Signal interference in a long cable

classic Classic list List threaded Threaded
9 messages Options
Reply | Threaded
Open this post in threaded view
|

Signal interference in a long cable

Michał Pleban
Hello!

I am fighting with signal interference in a long cable. I have clock and
data signals, plus some others too (they come from the C65 to the 1565
drive).

The signals interfere with each other, causing the clock signal to look
like this:

http://i68.tinypic.com/nqqth.jpg

The longer the cable, the worse the signal distortion. The distorted
clock signal confuses the shift register, making it shift some unwanted
bits and causing corruption of the received data.

I tried some simple stuff like termination but it didn't really change
anything. Passing the clock signal via some receiver circuit (for
example, 7414 Schmitt trigger gate, or a 75182 line receiver) helps
*somewhat*, but with a long enough cable (over 1 meter) the distortions
become too big to filter them out this way.

I am not very goot at analog stuff, so I would appreciate any pointers
to what can be done with this?

Regards,
Michau.







       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Gerrit Heitsch
On 12/21/2016 07:51 PM, Michał Pleban wrote:

> Hello!
>
> I am fighting with signal interference in a long cable. I have clock and
> data signals, plus some others too (they come from the C65 to the 1565
> drive).
>
> The signals interfere with each other, causing the clock signal to look
> like this:
>
> http://i68.tinypic.com/nqqth.jpg
>
> The longer the cable, the worse the signal distortion. The distorted
> clock signal confuses the shift register, making it shift some unwanted
> bits and causing corruption of the received data.
>
> I tried some simple stuff like termination but it didn't really change
> anything. Passing the clock signal via some receiver circuit (for
> example, 7414 Schmitt trigger gate, or a 75182 line receiver) helps
> *somewhat*, but with a long enough cable (over 1 meter) the distortions
> become too big to filter them out this way.
>
> I am not very goot at analog stuff, so I would appreciate any pointers
> to what can be done with this?

You have short spikes you want to get rid off. How about a capacitor
between the signal and GND. Yes, that will soften the signal in general
so you have to experiment if there is a cap that will work without
softening the rising and falling edges too much. Put the cap at the
input of the shift register or the 74LS14 if used.

Try 470pF and see what happens.

  Gerrit



       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Francesco Messineo
In reply to this post by Michał Pleban
On Wed, Dec 21, 2016 at 7:51 PM, Michał Pleban <[hidden email]> wrote:

> Hello!
>
> I am fighting with signal interference in a long cable. I have clock and
> data signals, plus some others too (they come from the C65 to the 1565
> drive).
>
> The signals interfere with each other, causing the clock signal to look
> like this:
>
> http://i68.tinypic.com/nqqth.jpg
>
> The longer the cable, the worse the signal distortion. The distorted
> clock signal confuses the shift register, making it shift some unwanted
> bits and causing corruption of the received data.
>
> I tried some simple stuff like termination but it didn't really change
> anything. Passing the clock signal via some receiver circuit (for

what kind of termination did you use?
What gate drives the signal?
Is the power to the driver and receiver chip very well bypassed?

> example, 7414 Schmitt trigger gate, or a 75182 line receiver) helps
> *somewhat*, but with a long enough cable (over 1 meter) the distortions
> become too big to filter them out this way.

if it's really an "interference" (the exact term here is crosstalk)
between two signals, you have to run a GND wire between each two data
(or clock or whatever else) wires. That's how original parallel SCSI
and HPIB bus cables were made. Of course these busses are also *very*
well terminated.
It might be a termination issue or a bypassing issue likely.
You should first try to add a 47 ohm resistor just before the cable
beginning at the driver side, on every signal (but VCC if any and
GND).
The take a picture of the scope with AC coupling, gnd clip on gnd pin
of the driver and probe tip on the VCC pin of the driver. The do the
same on receiver chip. This will show you if there's any ground or
power bounce (bad decoupling in this case).

Also, whenever you suspect a crosstalk issue, take a picture of both
signals, not just one :)

F

       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Michał Pleban
In reply to this post by Gerrit Heitsch
Hello!

Gerrit Heitsch wrote:

> You have short spikes you want to get rid off. How about a capacitor
> between the signal and GND. Yes, that will soften the signal in general
> so you have to experiment if there is a cap that will work without
> softening the rising and falling edges too much. Put the cap at the
> input of the shift register or the 74LS14 if used.
>
> Try 470pF and see what happens.

The closest I had was 1nF. The signal got visibly better, and I was able
to read the dist directory - big improvement!

What is the significance of 470pF? Is it related to the clock speed?

Regards,
Michau.


       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Michał Pleban
In reply to this post by Francesco Messineo
Hello!

Francesco Messineo wrote:

> what kind of termination did you use?

I tried a termination resistor. I used a potentiometer to try different
values from x Ohm to xxx Ohm, none had any visible impact.

> What gate drives the signal?

I believe it's a HMOS-II chip.

> Is the power to the driver and receiver chip very well bypassed?

I am not sure what that means :-(

> if it's really an "interference" (the exact term here is crosstalk)
> between two signals, you have to run a GND wire between each two data
> (or clock or whatever else) wires. That's how original parallel SCSI
> and HPIB bus cables were made. Of course these busses are also *very*
> well terminated.

That is a good idea, but it would mean I had to make my own cable
instead of using an off-the-shelf one. I can go there, but as a last resort.

> It might be a termination issue or a bypassing issue likely.
> You should first try to add a 47 ohm resistor just before the cable
> beginning at the driver side, on every signal (but VCC if any and
> GND).

Another good idea, but soldering resistors inside a C65 gives me chills
for some reason ;-) The receiver is also meant to work on a sock machine
without madifications.

