Oscilloscope bandwidth and PWM pulse width

Question

I've been working on a project which involves a PWM signal that operates at 30khz with a 10 bit resolution. I am generating the signal with a Teensy 4.0, which has a 600mhz processor, well above the ~31mhz required to generate the PWM frequency and resolution I need. I was measuring this signal with my oscilloscope, just to verify that everything was working, and I noticed that as soon as the PWM pulse width drops below 200ns, I stop seeing well-formed pulses. 200ns corresponds to ~5mhz, and my scope has a 100mhz bandwidth, so my sense is that is should be able to display the minimum pulse width of a 30mhz signal. Am I wrong about this, or is it possible that something else is affecting the my ability to either emit or measure pulse widths below 200ns (as far as I can tell—the actual point where it stops providing clear measurements might be lower). I checked my probe and it was set to 10x, as was the oscilloscope, so I don't think that's a factor.

Here's a 300ns pulse. I was actually able to measure a 200ns pulse by increasing the PWM frequency to 48khz (sorry about my fingers, but you can see the important bits)

This is the minimum pulse width, which should be about 32ns

Answer 1

The small B in the CH1 label below indicates that the BW limit is on. click on CH1 knob and select Bandwidth Limit Off

The real signal will look more rectangular most likely.

The bandwidth limit is offered by these scopes so you are not overly confused by the oscillations when not doing a proper differential measurement with low loop impedance return cable. It also helps to see mVs of ripple in the kHz range that can be buried in 10s of mV rms wideband noise.

Answer 2

Your scope should be able to reproduce rise and fall times down to around 3.5ns.

There is a relationship between the rise time of an edge and the signal bandwidth. The bandwidth of a square wave does not depend on the square wave frequency. It will always be higher than the frequency. The exact bandwidth in the signal will depend on the rise and fall time of the square wave.

What this means for an oscilloscope is that fast square waves may not be reproduced correctly. If the scope bandwidth is too low, the apparent rise and fall times on the scope will longer (slower) than the actual rise and fall times.

Here is the approximate relationship:

BW = 0.35 / Tr

Where BW is the bandwidth and Tr is the rise time from 10 to 90 percent. Solving for Tr you get:

Tr = 0.35 / BW

So for you, Tr = 0.35/100MHz = 3.5ns.

The scope probe can also limit rise times, but in your case you are using a 150 MHz probe, so it should be OK.

The last thing to double-check is the probe compensation. An under-compensated probe will cause the rise times to appear slower than they actually are.

If the probe compensation is OK, then most likely, the signal you are seeing on the scope is the real signal. Some RC time constant in the actual circuit may be limiting the rise and fall times.