CAN transceiver with high differential input impedance

[Mochuelo]

Hi,

I posted a similar question some time ago, but that was about RS485.
Today I discovered that CAN is much closer to what I want (since it
includes CSMA/CD+AMP).

Does anyone know of a CAN transceiver with megaohms of differential
input impedance? I am ok with 1 kbps of speed, or even less, but I do
need around 4000 devices connected to the bus. I haven't found any
transceiver with more than 100 kohm of Rdiff, and I don't see any
reason why they could not make them with tens of megaohms. I do need
ESD protection, but that could still be achieved. Has anyone
experimented with a 2-wire custom physical layer with a high
differential input impedance?

Thanks in advance.

 

[PeteS]

I don't think you are going to achieve 10M of differential impedance
with 4000 devices hanging off the bus. It may be possible, but with
ordinary devices I don't really see how (I'm open to being convinced,
of course

Apart from all the wire capacitance, there's the pin capacitive loading
of the devices. At a 1k data rate (500Hz line frequency), you would
need to keep the total capacitance below about 30pF to achieve 10M of
*capacitive reactance*, let alone impedance - tough, if not impossible
to do with so many physical devices.

CAN devices have to do a tradeoff in their input structure (see the
SN65HVD231 series datasheet from TI for a nice equivalent input
circuit) to balance loading against speed. No-one will buy the devices
if they aren't rated at least at 250kb/s (and most are rated at 1Mb/s).

As with all such busses, data rate is inversely proportional to cable
length, as your application obviously would be.

Just my $0.02

Cheers

PeteS

 

[Mochuelo]

I don't think you are going to achieve 10M of differential impedance
with 4000 devices hanging off the bus.

I was talking about 10 Mohm per transceiver.

It may be possible, but with
ordinary devices I don't really see how (I'm open to being convinced,
of course

I haven't found an ordinary device that exhibits that. That's why I
was asking whether someone else either knows of a commercial device
like that or has experimented with self-made devices.

Apart from all the wire capacitance, there's the pin capacitive loading
of the devices.

Wire capacitance should not be a problem at 1 kbps. Even if each one
of the 4000 devices had the capacitance of an ordinary device, which
is around 20 pF, that would make Ctot=80 nF, which, together with
Rline=(60 ohm||2500 ohm)=58.59 ohm makes time_constant=4.7 us.
Neglectable at 1 kbps, and so are reflections (max node-to-node
distance is lower than 300 m).

At a 1k data rate (500Hz line frequency), you would
need to keep the total capacitance below about 30pF to achieve 10M of
*capacitive reactance*, let alone impedance - tough, if not impossible
to do with so many physical devices.

CAN devices have to do a tradeoff in their input structure (see the
SN65HVD231 series datasheet from TI for a nice equivalent input
circuit) to balance loading against speed. No-one will buy the devices
if they aren't rated at least at 250kb/s (and most are rated at 1Mb/s).

I would buy them, because I don't care about speed. I care about input
impedance, and about satisfying dc thresholds.

Best.

 

[Mochuelo]

Wire capacitance should not be a problem at 1 kbps. [...]

That was actually device differential input capacitance. Add some wire
differential capacitance, and you still have a large margin, up to the
1 ms time per bit.

 

[PeteS]

On a per device basis, I don't think you'll find ultra high impedance
inputs on devices designed for differential *signalling* (such as can,
lvds, etc.,etc).

What you may wish to look at is LIN bus transceivers. They were
designed for lower speed applications (2.4kb/s - 20kb/s). A cursory
search of TI shows the TPIC1021
http://focus.ti.com/docs/prod/folders/print/tpic1021.html
with input leakage on the data pin (LIN is a single wire system rather
than differential) of +/- 5uA max. That's about 1.3M input resistance.
Although not as good as the 10M you wish for, it's better than the
differential transceivers you have looked at.

A number of companies make LIN transceivers.

Cheers

PeteS

 

[[email protected]]

Maybe I'm confused and haven't slept much but aren't most differential
signaling methods relatively low in differential impedance. i.e. by way
of relatively small termination resistors. Off the top of my head LVDS
in SATA is something like 150ohms (or is it 50?), ditto CAN, RS-485,
etc. There is some signal integrity reason for this relating to
preventing the RF waves from bouncing around.

Accordingly, most differential signaling methods are low impedance
almost by definition. I think what you're looking for is more along the
lines of a single ended system. LIN has been mentioned in this thread,
I don't know what the input impedance requirements of RS-232 are, CAN
has a 1 wire version, ditto K-Line, LIN, etc. The problem becomes that
high impedance systems are more prone to noise.

What's your application? If you want a decent way to hook up many items
to a network of some sort, your best bet may be to use Ethernet using
some topology of hubs.

Chris

 

[Mochuelo]

Maybe I'm confused and haven't slept much but aren't most differential
signaling methods relatively low in differential impedance. i.e. by way
of relatively small termination resistors. Off the top of my head LVDS
in SATA is something like 150ohms (or is it 50?), ditto CAN, RS-485,
etc. There is some signal integrity reason for this relating to
preventing the RF waves from bouncing around.

I never said I would not load the line with 60 ohm (120 ohm on each
side of the bus). I said I was looking for high-impedance
transceivers. The total line termination impedance does not depend on
the number of transceivers. The total transceiver impedance does.

Accordingly, most differential signaling methods are low impedance
almost by definition. I think what you're looking for is more along the
lines of a single ended system. LIN has been mentioned in this thread,
I don't know what the input impedance requirements of RS-232 are, CAN
has a 1 wire version, ditto K-Line, LIN, etc. The problem becomes that
high impedance systems are more prone to noise.

What's your application? If you want a decent way to hook up many items
to a network of some sort, your best bet may be to use Ethernet using
some topology of hubs.

But the price of 4000 Ethernet transceivers is very high, and if I
don't want to use coax cable, the price is even higher, since I need
to place 4000 cables and use some large number of switches.

All this is to interconnect 4000 sensors to be placed in a parking
lot. Each sensor detects presence/absence of vehicle on its
corresponding place, and informs a central computer.

 

[Uwe Bonnes]

On 27 Sep 2005 13:54:43 -0700, "[email protected]" <[email protected]>
wrote: ....
But the price of 4000 Ethernet transceivers is very high, and if I
don't want to use coax cable, the price is even higher, since I need
to place 4000 cables and use some large number of switches.

All this is to interconnect 4000 sensors to be placed in a parking
lot. Each sensor detects presence/absence of vehicle on its
corresponding place, and informs a central computer.

Think about splitting up the bus...