A measurement system is only as good as its weakest link. The bandwidth of an oscilloscope is always a key banner specification, but there is more to the measurement system than just the oscilloscope.
In fact, the oscilloscope is often not the weakest link in the measurement system. A measurement system also consists of probes, cables, connectors, and even fixtures. Each of these elements has the potential to cause more loss of bandwidth than the oscilloscope. While cables and connectors typically have very low loss, this is not the case for probes and probe accessories.
Oscilloscope vendors continue to have difficulty keeping probing bandwidths on par with oscilloscope bandwidth. The strongest evidence of this is that real time oscilloscope bandwidths have achieved up to 63 GHz of bandwidth (Agilent’s DSAX96204Q), yet probe bandwidths are limited to a maximum of 30 GHz (Agilent’s N2803A).
The other difficulty with probing bandwidth is that different vendors use different methods to correct for frequency response non-linearities. Probe correction is simply a filter applied to the oscilloscope at run time to make the frequency response of the probe flat.
A flatter frequency response leads to more repeatable and accurate measurements. The difficulty in designing high bandwidth probes as well as the variation in probe correction methods make probes prime candidates to be the weakest line in the measurement system.
As a result, for the best measurement accuracy, it is important to look beyond the oscilloscope bandwidth and know the impacts of the probe bandwidth and its frequency response.
When measuring probe frequency response, there are two different methodologies typically used by oscilloscope vendors (Vin/Vout and Vsrc/Vout). To understand how these are different, one must understand the key terms.
Vin: The voltage at the input of the probe as loaded by the probe
Vout: The voltage as seen by the oscilloscope via the probe
Vsrc: The voltage at the probe tip with an ideal probe, in other words the voltage at the source with no probe connected
The goal of Vin/Vout and Vsrc/Vout frequency response corrections is to keep the response perfectly flat to the bandwidth of the probe. This is done by correcting Vout to equal Vin for Vin/Vout correction or by correcting Vout to equal Vsrc for Vsrc/Vout correction. Figure 1 graphically shows this depiction for better understanding. The fundamental difference between these two methods is that one includes probe loading (Vin/Vout), while Vsrc/Vout ignores the loading effect of the probe in its correction. The Vsrc/Vout assumption is a safe one in a 50 ohm environment (50 ohms is the assumption of the device impedance) or if the person applying the probe correction knows the impedance of device being probed; but as the transmission line drifts from 50 ohms, it will cause this correction method to be less accurate, unless the correction can be changed with source impedance. Because Vin/Vout correction considers probe impedance, the correction will be correct regardless of the transmission line impedance. Fundamentally being able to stay accurate regardless of source impedance is one advantage for using the Vin/Vout correction.
If you are unsure what correction method to use, there is a very quick method to measure the response of your probe.
- Using the probe calibration/deskew fixture provided by the oscilloscope vendor (such as Agilent’s E2655A), connect the fixture to channel 3 of your oscilloscope.
- The E2655A provides an input and acts as a through when nothing is connected. Connect the input of the E2655A to a fast calibration edge. For instance, the Agilent 90000 X-Series provide a fast calibration edge on the front panel that works very well for this operation. Enable the AUX OUT of the oscilloscope to be “Fast Edge”, another example is Agilent’s N2806A calibration edge as it provides sub-10 ps edges.
-
Trigger on the edge and do the following in channel 3:
-
Turn the scope into averaging mode (recommend at least 1000 averages)
Figure 1: Fast edge ran through the deskew fixture and deeply averaged -
Turn on the differentiation math function of the fast step that is captured in step a, this creates a pulse.
Figure 2: Pulse created by differentiating the fast edge -
Turn on the Fast Fourier Transfer Function (FFT) of the pulse created in step b. This provides the full voltage of Vsrc (the voltage at the probe tip of an ideal probe).
