cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Observer
Observer
1,157 Views
Registered: ‎10-17-2018

Externally observe fast signals

Hi,

I am experimenting with some oscillation circuits (ring oscillators and self-timed ring oscillators). To verify the functionality of the circuits, I would like to observe the output signal using an Oscilloscope and a Logic Analyzer. The problem is I cannot get to the signals without using PMOD headers which seem uncapable of providing high frequency (>50MHz) output. When constructing a Ring-Oscillator whose frequency I estimated to be around 50MHz, and routing my clock signal to such a pin, the oscilloscope shows a signal that is not even remotely rectangular, althoug the frequency is in the range that I am expecting. The peak-peak voltage is somewhere around 1V and therefore does not trigger the Logic Analyzer. This particular design is one with the lowest frequencies I want to test. The highest would be around 800MHz.

I tried using the ILA core, but cannot clock it higher than 200Mhz, which is too slow to see the exact timing of the clock.

Is there anything I can do to verify these signals?

I have two boards available, the Digilent Nexys 4 DDR and the Digilent Genesys 2. The Genesys 2 has an FMC header that ought to be able to achieve speeds in the Gbit/s range, so that should work, right? But I have no way of connecting neither my Oscilloscope nor my Logic Analyzer to it (Touching the Pins with the Oscillator Probe yields in a result thats very similar to the one achieved when using the PMOD header).

Thanks!

0 Kudos
23 Replies
Highlighted
Teacher
Teacher
1,145 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

A few things,

Why are you getting such a low voltage on the output ?

    What is the FPGA , Even the slowest artix will give a good waveform at 50 MHz,

        You need to look at the cucruit diagram, see if anything else is on the pin your using,

           you need to check what the bank IO voltage is, 3v3 , 1v8 or what , You wpnt get more thanthat

                you need to look at what IO standard you have set on the FPGA pin, 

 

At 50 MHz, a nice clock very hard to see on a scope with just the  crock clip on the probe randomly conected to a "earth".   It needs to be short, and conected to a good earth near the source of the signal . For instance, if your meassuiring the signal at the Pmod,  try the Pmod connector ground,

 

 

 

 

 

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Highlighted
Observer
Observer
1,121 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

> Why are you getting such a low voltage on the output ?

I don't know. For faster clocks it is in 100mV range.

> What is the FPGA , Even the slowest artix will give a good waveform at 50 MHz

The Nexys Board has an Artix XC7A100T-CSG324 on it, the Genesys uses a Kintex XC7K325T-2FFG900C.

> You need to look at the cucruit diagram, see if anything else is on the pin your using

Can't see anything on the PMods...

https://reference.digilentinc.com/_media/reference/programmable-logic/nexys-4-ddr/nexys-4-ddr_sch.pdf

https://reference.digilentinc.com/_media/reference/programmable-logic/genesys-2/genesys-2_sch.pdf

> you need to check what the bank IO voltage is, 3v3 , 1v8 or what , You wpnt get more thanthat

It is 3.3V for the PMod Headers and 1.2V for the FMC.

> you need to look at what IO standard you have set on the FPGA pin

LVCMOS33 for PMod, LVCMOS12 for FMC.

> For instance, if your meassuiring the signal at the Pmod, try the Pmod connector ground

I used that one when using PMod, when just prying around in the FMC connector, I had the ground connected to the PMod Ground as well, as I cannot get in there with the clamp.


@drjohnsmith wrote:

A few things,

Why are you getting such a low voltage on the output ?

    What is the FPGA , Even the slowest artix will give a good waveform at 50 MHz,

        You need to look at the cucruit diagram, see if anything else is on the pin your using,

           you need to check what the bank IO voltage is, 3v3 , 1v8 or what , You wpnt get more thanthat

                you need to look at what IO standard you have set on the FPGA pin, 

 

At 50 MHz, a nice clock very hard to see on a scope with just the  crock clip on the probe randomly conected to a "earth".   It needs to be short, and conected to a good earth near the source of the signal . For instance, if your meassuiring the signal at the Pmod,  try the Pmod connector ground,

 

 

 

 

 


 

0 Kudos
Highlighted
1,114 Views
Registered: ‎06-21-2017

Re: Externally observe fast signals

Some boards, and the Digilent may be one, use voltage translator chips on the PMOD.  This will slow things down even more than the PMOD itself.  Your best bet may be to get a breakout board for the FMC connector.  Reiterating what others have said, you need a good short ground for your scope probe.  What is the bandwidth of your scope and probe?

