Actions
Forum Thread Details
  • Replies 21 replies
  • Subscribers 356 subscribers
  • Views 350 views
  • Users 0 members are here
Related

Observing 96MHz clock signal and its divisions leads to "strange" results

obones

Hello all,

I'm toying (pun intended) with ideas in the context of a future challenge and while I'm not settled on what the thing should do, I'm decided on using the MAX32666FTHR2 board that I was gifted with a couple of years back.

The "Light Up Your Life Challenge" gave me a push to finally use the 8*8 WS2812B matrix that I once bought because it looked nice and was not expensive, and so I'm trying to hook it up to the board.

As I can't seem to get the proper output signal to send to the LEDs, I decided to observe what's coming out of the MAX32666 pins.

My first experiment was to use a continuous timer, whose clock is half the system clock (96MHz) and which toggles the output pin every time the counter reaches the comparison value. Here is what I get with two relatively high values:

image image

The signal should be a nice square wave, it does not look that clean, but I know there's always some rise and fall time coming into play.

But things get worse when the comparison value for the timer is even lower, like here with 3:

image

After further fiddling with the various peripherals on the MCU, I was able to output the 96MHz clock signal, and this is what I get:

image

I had to stop the oscilloscope because it was not able to see a proper trigger level, let alone compute the frequency from such small signal.

Now, if I read the graph properly, there's about 10ns between two peaks, which is consistent with a 96MHz frequency for the signal.

However, what I'm really wondering about is the fact that the faster the signal, the more distorted I observe it, to the point where it won't even reach full voltage swing between 0V and 3.3V.

Out the top of my head, I can see three reasons for these results:

  1. Reaching the limits of the MAX32666 pin switching capabilities
  2. Reaching the limits of what the oscilloscope can see
  3. Reaching the limits of my own stupidity

For point 1, I tried looking for rise and fall time on output pins in the electrical characteristics for the MAX32666 but could not find anything related to this in the (somewhat confusing) datasheet.

For point 2, the DSO is said to be "100MHz" so as 96 is lower than 100, it should be fine. But maybe Shannon law should be taken into account and 100/2 being lower than 96, the results are expected.

For point 3, well, it could come on top of the two other points, and so I won't rule it out just yet.

I know there is a "Oscilloscope 101" webinar in the works which might actually answer those questions just fine. If that's the case, I'd gladly wait for it.

Thanks for your attention.

  • Some other question that come to my mind:

    - you never mention what scope you are using (from the probes it seems a Tektronix). This should not matter, but might give hints into what else could go wrong.

    - you also do not mention what that board is you are using. Is there a schematic for it?

    - you mention that you were 'able to output the 96MHz clock signal' - this hints at that this is not the actual oscillator, but another pin at the MAX32666, right? Are you _sure_ that there is nothing else connected to that pin? How do you probe to that pin? At 96MHz even moderate resistors or capacitors can wreak havoc with your signal...

  • in reply to hlipka

    The scope is a Multicomp Pro MP720012

    The board, I mentioned it in the original message, it's the MAX32666FTHR

    The 96MHz oscillator is internal to the MAX32666 so you can only see it if you output it on one of its GPIO pin, which I did via the Audio subsystem bit clock configuration.

  • in reply to obones

    I took the MAX32666FTHR for a part number, not the board name :-(

    Short calculation (well, I used calculator.academy/.../ ): at 96MHz, a capacitance of 10pF results in an effective impedance of 166Ohm. 100pF (which very well happens when your probe is in 1:1 mode) will be about 16Ohm. 8mA through 16Ohm are about 130mV.

  • in reply to hlipka

    MAX32666FTHR is part number (and name) of board. MCU part number is MAX32666GXMBL+. Schematics is in board datasheet. On board website, there are even gerbers.

  • in reply to obones

    Most oscilloscope probe bags have accessories including the spring ground - it should be fitted by removing the ground clip lead, removing the "hook" attachment and pushing over the coaxial ground contact collar which is around the probe tip as shown in the image at this link: https://electronics.stackexchange.com/questions/136123/how-do-you-attach-an-oscilloscope-ground-spring

    - Gough

  • in reply to obones

    Yeah. That's a bad idea. Probe loading is hefty at 1x. You should be using 10:1 for anything above around 8MHz in my experience.

    - Gough

  • in reply to obones

    The oscilloscope has a 100 Mhz bandwidth. If you probe a 100 MHz square wave signal with it, you 'll notice a few things - even with perfect probing techniques

    • The signal root wave (a 96 MHz sinus) level will be 3 dB down
    • The rest of the square wave is made up of frequencies above the scope's upper bandwidth level. You will not get those.

    So in general: if you use the x10 mode, you'll get an attenuated sinus-ish result at best.

  • in reply to Jan Cumps

    Agreed... and the shape change to me is more profound than the 3dB drop.  You first notice it in the edges... rather than a nice clean edge they become sine like.  And although mathematically they have nothing to do with triangles, to me, in terms of appearance, they do look like imperfect triangular waves... and ultimately when you get really close to the limit, they become miniscule sinewaves.  If the signal being probed was a pure sine wave I suppose the amplitude would be 70.7% at the bandwidth limit... but since, as you pointed out, that a square wave is composed of even higher frequency components, the amplitude will be much less than 70.7%.

  • in reply to Gough Lui

    Ah!
    So that's what this spring is about!
    I thought it was a spare for the spring inside the hook...

  • in reply to obones

    I quite often improvise 'springy clips', like this

    image

    That way, you can solder them, rather than hold the probe with one hand whilst working the controls with the other, though in this case it's quite precarious the way that I just have the tip touching the blob of solder (don't do that in the middle of a valuable board that you don't want to damage). The component between the blobs of solder is the SMD load resistor that I'm trying to probe as cleanly as possible. Sometimes, if I can be bothered, I'll do a little spiral for the probe tip to sit in which can also be soldered in place.

    My improvised ground clip was a lead offcut from a component (some old, cheap, but actually very good LEDs) and is plated steel rather than copper - copper is too soft and won't hold it's shape gripping the ground sleeve of the probe.

    If you want to see what I was doing with it, it's in the blog below; re-reading that, it's a right mess - I haven't explained what I was doing at all well. And I was wrong about the risetime. That was done with an old 200MHz Tek 2-series scope and the x10 passive probe that came with the scope, so the risetime you see in the waveforms is actually that of the scope and not the pin drivers on the Arduino Uno, nor the output of the transformer, which are both probably beating the scope, but it does show the clean results, without ringing, you can get in a low impedance circuit where the probe isn't loading it too much, and it does show how good the performance of a transmission-line pulse transformer can be, so not entirely wasted as a blog.

    https://community.element14.com/members-area/b/blog/posts/experimenting-with-a-transmission-line-transformer