Figure 3: FFT of the source through the deskew fixture - Save the FFT created in step c. as memory 1 on the oscilloscope
-
Turn the scope into averaging mode (recommend at least 1000 averages)
-
Connect the probe of your choice (preferably an active probe) to channel 1 of the oscilloscope and probe the tip on the E2655A fixture (see image 1).
Image 1: The probe connected to channel 1 and the tip connected to deskew fixture on channel 3 -
The signal you see on channel 1 is the fast edge loaded by the probe. In this step, the fast edge signal has both the loss of the probe and the probe loading impacting the measurement. Repeat step 3, but using this channel 1 signal (and save to waveform memory 2). You should now have a new frequency response, which is the response of the probe or Vout.
In steps 3 and 4 you have measured Vout and Vsrc. Only one more voltage is needed – Vin. This is the voltage of the probe at the tip of the probe as loaded by the probe. The nice part of this process is that you already have this measurement easily available to you. The through you have from the E2655A is now showing Vin as the signal has changed due to the loading of the probe tip. - To measure Vin, repeat step 3 using channel 3 again, but this time with the probe connected to the fixture. Save the new waveform to memory 3.
Now that you have all three signals, simply use the divide function in the oscilloscope to see the probe correction or the response of the probe. To measure Vin/Vout, divide waveform memory 3 by memory 2. To measure Vsrc/Vout, divide memory 1 by memory 2.
Again, the goal of an oscilloscope, and ultimately a probe vendor, is to correct the probe response (Vout) such that Vin is equal to Vout for Vin/Vout correction or Vsrc to equal Vout for Vsrc / Vout correction. The reason the vendor wants Vout to resemble Vin (for Vin/Vout correction) is that the vendor does not want its customers to see dramatic differences in measurements caused by the loss of the cable between a measurement that only has probe loading as part of it (Vin) and what the oscilloscope actually sees after probing (Vout). When you look at Vin divided by Vout, the closer this response is to flat, the better the correction and the better the signal depicted on the oscilloscope reflects reality.
The intent of this paper is not to discuss the advantages and disadvantages of Vin/Vout and Vsrc/Vout probe correction, but both correction methods have their advantages. Ultimately Vsrc/Vout corrects the response of the probe with the assumption that the probe has no loading. This correction requires some knowledge of the transmission line being probed. To correct for Vsrc/Vout, the oscilloscope vendor must make an assumption about what your transmission line is for its impedance. The assumption is that it is a 50 ohm line. While this tends to be the standard, transmission lines vary and once this assumption breaks down, the probe correction becomes less accurate. The other method is Vin/Vout. This method includes the loading of the probe and because of this; its correction is independent of the source impedance.
As explained earlier, each oscilloscope vendor chooses for the user the method of probe correction that is applied to their oscilloscope and typically end users do not have control over their correction method. One emerging alternative is software that allows users to choose their method of correction. Agilent Technologies provides this functionality through its PrecisionProbe software. The software allows users to measure the frequency response of the probe and then correct this frequency response. PrecisionProbe software also enables users to choose the correction method. Another key feature is that if your transmission line is different than 50 ohms, PrecisionProbe provides the ability for you to choose the source impedance. In other words, it removes the assumption of a 50 ohm environment and makes the Vsrc / Vout method accurate regardless of the source impedance.
PrecisionProbe is an emerging alternative to probe correction as it gives some control of the probe response to the end user. Oscilloscope vendors spend years designing high bandwidth probes to give them high bandwidth and high accuracy. Unfortunately, the hardware design is not enough and probes need digital signal processing (DSP) correction to make them more accurate. However, even with this correction, the probe may still be the weakest link in the measurement system. To understand if this is a problem in your measurement system, you need to understand how to measure a probe response and what correction is applied to the probe. Today, tools such as Agilent’s N2809A PrecisionProbe software are available to make measuring the probe frequency response easy and also give oscilloscope users capabilities that previously were not possible. Understanding this key component in the measurement system will give you more repeatable and accurate measurements.