3
\$\begingroup\$

I did a bit of a search in the forum but despite there being a few related topics related I couldn't find the specific answer for this question.

Using a 50Mz scope I connected the probe to an FPGA output pin where there's an output of a 3.3Vp@25MHz square wave signal. I did make sure to connect the grounds as short as possible and close to that clock signal.

Using 1x and 10x scope probe options I got the following results:

1x probe: oscilloscope screen showing roughly a sine wave shape 10x probe: oscilloscope screen showing a square wave with ringing

Well, for the 1x I think that I'm getting attenuation given by the low pass filter of the 50MHz scope for a 25MHz signal, and that could explain as well why the signal is not square and the amplitude is lower...

But the part I don't understand is the result of the 10x option: why is that ringing happening? And why is it so different to the 1x?

Obviously that overshooting is causing that the signal to increase the amplitude to 5.39Vp...

\$\endgroup\$
  • \$\begingroup\$ You already received many answers, but I would like to add this link: youtube.com/watch?v=aJsJibDNg9M (It's a video by Jack Ganssle about the perils of probing and how the impedance of your probe changes drastically - really drastically - with frequency). \$\endgroup\$ – Sredni Vashtar Apr 9 '17 at 23:04
  • \$\begingroup\$ You should check the datasheet of the probe. Usually the 10x attenuation has the bandwidth spec while the 1x has a much lower bandwidth (e.g. 0.1x of the bandwidth spec). So something you see at 10x, you won't see at 1x. In this case I believe the waveform from the 10x settings is closer to the true waveform and the ringing comes from your device under test. \$\endgroup\$ – user3528438 Apr 9 '17 at 23:04
6
\$\begingroup\$

As @user2233709 said, first adjust your probe's compensation. If the compensation is not set properly, you won't see a sensible trace on the oscilloscope.

Second, a 50MHz scope won't do a good job of showing a 25MHz square wave; it doesn't have enough bandwidth. If you look at the Fourier series of a 25MHz square wave, it has all the odd multiples of 25MHz: 75MHz, 125MHz, etc. A 50MHz scope will roll off those high frequency signals, and the result will look closer to a sine wave.

Third, since the signal with the 1X probe looks pretty much like a sine wave, it suggests that your circuit is getting loaded down by that probe. That's why you switch to a 10X probe, but even then, it may be too big a load, and just looking at the signal may be distorting it.

\$\endgroup\$
5
\$\begingroup\$

Most likely you are ignoring the mismatched probe impedance with a long ground wire to the clip. Any rise times < 50ns must be done without a ground clip and tip using the tip and barrel , otherwise the probe inductance is added to the signal and will ring at say 50MHz for a typical long probe ground.

This assumes that the Risetime and overshoot is already calibrated in the probe and this 1st order filter does not affect this ringing, other than slight gain change with Cadj.

A 1mm wire diam with length 100mm is about 100nH and 2mm ( or 2% of length) diam is about 10% less and 0.1mm thick ( D=0.1% of length) is about 150nH.

While two parallel L's are always 1/2 inductance of same D/l ratio, of a single wire.

It is the cable capacitance and ground wire inductance and your test method at fault..

Proper 200MHz measurement method for textbook waveforms from ideal source.

enter image description here

Twisted pair of about 120 Ohms is another method. Impedance is the ratio of conductor diameter to gap ground or adjacent signal. e.g. track width/gap can be around 50 Ohms with e=4 dielectric.

Of course long breadboard wires will do the same thing.

N.B. It is NOT the probe differential capacitance at the tip with 10M , but the entire coaxial cable to scope at some 20pF/ft or 60pF/m so a 1m coax probe cable of ~60pF and 100nH 10cm ground clip will ring about 60MHz (mental calc, U check)

\$\endgroup\$
3
\$\begingroup\$

Good probes have a variable capacitor to compensate the internal capacitance of the oscilloscope’s input. You should adjust it, using the test signal good oscilloscopes provide.

The oscilloscope capacitor behaves like a 1MΩ resistor paralleled with a few-pF capacitor. A 10× probe is a 9MΩ series resistor that build a 1:10 voltage divider with the oscilloscope’s input. But it also build a low-pass filter with the internal capacitance. So a compensation capacitor is added, paralleled with the 9MΩ resistor, so that the voltage divider is frequency-neutral.

If that compensation capacitor is too low, you get a low-pass filter. If it’s too high, you get a high-pass filter (what you get).

\$\endgroup\$
2
\$\begingroup\$

Is this a 1x/10x switchable probe? If so, your probe likely doesn't get the rated bandwitdh in 1x mode, giving you only the fundamental, and a quite attenuated one at that. See this video by EEVblog Dave L. Jones for more information:

https://www.youtube.com/watch?v=OiAmER1OJh4

For the 10x: Fourier analysis teaches us that there is a lot of high-frequency content in a square wave. This means that your 25MHz square wave also have frequency components at 75, 125, 175, ... MHz. This could also lead to a quite distorted signal - A quick glance seems to indicate that you ahve the first and third harmonic only (square wave has odd harmonics). On top of that, make sure your probe is well adjusted. Most probes have adjustable compensation capacitors. You need to tweak this capacitor (ideally with a teflon or nylon screwdriver) using the compensation terminals on the scope - explaining the entire procedure here would be outside the scope (hah, see what I did there) of this page.

\$\endgroup\$
  • \$\begingroup\$ e.g. lgstatic.transcat.com/media/pdf/rigol-PVP2350-User-Guide.pdf For Rigol PVP2350, the bandwidth at 1x settings is 1/10 of the 10x setting. \$\endgroup\$ – user3528438 Apr 9 '17 at 23:08
  • \$\begingroup\$ You would need a very steep filter for that ringing to come out as a result of lack of harmonics. Nothing that arise from a band limited probe+scope. \$\endgroup\$ – pipe Apr 10 '17 at 12:41
1
\$\begingroup\$

Here is model of 10X scope probe (using just LRC model);

enter image description here

The IC has 31 Ohms Rout, providing the dampening.

The probe is 15pF and 100nH (that ground lead).

The ringing is about 300MHz.

In your case, I suspect the FPGA has VDD & GND ringing, which the output driver Rout filters when loaded by a 200pF? X1 probe.

Can you examine one of the outputs set to logic "0" to see GND, or set to logic "1" to examine internal VDD, as the FPGA continues its clocking?

\$\endgroup\$
0
\$\begingroup\$

The ringing you're seeing on the 10:1 probe mode is what the signal actually looks like. The 1:1 setting is generally severely bandwidth limited, so it won't pass those higher frequency components to the scope.

If you want to learn more about why that would matter, you can check out the Keysight Oscilloscopes YouTube channel, I've talked about both of these issues in recent videos. https://www.youtube.com/keysightoscilloscope

\$\endgroup\$
  • \$\begingroup\$ Actually the ringing is from the ground wire L and the coax C and low ESR source giving high Q. Reducing length of L to zero , will get you text book waveforms in a good matched probe up to the rise time limits. \$\endgroup\$ – Sunnyskyguy EE75 May 15 '17 at 19:21

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.