Voltage protection for 12MHz data line

[Jim]

In order to cater for erroneous connection to our unit, we need to limit the
maximum voltage on a particular input line to 3.3V max. The line carries a
serial bitstream running at up to 12MHz. The clock signal is recovered from
the bitstream, so we cannot afford to round off the squarewave much at all.

In the past we have used zener diodes to clamp voltages on certain lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor, so
we don't want to spend more than $1.50 on this protection if at all
possible.

Thanks,

Jim





-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 100,000 Newsgroups - 19 Different Servers! =-----

 

[John Woodgate]

I read in sci.electronics.design that Jim <me@privacy.net> wrote (in
<4055d07c_8@corp.newsgroups.com&gt about 'Voltage protection for 12MHz
data line', on Mon, 15 Mar 2004:

In order to cater for erroneous connection to our unit, we need to limit
the maximum voltage on a particular input line to 3.3V max. The line
carries a serial bitstream running at up to 12MHz. The clock signal is
recovered from the bitstream, so we cannot afford to round off the
squarewave much at all.

Can you not capacitively-couple it to wherever you might get a higher
voltage from? This is much simpler than arranging low-cost over-voltage
protection for a high-speed line at 3.3 V.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Jeffrey Turner]

Jim wrote:

In order to cater for erroneous connection to our unit, we need to limit the
maximum voltage on a particular input line to 3.3V max. The line carries a
serial bitstream running at up to 12MHz. The clock signal is recovered from
the bitstream, so we cannot afford to round off the squarewave much at all.

In the past we have used zener diodes to clamp voltages on certain lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor, so
we don't want to spend more than $1.50 on this protection if at all
possible.

Well if it's not really important, why bother? I'm pretty sure there
are high speed zeners out there, but even if I only billed you
$30/hour it would take more than $1.50 to find out.

--Jeff

--
A man, a plan, a cat, a canal - Panama!

Ho, ho, ho, hee, hee, hee
and a couple of ha, ha, has;
That's how we pass the day away,
in the merry old land of Oz.

 

[Jim]

"Jeffrey Turner" <jturner@localnet.com> wrote in message
news:105bll9ag8gi070@corp.supernews.com...

Jim wrote:

In order to cater for erroneous connection to our unit, we need to limit
the
maximum voltage on a particular input line to 3.3V max. The line carries
a
serial bitstream running at up to 12MHz. The clock signal is recovered
from
the bitstream, so we cannot afford to round off the squarewave much at
all.

In the past we have used zener diodes to clamp voltages on certain
lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor,
so
we don't want to spend more than $1.50 on this protection if at all
possible.

Well if it's not really important, why bother? I'm pretty sure there
are high speed zeners out there, but even if I only billed you
$30/hour it would take more than $1.50 to find out.

--Jeff

--
A man, a plan, a cat, a canal - Panama!

Ho, ho, ho, hee, hee, hee
and a couple of ha, ha, has;
That's how we pass the day away,
in the merry old land of Oz.

Hi Jeff,

I know I can do this with pulse transformer for $1.40, but would prefer to
use less PCB real estate if possible for the same kind of price or less -
standard zeners are a lot less than that.

Jim





-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 100,000 Newsgroups - 19 Different Servers! =-----

 

[Jim]

"John Woodgate" <jmw@jmwa.demon.contraspam.yuk> wrote in message
news:nOM81EHwQdVAFwGL@jmwa.demon.co.uk...

I read in sci.electronics.design that Jim <me@privacy.net> wrote (in
4055d07c_8@corp.newsgroups.com&gt about 'Voltage protection for 12MHz
data line', on Mon, 15 Mar 2004:
In order to cater for erroneous connection to our unit, we need to limit
the maximum voltage on a particular input line to 3.3V max. The line
carries a serial bitstream running at up to 12MHz. The clock signal is
recovered from the bitstream, so we cannot afford to round off the
squarewave much at all.

Can you not capacitively-couple it to wherever you might get a higher
voltage from? This is much simpler than arranging low-cost over-voltage
protection for a high-speed line at 3.3 V.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

Thank you for your feedback, John.

To explain further - the input line comes from a socket on the chassis of
the unit. The input line connects to the centre of the socket, and the
outside of the socket is grounded to the chassis.

The input line connects to a pin of an IC that is rated at 3.6V max. The
IC's ground plane is connected to the chassis.

