I'm seeing unexpected noise on my ocsilloscope.
If you plug nothing into the scope (or a clean 0 volt signal) and measure it on the highest resolution scale, the noise appears to be: ~1.4mV peak-to-peak.
Then if you measure it on a greater scale, say 500 mV/divisions, the noise appears to be: ~100 mV p-p
The noise is still really ~ 1.4 mV p-p or less, but the digitizer always forces a signal noise to encompass several bits, even if it is actually clean enough to be within only 2 bits.
What you are measuring is random digitizer noise which is common to any analog to digital converter (A/D). In the case of higher sampling rate A/D's, the effect is typically more pronounced. We have an internal spec for this effect for the TDS7404 as follows:
<= 300uV + (6.5% of V/div setting) >= 10mV/div
<= 320uV + (6.5% of V/div setting) < 10mV/div.
The random noise is specified in RMS values. This is because peak to peak measurements have no real meaning in random noise. The longer you measure random noise the higher the peak to peak value. If you could theoretically measure noise for an infinite amount of time, you would have an infinitely large noise measurement. The most accurate way to measure noise in RMS is to measure it using a histogram measurement. The RMS of noise is equal to the standard deviation of the signal. See Figure 1.
Figure 1: Noise at 1mV/div
The RMS noise at 1mV/div is ~204uV, which is under the specification of (320uV+6.5%*1mV) = 385uV
Figure 2: Noise at 100mV/div
The RMS noise at 100mV/div is ~2.8mV, which is under the specification of (300uV+6.5%*100mV) = 6.8mV.
Figure 3: Noise at 500mV/div
The RMS noise at 500mV/div is ~14.3mV which is under the specification of (300uV+6.5%*500mV) = 32.8mV.
Testing method overview:
Testing is done at center screen with 0V offset at 2.5GS/s with a record of length 2500 points, after performing Signal Path Compensation on the oscilloscope. The RMS noise is measured by calculating the standard deviation of the waveform record. The input is ground coupled.
常见问题 ID 59601查看所有常见问题 »