the NRZ spectrum, you can detect it with a pulse detector, diode
pump, or even a tuned circuit

Any decent UARt will ignore it, or you can filter it out using a low
pass.

 

[Klaus Kragelund]

Klaus,



Are the two data streams as simplistic as shown on your drawing? Or

can they assume other patterns?

The datastream is like the one from an UART TX channel. 11 bits, start bit,data and stop bits. Baudrate can be anything from 2400 to 115200.

The second signal that needs to be combined with the TX signal just needs aresponse time of max 1ms.

Cheers

Klaus

 

[Ralph Barone]

Sorry for my bad English, by "you" (plural) I was referring to the
original situation.


When working with power plant people, they always demand fibers, even
if 1 kV galvanic optoisolation would be quite sufficient.

For the stuff I work with (power system protective relaying), copper based
communications is only allowed inside a panel. If it leaves the panel, it
gets converted to fibre.

 

[Ralph Barone]

The datastream is like the one from an UART TX channel. 11 bits, start
bit, data and stop bits. Baudrate can be anything from 2400 to 115200.

The second signal that needs to be combined with the TX signal just needs
a response time of max 1ms.

Cheers

Klaus

Can you do a minor modulation of the width of the start or stop bits that
is small enough that it doesn't affect the downstream UART. At 115 kbps
there should be enough packets to send one bit every ms with a goodly pile
of redundancy. Unfortunately, it might take more space than a second
optocoupler.

 

[[email protected]]

Can you do a minor modulation of the width of the start or stop bits that
is small enough that it doesn't affect the downstream UART. At 115 kbps
there should be enough packets to send one bit every ms with a goodly pile
of redundancy. Unfortunately, it might take more space than a second
optocoupler.

....and money and power. I think these avenues have already been
explored.

 

[Klaus Kragelund]

...and money and power. I think these avenues have already been

explored.

The alternative would be to try another direction removing the iCoupler andOpto and using a transformer instead (the Pulse PH9185 is approved for thecorrect isolation and has the same area as the iCoupler)

Signaling from primary to secondary would then be at two different frequencies with some filters to "decode" the signals and communication the other way could be a current modulation scheme or injection of a third frequency.

It does however quickly become complex, but the upside is that current consumption may possibly be reduced.

Immunity to common mode signals is another matter, the ADUM2201m has immunity of 35kV/us. The PH9185 has 8pF capacitance from primary to secondary, soa dV/dt of this rate will inject more than 100mA into the rail of the primary side which could affect the communication related to how well the capacitance is coupled to the primary side and if it is symmetric coupled

Anyony been down this road, using a transformer for data transmission and observations on the effects of common mode transients?

Cheers

Klaus

 

[Klaus Kragelund]

The alternative would be to try another direction removing the iCoupler and Opto and using a transformer instead (the Pulse PH9185 is approved for the correct isolation and has the same area as the iCoupler)



Signaling from primary to secondary would then be at two different frequencies with some filters to "decode" the signals and communication the otherway could be a current modulation scheme or injection of a third frequency..



It does however quickly become complex, but the upside is that current consumption may possibly be reduced.



Immunity to common mode signals is another matter, the ADUM2201m has immunity of 35kV/us. The PH9185 has 8pF capacitance from primary to secondary, so a dV/dt of this rate will inject more than 100mA into the rail of the primary side which could affect the communication related to how well the capacitance is coupled to the primary side and if it is symmetric coupled



Anyony been down this road, using a transformer for data transmission andobservations on the effects of common mode transients?

Lower capacitance and space reduction can be athieved with a coreless PCB xformer, although it needs to be subjected to a higher frequency to reduce magnetizing losses.

Cheers

Klaus

 

[[email protected]]

The alternative would be to try another direction removing the iCoupler and Opto and using a transformer instead (the Pulse PH9185 is approved for the correct isolation and has the same area as the iCoupler)

Including drivers, receivers, and DC restore?