If the user's equipment is fully compliant (S/PDIF) then the maximum voltage
we should get is 1V. Unfortunately, some equipment, such as some computer
soundcards, output 5V TTL levels. We want to prevent damage to the IC if
this happens (and allow it to work if at all possible).

Thanks,

Jim





-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 100,000 Newsgroups - 19 Different Servers! =-----

 

[Jim]

"Jim" <me@privacy.net> wrote in message
news:4055e6e4$1_3@corp.newsgroups.com...

I know I can do this with pulse transformer for $1.40, but would prefer to
use less PCB real estate if possible for the same kind of price or less -

Sorry - I'm tallking rubbish above. We were considering using a 2:1
transformer to provide DC, and AC protection (up to 7.2V) but this will make
the lower signal levels fall below spec, so no solution yet.

Jim





-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 100,000 Newsgroups - 19 Different Servers! =-----

 

[John Woodgate]

I read in sci.electronics.design that Jim <me@privacy.net> wrote (in
<4055e6e6$1_3@corp.newsgroups.com&gt about 'Voltage protection for 12MHz
data line', on Mon, 15 Mar 2004:

If the user's equipment is fully compliant (S/PDIF) then the maximum
voltage we should get is 1V. Unfortunately, some equipment, such as some
computer soundcards, output 5V TTL levels. We want to prevent damage to
the IC if this happens (and allow it to work if at all possible).

Then a low-capacitance signal diode will do. Just connect its cathode to
3 V, its anode to the signal and put a few hundred ohms between that
junction and the centre of the input socket. That will clamp an
overvoltage signal to 3.6 V quite well, while the diode will normally be
biased off.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Frank Bemelman]

"Jim" <me@privacy.net> schreef in bericht
news:4055d07c_8@corp.newsgroups.com...

In order to cater for erroneous connection to our unit, we need to limit
the
maximum voltage on a particular input line to 3.3V max. The line carries a
serial bitstream running at up to 12MHz. The clock signal is recovered
from
the bitstream, so we cannot afford to round off the squarewave much at
all.

In the past we have used zener diodes to clamp voltages on certain lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor, so
we don't want to spend more than $1.50 on this protection if at all
possible.

I just read another message on this group, and one suggestion
was to put a low capacitance diode in series with the zener.

--
Thanks, Frank.
(remove 'x' and 'invalid' when replying by email)

 

[Tim Shoppa]

"Jim" <me@privacy.net> wrote in message news:<4055d07c_8@corp.newsgroups.com>...

In order to cater for erroneous connection to our unit, we need to limit the
maximum voltage on a particular input line to 3.3V max. The line carries a
serial bitstream running at up to 12MHz. The clock signal is recovered from
the bitstream, so we cannot afford to round off the squarewave much at all.

In the past we have used zener diodes to clamp voltages on certain lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor, so
we don't want to spend more than $1.50 on this protection if at all
possible.

Thanks,

For 3.3V, 5 series silicon forward-biased diodes may be a good solution.

Sidactors are available down to 25V or so but that's not low enough for you.

Teccor also makes something they call a "Battrax" that may work but I
have no experience with them and they may be too expensive.

Are you really sure that you need to stop at 3.3V and cannot possibly
go more? Line receivers that I'm familiar with are often specced
up to 15V and many are specced to 25V.

Tim.

 

[Jim]

"Frank Bemelman" <f.bemelmanx@planet.invalid.nl> wrote in message
news:40560762$0$90762$c3e8da3@news.astraweb.com...

"Jim" <me@privacy.net> schreef in bericht
news:4055d07c_8@corp.newsgroups.com...
In the past we have used zener diodes to clamp voltages on certain
lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor,
so
we don't want to spend more than $1.50 on this protection if at all
possible.

I just read another message on this group, and one suggestion
was to put a low capacitance diode in series with the zener.

--
Thanks, Frank.
(remove 'x' and 'invalid' when replying by email)

Thanks for that Frank - I'll look into it.

Jim

 

[Jim]

"Tim Shoppa" <shoppa@trailing-edge.com> wrote in message
news:bec993c8.0403151206.3581000c@posting.google.com...

For 3.3V, 5 series silicon forward-biased diodes may be a good solution.

Sidactors are available down to 25V or so but that's not low enough for
you.

Teccor also makes something they call a "Battrax" that may work but I
have no experience with them and they may be too expensive.

Are you really sure that you need to stop at 3.3V and cannot possibly
go more? Line receivers that I'm familiar with are often specced
up to 15V and many are specced to 25V.