Highlighted
Observer
Observer
1,107 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals


@bruce_karaffa wrote:

Some boards, and the Digilent may be one, use voltage translator chips on the PMOD.  This will slow things down even more than the PMOD itself.  Your best bet may be to get a breakout board for the FMC connector.  Reiterating what others have said, you need a good short ground for your scope probe.  What is the bandwidth of your scope and probe?


> Some boards, and the Digilent may be one, use voltage translator chips on the PMOD

Shouldn't that be visible on the schematic?

> Your best bet may be to get a breakout board for the FMC connector

I also thought about that, and the best fitting for my purpose that I could find was this one:

HW-FMC-XM105-G Datasheet 

It features two SMA connectors which I could use for the oscilloscope (they are listed as clock-inputs, but I guess they can be used for probing too, right?) and a Mictor connector for the Logic Analyzer. I think it would be perfect, but I have to convince my university to buy one.

> What is the bandwidth of your scope and probe?

I am using a 200 MHz, 2 GS/s scope with a 200MHz probe in my office, but also tried with a 8 GHz, 40 MS/s scope in our lab a few days ago with the same results. In the lab I did not take note of the probe bandwith, but I was unable to use the "good" probes with the SMA connectors as I do not have such a connector on either of my boards.

0 Kudos
Highlighted
Teacher
Teacher
1,101 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

100 mV output swing is wrong,

Are you certain your in control of the fpga , i.e. your code is working,

Your FPGA has a clock input,
do a version of the FPGA, that takes that clock in, deivdes it by say 1024, and outputs the top bit of the counter onto the PMOD pin you using,

That will be a good 3v3 logic signal,

If your design does not put this out, you have a more fundamental problem you need to solve first,


<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
Highlighted
Observer
Observer
1,097 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals


@drjohnsmith wrote:
100 mV output swing is wrong,

Are you certain your in control of the fpga , i.e. your code is working,

Your FPGA has a clock input,
do a version of the FPGA, that takes that clock in, deivdes it by say 1024, and outputs the top bit of the counter onto the PMOD pin you using,

That will be a good 3v3 logic signal,

If your design does not put this out, you have a more fundamental problem you need to solve first,



I am assuming my design works as I can see an oscillation on the ILA. As that clock is slower it is sampled, but I can see it. A colleague of mine also suggested scaling the clock down, let me see if that helps.

0 Kudos
Highlighted
Observer
Observer
1,071 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

Downscaling did not work either, so in desperation I tried using a different PMod Header. Using PMod-C instead of PMod-A results now look different.

I generated a Ring-Oscillator with 253 stages. It oscillates nicely at almost 10MHz:

slow.jpg

Scaling that same design down with an 8-bit counter, it oscillates at 565kHz with a very pretty waveform:

downscaled.jpg

Ring-Oscillator with 53 stages, oscillating at 55MHz? No problem:

53.jpg

25 stages, 118MHz? Starts to look more like a sine-wave and peak-peak voltage is down to 2.6V.

25.jpg

19 stages, 167 MHz? peak-peak down to 1V.

19.jpg

I can't go any faster without reaching the bandwiith limitations of my desk-scope, but it looks like the problem still persists with PMod-C, it will just occur later.

So the design seems to be ok, but I still run into PMod limitations. I guess I will have to talk someone into buying me a FMC card.

Thanks to you both for your insightfull input. If you have any other ideas, please let me know!