Signaling from primary to secondary would then be at two different frequencies with some filters to "decode" the signals and communication the other way could be a current modulation scheme or injection of a third frequency.

Again, we come back to space and power.

It does however quickly become complex, but the upside is that current consumption may possibly be reduced.

Yes, and complexity = space and power. It's very doubtful that you're
going to beat an integrated solution with discretes.

 

[Joerg]

On Thu, 17 Jan 2013 03:14:00 GMT, Ralph Barone

[...]

...and money and power. I think these avenues have already been

explored.

The alternative would be to try another direction removing the
iCoupler and Opto and using a transformer instead (the Pulse PH9185
is approved for the correct isolation and has the same area as the
iCoupler)

Signaling from primary to secondary would then be at two different
frequencies with some filters to "decode" the signals and
communication the other way could be a current modulation scheme or
injection of a third frequency.

It does however quickly become complex, but the upside is that
current consumption may possibly be reduced.

Immunity to common mode signals is another matter, the ADUM2201m has
immunity of 35kV/us. ...

I don't remember which one of the AD couplers it was but one client
experienced major EMC issues with one of them. They used it for
isolating an RS232 link. The noise pollution was in the VHF range and
tough to filter out because on the isolated side we couldn't use much in
terms of caps to ground because of agency certs and such.

... The PH9185 has 8pF capacitance from primary to
secondary, so a dV/dt of this rate will inject more than 100mA into
the rail of the primary side which could affect the communication
related to how well the capacitance is coupled to the primary side
and if it is symmetric coupled

Nothing there that a TVS plus maybe a TL431-style rail clamp couldn't
handle

Anyony been down this road, using a transformer for data transmission
and observations on the effects of common mode transients?

Yes, but mostly we made our own transformers because of strict agency
approvals that commercial vendors could not accommodate, or where they
said it takes forever to get us "the papers". Also because I often need

50MHz in BW. If you have a comparator with hysteresis on the other side

that might just take care of the whole DC status restore. Only the first
status after power-up is indeterminate, once you have sent the first
transition it's "locked" into the correct signal polarity.

Coupled PCB inductors are another option if you slip a core through
that. Such flat cores are now quite mainstream. But you need a
controlled PCB fab process, usually with the proper certs and all that.
So with all things considered this may not actually save money. I also
have a hard time trusting those cores not leaving their work places in
cases where the unit could take a hard fall.

Of course, whenever I can I use optocouplers because that is cheaper and
often also smaller in footprint. They come in multi-packs as well.

For super high isolation there is also the McGyver method: LED and
photodiode, plus a snippet of thin hose.

 

[[email protected]]

On Thu, 17 Jan 2013 03:14:00 GMT, Ralph Barone

[...]

Can you do a minor modulation of the width of the start or stop
bits that is small enough that it doesn't affect the downstream
UART. At 115 kbps there should be enough packets to send one bit
every ms with a goodly pile of redundancy. Unfortunately, it
might take more space than a second optocoupler.


...and money and power. I think these avenues have already been

explored.

The alternative would be to try another direction removing the
iCoupler and Opto and using a transformer instead (the Pulse PH9185
is approved for the correct isolation and has the same area as the
iCoupler)

Signaling from primary to secondary would then be at two different
frequencies with some filters to "decode" the signals and
communication the other way could be a current modulation scheme or
injection of a third frequency.

It does however quickly become complex, but the upside is that
current consumption may possibly be reduced.

Immunity to common mode signals is another matter, the ADUM2201m has
immunity of 35kV/us. ...

I don't remember which one of the AD couplers it was but one client
experienced major EMC issues with one of them. They used it for
isolating an RS232 link. The noise pollution was in the VHF range and
tough to filter out because on the isolated side we couldn't use much in
terms of caps to ground because of agency certs and such.

I used some of them (nine per board) that included the integrated,
isolated, power supply. Passing the EMI tests was a piece of cake,
probably because I was scared shitless of the 300MHz coming from the
inverter, so took precautions. I was using it for an isolated RS422
link. I needed the isolated power to supply the RS422 drivers.

