0
\$\begingroup\$

I am having problems figuring out the best way to isolate sense lines, filter noise, and avoid ground loops. Here is the situation:

  • Sense multiple points on a DUT board (extremely noisy environment).
  • Drive the sense signals along ~50ft of 50ohm coax.
  • Measure the signals with Scope(s). Using standard BNC input, differential probes are not an option.

I am using a single ended Hi-Speed buffer LH0063 (very high slew-rate and very high input impedance), click here for datasheet

I have to use this part. I cannot swap it for a different signal driver. The driver(s) can go unstable. If there are ground loops or multiple drivers not correctly decoupled or driver lines to close to sense lines, the output will oscillate.

Here are 3 scenarios I came up with. I don't know which is best or if I should do a combination, or what changes I should make. Any feedback would be helpful.

  • ALL TRANSMISSION LINES ARE 50ohm BNC cables

  • SENSE SIGNALS ARE IN ~10MHz frequency range

Scenario 1 Scenario 1

Scenario 2 Scenario 2

Scenario 3 Scenario 3

\$\endgroup\$
5
  • \$\begingroup\$ LH0063 was obsolete years ago so why still use it. I don't mind but what does concern me is the lack of a believable data sheet and your link falls into that category. Also your basic transmission from buffers to control room won't be that great and it won't isolate. Also the arrow symbols pointing down in the control room tells me nothing about where the screen terminates. \$\endgroup\$ – Andy aka Jan 29 at 19:37
  • \$\begingroup\$ If you are really using coax, then the outer shield of those cables need to be terminated to chassis ground, not signal ground as you seem to be showing. Also, you should consider going differential analog from your sense boards to the control room, if you really want that interface to be analog. \$\endgroup\$ – SteveSh Jan 29 at 19:56
  • \$\begingroup\$ The arrows in the 'control room' region are the ground connection on the scope terminals; LTspice did not have an 'earth' symbol. I want to use the LH0063 because it is what the facility currently has ready to go. If I need to make changes or swap things while I'm there it is much easier. I did not know that the outer shield needs to be chassis ground. I have to research how and where to connect signal-ground and chassis-ground. \$\endgroup\$ – Tony Jan 29 at 21:16
  • \$\begingroup\$ You have asked 23 questions and received good answers to several so, maybe you were not aware that you should formally accept the best answer to questions if that answer is good enough. If the best answer needs clarification you should seek clarification. Take the tour because it explains how you accept answers. You've been a member for five years so I'm a bit dumbfounded how you missed this piece of information..... \$\endgroup\$ – Andy aka Jan 31 at 14:21
  • \$\begingroup\$ ....Not formally selecting an answer isn't wrong but, in my opinion, not selecting an answer will portray you as being mean and people will notice this and choose not to help you. You come here for help so, if you "play the game" you'll maximize your chances of getting the help you need. Here's an example where I gave good information on the subject of connecting signals via coax yet, you don't appear to have taken it fully on-board in this question. \$\endgroup\$ – Andy aka Jan 31 at 14:22
1
\$\begingroup\$

None of the above mitigates differences in ground noises which degrades the SNR.

If your noise is induced into the buffer, which only needs to be to a GBW say twice your requirement.

The termination must be differential across 50 Ohms if the noise is being induced on the 50 Ohm Cable. This can be tested simply with a 50 ohm source replacing the buffered signal to see what your noise pollution is on the cable before you see how much is getting in to validate your assumption. A series of CM chokes must be added to the Coax or simply use the 20MHz filter on the DSO unless saturation occurs on the CM signal.

Each connection from one ground to the next by cable needs a Pi filter with CM choke and caps to local ground. I suggest 1kOhm @ 100MHz DCR=1

Clam-shell CM chokes are inductive over 1 to 2 decades in frequency , so more than one of different spans may be needed on the sensor cable.

The cost of clamshell beads varies from about US$0.50 to US$10.00each depending on the source. Smaller coax can be wound several turns to raise the CM impedance. Then RF shunt caps to ground will increase CM attenuation greatly while limiting differential impedance by chosen limit. >= 20MHz

Some well-known manufacturers of ferrite beads are: Manufacturers
Fair-Rite Products Corp. http://www.fair-rite.com/newfair/index.htm
Ferrishield http://www.ferrishield.com/
Ferroxcube http://www.ferroxcube.com/
Murata http://www.murata.com/products/emc/index.html
NEC/Tokin http://www.nec-tokin.com/english/product/dl_emc.html
Parker Chomerics http://www.chomerics.com/products/emi/cableconnector.html
Laird (Steward) http://www.lairdtech.com/Products/EMI-Solutions/Ferrite-Products/
TDK http://www.tdk.com/ferrites.php
Würth Electronik http://www.we-online.com/web/en/passive_bauelemente_standard/willkommen_pbs/Welcome.php

\$\endgroup\$
2
  • \$\begingroup\$ Thank you for the feedback...The noise on transmission lines from 'SENSE-BOARD' to 'CONTROL-ROOM' is not the biggest problem (though not great). I should have been more clear. The major source of noise is from the 'CIRCUIT-UNDER-TEST' region. There is no-way to mitigate it, noise-levels are measured before hand as a control... Currently, the noise is greater than the signals I am trying to measure. How can I bring a noisy (Vsense-Vground) measurement to the buffer region without differential inputs? Is it possible, maybe something with AC coupling? \$\endgroup\$ – Tony Jan 29 at 22:02
  • 1
    \$\begingroup\$ define your CMRR that you need then design a CM filter with matched caps and a good CM choke . Compute impedance slope multiply by CM noise signal spectrum. derive SNR you want \$\endgroup\$ – Tony Stewart EE75 Jan 29 at 22:08

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.