下載
下載

下載手冊、產品規格表、軟體等等:

下載類型
機型或關鍵字

Spectrum View: A New Approach to Frequency Domain Analysis on Oscilloscopes

Spectrum View: A New Approach to Frequency Domain Analysis on Oscilloscopes

Spectrum View is a new way of performing spectrum analysis on an oscilloscope. This application note shows and explains how Spectrum View operates and how it differs from traditional oscilloscope FFT functions.

下載檔案

Spectrum View: A New Approach to Frequency Domain Analysis on

Oscilloscopes

––

APPLICATION NOTE

FIGURE 1. Spectrum View enables simultaneous analog and spectrum views with independent controls in each domain.

Debugging embedded systems often involves looking for clues that are hard to discover just by looking at one domain at a time. The ability to look at time and frequency domains simultaneously can offer important insights. Mixed domain analysis is especially useful for answering questions such as:

What’s going on with my power rail voltage when I’m transmitting wireless data?

Where are the emissions coming from every time I access memory?

How long does it take for my PLL to stabilize after power-on?

Mixed domain analysis can help answer questions like these by providing views of time domain waveforms and frequency domain spectra in a synchronized view. Up until recently, the Tektronix MDO4000C mixed domain oscilloscope has been the only oscilloscope to offer synchronized time and frequency domain analysis with independent control over waveform and spectrum views.

To address this need, the 4, 5 and 6 Series MSO mixed signal oscilloscopes offer an analysis tool called Spectrum View. It is an option in the 4 Series MSO and a standard feature in the 5 and 6 Series MSOs. It delivers several key capabilities:

Enables the use of familiar spectrum analysis controls (Center Frequency, Span and RBW)

Allows optimization of both time domain and frequency domain displays independently

Enables a signal to be viewed in both a waveform view and a spectrum view without splitting the signal into different inputs

Enables accurate correlation of time domain events and frequency domain measurements (and vice versa)

Significantly improves achievable frequency resolution in the frequency domain

Improves the update rate of the spectrum display

 

 

spectrum view domain analysis

 

FIGURE 2. Digital down converters implemented on a custom ASIC enable simultaneous waveform and spectrum views with independent controls in the Tektronix 4, 5 and 6 Series MSOs.

A new architecture

Spectrum View uses patented hardware built into the instruments. To understand how it works, it is important to note that digital oscilloscopes generally run their analog-to- digital converters (ADC) at the maximum sample rate. The stream of ADC samples is then sent to a decimator that keeps every Nth sample. At the fastest sweep speeds, all samples are kept. At slower sweep speeds, it’s assumed the user is looking at slower signals and a fraction of the ADC samples are kept. In short, the purpose of the decimator is to keep the record length as small as possible while still providing adequate sample rate to view signals of interest in the time domain.

In the 4, 5 and 6 Series MSOs, behind each FlexChannel input is a 12-bit ADC inside a custom ASIC. As shown in Figure 2, each ADC sends high-speed digitized data down two paths. One path leads to hardware decimators which determine the rate at which time domain samples are stored. The second path leads to digital down-converters also implemented in hardware. This approach enables independent control of the time domain and frequency domain acquisitions, allowing optimization of both waveform and spectrum views of a given signal. It also makes much more efficient use of the long but finite record length available in these instruments.

FIGURE 3. With the time domain optimized using conventional FFTs, frequency domain detail is lacking on this spread-spectrum clock signal.

Spectrum View with independent controls vs. conventional FFT

Although spectrum analyzers are designed specifically for viewing signals in the frequency domain, they are not always readily available. Scopes, on the other hand, are almost always nearby in the lab so engineers tend to rely on scopes as much as possible. For this reason, oscilloscopes have included math-based FFTs (fast Fourier transforms) for decades. However, FFTs are notoriously difficult to use for two reasons.

First, for frequency domain analysis, spectrum analyzer controls like center frequency, span and resolution bandwidth (RBW) make it easy to define the spectrum of interest. In most cases, however, oscilloscope FFTs only support traditional controls such as sample rate, record length and time/div, making it difficult to get to the desired view.