 

[Klaus Kragelund]

See...



http://www.analog-innovations.com/SED/Klaus_Combiner.pdf



Uses noise removal gimmicks I first used for ignition systems...

screened out the spark radiation into the pickup

Very nice, thanks. I'll work further along this idea to check it out

Regards

Klaus

 

[Klaus Kragelund]

Including drivers, receivers, and DC restore?

Drivers can be directly from the microcontroller, using a high frequency PWM timer

Again, we come back to space and power.

Yes. Could be a single pulse with restoration on the secondary side to recover the signal to save power.

Cheers

Klaus

 

[Klaus Kragelund]

On Thu, 17 Jan 2013 03:14:00 GMT, Ralph Barone

[...]

Can you do a minor modulation of the width of the start or stop
bits that is small enough that it doesn't affect the downstream
UART. At 115 kbps there should be enough packets to send one bit
every ms with a goodly pile of redundancy. Unfortunately, it
might take more space than a second optocoupler.


...and money and power. I think these avenues have already been

explored.

The alternative would be to try another direction removing the

iCoupler and Opto and using a transformer instead (the Pulse PH9185

is approved for the correct isolation and has the same area as the


Signaling from primary to secondary would then be at two different

frequencies with some filters to "decode" the signals and

communication the other way could be a current modulation scheme or

injection of a third frequency.

It does however quickly become complex, but the upside is that

current consumption may possibly be reduced.

Immunity to common mode signals is another matter, the ADUM2201m has

immunity of 35kV/us. ...

I don't remember which one of the AD couplers it was but one client

experienced major EMC issues with one of them. They used it for

isolating an RS232 link. The noise pollution was in the VHF range and

tough to filter out because on the isolated side we couldn't use much in

terms of caps to ground because of agency certs and such.

We have had the same problem. AFAIR the problem was with the first versions of the Analog iCouplers, it has now been solved.

Nothing there that a TVS plus maybe a TL431-style rail clamp couldn't

handle

Yes, should be do-able.

Yes, but mostly we made our own transformers because of strict agency

approvals that commercial vendors could not accommodate, or where they

said it takes forever to get us "the papers". Also because I often need


that might just take care of the whole DC status restore. Only the first

status after power-up is indeterminate, once you have sent the first

transition it's "locked" into the correct signal polarity.



Coupled PCB inductors are another option if you slip a core through

that. Such flat cores are now quite mainstream. But you need a

controlled PCB fab process, usually with the proper certs and all that.

So with all things considered this may not actually save money.

Yes, that is certainly true. We need the approval anyway, since UL are going to look into other stuff.

I also

have a hard time trusting those cores not leaving their work places in

cases where the unit could take a hard fall.

Well, only for planar. If its coreless it can take the punishment from an elephant

Cheers

Klaus

 

[[email protected]]

On Thu, 17 Jan 2013 13:15:08 -0800 (PST), Klaus Kragelund

Would you Google groupies dump that shit and get a real
newsserver/reader?! What a PITA!

Drivers can be directly from the microcontroller, using a high frequency PWM timer

Microcontrollers take zero space, power, and cost?

Yes. Could be a single pulse with restoration on the secondary side to recover the signal to save power.

Using the same zero space, power, and cost microntroller, I presume.

A similar and even simpler solution has already been proposed. The OP
has already rejected it as being unworkable.

 

[Klaus Kragelund]

On Thu, 17 Jan 2013 13:15:08 -0800 (PST), Klaus Kragelund




Would you Google groupies dump that shit and get a real

newsserver/reader?! What a PITA!



<...>

Sorry, will get a new reader shortly

Microcontrollers take zero space, power, and cost?

Yes, when we have one in there allready. (only on the primary side)

Cheers

Klaus

 

[[email protected]]

Sorry, will get a new reader shortly

Thank you! ;-)

Yes, when we have one in there allready. (only on the primary side)

How do you deal on the secondary? Smaller than a dual isolator?