0

I have an RF signal (5ns pulses every 200ns) coming out from a SMA connector on FPGA board (Spartan 3E Starter Kit). I want to send this signal back to the FPGA using one of the on-board pins.

So, I cut of one end of the SMA to SMA connector cable, hoping to use it for inputing the signal back to FPGA. I tested the waveform on the oscilloscope to find that the output signal was in mV range instead of around 3.3V range (input signal).

So, have I made a mistake? Is there any alternative way which I could follow?

Neel Mehta
  • 629
  • 2
  • 10
  • 21
  • 2
    Can you give us a schematic of the circuit that drives the SMA output port -- is it just a straight connection from the FPGA to the SMA center pin, or are there some other components in between? Second, what oscilloscope are you using and what is its analog bandwidth? How are you connecting the oscilloscope to the cut end of the cable --- if you're using a probe, what's its capacitance? If you're somehow connecting directly to the oscilloscope input, what's the scope input capacitance? And are you using the 50 Ohm or 10 MegOhm input setting? – The Photon May 18 '12 at 15:41
  • 2
    Also, how is the FPGA output configured? (for example, if you're using Xilinx, what IOSTANDARD and SPEED did you specify for the output pin)? – The Photon May 18 '12 at 15:44
  • @ThePhoton SMA output port is onboard the FPGA. So, the FPGA is driving the output to the SMA. There are no components in between as of now. I am using [R&S®RTO1022](http://www2.rohde-schwarz.com/en/products/test_and_measurement/industries/aerospace_and_defense/test_and_measurement_instruments/oscilloscopes_for_aerospace_and_defense/RTO-%7C-4978.html) digital oscilloscope from Rhode and Schwarz which has a bandwidth of 2 GHz and has 2 channels. I am connecting the other end of the cable using the probes that came with the DSO. I am not sure of the capacitance. Input setting is 50 ohm. – Neel Mehta May 18 '12 at 18:51
  • @ThePhoton The IOSTANDARD is LVCMOS33 (corresponds to 3.3V). I am not sure what you mean by SPEED but SLEW TYPE is SLOW and DRIVE STRENGTH is 12. – Neel Mehta May 18 '12 at 18:57
  • Yes, SLEWTYPE is the spec I was thinking of...It's worth trying fast slew rate and maximum drive strength. But CMOS is not well suited to driving a coaxial cable -- you may need an external buffer chip to provide drive capability for your cable. How long is the cable? – The Photon May 18 '12 at 19:13
  • @ThePhoton The output from the SMA when connected directly to the DSO (using SMA) gives a good signal. Its only when I cut off one end of the SMA and test it with the DSO probe, I find the voltage to drop down. The cable is around 1m long. – Neel Mehta May 18 '12 at 20:45
  • When you cut of the connector you have just made a simple antenna, so the reason for the weaker signal should be sine the signal has been radiated into the air. (And if you are using a coax cable then you are close to one of these http://en.wikipedia.org/wiki/Coaxial_antenna ) – Jontas May 18 '12 at 21:28
  • @Neel, What is the model number of your probe? How are you connecting the ground of the probe to the o.c. of the cable? A photo would be very helpful. – The Photon May 18 '12 at 21:51
  • If you cut the cable and soldered the centre conductor to the board/pin but didn't solder the outer conductor then you wouldn't expect as much signal. At 200 MHz, the signal is in the dielectric, not the conductors. You need to get the signal out of the cable dielectric and back into the fibreglass of the board. Usually means soldering both inner and outer conductor of the coax. Also you may want to consider a termination resistor at the load end of the cable. –  May 18 '12 at 23:02
  • I have just connected the probe to the internal metal wire. I haven't connected anything else to the ground/shielding wires or the dielectric. Is that the mistake that I am making? – Neel Mehta May 19 '12 at 05:54
  • Proper probe technique is part of it. Another problem is that you're probably using a typical 10Mohm passive probe, which means you can't use the 50 ohm setting on the scope. You should also terminate the transmission line with a ~50 ohm resistor so that reflections don't mess up your signal. – mng May 19 '12 at 18:02

1 Answers1

1

This isn't answering the whole question as asked, but I think it addresses the problem we found through discussion in the comments:

For the voltage on your circuit under test to have any way to affect the behavior of the 'scope, there must be a complete circuit --- there must be a return path from the oscilloscope back to your circuit's ground. If you don't connect the probe's ground to your circuit's ground, the return path will have to be through some long route back through the mains, or through some unknown capacitive path.

To get good measurements, you must connect the scope ground to your circuit. For best measurements, connect it with the shortest lead possible. This becomes more and more important as you try to measure faster and faster waveforms.

The classic long tail ground lead is adequate up to a few 10's of MHz(?):

enter image description here

The image is from Doug Smith's webpage. You can see his plan for a much higher frequency probe here --- note the emphasis on keeping the ground connection short.

One classic way to get a good probe ground connection was with a special coaxial fixture that required a dedicated footprint on the board being tested:

enter image description here

Another was the "pig-sticker" probe:

enter image description here

Much of the improvement in scope probe technology I've seen over the past few years has involved ever-trickier (and more expensive) accessories to provide an extremely short ground path between the probe and the circuit under test:

enter image description here

For your measurement you will want to connect the ground shield of the coax from as close to the cut end as possible to your probe's ground, by the shortest path possible.

The Photon
  • 126,425
  • 3
  • 159
  • 304