Why is my clock signal overshooting when using a 10x probe?

[Tweedle_Dee]

I'm working on a microcontroller project where I'm sending out a clock signal at 153kHz and then converting it to 0 to 12V levels using a n-channel MOSFET. The wires are approximately 2 feet long after the conversion. I was disturbed by the excessive overshoot I was seeing on my USB scope. I then changed the probe from 10x to 1x and the overshoot seemed to go away for the most part. As I understand it, setting your probes to 10x will increase the input impedance by the same amount. I'm not sure what effect there is on capacitive loading. The 1x probe plot looks like there is more capacitance. I'm using a passive generic crummy probe that came with my USB scope. Can anyone help me understand what is going on? Thanks.

http://photos.imageevent.com/crossfamily/campingatsunsetbay/large/Overshoot_clk_10xProbe.PNG

http://photos.imageevent.com/crossfamily/campingatsunsetbay/large/Overshoot_clk_1xProbe.PNG

 

[Tweedle_Dee]

Just noticed that my probe shows 115pF on 1x and 16pF on 10x. However, when I put a 22pF or 100pF cap to ground in parallel with my probe set to 10x, the overshoot increases??

 

[Bob S]

Nice USB 'scope. What is the pullup resistor on the n-channel MOSFET drain? Looks like it is fairly large. For high frequency work, where the rise and fall times are sub microsecond, you should use coax cables, with the cables terminated with a resistance that equals the characteristic impedance of the cable (usually 50 to 75 ohms).

Bob S

 

[Tweedle_Dee]

1k resistor on drain. I didn't think 150k was high enough frequency to worry about reflections although I know as the wire length increases, it's more likely to become an issue.

 

[uart]

Hi TweedleDee. The overshoot is not necessarily real but may be an anomaly within your probe, this is actually pretty common. That's why most good probes have a small adjustment screw to tweak the rise/fall times versus overshoot. You normally calibrate them by connecting to a pulse source with known clean fast edges and then adjust the calibration screw (typically a variable capacitance) until you get the best shape clean pulses displayed on the scope.

Look carefully at the probe (look both at the probe end and at the connection end) and you may be able to find such a calibration screw (usually well recessed and out of sight). If you can't find one then your probe might be too cheap to include this feature. In that case then you're stuck with an out of calibation probe that will always show overshoot (on fast edges), even where there really isn't any.

 

[The Electrician]

A typical passive probe will have only about 6 to 10 MHz frequency response on the 1X setting, whereas the response on 10X will probably be 100 MHz or more.

The overshoot just isn't visible on 1X.

It looks like your USB scope is a two channel model. Connect the other probe to channel 2 and set channel 1 to 1X and channel 2 to 10X. With both probes connected to the same point, compare the two traces and see if you don't get more apparent overshoot with the 10X probe.

 

[uart]

The wires are approximately 2 feet long after the conversion.

Whoops I'd previously missed that point about the 2' wires. In that case the overshoot is very likely real and "the electricians" explanation above is the most likely.

You've probably got something like 2 uH and about 30 to 50 pF loading their, so oscillation at about 15 to 20MHz is to be expected. The reason that the oscillation is only on the trailing edge is because the pull up resistor is large enough to make this well and truly overdamped when the mosfet turns off. When the mostfet turns on however the resistance is probably only a few ohms so the parasitic LC circuit is underdamped and the oscillations result.

 

[Tweedle_Dee]

A typical passive probe will have only about 6 to 10 MHz frequency response on the 1X setting, whereas the response on 10X will probably be 100 MHz or more.

The overshoot just isn't visible on 1X.

Ah yes, that makes sense. Hadn't thought of that.

You've probably got something like 2 uH and about 30 to 50 pF loading their, so oscillation at about 15 to 20MHz is to be expected. The reason that the oscillation is only on the trailing edge is because the pull up resistor is large enough to make this well and truly overdamped when the mosfet turns off. When the mostfet turns on however the resistance is probably only a few ohms so the parasitic LC circuit is underdamped and the oscillations result.

Okay, this is starting to make sense. So, would a possible fix be to add a 1k resistor to the output causing it to be loaded at all times? 1k load when turning on, 2k load when turning off.

 

[The Electrician]

You haven't told us whether your application requires fast edges or not. If you slow down the edge rate, that will add skew to the clock signal you're converting to 12 volt level.

Assuming you don't care about the slowing of the edge rate, try putting a trimpot of several kilohms in series with the hot side output (not the ground, in other words) of the MOSFET to the 2 foot wires. Adjust the resistance of the trimpot until the overshoot goes away.

You could also add a clamp diode at the far end of the wires to prevent the voltage from going below ground more than a diode drop; this would preserve the fast negative going edge rate, which may or may not be desirable.

 

[Tweedle_Dee]

You haven't told us whether your application requires fast edges or not. If you slow down the edge rate, that will add skew to the clock signal you're converting to 12 volt level.

As far as I know, fast edges are not important so I think this fix will be acceptable. I'll experiment with it tonight and report back. Thanks a ton.

 

[Tweedle_Dee]

That helped a lot. Thanks for all the input. This is a great forum!

http://photos.imageevent.com/crossfamily/campingatsunsetbay/large/ConsultClk_w_470ohms.PNG