Tim.

Hi Tim,

Yes, that's what I'm thinking - this IC's inputs are spec'ed to Vcc (3.3V) +
0.3V, but I wonder if 5V would actually do harm? But even if it worked on
our samples, it might not work on future versions of it, and we wouldn't
have a lot of comeback!

Thank you for the other ideas too; we'll take a look at those.

Jim

 

[Mac]

On Mon, 15 Mar 2004 11:15:34 +0000, Jeffrey Turner wrote:

Jim wrote:

In order to cater for erroneous connection to our unit, we need to limit the
maximum voltage on a particular input line to 3.3V max. The line carries a
serial bitstream running at up to 12MHz. The clock signal is recovered from
the bitstream, so we cannot afford to round off the squarewave much at all.

In the past we have used zener diodes to clamp voltages on certain lines,
but I understand that a standard zener would have create too much
capacitance for this line. Could someone please suggest an alternative
approach, or are there high speed zeners around? BTW, cost is a factor, so
we don't want to spend more than $1.50 on this protection if at all
possible.

Well if it's not really important, why bother? I'm pretty sure there
are high speed zeners out there, but even if I only billed you
$30/hour it would take more than $1.50 to find out.

--Jeff

Don't you think you're being pretty harsh? For one thing, when he said
$1.50, I'm pretty sure he meant $1.50 per board. You are acting like it
was a cap on up-front design time. And where did he say that it is not
really important? You quoted his whole post, and I don't see that phrase
anywhere.

Maybe there's something I'm missing.

regards,
Mac

 

[w_tom]

See this manufacturer for such devices:
http://www.semtech.com/html/tvs_low_capacitance.html

Jim wrote:

In order to cater for erroneous connection to our unit, we need to
limit the maximum voltage on a particular input line to 3.3V max.
The line carries a serial bitstream running at up to 12MHz. The
clock signal is recovered from the bitstream, so we cannot afford
to round off the squarewave much at all.

In the past we have used zener diodes to clamp voltages on certain
lines, but I understand that a standard zener would have create
too much capacitance for this line. Could someone please suggest an
alternative approach, or are there high speed zeners around? BTW,
cost is a factor, so we don't want to spend more than $1.50 on
this protection if at all possible.

 

[Tim Shoppa]

"Jim" <jim@nospam.com> wrote in message news:<w9o5c.517$3D1.446@newsfep3-gui.server.ntli.net>...

"Tim Shoppa" <shoppa@trailing-edge.com> wrote in message
news:bec993c8.0403151206.3581000c@posting.google.com...
For 3.3V, 5 series silicon forward-biased diodes may be a good solution.

Sidactors are available down to 25V or so but that's not low enough for
you.

Teccor also makes something they call a "Battrax" that may work but I
have no experience with them and they may be too expensive.

Are you really sure that you need to stop at 3.3V and cannot possibly
go more? Line receivers that I'm familiar with are often specced
up to 15V and many are specced to 25V.

Tim.

Hi Tim,

Yes, that's what I'm thinking - this IC's inputs are spec'ed to Vcc (3.3V) +
0.3V, but I wonder if 5V would actually do harm? But even if it worked on
our samples, it might not work on future versions of it, and we wouldn't
have a lot of comeback!

Thank you for the other ideas too; we'll take a look at those.

If you have the PCB real estate to put in a simple differential line
receiver (an 8-pin SOIC holds a transceiver) you'll win in more ways than just
the V_in specs. I'd be critical about sending 12MHz at 3.3VCMOS or TTL levels
unterminated for more than a few feet, but there's no problem at all
sending it over a terminated differential pair for a mile.

Tim.

 

[Fred Bloggs]

John Woodgate wrote:

Can you not capacitively-couple it to wherever you might get a higher
voltage from? This is much simpler than arranging low-cost over-voltage
protection for a high-speed line at 3.3 V.

I'm not going to even grace that pretentious troll with a reply, but
he's talking about a simple S/PDIF digital audio channel IEC958 where
some CD's and other garbage producing devices have contorted the signal
level from the standard 0.5Vpp ac-signal at 75 ohm bi-phase coding at
6MHz maximum to std TTL with at most TTL buffered drive. A suitably
protected receiver that may receive both is:

Please view in a fixed-width font such as Courier.





0.1u
ctr >----||------+----+--------+----->
| | |
| | | +.25V - --
--- --- / | | | |
RCA Jack / \ \ / 75 0V---| |---| |-----
--- --- / | | | |
1N4148 | | \ -.25V - --- -
| | |
shield>------------+----+--------+-----+
|
-+-
///

obvious extension for coupling onto rcvr IC DC threshold

 

[John Woodgate]

I read in sci.electronics.design that Fred Bloggs <nospam@nospam.com>
wrote (in <4057240F.7060001@nospam.com&gt about 'Voltage protection for
12MHz data line', on Tue, 16 Mar 2004:

John Woodgate wrote:


Can you not capacitively-couple it to wherever you might get a higher
voltage from? This is much simpler than arranging low-cost over-voltage
protection for a high-speed line at 3.3 V.

I'm not going to even grace that pretentious troll with a reply,

What does the 0.1 uF cap do in your circuit?
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Jim]

"Mac" <foo@bar.net> wrote in message
newsan.2004.03.16.03.00.25.137919@bar.net...

On Mon, 15 Mar 2004 11:15:34 +0000, Jeffrey Turner wrote:

Well if it's not really important, why bother? I'm pretty sure there
are high speed zeners out there, but even if I only billed you
$30/hour it would take more than $1.50 to find out.

--Jeff

Don't you think you're being pretty harsh? For one thing, when he said
$1.50, I'm pretty sure he meant $1.50 per board. You are acting like it
was a cap on up-front design time. And where did he say that it is not
really important? You quoted his whole post, and I don't see that phrase
anywhere.

Maybe there's something I'm missing.

regards,
Mac

Thanks for coming to my defence, Mac! I admit this is our first consumer
product and we are on a steep learning curve trying to productize this, so
some of my queries may seem rather basic or plain stupid.

You are right - maybe it wasn't clear but I was suggesting $1.50 per board
of course. The unit is in a pretty cost-sensitive area (as are many consumer
goods), and so we have to save money wherever possible without compromising
performance or robustness too much. The protection we need is vital to
prevent a relatively large proportion of returns and unhappy customers. But
it could make the product unviable if the protection required cost a
significant amount compared to the rest of the costs - the other ICs and
semiconductors in total come to about $30 or so.

Jim

 

[Fred Bloggs]

John Woodgate wrote:

I read in sci.electronics.design that Fred Bloggs <nospam@nospam.com
wrote (in <4057240F.7060001@nospam.com&gt about 'Voltage protection for
12MHz data line', on Tue, 16 Mar 2004:

John Woodgate wrote:


Can you not capacitively-couple it to wherever you might get a higher
voltage from? This is much simpler than arranging low-cost over-voltage
protection for a high-speed line at 3.3 V.

I'm not going to even grace that pretentious troll with a reply,


What does the 0.1 uF cap do in your circuit?

Is that a trick question? The "productizing" troll is gone....

 

[Mac]

On Tue, 16 Mar 2004 05:35:24 +0000, Tim Shoppa wrote:

"Jim" <jim@nospam.com> wrote in message
news:<w9o5c.517$3D1.446@newsfep3-gui.server.ntli.net>...
"Tim Shoppa" <shoppa@trailing-edge.com> wrote in message
news:bec993c8.0403151206.3581000c@posting.google.com...
For 3.3V, 5 series silicon forward-biased diodes may be a good
solution.

Sidactors are available down to 25V or so but that's not low enough
for
you.

Teccor also makes something they call a "Battrax" that may work but I
have no experience with them and they may be too expensive.

Are you really sure that you need to stop at 3.3V and cannot possibly
go more? Line receivers that I'm familiar with are often specced up
to 15V and many are specced to 25V.

Tim.

Hi Tim,

Yes, that's what I'm thinking - this IC's inputs are spec'ed to Vcc
(3.3V) + 0.3V, but I wonder if 5V would actually do harm? But even if
it worked on our samples, it might not work on future versions of it,
and we wouldn't have a lot of comeback!

Thank you for the other ideas too; we'll take a look at those.

If you have the PCB real estate to put in a simple differential line
receiver (an 8-pin SOIC holds a transceiver) you'll win in more ways
than just the V_in specs. I'd be critical about sending 12MHz at
3.3VCMOS or TTL levels unterminated for more than a few feet, but
there's no problem at all sending it over a terminated differential pair
for a mile.

Tim.

I agree about the dangers of sending single-ended logic signals over long
cable to an unterminated receiver. On the other hand, if the receiver IS
terminated with the characteristic impedance of the cable, and the driver
can drive that impedance, then the signal will be very clean at the
receiver, even with 100+ feet of cable.

