You need to test, we're here to help.

You need to test, we're here to help.

04 May 2021

How to Use Memory Properly


In a recent post, we addressed setting sample rate for serial data acquisition, but let’s look again at how time per division (time/div), memory length and sample rate all interact, and what you can do to optimize your use of oscilloscope capture memory when setting up your timebase.
Figure 1: Sample rate as a function of time/div for three different memory lengths. Longer memory extends the range of time/div settings that support the highest sample rate.
Figure 1: Sample rate as a function of time/div for three different memory lengths.
Longer memory extends the range of time/div settings that support the highest sample rate.

The relationship between time/div, memory length and sample rate can be expressed using either of the following equations:

Capture Time = 10*Time/div = Memory/Sample Rate

Or

Sample Rate = Memory/Capture Time = Memory/10*Time/div 

The time/div setting determines the oscilloscope’s capture time window, which is always 10 times the time/div setting. A very long capture time immediately constrains your sample rate, unless you have significant available memory to support it.

Sample rate is important because it affects the digital bandwidth of the oscilloscope. An oscilloscope’s effective bandwidth is the lesser of its analog bandwidth or one half of the sampling rate.  The analog bandwidth is set by the front-end analog amplifier and is the -3 dB point of its frequency response. The digital bandwidth is an expression of the Nyquist criteria for digitized instruments. Time/div sets the capture time window, but memory depth and sampling rate define effective digital bandwidth.  

Figure 2: Broadband signals with fast edges require a higher sample rate to assure adequate time resolution on the edges.
Figure 2: Broadband signals with fast edges require a higher
sample rate to assure adequate time resolution on the edges.
You may have no choice but to use a high sample rate given the bandwidth of the signal you must acquire. It is important to realize that while a digital bandwidth of one half the sample rate will work for narrow-band waveforms like sine waves, broadband signals with fast edges may require sample rates that oversample the edge by four or more times. This is illustrated in Figure 2 where a 5 MHz square wave is acquired at both 250 MS/s and at 2.5 GS/s. The 250 MS/s waveform has a 4 ns sample period but only manages to acquire three samples per edge, and the pulse edge is poorly defined. The 2.5 GS/s waveform results in only 400 ps of time, but a much better definition of the edge shape.

Sample rate and memory depth also determine your ability to simultaneously analyze high-frequency and low-frequency phenomena, such as when making power integrity measurements of high-frequency noise on a low-speed signal, or analyzing high-speed data packet content on a serial data stream with low-speed inter-packet intervals.

In these situations, it is important to make smart use of your available memory in order to support a higher sample rate. The greater the oscilloscope’s memory depth (available number of sample points), the larger the number of time/div settings it can maintain at it’s maximum sampling rate, as shown in Figure 1.

Ask yourself these questions:

Do I need time, or do I need sample rate/bandwidth?

If you must set a specific capture time, use the oscilloscope’s Set Maximum Memory setting. In this mode, you should carefully monitor the sample rate to prevent accidental aliasing. Sequence mode sampling can help extend the period of acquisition without sacrificing too much sample rate. It may also be that you can make do with the lower sample rate, depending on your use of the acquisition, to buy yourself more time.

If you need to assure a desired digital bandwidth, use the Fixed Sample Rate setting rather than the Set Maximum Memory setting. Fixed Sample Rate mode will guarantee the selected sample rate, while varying the memory length as time/div is changed. It will also restrict the time/div setting to prevent changing the sample rate. 

Do I need all my channels?

If you need more memory to support a higher sample rate and the application does not require all the oscilloscope analog input channels, you can double—or quadruple—the available memory by decreasing the number of active channels. This will likewise increase the maximum available sampling rate.

In general, you only ever want to keep active those channels you are actually using. If you’re using only one input channel, set Auto Active Channels so the oscilloscope can allocate maximum memory to that channel, or follow the instructions in your oscilloscope Operator’s Manual to interleave channels appropriately.

And remember, you can halve the number of input channels you’re using, and gain memory, if it is feasible for you to probe differential rather than single-ended. 

Do I simply need more memory?

Generally, the greater the memory depth, the more flexibility the oscilloscope has in achieving the desired capture time and sampling rate to match the measurement application. Most Teledyne LeCroy oscilloscopes have memory upgrade options that can significantly increase the oscilloscope’s acquisition capabilities, without requiring the purchase of an entirely new oscilloscope.

Watch Steve Murphy give more tips on Setting Up Your Timebase and Using Memory Correctly.

See also "Take a Coffee Break and Learn...":

How to Use Measurement Statistics to Set Up Triggers

How to Layer Measurement Tools



No comments: