2 different scope representations of the same RF signal

[kronos]

Hello,
I have a strange phenomenon and I do not know how to explain it.

I generate a 125 MHz signal with a Si5351 board. I can view it in the scope (Rigol, 100 MHz) in 2 ways:
1) I view it with the x10 probe; I use the spring connector (not the long ground lead) directly at the output SMA connector. I see this image, which is sort of expected: it is not square, because the BW of the scope is not enough.



2) I connect a short RG58 coax to the output SMA, and connect it to the scope; the connection is via a T-connector, with a 50 ohm load at the third end. To my surprise, I see this image: much less amplitude, but much better shape.



I cannot explain why both are so different in both aspect and amplitude. I do not understand why the signal looks better with the coax, it is even a square wave, though much smaller. It has even a rise time of 1.24 ns. How is this possible in a 100 MHz scope?

[radiolistener]

I cannot explain why both are so different in both aspect and amplitude. I do not understand why the signal looks better with the coax, it is even a square wave, though much smaller. It has even a rise time of 1.24 ns. How is this possible in a 100 MHz scope?

si5351 has pretty fast rise time 1 ns. It means that signal cosists of harmonics about 1 GHz. 1:10 probe cannot handle it, because it has too high input capacitance and too small bandwidth.

When you use probe, it just cutoff all harmonics above 50-100 MHz and reduce amplitude for the signal above 50 MHz. And you will see just low frequency components of the signal. This is why it looks more like sine wave.

Such behavior is very predictable and common for any probe, because almost all probes have limited bandwidth and cannot handle 1 GHz without significant attenuation.

So the only way is to use good coax cable with 50 Ω pass-through dummy load on oscilloscope side.


Note: in order to avoid invalid amplitude, you're needs to put 50 Ω pass-through dummy load at oscilloscope connector. si5351 should be connected to coax cable with no load, because it's output is close to 50 Ω (it's about 55-60 Ω)

[MarkL]

If I understand correctly, you are using the x10 probe directly on the SMA connector with no 50ohm load on it.  With your coax scenario, you are driving a 50ohm load.  The amplitude with the x10 probe is more because it is driving less load.

If you put a 50ohm terminator or load directly on the SMA connector and then measure the SMA pins with the x10 probe, you should get more comparable results.

A scope will have a rolloff in frequency response.  It doesn't stop instantly at its rated bandwidth, and most scopes will surpass their rated bandwidth.  So you should still see some squaring off of the waveform.

You didn't say what scope you're using, but if it's a Rigol DS1054Z upped to 100MHz, testing has shown it to have a bandwidth of around 130MHz.

[kronos]

si5351 has pretty fast rise time 1 ns. It means that signal cosists of harmonics about 1 GHz. 1:10 probe cannot handle it, because it has too high input capacitance and too small bandwidth.
When you use probe, it just cutoff all harmonics above 50-100 MHz and reduce amplitude for the signal above 50 MHz. And you will see just low frequency components of the signal. This is why it looks more like sine wave.

Yes, this was my expectation, and indeed the result with the x10 probe seems logical to me.

So the only way is to use good coax cable with 50 Ω pass-through dummy load on oscilloscope side.

Note: in order to avoid invalid amplitude, you're needs to put 50 Ω pass-through dummy load at oscilloscope connector. si5351 should be connected to coax cable with no load, because it's output is close to 50 Ω (it's about 55-60 Ω)

And this is my problem. With the coax, I see a square wave and much less amplitude. Not expected. The scope input has 1Mohm and 13 pF, so that reduces to 13 pF input in parallel with the 50 ohm termination, after the 50 ohm coax.
How can the analog front end be able to display a square wave? The 3rd harmonic is at 375 MHz, way above the BW limit. I should see a sine wave, coax or not. It even measures 1.24 ns risetime, this corresponds roughly to a scope BW of 280 MHz, it should not be able to measure this risetime!

[bdunham7]

You can measure that signal 10 different ways on 5 different 100MHz scopes and you'll get 50 different pictures.  There is no 'real' signal because the scope, probes and connections become part of the circuit.  A fast (at least 500MHz) scope with a 50R input might give you a better picture, but even that won't be perfect.

A hacked Rigol 1054Z has a rather unpredictable response at these frequencies.  The measured risetime of a known fast edge will vary quite a bit, from 3.5ns down to below 2ns, depending on the vertical gain (that's a bug, not a feature) and the bandwidth roll off is gradual, so there is still a significant response up to 400MHz or so.  There's nothing unusual about what you are seeing but trying to figure out every detail of why it looks that way thinking that you can extract usable information is probably not all that helpful.   

[kronos]

If I understand correctly, you are using the x10 probe directly on the SMA connector with no 50ohm load on it.  With your coax scenario, you are driving a 50ohm load.  The amplitude with the x10 probe is more because it is driving less load.

Correct, yes, x10 probe with no load. Well, the capacitance of the probe is loading the Si5351, isn't it? At 125 MHz, that is 100 ohm.

If you put a 50ohm terminator or load directly on the SMA connector and then measure the SMA pins with the x10 probe, you should get more comparable results.

I tried, and this is the result.



I do get the same amplitude, yes! It also looks more like a square wave. Not as good as with the coax, though.
What is the interpretation? was I seeing the complete amplitude as there was no load at the Si5351? So with 50 ohm load, if its output impedance is 60 ohm, I should see ca. 1.3V, if its output voltage is 3V.

