Termination for digital output using 7811R and SCH68-68: output/characteristic impedance

[ds314]

Hello,

I want to use the PXI-7811R with the SCH68-68 cable in order to control some custom circuitry. I will use the 7811R for now as an output only. The signal integrity and timing are important so I would like to determine the best type of termination to use in the custom circuitry in order to minimise relfections. I could not find any published specification for the characteristic impedance of the SHC68-68 cable (I found some information on some of the other 68-pin cables, however not specifically the SHC68-68), additionally there was nothing about the output impedance of the 7811R, except for an indirect mention in the following KnowledgeBase article: http://digital.ni.com/public.nsf/allkb/D58C269F7771770E86256F490082E6C8

"The output impedance for the R Series devices is non-linear due to the circuitry used to provide 5V tolerance. The output impedance is low (<50 ohms) for output voltages near 0V. As the output voltage increases to around 3.3V, the impedance rises to approximately 100 ohms. As the voltage rises beyond 3.3V, the impedance rises very quickly to megohms."

I would like to find out what the characteristic impedance of he SHC68-68 is, and to verify that the above statement in the KnowledgeBase definitely applies to the 7811R, and if so then whether more precise information is available - i.e. how much exactly is meant by "<50 ohms", as this in combination with the cable's characteristic impedance will directly relate to how much signal is reflected for a second run towards the receiver in case the receiver is not properly matched.

Of course I could determine these answers by careful experimentation, however I was quite surprised that this information was not published so I wanted to see if these figures are at all available before spending time determining them myself.

Many thanks in advance!

[johnsold]

Any digital circuit using nonlinear technologies (TTL, CMOS) will not have a well defined output impedance is as clearly described by the KB article you quoted. Since the SHC68-68 is designed to be used with such digital signals and relatively low frequency analog signals, I doubt that it has a well defined characteristic impedance either.

If you want to avoid signal integrity problems due to impedance mismatch, then you may need to put a buffer with a constant output impedance (such as 50 ohms) between the 7811R and your external device. Connect the buffer to your external device with a quality coaxial cable matching the buffer output impedance. If the signal at the input of the buffer is already degraded, consider a good high speed comparator before the buffer.

Lynn

[ds314]

Hi Lynn,

Thanks  for your input! I understand the reasons why the output impedance of the 7811R is nonlinear, and that is not what bothered me about the specification - rather, I was wondering if there is a more precise figure available than '<50 ohms' when the output is in 'low' state, as without any additional information could mean anything from 50 down to almost 0, and considering the typical impedances of cables this could be either a reasonable match or a huge mismatch. Perhaps I should mention that unless something changes about the plan for this project, I will be using mostly very short pulses (down to the 25ns limit for the 7811R) as output signals, which means there will be long intervals of time during which the outputs will be fixed in 'low' state (or 'high' if I decide to invert my logic).

Maybe it's my misunderstanding, but I am not totally clear on what you mean when you say that the characteristic impedance of the SHC68-68 may not be well defined. Are you talking about the frequency-dependence of the impedance, or maybe something different? I saw that NI have a published 'detailed specification' for a different cable, SH68-68-EP, found here: http://digital.ni.com/public.nsf/allkb/36220E8F02BF21288625686900766E27 - however, a similar specification does not seem to exist for the SHC68-68 and I was wondering if there are any details like this available at all.

Thanks also for the suggestion about buffering before my external circuitry. However, I'd like to avoid putting anything additional in between my external circuitry and the 7811R, especially if that involves additional cables or more external devices (mainly for the sake of simplicity and cost and the time I have to get this finished) - so what I'm trying to find out with this enquiry is what kind of termination is optimal for my external circuit to minimise reflection back towards the 7811R - i.e. should I terminate with a resistor, if so, what value, and if there's anything else I should be aware of. The main point is the matching between the cable and my external circuit; the question about the 7811R impedance is purely for me to assess how bad of an issue it will be if the termination on the external circuitry does not function perfectly and reflects something back, at which point I'd want to know how much in turn would be reflected back from the 7811R due to the mismatch between the cable and 7811R impedance.

[johnsold]

Digital devices typically operate in saturation and that makes defining the impedance difficult.

Considering that the impedance of the digital twisted pairs in the two cables for which NI publishes detailed specifications is 107 ohms and 100 +/- 15 ohms it might be reasonable to  place a resistor in series with the source, perhaps 68 or 82 ohms to start, and 100 ohms in series with the load. The device cannot drive a 100 ohm load to ground so the typical shunt termination will not work.  That will result in a time constant ~ 5 ns. The delay through the cable is also ~ 5 ns (assuming the 1 m cable).

What is the impedance of your load? Will your signals be 3.3 V or 5 V? If you can work at 3.3 V, you will probably get much cleaner behavior because the output is actively driven (=> low impedance) to that level.