> The take a picture of the scope with AC coupling, gnd clip on gnd pin
> of the driver and probe tip on the VCC pin of the driver. The do the
> same on receiver chip. This will show you if there's any ground or
> power bounce (bad decoupling in this case).

I will do that tomorrow, thanks.

> Also, whenever you suspect a crosstalk issue, take a picture of both
> signals, not just one :)

There are six signals in the cable, plus GND and reset. I am not sure
what crosstalks with what :-(

Regards,
Michau.



       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Gerrit Heitsch
In reply to this post by Michał Pleban
On 12/21/2016 10:15 PM, Michał Pleban wrote:

> Hello!
>
> Gerrit Heitsch wrote:
>
>> You have short spikes you want to get rid off. How about a capacitor
>> between the signal and GND. Yes, that will soften the signal in general
>> so you have to experiment if there is a cap that will work without
>> softening the rising and falling edges too much. Put the cap at the
>> input of the shift register or the 74LS14 if used.
>>
>> Try 470pF and see what happens.
>
> The closest I had was 1nF. The signal got visibly better, and I was able
> to read the dist directory - big improvement!
>
> What is the significance of 470pF? Is it related to the clock speed?

It was just a number that looked OK for the frequency since you don't
want to distort the real signal too much. You can also create 500pF by
using two 1nF caps in series.

Also, you could try a small resistor (< 100 ohm) between output and
input but before the cap. Like this:

      ___
  ---|___|----*-----o shift register input
              |
              |
             ---
             ---
              |
              |
             ---
             GND


Play around with the values some and see what works best. Then try a
different cable. :)

  Gerrit





       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Segher Boessenkool
In reply to this post by Michał Pleban
On Wed, Dec 21, 2016 at 10:15:51PM +0100, Michał Pleban wrote:
> What is the significance of 470pF? Is it related to the clock speed?

With an equivalent resistance of the cable of about 100R that gives an
RC constant of about 5MHz (so the "knee" of your lowpass filter is there,
the signal should be at lower frequencies).  As always you'll have to
experiment to see what works best, but this gives a reasonable starting
point.


Segher

       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

lebbatdot
In reply to this post by Michał Pleban
Depending on what you are up to you might want a completely different design...balenced lines and twisted pairs, do a Google search and some circuits should pop up


* Micha?? Pleban ([hidden email]) wrote:

> Hello!
>
> Francesco Messineo wrote:
>
> > what kind of termination did you use?
>
> I tried a termination resistor. I used a potentiometer to try different
> values from x Ohm to xxx Ohm, none had any visible impact.
>
> > What gate drives the signal?
>
> I believe it's a HMOS-II chip.
>
> > Is the power to the driver and receiver chip very well bypassed?
>
> I am not sure what that means :-(
>
> > if it's really an "interference" (the exact term here is crosstalk)
> > between two signals, you have to run a GND wire between each two data
> > (or clock or whatever else) wires. That's how original parallel SCSI
> > and HPIB bus cables were made. Of course these busses are also *very*
> > well terminated.
>
> That is a good idea, but it would mean I had to make my own cable
> instead of using an off-the-shelf one. I can go there, but as a last resort.
>
> > It might be a termination issue or a bypassing issue likely.
> > You should first try to add a 47 ohm resistor just before the cable
> > beginning at the driver side, on every signal (but VCC if any and
> > GND).
>
> Another good idea, but soldering resistors inside a C65 gives me chills
> for some reason ;-) The receiver is also meant to work on a sock machine
> without madifications.
>
> > The take a picture of the scope with AC coupling, gnd clip on gnd pin
> > of the driver and probe tip on the VCC pin of the driver. The do the
> > same on receiver chip. This will show you if there's any ground or
> > power bounce (bad decoupling in this case).
>
> I will do that tomorrow, thanks.
>
> > Also, whenever you suspect a crosstalk issue, take a picture of both
> > signals, not just one :)
>
> There are six signals in the cable, plus GND and reset. I am not sure
> what crosstalks with what :-(
>
> Regards,
> Michau.
>
>
>
>        Message was sent through the cbm-hackers mailing list

       Message was sent through the cbm-hackers mailing list
Reply | Threaded
Open this post in threaded view
|

Re: Signal interference in a long cable

Marko Mäkelä
In reply to this post by Michał Pleban
On Wed, Dec 21, 2016 at 10:22:33PM +0100, Michał Pleban wrote:
>>Is the power to the driver and receiver chip very well bypassed?
>
>I am not sure what that means :-(

Did you measure the Vcc and GND at both chips, relative to the ground
plane? If there is noise in the power supply, all sorts of strange
things can happen. Basically, whenever there is a problem with digital
electronics, one of the first things that you should check is the power
supply to each chip.

I suppose that "well bypassed" means that there is both a large
electrolytic capacitor and a smaller capacitor between the Vcc and GND
of the chips.

>> Also, whenever you suspect a crosstalk issue, take a picture of both
>> signals, not just one :)
>
>There are six signals in the cable, plus GND and reset. I am not sure
>what crosstalks with what :-(

Is there a transition on any of the other signals where you see the
spikes? If all signals are stable during the spikes, then the source of
the distortion should be something else.

Looking at the picture http://tinypic.com/view.php?pic=nqqth&s=9 again,
both spikes seem to correlate with the high-to-low transition of the
clock signal itself. The small negative spike occurs just before the
transition, and the positive spike occurs almost 1/4 clock cycles after
the transition.

I guess the negative spike is not a problem, but the positive one is.

Causation is not correlation of course. I would expect there to be some
transition in some line some time after the high-to-low transition of
the clock signal. Can you check the other 5 signals, one at a time,
together with the clock signal?

        Marko

       Message was sent through the cbm-hackers mailing list