7
\$\begingroup\$

We are busy designing a large system with around four PCBs slotting into a backplane. Each PCB could contain a DSP as well as a large FPGA. We would like to run 2 JTAG chains through the system, one chain for the DSPs and one for the FPGAs.

My worry is that there are a lot of high frequency comms/data running through the backplane and that this might interfere with our two JTAG chains. The back plane will be used to transport multiple lvds data signals at around 500MHz.

I would like to be on the safe side with the JTAG chains. The question is if it is possible to run JTAG over a differential standard and if so, are there any pitfalls to watch out for.

\$\endgroup\$
12
\$\begingroup\$

This is a crosstalk question as I understand it.

JTAG is single-ended typically 3.3V signal swing.

This can couple to your differential signals (which are most likely LVDS with much lower voltage swing) and create logic errors.

The other way around is also possible (the LVDS signal coupling to the single-ended JTAG signal). As I understand it this is what you worry about.

Obviously there is also the LVDS-to-LVDS coupling.

If your back-plane is an off-the-shelf type, you need to examine those 3 crosstalk scenarios and see if it is okay or not before you start designing something more complex.

If you are designing the back-plane, it should be relatively easy to do this in a way that will make crosstalk a non-issue.

My guess is you will have more trouble with the JTAG line disturbing the LVDS than the other way around, but don't listen to my blind guessing. Let the numbers talk.

We can set up the simulation in SigXplorer like this:

enter image description here

The JTAG is driven here by an IBIS model of the SPARTAN 6 FPGA, 3.3V, 24mA drive strength, Fast. This is likely a slightly more hefty output than your JTAG output driver, but better be on the safe side. The other line is a passive line here driven by a 3.3V LVDS output from a SPARTAN 6 in the low state.

Simulating the crosstalk show this for 5 different trace-to-trace spacings (0.1mm to 0.5mm):

enter image description here

So you can pick 10mV crosstalk just by using 0.5mm spacing and that will be very quiet for the LVDS lines.

Now let’s try this the other way, where the LVDS driver is active and the JTAG is quiet low.

enter image description here

As you can see, even with the closest spacing of 0.1mm trace-to-trace, the LVDS signal does only couple about 40mV of signal onto the JTAG lines. Increase the spacing a bit and you won’t even be able to see it.

The conclusions are:

  • Maybe you should worry more about the JTAG line disturbing the LVDS
    lines than the other way around.
  • If you design the back-plane, you should easily be able to arrange
    spacing to have very little crosstalk between JTAG and LVDS.

All this is done with an IBIS simulator and I strongly recommend you spend a few hours doing this before committing to layout. Your situation will be different - so you can not just use my simulations, although they should give you some idea. Simulating before layout can save you so much time later.

Full disclosure: I do conduct training and consulting in signal integrity using this software and the software vendor (Cadence) often sponsors use of software for those events, but other than that I am not affiliated.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Thanks for the very detailed answer! We do not design the backplane ourselves but do at least have some influence in the process. I'll investigate the IBIS models as well. \$\endgroup\$ – Neville Bamshoe May 30 '13 at 12:16
  • \$\begingroup\$ If you don't design the back-plane, you should find out how the trace-to-trace coupling is. Could be given as S-parameters. Could be shown simply as a board cross section. But without that information, don't rely on that back-plane. Then do the analysis shown for the given back-plane. \$\endgroup\$ – Rolf Ostergaard May 30 '13 at 12:19
  • \$\begingroup\$ @user21976 IMHO This isn't a place to announce the trainnings you offer. \$\endgroup\$ – Jesus Castane May 30 '13 at 22:20
  • \$\begingroup\$ @JesúsCastañé he just disclosed his position in relation to the software, which is advised from the FAQ and totally correct in my opinion. \$\endgroup\$ – clabacchio May 31 '13 at 11:20
  • \$\begingroup\$ @JesúsCastañé I agree. And will gladly remove that section. I just felt I sounded a bit like a Cadence sales person. BTW: Were you happy with my answer to your question here: electronics.stackexchange.com/questions/68025/… \$\endgroup\$ – Rolf Ostergaard May 31 '13 at 11:42

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.