A scope will have a rolloff in frequency response.  It doesn't stop instantly at its rated bandwidth, and most scopes will surpass their rated bandwidth.  So you should still see some squaring off of the waveform.

You didn't say what scope you're using, but if it's a Rigol DS1054Z upped to 100MHz, testing has shown it to have a bandwidth of around 130MHz.

It is a DS1102. Even assuming 130 MHz BW, it should not be able to display a square wave at 125 MHz (the 3rd harmonic is at 375 MHz), and it should not measure below about 2.6 ns risetime. Which it does now also with the x10 probe.

[kronos]

You can measure that signal 10 different ways on 5 different 100MHz scopes and you'll get 50 different pictures.  There is no 'real' signal because the scope, probes and connections become part of the circuit.  A fast (at least 500MHz) scope with a 50R input might give you a better picture, but even that won't be perfect.

A hacked Rigol 1054Z has a rather unpredictable response at these frequencies.  The measured risetime of a known fast edge will vary quite a bit, from 3.5ns down to below 2ns, depending on the vertical gain (that's a bug, not a feature) and the bandwidth roll off is gradual, so there is still a significant response up to 400MHz or so.  There's nothing unusual about what you are seeing but trying to figure out every detail of why it looks that way thinking that you can extract usable information is probably not all that helpful.

OK, understood. It is a DS1102, not hacked. Still I find it very strange that I can measure a risetime of 1.24 ns. You are right, I was trying to understand what I see, specially as they are very different images. I now found the major source of discrepancy, as measuring with the x10 probe but loading with 50 ohm gives a comparable result. Which I still do not fully understand.

What would experts recommend in cases like this? Measuring with the x10 probe or with coax?

[Tom45]

What would experts recommend in cases like this? Measuring with the x10 probe or with coax?

The probe has a bandwidth limit of its own. So the overall displayed signal is affected by the combination of the probe's bandwidth and the scope's bandwidth.

A higher bandwidth probe would help.

Given the output impedance of your signal source, the coax with 50 ohm termination at the scope input would be the best bet. Unless you buy a much higher bandwidth probe. Best would be to get a 50 ohm feedthrough terminator if you don't already have one since the Rigol doesn't have a 50 ohm input option.

Or, you can make your own low impedance Z0 probe. See: https://entertaininghacks.wordpress.com/library-2/scope-probe-reference-material/ for links to some articles.  With a little bit of work you can make a much faster probe that would be well suited to what you are doing.

[bdunham7]

OK, understood. It is a DS1102, not hacked. Still I find it very strange that I can measure a risetime of 1.24 ns.

AFAIK, that is the same scope, just hacked (Full BW enabled) by Rigol for fee.  It still lacks predictable over-bandwidth response characteristics--IOW, it turns into garbage rather than rolling off nicely.  The displayed rise-time issue depends on the scopes measurement algorithm, the effects of distortion and group/phase delays as well as the possibility that the front end is being overdriven. 

What would experts recommend in cases like this? Measuring with the x10 probe or with coax?

Using a scope with sufficient rated bandwidth to see your signal would be the first thing.  Then you have to consider circuit loading from whatever your connection is.  If you can have the source connected directly to the scope input, like the Leo Bodnar fast edge device, that works pretty well, although there aren't many cases where you can do that.  If your device has a 50R output impedance, then a coax terminated at the scope is probably best.  If your source is low impedance but not 50R, then a properly compensated 10M/10X probe may work, but will further limit bandwidth.  If your source is high impedance, you need an active probe of some sort or you need to figure out what your circuit loading is going to do to your signal.  Circuit loading issues with normal 10X probes start well below 100MHz--at 100MHz, the lowest input impedance passive probe that I have is actually a 10X/500R/1pF Tek 6156 that requires a 50R scope input.  There's no easy way out.

[David Hess]

It looks to me like your have the volts/div set to 100mV/div for the probe and 200 mV/div for the coaxial cable.  Before assuming a difference in probe response or loading on the source, double check that the oscilloscope response is not changing with volts/div and separately signal amplitude.  We know that the Rigol DS1000Z series suffers from problems in this area.  An RF stepped attenuator or set of RF attenuators is useful for this.

When I compare probe and coaxial cable measurements on my 100 MHz Tektronix 2232 from a much faster edge source, I get practically identical results even with one of those cheap x1/x10 P6100 probes that you can find anywhere.

[Simon]

you are driving a scope with a signal it is not meant to display properly. Why? the answer to the response of the scope is either a university thesis or you go and use the correct kit.

[David Hess]

you are driving a scope with a signal it is not meant to display properly. Why? the answer to the response of the scope is either a university thesis or you go and use the correct kit.

But a better oscilloscope will at least deliver consistent results within its measurement capabilities.

[MarkL]

...
I do get the same amplitude, yes! It also looks more like a square wave. Not as good as with the coax, though.
What is the interpretation? was I seeing the complete amplitude as there was no load at the Si5351? So with 50 ohm load, if its output impedance is 60 ohm, I should see ca. 1.3V, if its output voltage is 3V.
...

The Si5351 has a 50 ohm output impedance, so unless the board has some series resistance, you should get closer to 50% when loading it with 50 ohms.

The amplitude you're seeing is likely limited by the scope/probe's bandwidth.  To see a more accurate amplitude, you could slow down the Si5351 output until the waveform has a well defined flat top and flat bottom (where the amplitude has settled to values measurable with your equipment).