1
\$\begingroup\$

I need to transmit several analog signals from one board to another (maximum 1m distance) in a very noisy environment (power electronics converter operating at 20kHz). The range of frequencies of the analog signals is [0,20 MHz]. Presently I opted for differential signals, and designed a differential driver AD8132 on the transmit side. The signal from the sensor is 300mV p-p (current shunt if it matters). On the receiver side, I am using a difference amplifier AD8270. I expect this to do sufficiently well since the common mode signal will be small (grounds of both boards are tied at single point) The circuit is illustrated in the image below (i simulated it in spice with different components earlier but the circuit layout is the same. In the meantime i changed the amplifiers but the functionality should stay the same). transmitter and receiver I opted for simple difference amplifier. The rationale is that the common mode voltage is expected to be low which in combination with matched resistor network of AD8270 should give decent performance.I plan to use a twisted pair to transmit the signals.

  1. Should I be worried about reflections on the cable?
  2. Would a coaxial cable be a better/simpler solution in this situation?

Any thoughts/suggestions are appreciated.

\$\endgroup\$
  • \$\begingroup\$ Are you saying that you're measuring currents of 20MHz signals? \$\endgroup\$ – JimmyB Apr 12 '18 at 12:27
  • \$\begingroup\$ Yes, it is some fancy control that requires large bandwidth current measurement. \$\endgroup\$ – ercegovac Apr 12 '18 at 12:28
  • \$\begingroup\$ Are the measuring resistors grounded at one side? This makes a big difference to the circuit you can use. \$\endgroup\$ – Toni Homedes i Saun Apr 13 '18 at 8:06
  • \$\begingroup\$ Yes. But what I can tell for sure is that the power ground where resistor are attached will be oscilating several volts with respect to the ground of the control board. The di/dt of the current through the resistor is more than 100 A/us. \$\endgroup\$ – ercegovac Apr 13 '18 at 8:13
  • \$\begingroup\$ What Signal-noise-ratio must you preserve? is 1mV or 10mV or 100mV additive trash too much? \$\endgroup\$ – analogsystemsrf Apr 14 '18 at 3:37
3
\$\begingroup\$

Should I be worried about reflections on the cable?

At 20 MHz the wavelength is 15 metres so a 1 metre cable is starting to get close to the problem area but I'd be more concerned that the LTC6363 is not going to deliver the goods at 20 MHz with a gain of 10 (20 dB). Look at the DS - it shows that it's fairly flat in gain up to about 1MHz then peaks up about 5 dB at 4 MHz then starts rolling off at about 6 MHz.

This effect will overshadow any relection nuances in my opinion.

enter image description here

The AD8132 that I now notice you are using (shame on you for not putting this on the schematic!) is a lot better but, still, at 20 MHz and with a gain of 20 dB there might be 2 or 3 dB attenuation and this will be of about the same order as the effect of relections.

Would a coaxial cable be a better/simpler solution in this situation?

Use shielded twisted pairs for best signal integrity and attach the shield to ground at the receiving end.

\$\endgroup\$
  • \$\begingroup\$ You are right, i'm sorry. As I explained, the model was already there and I just used it for illustration purposes. I did not expect it to cause confusion. I'm building an experimental setup so it is not necessary to have everything operating perfectly. It just need to be good enough. I'm basically making trade-offs between time invested and the quality. \$\endgroup\$ – ercegovac Apr 12 '18 at 13:02
1
\$\begingroup\$

I suggest(ed) this

schematic

simulate this circuit – Schematic created using CircuitLab

Have very low Zsource and Zload, to reduce the effects on any electric fields entering the cable in im-balanced manner.

This approach does assume extending the shunts' leads off that PCB is OK.

============================

What SNR Signal Noise Ratio must you achieve? To Be Determined? by what causes TOO MUCH JITTER in the control loop?

A successful differential-signal movement of the shunt-voltage will require balanced impedances in the two-wire signal between the 2 PCBs.

Does the differential-driver have sufficient Common Mode Rejection to not trash the differential signal?

I'm tempted to suggest a discrete approach. So here tis. Purpose is to remove the common mode trash at the Power Board, and only ship offboard two differential signals that are very clean (there will be injected charge, which I've computed as 2mA)

schematic

simulate this circuit

\$\endgroup\$
  • \$\begingroup\$ I was thinking initially of this but due to lack of experience I did not know what to expect. Basically I have a differential signal of 200-300mV (peak) riding on several volts common mode signal. The shunt resistors are grounded but the power ground oscillates at switching frequency with several volts amplitude with respect to the control board ground. \$\endgroup\$ – ercegovac Apr 13 '18 at 8:33
  • \$\begingroup\$ Would CAT7 SFTP cable do the job? \$\endgroup\$ – ercegovac Apr 13 '18 at 17:32
  • \$\begingroup\$ CAT7 SFTP is a pretty decent cable by my measurements. Alternatively, try SATA cabling - admittedly it's not twisted which might be a downside. \$\endgroup\$ – frr Apr 14 '18 at 5:23

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.