Lynn

[ds314]

Thanks again. I'm quite glad you pointed out that the 7811R would not be able to handle a low impedance termination, when looking through the datasheet I noticed the 5 mA limit - however, quite carelessly, I forgot to account for this when I started thinking about termination!

Just to supplement your list of known figures about other NI cables, there is another cable for which impedance seems to be unofficially known as 90 ohms (SHC68-68RDIO), I found this in another forum thread from 2012 ( http://forums.ni.com/t5/Real-Time-Measurement-and/I-challenge-NI-to-show-us-a-impedance-matched-1m-R... ) - not far off from the figures you quoted for the other cables, so it's tempting to assume hat the SHC68-68 is somewhere in the same region, though not really totally conclusive! It could be that I want an accurate figure I might have to measure it myself, though in context of the current limitation stated above this figure might not be as directly useful to me as I initially thought.

As for the other points  - yes, I was planning to go with the 3.3V output level. As said, I wanted to avoid putting anything in between the 7811R and its end of the cable, and rather try to find a solution on the other end which fits in my custom circuitry. As for the load itself, this has not been determined as the circuit is in early design stages. The circuit will drive a set of PCB-mounted high-voltage (100-400V) pulsers (either ICs or discrete, I haven't settled on this yet), so in between the digital inputs coming from the 7811R and the pulsers themselves will be some type of buffering capable of driving the pulsers (and depending on some other design choices that I have not yet finalised, at most very rudimentary logic functions). The system I'm putting together is a basic ultrasonic imaging setup, so my concerns about signal integrity are really to ensure timing of the control pulses is precise and to prevent erroneous firing or misfiring of the ultrasonic transducers. As this is an academic project with very little available funding remaining for me to generate some results, I'm looking for the cheapest and quickest solution - though 'cheap' here also meaning that it has to work the first time!

[johnsold]

Thanks for the description of your project.

I think I would put high speed comparators at the custom device end. Set the threshold to about the midpoint where the reflections are likely to have minimal effect.  There will be some delay due to propagation delay in the comparators and probably some due to the cable and mismatches, but those should be constant and fairly similar from channel to channel. Then you have clean signals to drive the pulsers.

Designing a 400 V 25 ns switch sounds like fun. Do you need to control the amplitude or just switch an available voltage? What kind of loads will be connected to the pulsers?

Remember the old saying: "Cheap, fast, good - pick two!"

Lynn

[ds314]

Your comparator suggestion certainly seems like a valid plan, thanks for this! This will also allow me to scale the control signal as I need to drive the pulsers. I'm not very worried about propagation delay as long as it's consistent.

Yes, the pulser design is certainly not simple, and as I haven't spent too much time on that end of things I'm not sure what kind of switching time I can practically achieve - however I wanted to make sure I do not inherently introduce limitations on the control signal on top of the 25 ns limitation of the 7811R - so that the limiting component is the pulser itself. I'm hoping to have a decent amount of energy in the pulse up to somewhere between 15-20 MHz.

I don't have to control the amplitude of the pulses - it would be great if I can and it would be on the list of ideal features, however it's not a priority as the priority is to have the basic pulsing capability and have it work properly. So, 100-400V means the pulsers will run off a constant voltage somewhere in this range (to be determiend), and 400V is the maximum limit I would consider but I imagine 100-200V is more likely. Obvisouly, lower voltages will release some of the other constraints but there's a number of factors to balance so I don't have a final answer yet. I would also prefer to have negative as well as positive pulses (meaning a negative plus postive supply) however it's quite likely this will be omitted to make the board cheaper. There will be 8 channels with a load of 1 ultrasonic transducer per channel, the impedances of these transducers are in the process of being acquired, and will be frequency dependent - however in combination with the matching network built into the transducer I anticipate that there will be a range of frequencies of interest where the impedance is close to 50 Ohms (or possibly some other average value without a huge imaginary part).

[johnsold]

I have designed a driver for a piezo valve operator. The piezo device looks like a capacitor with a resistive component representing the power delivered to the load. Because piezo devices physically expand when charge is applied, the capacitance also changes with the terminal voltage. I saw changes of several percent with applied voltages ~100 V. (Working from memory - do not have immediate access to the data). The key point is that the load capacitance is not constant. Also the load was nowhere near 50 ohms. If the transducer is designed to be driven by square pulses, I would not expect a nice clean impedance. The device capacitance was ~2 uF.

Speeds were limited by how fast the capacitances could be charged through the resistances of the switches. Our speeds were on the order of 10s to 100s of microseconds. The valve mechanism could not move much faster than that although I suspect that the piezo component was considerably faster.  

If you would like to discuss the circuit design aspects off line, PM me with your contact information.  That discussion is not particularly relevant to this thread.

Lynn

[ds314]

Thanks again for offering to share your experience, it's much appreciated! From what I know about our transducers our requirements are a bit different (e.g. the capacitance is much smaller), however I will message you if I have any questions. And yes, this thread is about the digital interfacing so don't want to take it off topic!