Prev: An oscillator for Mr. Williams
Next: JOB: Client Needs QA Test Lead (Ft.Worth, TX)
From: keithw86 on 15 Jul 2010 14:44
On Jul 15, 10:25 am, "markp" <map.nos...(a)f2s.com> wrote:
> >> Hi All,
> >> I have an application that requires about 12 sensor devices over a RS485
> >> cable, evenly spaced over about 20m with a controoler at the end.
>
> >> I'd like to use galvanically isolated RS485 buffers such as IL3185 on
> >> each
> >> sensor, but would also like to not have to put isolated DC-DC converters
> >> to
> >> supply the power to the cable side of the buffers.
>
> >> So can I supply +5V up the cable to all the devices to power their cable
> >> side buffers? Since it is RS485 only one sensor will be transmitting at a
> >> time so as long as the input buffers don't take much current it would
> >> seem
> >> practicable (with some capacitance and possible ferrite filtering where
> >> the
> >> power comes in to each sensor).
> > Sure.  Typically this is done with a pair of wires (one each for
> > voltage and ground) Transformer couple the RS485 data and apply the
> > power to the center taps.  In a similar way power is pulled off the
> > center taps at the other end.  Power over Ethernet works the same way..
> >> First, is this a pratical solution? If so, assuming CAT5 cable with a
> >> pair
> >> dedicated to the 2 wire RS485 signals, a pair to +5V and a pair to
> >> ground,
> >> how long a cable would be possible, or is it best to use a higher voltage
> >> and use a cheap LDO on the receivers to generate +5V?
> > We power our remote widget over 1500' of CAT-5.  We start out with 24V
> > and use a buck regulator at the far end.  We could go further but the
> > drop-out on the regulator is pretty bad.  1500' is far enough for our
> > application so we haven't bothered fixing it.
>
> I guess I'm concernred about the voltage drop across a 20m cable due to the
> resistance of the cable when the worst case of all 12 sensor receivers are
> powered and one sensor's transmitter is on. The caps would smooth things
> locally so it's down to DC losses.

20M isn't very much cable. The I2R drop is pretty easy to calculate
at .08ohms/meter. The transformer's DCR will swamp that. Don't forget
that there are two wires and the two transformer windings are in
parallel (though there are two transformers in series).

> As a matter of interest, what drop-out voltages were you seeing (I'm
> assuming that's down to cable losses), and how much current were you
> supplying? At the moment I've got a 5V regulated supply available, but the
> chips require 4.5V.

Something like 70mA. 1500' of cat-5 is 15ohms (two wires) and another
3 ohms or so for the transformer, so the drop is about 1.25V.

> Also I guess since power has return currents I should use a twisted pair for
> ground and power to reduce noise, maybe tripling up as I have 4 wire pairs
> available and one is taken by the signal pair (?).

Ideally, but if they're used for data, not so much. The POE
connection I described works quite well. Cat-5 pairs are twisted
together, so it's not a disaster to have a differential current
between pairs. The thing to watch with all UTP connections is common
mode EMI coupling. Switchers tend to make this a real problem.
Passing the cable through a ferrite core once or twice tends to knock
a lot of that down.

From: Jamie on 15 Jul 2010 18:55
Paul Keinanen wrote:

> On Thu, 15 Jul 2010 15:25:18 +0100, "markp" <map.nospam(a)f2s.com>
> wrote:
>
>
>>Hi All,
>>I have an application that requires about 12 sensor devices over a RS485
>>cable, evenly spaced over about 20m with a controoler at the end.
>>
>>I'd like to use galvanically isolated RS485 buffers such as IL3185 on each
>>sensor, but would also like to not have to put isolated DC-DC converters to
>>supply the power to the cable side of the buffers.
>>
>>So can I supply +5V up the cable to all the devices to power their cable
>>side buffers? Since it is RS485 only one sensor will be transmitting at a
>>time so as long as the input buffers don't take much current it would seem
>>practicable (with some capacitance and possible ferrite filtering where the
>>power comes in to each sensor).
>>
>>First, is this a pratical solution? If so, assuming CAT5 cable with a pair
>>dedicated to the 2 wire RS485 signals, a pair to +5V and a pair to ground,
>>how long a cable would be possible, or is it best to use a higher voltage
>>and use a cheap LDO on the receivers to generate +5V?
>
>
> While not exactly RS-485, the CAN bus based DeviceNet standard
> transfers the actual data in one pair and the power for the
> transceivers and small loads in an other pair. You might find some
> interesting information about power arrangements etc.
>
> If you have full control of both the master and slave side, have you
> considered the good old current loop system, which is quite easy to
> optoisolate (no local floating power needed at slaves) ?
>
> Connect all devices, both master as well as slaves, both receivers
> (photo transistors) and transmitters (optoisolator LEDs) into a single
> 20 mA current loop. When idle, all transmitter transistors are
> conducting and the 20 mA current flows through the loop and when the
> active station (master or selected slave) wants to send the "0" bit,
> it will cut the loop current. Since the active station will hear its
> own transmission on the receiver side, it must contain some echo
> cancellation in software.
>
> The problem with this arrangement, especially with optoisolators with
> bipolar transistors, is the large voltage drop at each station, nearly
> 2 V for the receiver LED and 1 V for the Vce of the transmitter. With
> 12+1 stations the loop current loop would have to provide nearly 40 V,
> which might be too much for some optoisolator transistors.
>
> One way around this problem is to use optoisolators with FET output
> stage and hence a low voltage drop across the transmitter, reducing
> the loop voltage requirement.
>
> An other approach is to use two loops, one containing the master
> transmitter and all the slave receiver LEDs in series, the other
> containing the master receiver LED and the slave transmitter
> transmitters (bipolar or FET) in series. 24 V should be sufficient for
> both loops. In this configuration, echo cancellation is not needed.
>
Funny you would talk about this, we have some Red Lion devices that use
the current loop and if you don't have more than one slave unit on the
loop, in hot operating conditions, the crappie current source circuit
used will bias itself more than 20 ma's and over time burn out the
optical input on the slave device!. So, what I have done to correct the
issue for now is to use a R in series. This really should be documented
in the Red Lion manual but I haven't see it..

The current source that is being used is your first year 101 basic
PNP transistor being biased to 20 ma. there is a diode on the base to
the (+) rail to help out but its not enough, it still gets out of spec..
I've see it reach up to 35 ma when thing get warm..

Jamie.


From: Uwe Hercksen on 16 Jul 2010 03:50


markp schrieb:
>
> First, is this a pratical solution? If so, assuming CAT5 cable with a pair
> dedicated to the 2 wire RS485 signals, a pair to +5V and a pair to ground,
> how long a cable would be possible, or is it best to use a higher voltage
> and use a cheap LDO on the receivers to generate +5V?

Hello,

I would look for the cable data, especially the resistance per length of
one wire and calculate the voltage loss over a 20 m long cable with the
necessary current for the tranceivers.

Bye

First  |  Prev  | 
Pages: 1 2
Prev: An oscillator for Mr. Williams
Next: JOB: Client Needs QA Test Lead (Ft.Worth, TX)