Getting the most out of the accuracy of your oscilloscope is not hard but it takes some attention to detail.
Getting the most out of the accuracy of your oscilloscope is not hard but it takes some attention to detail. This article looks into a number of ways to improve your measurements.
Use multiple display grids to maintain dynamic range
Consider a common practice that throws away one-quarter of the oscilloscope’s dynamic range, attenuating signals to make them fit on a single common screen as in Figure 1.
Figure 1 If you attenuate acquired signals to make them fit on a single display grid you are throwing away dynamic range.
Digital oscilloscopes are set up to map the full range of their analog to digital converter’s (ADC) to the full or almost full display grid. If you attenuate the signals from each channel to make them fit in one-quarter of the grid you have thrown away two digital bits of dynamic range. Additionally, offset has to be added to display the trace within the selected two divisions. There is an offset accuracy specification which adds an additional error source to the readings.
What you should do is use a multiple grid display and place each channel in its own grid. No attenuation or voltage offset is required, and the signals are displayed at full dynamic range as shown in Figure 2 where a quadrature display shows four grids each at full dynamic range.
Compare the top traces in each figure; notice the lower noise on each signal in Figure 2. Attenuating the input signal by acquiring it at a higher vertical scale reading reduces the signals vertical displacement but it does not reduce the internal noise in the oscilloscope channel.
Figure 2 A four grid display with a single channel in each grid displays the waveforms at full dynamic range.
The result is a lower signal to noise ratio. Look at an overlaid comparison of the signal on channel 1 shown in Figure 3.
Figure 3 Overlaying a signal acquired at 50 mV/division (green trace) with one acquired at 200mV/division (yellow trace). The yellow trace is broader, less well defined, and noisier.
The trace which was attenuated by a factor of four was acquired at 200 mV/division. When it is displayed at 50 MV/division it appears with more noise and less resolution. The unattenuated trace, in green, shows a small perturbation riding on the sine wave that is not visible in the attenuated signal that is due to improved dynamic range.
Not only does the display exhibit this loss of amplitude resolution but the loss affects other measurements as well. The peak-to-peak measurement is very sensitive to noise. Note that the peak-to-peak reading of the attenuated signal is reading that higher noise level and is 44 mV higher than the signal acquired a 50 mV/division. The rms levels are very close. This is because the rms process integrates the signal reducing the measured noise level.
The higher noise level of the attenuated signal also affects the measurement of its frequency. Note that the measured uncertainty, as expressed by the standard deviation of the measurement, is twice as great for the attenuated signal.
Just say no to attenuating signals to fit them on a single grid—use multiple grids display and show each signal at full dynamic range.
More accurate cursor reading
There are three measurement tools available in an oscilloscope: the screen graticule, cursors, and measurement parameters. Cursors are markers that can be moved over a displayed waveform and record the cursor’s location in time and the amplitude of the waveform at the intersection with the cursor. The accuracy of cursor measurement depends on the user’s ability to place the cursors accurately on the desired point of the waveform.
You can improve the accuracy of cursor placement with several simple tricks. First is the stop the acquisition while placing the cursors. The waveform varies from acquisition to acquisition, and you will find it easier to place the cursors if the waveform isn’t changing all the time. The second, and more important hint is to turn on a zoom trace or traces. Cursors track in the zoom region and the larger display makes it easier to place the cursors as shown in Figure 4.
Figure 4 Cursors track in the zoom traces, use zoom traces, which show an expanded view of the waveform, for more accurate cursor placement.
Not only does the expanded display in the zoom make it easier to see the fine details on the waveform but the rate of change of cursor movement is reduced when the cursor enters the zoomed area. The slower rate of change provides greater control in placing the cursors.
In the example shown the cursors are to be placed at the zero crossings of the sine wave. Two zoom views are displayed, one for each crossing. Cursors are placed manually while monitoring the cursor amplitude in the channel annotation box until the cursor amplitude reads 0 V.
Note that the measurement parameter P1 measures the mean period of the sine wave as 99.9999 ns compared with the cursor value DX = 100.04 ns. The measurement parameter has a much higher resolution because it applies dual interpolation operations to the determination of the period. In general, the measurement parameters provide the most accurate measurement results. Cursors, however, offer a more general measurement capability, there is not a measurement parameter for every measurement event. Where a measurement parameter exists, it will produce a more accurate result than cursors.
Selective measurements for parameters
Figure 5 shows an example of a waveform that is hard for a standard measurement parameter to make without a little help. The waveform is the gated clock for an I2C serial interface. The two waveforms are identical, and the frequency parameter will be used to measure the clock frequency using different measurement setups.