0 Kudos
Highlighted
1,066 Views
Registered: ‎06-21-2017

Re: Externally observe fast signals

You have already reached the bandwidth limit of your scope.  A since wave is made up of th4e fundimental frequency and all of the odd harmonics of that frequency.  See https://en.wikipedia.org/wiki/Square_wave   With a 200MHz scope, you will see attenuation of high order harmonics with higher frequency square waves.  This will start to make your square wave look like a sine wave.  This is not to say that the PMOD connector isn't also attenuating higher frequency signals.

Highlighted
Observer
Observer
1,026 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

That's right, but I don't really care about the form of the curve. Whats important to me is the frequency and the amplitude. As the harmonics do not affect either of them, it's ok for me. The problem arises if the frequency of the fundamental is higher than the bandwith.
I fully expect the deformation of the rectangular function to come from effects from either the connection or the scope (before your comment I did not think of the scope though), so if I see a sine with a certain frequency at the output, I expect there to be a rectangular function with that same frequency on the chip. Now one might say that I can assume the same about the voltage, but the thing is if I can't get high-enough voltages out of the chip, my logic analyzer will not trigger and I would like to use the logic-analyzer to evaluate the timing (clock jitter).
0 Kudos
Highlighted
Teacher
Teacher
1,020 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

Your scope bandwidth is probably limiting this,
The square wave amplitude is made up of the power of all the frequencies,
as you remove them the amplitude will reduce.

Also look at your probe bandwidth , it has to be a good 500 MHz one to get a 100 MHz square wave ,

at 100 MHz, you should also be looking into termination of the signal.
send us a picture of your probing of the pmod .

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Highlighted
Observer
Observer
1,001 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

> The square wave amplitude is made up of the power of all the frequencies

I don't get it. The amplitude of the fundamental oscillation of the square wave is 4/pi. That is more than one so I should actually see higher voltages, shouldn't I?

0 Kudos
Highlighted
Observer
Observer
999 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

Oh and I am not in the office anymore, as it's 8pm where I am. I will check with a better scope and probe tomorrow, although I don't know if there are better probes that do not need and SMA connector.

0 Kudos
Highlighted
Teacher
Teacher
988 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

In theory yep, but in real scopes , that dont happen,

Its either your scope / probe or its the connections,

Its strange that things changed when you use a different Pmod,

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Observer
Observer
896 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

Interesting, thanks for the clarification.

Unfortunately I cannot get into the lab today, I'll have to wait till monday.

In the meantime I am seriosely thinking about buying an FMC card. However, the one I listed earlier (hw-fmc-xm105-g) seems to be unfitting, as the SMA connectors are only for clock input (M2C, mezzanine to carrier). Did I read the datasheet correctly oder can I use those SMA connectors to output my clock from the FPGA?

If I cannot use that board, another one I found would be this one: FMC_SMA_LVDS. This board provides 8 pairs of differential C2M lines on SMA connectors. This should definitely work, right?

I found those two via google, does any other card that I did not list come to mind?

0 Kudos
Highlighted
Teacher
Teacher
869 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

Before you spend any money 

   lets get the meassurments sorted

You have this board ?

https://reference.digilentinc.com/reference/programmable-logic/nexys-4-ddr/reference-manual

 

This has a 100 MHz oscilator on it, 

    you can get to the clock pin, just on the side of the SMD oscilator , 

           get the scope on that, see what it looks like, 

What is comfusing by your measurments is that slow signals, have lovely fast rise times,

     which indicates itshould work at speed,

      yet when you speed up the signal, you get horrible round waves.

 

It also why I suggested that you take in the 100 MHz,

    pump that into the MMCM on the chip, to say 200 MHz, 

      and then make a counter, put out to your pin,

 

It ocud be that your signal inside the chip has a terrible mark space ratio, or may be as its an internal oscilation, running outside data sheet specification,

 

There is something funnny going on here, you need to get to bottom of this first before you do anything else.