Second, even if the scope offers spectrum analyzer style controls, the FFT is driven by the same acquisition system as that used for the analog time domain view. Changing the center frequency, span, or resolution bandwidth will change the scope’s horizontal scale, sample rate and record length in unanticipated and undesired ways. Once the desired frequency domain view is achieved, the time domain view of other signals is no longer usable. When adjustments are made to horizontal scale, sample rate, or record length to again achieve the desired time domain view, the FFT view is no longer usable. For example, the next two screenshots taken from an MDO3000 illustrate the time domain and FFT views of a spread spectrum clock that moves from 97 MHz to 100 MHz. In Figure 3, the time domain view enables easy visualization of the clock, but the FFT doesn’t have adequate resolution to be useful. In Figure 4, the FFT shows the spread spectrum nature of the clock, but the time domain view is no longer helpful.

FIGURE 4. With the FFT view optimized, the time domain view of the clock signal is now not useful.

Spectrum View provides the ability to adjust the frequency domain using familiar center frequency, span and RBW controls.

And because these controls do not interact with the time domain scaling, it is possible to optimize both views independently as shown in Figure 5.

FIGURE 5. Looking at the same spread-spectrum clock signal as Figures 3 and 4, Spectrum View enables optimized views of both time and frequency domains at the same time.

Spectrum Time

An on-screen indicator called Spectrum Time is used to indicate where in time the spectrum shown in the Spectrum View window came from. The width of the Spectrum Time indicator is simply the Window Factor divided by the Resolution Bandwidth. (See the appendix for more on window factors.) You can move Spectrum Time throughout the acquisition to see how the frequency domain view changes over time. You can even do this on a stopped acquisition.

In Figures 6 through 9, we’ve captured the startup sequence of the spread spectrum clock discussed earlier. The Spectrum Time indicator appears very narrow in the screenshots and is highlighted by a red box to assist the reader. In this case, Spectrum Time is 1.9 (Window Factor) / 10,000 (RBW) = 190µs wide.

FIGURE 6. Spectrum Time (highlighted with a red box) is placed early in the acquisition, before the trigger event. As expected, there aren’t any strong signals in the frequency domain as the clock has not turned on yet.

FIGURE 7. Spectrum Time placed approximately 20 ms after the clock turned on. Notice that the clock is not spread spectrum yet, it’s merely sitting at 94 MHz.

FIGURE 8. Spectrum Time placed approximately 300 ms after the clock turned on. Notice that the clock is now exhibiting spread spectrum behavior, but it’s using more of the spectrum than intended. The cursors show the expected spread spectrum width.

FIGURE 9. Spectrum Time placed approximately 324 ms after the clock turned on. Notice that the clock is now exhibiting spread spectrum behavior and is within the intended spectrum operating range.

RF vs. Time Waveforms

The underlying I&Q data that’s used to create the spectrums shown by Spectrum View can also be used to calculate RF vs. Time waveforms that show how various characteristics of the RF waveform vary over the entire acquisition, not just where Spectrum Time is positioned. Three types of waveforms are available:

Magnitude – The instantaneous magnitude of the spectrum vs. time

Frequency – The instantaneous frequency of the spectrum relative to the center frequency vs. time

Phase – The instantaneous phase of the spectrum relative to the center frequency vs. time

Each of these traces can be turned on and off independently, and all three can be displayed simultaneously.

These waveforms are shown in Figures 10-12 below and reveal additional information about the signal of interest. There are four time-domain waveforms shown in the Waveform View

in each of the images. The top one is the analog view of the signal. Next is the RF Magnitude vs. Time waveform, then the RF Frequency vs. Time waveform and finally the RF Phase vs. Time waveform.

FIGURE 10. It’s very easy to see what’s going on with this spread spectrum clock by looking at the Magnitude and Frequency vs. Time waveforms. The Magnitude vs. Time trace shows the signal turning on at the trigger point at a very low level and the Frequency vs. Time trace shows the signal staying at a single frequency for the first ~300ms. At that point, we see the amplitude of the signal increase significantly and the frequency begin changing.

FIGURE 11. We’ve now zoomed in on the period of interest (roughly 300-320 ms after the trigger event). Notice that we can clearly see the amplitude fluctuating and the frequency changing over a broader range than it should be.

FIGURE 12. We’ve now zoomed in even further and can easily view the triangular frequency modulation in use and can confirm through the automatic measurement in the results bar that we are getting the correct modulation rate of 39.07 kHz.