Also, changing to differential doesn't solve the OP's basic problem: users
might connect TTL signals to it. Unless, of course, you get a differential
line receiver which can handle the high voltages. They exist, but are
somewhat expensive, IIRC.

So switching to differential sensing is probably a good idea, but doesn't,
by itself, solve the problem.

--Mac

 

[Terry Given]

"Mac" <foo@bar.net> wrote in message
newsan.2004.03.17.04.01.29.286707@bar.net...

On Tue, 16 Mar 2004 05:35:24 +0000, Tim Shoppa wrote:

"Jim" <jim@nospam.com> wrote in message
news:<w9o5c.517$3D1.446@newsfep3-gui.server.ntli.net>...
"Tim Shoppa" <shoppa@trailing-edge.com> wrote in message
news:bec993c8.0403151206.3581000c@posting.google.com...
For 3.3V, 5 series silicon forward-biased diodes may be a good
solution.

Sidactors are available down to 25V or so but that's not low enough
for
you.

Teccor also makes something they call a "Battrax" that may work but I
have no experience with them and they may be too expensive.

Are you really sure that you need to stop at 3.3V and cannot possibly
go more? Line receivers that I'm familiar with are often specced up
to 15V and many are specced to 25V.

Tim.

Hi Tim,

Yes, that's what I'm thinking - this IC's inputs are spec'ed to Vcc
(3.3V) + 0.3V, but I wonder if 5V would actually do harm? But even if
it worked on our samples, it might not work on future versions of it,
and we wouldn't have a lot of comeback!

Thank you for the other ideas too; we'll take a look at those.

If you have the PCB real estate to put in a simple differential line
receiver (an 8-pin SOIC holds a transceiver) you'll win in more ways
than just the V_in specs. I'd be critical about sending 12MHz at
3.3VCMOS or TTL levels unterminated for more than a few feet, but
there's no problem at all sending it over a terminated differential pair
for a mile.

Tim.

I agree about the dangers of sending single-ended logic signals over long
cable to an unterminated receiver. On the other hand, if the receiver IS
terminated with the characteristic impedance of the cable, and the driver
can drive that impedance, then the signal will be very clean at the
receiver, even with 100+ feet of cable.

depends on the cable of course. I once debugged a 10Mbps RS485 setup that
never worked reliably - the root cause was the "engineer" who did the
"design" - he insisted on using 50 feet or so of ribbon cable for his
testing, even though the actual cable length was about 4 feet, and we were
planning on using cat-5 stp cable. He even went so far as to strip out all
but the 2 required cores from his bit of ribbon, then attempt to twist it.
The RS485 cct used a crazy "protocol" (i forget the ISO OSI model
nomenclature) involving 20ns (hopefully) rectangular pulses, and the
demodulator was edge-triggered (noise sample-and-hold). He found that
terminating resistors made no difference whatsoever, at either (or both, or
neither) end.

The actual solution was quite simple:
1) remove his 12" scope ground lead, and use a nice, tight coax connection
instead - instantly most of the so-called mess disappeared, ie his
measurements were mostly scope probe artifacts (read as bullshit)
2) chop the crappy bit of ribbon cable into tiny pieces, and put it in the
bin (if we werent on the 7th floor of an office block, I would have taken it
outside and set it alight), and use the right cable, terminated at both
ends.

Voila - the comms link worked perfectly (despite the daft protocol), and has
been in production ever since. This guy had been working on it for 3 months
or so, and it took about 3 days of arguing to get him to accept (despite the
fact the darned thing was now working) that my solution was OK (oh no, its
"crappy" maxim RS485 transceivers...lol)

although as an interesting aside, we passed power & ground thru the same
cat-5 cable, feeding (originallY) a set of buck converters. If the 0V wire
fell off (more crappy test setups), 0V current would flow thru the 485
transceiver clamp diodes, frying them - particularly embarassing when it
happened during a demo to an investor, killing the comms link (oops, no more
485 chips). solution: dont try and attach 10 1mm^2 wires to a single (albeit
loose) binding post on a bench psu - use banana plugs *sigh*

Also, changing to differential doesn't solve the OP's basic problem: users
might connect TTL signals to it. Unless, of course, you get a differential
line receiver which can handle the high voltages. They exist, but are
somewhat expensive, IIRC.

So switching to differential sensing is probably a good idea, but doesn't,
by itself, solve the problem.