Parameter P1 is measuring the frequency of the top trace. It sees 162 cycles and measures frequencies from 68.518 kHz to 100.298 kHz. This is not surprising because the waveform’s timing is not uniform. The value reading of P1 is the frequency of the of the last acquired cycle in the waveform and shows a frequency of 73.281 kHz. Looking at the last cycle in the M2 waveform (green trace) you can see that it is longer than most of the other cycles which accounts for the lower frequency.
Figure 5 The use of measurement gates to selectively measure the frequency of the gated clock waveform.
To solve this issue there are several techniques available in the parameter setup. The first is gating, as the name implies it allows measurements to be made only between user positioned gates. Some oscilloscopes use the measurement cursors to gate the parameters measurements. This Teledyne LeCroy oscilloscope uses a separate set of gate markers shown as dashed lines on the lower (yellow) trace. The gates are set about the first clock burst and measure the frequency over the eight complete cycles contained. The frequency parameter, in this case, reads from 99.914 to 100.109 kHz. The use of measurement gates has successfully limited the range of measurements to just those of the clock and ignored the gaps. Gating makes the measurement parameters a bit more flexible.
The second measurement tool is acceptance criteria. This tool allows the parameter to measure all the values but only displays those within a user entered range as shown in Figure 6:
Figure 6 Using parameter acceptance criteria to post only frequency values that are between 99 and 101 kHz.
Acceptance criteria is set up to display only measured frequency values between 99 and 101 kHz in parameter P2. The number of measurements within the range is 144 compared to the 162 listed in P1 which shows all the measured values. The frequency parameter measures starting with the first positive going edge so there are eighteen gaps in the clock waveform that are included in the measurement which is equal to the difference between the total measured values and the accepted values. It is valid to ask how you know the range of values to use as the acceptance criteria, the next section shows how you can do that.
Track and histograms
A track is a mathematical function, available in some oscilloscopes, which plots parameter values versus time. In this example it is useful to see what parts of the waveform are associated with the different waveform events. Figure 7 shows a track based on the frequency parameter.
Figure 7 Examples of the track (F2) and histogram (F1) functions based on the measured frequency parameters.
The track of the frequency parameter appears in trace F2 (red). The vertical scale of the track is in hertz, units of frequency versus time. This trace is time synchronous with the source waveform the I2C clock. It shows a pulse waveform with top values of 100.3 kHz and base values of 68.52 kHz. The top values are uniform in values but the base shows slightly greater values at each end of the track corresponding to the beginning and end of the source waveform. The track shows where the variation in frequency occurs. Note that the width of the 100 kHz segments is wider than the width corresponding to the 70 kHz and under sections. There are more clock pulses in the 100 kHz group.
The trace F1 is a histogram of the frequency parameter. Histograms are graphical plots showing the number of occurrences within a small range of measured data values plotted against the data value. It is an estimate of the probability of a measured value occurring. The data values used in the histogram can be acquired sample amplitudes, timing values, or measured parameters. The histogram of the P1 frequency measurements from the previous section is shown as trace F1 (yellow) in Figure 7.
The histogram is in the bottom grid. The horizontal axis is the measured value, in this case the frequency. The vertical axis is the number of measured values within a small bin. The bin size is users selectable. In this example the horizontal axis is broken into 1000 bins. So, for the roughly 100 kHz range the bin size is about 100 Hz. The histogram shows two obvious peaks and two much smaller peaks. The largest peak is at 100 kHz representing the clock the peak count in the 100 kHz bin is 34 with smaller counts in the adjacent bins, all together the count will be 144. The one furthest to the left is at 68.6 kHz, this is the frequency of the cycles which include the gaps. Two smaller peaks are also gap frequencies with values of about 72-73 kHz, which are associated with the parameter measurements at either end of the clock signal.
The structure of the histogram and track, with the bulk of the readings about 100 kHz, provides the information necessary to chose and set the acceptance limits in the parameter measurement. Values below 100 kHz are associated with the gaps in clock bursts and should be excluded from the measurement of the clock frequency.
The track and histogram functions offer tremendous insight into your measurements showing where the measured values original and how the values are distributed by value.
This article has shown several techniques to help improve measurements made on your oscilloscope including maximizing display resolution, cursor placement, and measurement parameters. For those who are interested, the oscilloscope used was a Teledyne LeCroy LabMaster. Other oscilloscopes have similar capabilities, but you will have to consult your user’s manual to find the corresponding features. Track and histogram are most often associated with jitter measurement tools.
Arthur Pini is a technical support specialist and electrical engineer with over 50 years experience in electronics test and measurement.
New products & solutions, whitepaper downloads, reference designs, videos
Register, join the conference, and visit the booths for a chance to win great prizes.