6
\$\begingroup\$

I have a board that tests a wiring harness' continuity by outputting a set of test vectors onto the harness and reading it back. The wires in the harness can be as little as 1 m and as large as 10 m.

As far as I know, a termination scheme will not work for this because the characteristic impedance of the wiring harness isn't under my control. It will also change from harness to harness. So, my aim is to slow the rise/fall times of my signals for two reasons:

  1. Improve signal integrity (I don't have any trouble with signal integrity so far)
  2. My falling edge undershoots the baseline voltage by -2V. I fear this may damage my CPLD because the max. negative voltage allowable is just -0.5V (-2V is OK if the current is less than 100mA and if the duration is less than 50 ns according to datasheet).

Here's a picture of the falling edge:

enter image description here

This is with the following circuit:

enter image description here

In order to increase the fall time I increased R1 to 650 Ohms and it produced the following waveform:

enter image description here]

My question is: Is my approach OK? Since termination isn't possible and I'd like to avoid damage to my device from signal undershoots and overshoots I felt it was best to slow down my fall times. I wanted to add, I have a LOT of time to sample the signal - as much as 800 uS before the signal rises again. You can see that here:

enter image description here

\$\endgroup\$
4
\$\begingroup\$

I'd tend to ensure that the setup and hold times were OK and not worry too much about the edges you are seeing from a timing point of view.

I'd be more worried about the 2V overshoot BUT this may not be real. At that sort of edge timing it is very possible for your instrumentation to interact substantially with the system you are measuring. (Heisenberg would feel right at home :-) ). What are you measuring this with? If you have good practical experience in measuring 10 nS rise and fall times then these traces may be correct. If not then they may not. robe tip without sleeve right at signal point. Ground from probe tip to ground on IC nearest signal point and with connection length of about 0.00 mm :-). 1000 MHz probe. Tektronix, HP, Agilent. Others maybe :-).
You circuit doesn't have to look like this BUT this is an example of Jim William's scope connections at << 100 MHz.

enter image description here

AN47 Liner Technology - High Speed Amplifier techniques and see refs at end.

BUT

I would strongly consider making the system so it could not make signals which could damage your CPLD. Using Schottky diode clamps to either power supply rail would help ensure this. On each wire a reversed biased diode is connected from circuit to V+ and circuit to ground. When signals are inside the rails the diodes do not conduct. If the signal "rings" outside the "rails" then the diodes clamp the signals. This is a standard method and effective. In some cases you will get interactions between what the diodes do and what you are trying to do BUT if this happens it is probably an indication that you need to fix what you are doing.

enter image description here

Excellent related Analog Devices app note Switch and Multiplexer Design Considerations for Hostile Environments

Diagram below from the above - sampler only - see app note.

enter image description here


I believe the following advice is good enough [tm] but my experience in such areas is relatively limited and relatively long ago - others with more and more recent practical experience may wish to suggest alternatives.

  • I'd also consider adding a small series drive resistor of the order of the lowest line impedance that you expect to drive.

  • Passive pullup gives you relatively informal control over your driver - is there any reason not to use an active high/low drive - probably using a buffer IC. This is liable to be as or more compact, cheaper and more designable.

  • You could consider a pullup/pulldown far end resistor termination to provide both ~~~ correct termination and correct mean level. eg 220r pullup and 33r pulldown or whatever.

References:

AN6039 - Board Layout Techniques for High-Performance Amplifiers

Analog Devices - HIGH SPEED HARDWARE DESIGN TECHNIQUES

Potential relevance to your circuit. Variable characteristic wiring loom drive makes this harder.

enter image description here

TI - High-Speed Amplifier PCB Layout Tips

Interesting - Review of Jim Williams: AN47 High Speed Amplifier Techniques

\$\endgroup\$
  • \$\begingroup\$ I'm afraid I don't have experience in measuring high speed signals. I'll do as you suggested and post the results. However, I have a feeling the undershoot still might be there because of the long length of the wiring - it may not be substantial at 1 m but it would would be at 10 m. There is a reason why I can't actively drive the signal high - the reason is that, as I mentioned, the board tests wiring harnesses. \$\endgroup\$ – Saad Apr 16 '12 at 11:17
  • \$\begingroup\$ If two wires on the harness are short circuited, it would cause bus contention if one of them is actively driven low and the other is actively high. With a open-drain MOSFET configuration, I avoid this issue. \$\endgroup\$ – Saad Apr 16 '12 at 11:17
  • \$\begingroup\$ Also, could you elaborate on the schottky idea? \$\endgroup\$ – Saad Apr 16 '12 at 11:20
  • \$\begingroup\$ @Saad - see ammended text re Schottky diodes. \$\endgroup\$ – Russell McMahon Apr 16 '12 at 11:26
  • \$\begingroup\$ Thank you for that answer. Just to be sure that I follow you, is this what you meant by diode clamping? i.imgur.com/xM26y.png \$\endgroup\$ – Saad Apr 16 '12 at 11:33
8
\$\begingroup\$

The ringing probably isn't real. Note that you're measuring fast signals (keep an eye on the scope's timescale!), and then the inductance of the probe's ground becomes important. It doesn't help to connect the probe's ground a few mm from where you probe the signal if the wire forms a 20 cm loop. You may eliminate possible measurement problems like this using a much shorter ground connection:

enter image description here

\$\endgroup\$
5
\$\begingroup\$

First of all, many of those ringings are just due to an unappropriate ground connection for your scope. Make sure that your are connecting the ground of your scope to the large plane or volume that is your 0 V reference. Also, twist the ground wire of your scope probe as much as you can around the live conductor of the probe, so that the multiple small areas that you create have different signs, and couplings to each one of those small areas cancel each other. All this makes a huge difference, in some cases.

Other than reducing the extra ringings that the bad connection of your scope may be introducing, to reduce the real ringings present in your circuit (the ones that exist before connecting any scope), you can try either one of these:

  • As you say, add a resistor in series with the gate of the MOSFET. That's used in many circuits.
  • Add a resistor in series with the drain (upper terminal) of the MOSFET. That will increase the fall time (and reduce the ringings), without modifying the rise time (I personally prefer this one). Actually, make it 1 kohm, if you want equal fall and rise times, and you don't mind both of them being long.
  • Add a small capacitor in parallel with your load, just where you were measuring its voltage, with the scope.
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.