Filtering an oscillator's output

Thread Starter

cmartinez

Joined Jan 17, 2007
6,886
I built a special kind of controller that uses seven MCU's working in parallel, and I'm quite pleased with the result. The arrangement works using a single oscillator whose output is connected in series to each MCU, and so far things are working nicely. The thing's supposed to generate a 0-5V 14.7456 MHz square wave, but I was surprised to see a horrible amount of under and overshoot when I scoped it.

upload_2018-12-29_12-41-43.png

The components are arranged so that the oscillator is installed on one side, and then a single line is connected sequentially to each MCU, which are separated about 2-1/4" inches from each other. That is, the total trace length from the oscillator to the last MCU is about 15-1/2". And that is a looooong trace to be used for this purpose, I know.

I'm guessing this dirty reading might be due to one of two reasons:
1.- My scope sucks. It's a humble digital hantek dso-2250 usb 100 MHz scope.
2.- The long trace is causing an inductive load with the above results.​

Any suggestions on how I might clean that signal a bit, making it more square-like? ... other than redesigning the entire PCB, of course... Or is it a non-issue, since it's working fine anyway?
 
Last edited:

crutschow

Joined Mar 14, 2008
24,716
Are you connecting the scope probe ground lead on the ground next to MCU's?
If yes, than the overshoot is likely due to the long unterminated trace length.
Is there a ground plane beneath the trace?

You might try adding about a 100Ω resistor in series with the oscillator output.
That should provide a rough match to the trace impedance and reduce the overshoot and ringing.
You can experiment with the value to see what works best.
Alternately you can try a ≈300Ω resistor (the minimum the clock can drive) to ground at the end of the trace (far end from the oscillator).

But if the circuit is working okay, then you may not need to do anything.
 
Last edited:

Thread Starter

cmartinez

Joined Jan 17, 2007
6,886
Are you connecting the scope probe ground lead on the ground next to MCU's?
If yes, than the overshoot is likely due to the long unterminated trace length.
Is there a ground plane beneath the trace?

You might try adding about a 100Ω resistor in series with the oscillator output.
That should provide a rough match to the trace impedance and reduce the overshoot and ringing.
You can experiment with the value to see what works best.
Alternately you can try a ≈300Ω resistor (the minimum the clock can drive) to ground at the end of the trace (far end from the oscillator).

But if the circuit is working okay, then you may not need to do anything.
Darn! I had forgotten about the resistor in series! ... I normally always use one in all of my designs. I just installed a 100Ω one with the following result:

upload_2018-12-29_13-50-15.png

It's quite an improvement. And it even seems to be overcompensating. I'm going to try it again with 75Ω and 47Ω and let you know how it went.

Many thanks!
 

Thread Starter

cmartinez

Joined Jan 17, 2007
6,886
Yeap ... 47Ω did the trick:

upload_2018-12-29_14-26-30.png

There's still an itsy-bitsy amount of undershoot present in the wave, but hey, what's a few millivolts between friends? :p

Thanks again, Crutschow!
 
Last edited:

MisterBill2

Joined Jan 23, 2018
5,763
I was going to ask about if you had checked the compensation adjustment on your scope probe, but it seems that you are not using a standard 10:1 PROBE. But it does matter. Glad he problem is solved.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
6,886
I was going to ask about if you had checked the compensation adjustment on your scope probe, but it seems that you are not using a standard 10:1 PROBE. But it does matter. Glad he problem is solved.
Yeah, that's one of the first things I checked. And in fact my probe was very slightly off. But not by as much as it appeared to be at first.

The type of probe I used is a selectable 1x to 10x standard one.
 
Top