Oscilloscope and high speed signal

Thread Starter

cam0

Joined Mar 23, 2010
25
Hi,

recently I wanted to see a particular signal driven on a bus between a microprocessor and memory. The clock signal has a pretty high speed (~100MHz) and I noticed that when I plugged the oscilloscope, the communication between the uproc and the memory became faulty. Voltage drop ? Shunt capa ? I was wondering whether simply "adapting" the oscilloscope to the bus would be enough..

Thanks !
 
Last edited:

beenthere

Joined Apr 20, 2004
15,819
Were you using a 10X probe and the shortest possible ground clip? If not, the loading may have corrupted the pulse train.
 

Thread Starter

cam0

Joined Mar 23, 2010
25
Yes, it was a 10x probe and for the ground clip, it wasn't the "shortest possible", but was like 3cm, so not really big. Unfortunately, I don't have the hardware in front of me, so I cannot tell you with probe it was exactly.

Thanks
 

SgtWookie

Joined Jul 17, 2007
22,230
Don't touch the trace with the probe. You'll add a significant amount of capacitance (25-60pF) which will change the electrical characteristics of the signal path.

Hold your probe very close to the trace; the signal will be capacitively coupled to the probe, and be of very low amplitude. You'll need a very high bandwidth 'scope and probe to see anything resembling what the actual signal looks like. A 250MHz BW scope will show a very distorted signal. A 1GHz 'scope and probe will have a fairly reasonable representation of the signal.
 

Thread Starter

cam0

Joined Mar 23, 2010
25
Thanks to both of you. SgtWookie, this is a very interesting trick you are giving here. For curiosity, why do we need a higher bandwidth oscilloscope ?

Thanks again,

c0
 

SgtWookie

Joined Jul 17, 2007
22,230
If you were looking at a 100MHz sinewave, a scope with a 100MHz bandwidth would be marginally OK.

But, you are looking at square waves. Ideal square waves are composed of the fundamental frequency, plus ALL of the odd harmonics of the fundamental frequency. This implies that unlimited bandwidth is required to reproduce an ideal square wave, which is why they do not exist in the real world.

But if you have 10x the bandwidth of the square wave being viewed, you can get a reasonable approximation of it. You just have to understand what you are observing.

See the animated graphic about halfway down the page in this Wiki entry:
http://en.wikipedia.org/wiki/Square_wave
Click on the image to see it run.
 

someonesdad

Joined Jul 7, 2009
1,583
Since I happened to be working with a function generator connected to a scope, I thought I'd give a measured example of what SgtWookie mentioned. I set the generator to give a 1 volt RMS square wave at 6 MHz (this is 1/10th the scope's bandwidth). That's the top trace. The bottom trace is after the generator was set to 50 MHz (its highest frequency). You can see the loss of most of the square wave's harmonics and the RMS voltage has dropped by about 25%. The generator's output is 50 Ω and the cable was properly terminated at the scope's input.
 

Attachments

Thread Starter

cam0

Joined Mar 23, 2010
25
someonesdad, thanks for having taken the time to do the experiment. Just to clarify, were you "touching the traces" or is it retrieved by coupling ?
 

someonesdad

Joined Jul 7, 2009
1,583
The generator was connected directly to the scope with a 50 ohm coax cable and a 50 ohm terminator. As the scope and generator were on again, here's a similar set of traces using a 200 MHz 10X probe (properly compensated). The top trace is a 5 MHz square wave; the ringing is because it's coming down a 50 ohm cable and is unterminated. The scope was untouched; then I changed the frequency to 50 MHz. You're really just seeing the fundamental, as the next harmonic is at 150 MHz and is terminated too much because of the scope's 60 MHz input bandwidth.

I have another generator that puts out up to 80 MHz sine waves and I could see from about 45 to 80 MHz, the scope's response seemed to increase (it could, of course, have been the generator too). But you'll note the 50 MHz square wave from the other generator had a slightly higher RMS value (1.08), so that's consistency between the two generators; hence, it's probably the scope's behavior.

The scope is a B&K 2534 60 MHz scope and is thus, from these measurements, rated a bit conservatively -- I didn't see the amplitude drop by much until above about 70-75 MHz. My old Phillips 25 MHz scope's measured 3 dB point is 31 MHz, so it was rated a bit conservatively too.

I tried SgtWookie's suggestion of trying to pick up the signal by capacitive coupling, but this is no doubt too low of a frequency. I was either touching the conductor with the full signal or had no signal at all.
 

Attachments

kingdano

Joined Apr 14, 2010
377
this is a very interesting thread.

the trick of coupling the signal to the probe as opposed to direct connection is a good one - i have run into problems like this before at work and not known what to do.
 

Thread Starter

cam0

Joined Mar 23, 2010
25
I am learning a bunch of new stuff, thanks to everybody! This brings me to another question then, why not using a good amplifier to *try* to have a better signal when using the trick of the coupling signal ? Would it be a good idea or would we have too much distortion & error ?
 

SgtWookie

Joined Jul 17, 2007
22,230
At 100MHz, you would NEED a "good amplifier". Preferably one with infinite input impedance and infinite bandwidth, and zero capacitive loading.

It's the added capacitance that causes the problem. 30pF doesn't sound like much, but realize that the harmonics are much higher than 100MHz; at 500MHz, 30pF gets to be a pretty big deal.
 

retched

Joined Dec 5, 2009
5,207
Yeah, I need to pick up some of those amps with infinite impedance and bandwidth. I am having trouble finding the zero cap loading.. ;)

High frequencies come require high levels of equipment and patience. (Even though they are faster)
 
Top