 

 

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
Highlighted
Observer
Observer
836 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

This is what I see when I probe the pin of the oscillator:

oscillator.jpg

This is the output of a mmcm set up to 200MHz:

mmcm.jpg

This is when I put that on a counter and take the LSB:

half_mmcm.jpg

One bit higher:

quarter_mmcm.jpg

One bit higher again:

eight_mmcm.jpg

So the curve is slowly approaching a rectangular form, as expected. Voltages rise with sinking frequencies as before.

Whats suprising for me is the huge difference between the peak-peak voltage of the two 100 MHz signals. The one from the oscillator is 5.42V, even though it should be 0.8*VDD (which is 3.3V) according to the datasheet. The one from the scaled-down mmcm output is only 2.28V. How can that be? Other than for slightly different rise- and fall-times those two signals should be the same, right?

Is there anything I can learn from this?

0 Kudos
Highlighted
Teacher
Teacher
823 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

A few things,
The data sheet shows max / min,
so the minimum Vout of the oscillator is 0.8 * Vdd.
It should be getting to 3v3 as the FPGA is very small load,

The reason I asked you to measure the oscillator output, is that is independent of the FPGA.
If you have the spec for the oscillator, what is the rise / fall time ?

You now have two sources of square wave the Oscillator "chip" and the fpga pins, which show similar results,

The fact you have seen such a terrible square wave from the oscillator into the scope, indicates to me your scope measurement is just not capable of measuring more than 50 MHz

I suspect a bad probe, is it set to 10:1 or 50 Ohm ?

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Highlighted
Observer
Observer
748 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

> If you have the spec for the oscillator, what is the rise / fall time ?

1.2-1.3ns rise time and 1.2ns fall time. (Datasheet DSC1033CC1)

It's a LeCroy PP010, fixed to 10:1. I also tried an Agilent N2862A 10:1 probe (only capable of 150MHz though) and a generic one with no label that can be set to either 10:1 or 1:1. Results were very similar to the original LeCroy probe, still far from a rectangular signal.

0 Kudos
Highlighted
Teacher
Teacher
741 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

Looks like your best probe is only capable of 200 MHz area
Which explains the 4 ns rise / fall times you see on the oscilator output,

You idealy need to use the spring clip on the probe, not the crock clip thing,

You also need some decent, at least 500 MHz probes to match you scope,

Ive not used them, but an example:

https://www.digikey.co.uk/product-detail/en/cal-test-electronics/CT4207/CT4207-ND/9645864

Dont spoil a good scope with bad probes.
<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
Highlighted
Observer
Observer
733 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

> You idealy need to use the spring clip on the probe, not the crock clip thing

you mean for the ground connection?

> You also need some decent, at least 500 MHz probes to match you scope

Did not realize the probes need to have a larger bandwith compared to the scope. Why is that?

I was planning on checking the oscillator output using the 8GHz Oscilloscope in our Lab, but the probes on that thing have SMA connectors and I cannot attach them to my board. So I am stuck using "cheap" probes and therefore cannot expect better results, is that correct?

0 Kudos
Highlighted
Observer
Observer
321 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

I finally solved my problem by switching to a Virtex Developement Board (VCU118). That board has SMA connectors for clock output that can handle the high frequencies without issues.

Thanks for your help.

I will not accept this as a solution as it is unfeasable for everybody who cannot change to such an expensive board. In such a case, your comments are way more helpfull.

Highlighted
Teacher
Teacher
311 Views
Registered: ‎07-09-2009

Re: Externally observe fast signals

sorry missed this

Yes , your probes are very important to the signal bandwidth you see.
The signal bandwidth you cna monitor is the lowest of the scope and the probes,

if you put a 10 MHz probe on a 5 GHz scope ( as I have seen ) you willl not be able to see a GHz signal. To a first order, probes act as a low pass filter.

Add to this the effect of bad grounding on probes, and you have to be very careful to interpret your scope picture.

https://www.allaboutcircuits.com/technical-articles/an-introduction-to-oscilloscope-probes/
<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Highlighted
Observer
Observer
299 Views
Registered: ‎10-17-2018

Re: Externally observe fast signals

That's a very interesting article you linked, thanks!

0 Kudos