What are the benefits of deep memory oscilloscopes?

Oscilloscope manufacturers usually regard memory depth as one of the most important indicators of an oscilloscope. It refers to the number of sampling points collected into the memory. Memory depth is usually indicated by channel. Some oscilloscopes use an interleaved memory architecture. For oscilloscopes with this structure, when all channels are turned on, the memory depth is a certain value; when only half of the channels are used, the memory depth will double.

There are two types of memory depth, one is standard and the other is optional. Although customers consider acquisition memory depth as a key indicator when purchasing an oscilloscope, a deep understanding of the pros and cons of memory can help customers make the best use of it. Deep memory brings value to users in three aspects:

●Longer waveform capture time

The most obvious advantage of deep memory is that it can capture waveforms for a longer period of time at a fixed sampling rate. For example, for waveforms with a long interval between cause and effect events, a very deep memory depth is required to capture both the cause and effect at once, allowing users to obtain all the information at once. Otherwise, the required information may never be obtained, or it may take a long time to find the relationship between the "effect" and the "cause".

The figure above shows that when the memory depth of each channel is set to 1Mpts, the oscilloscope can capture a 1ms waveform at a sampling rate of 10GSa/s.

The figure below shows that the same oscilloscope, but with the memory depth of each channel set to 100Mpts, can capture a 10ms waveform under the same sampling rate conditions.

Deep memory enables longer signal capture, which is required for many applications, especially those that focus on data post-processing. Deep memory provides engineers with greater flexibility. When not needed, it can be set to shallow memory, and when needed, it can be set to deep memory, just in case.

For high-end oscilloscopes, most oscilloscope manufacturers provide a usage mode that allows the memory to be divided into many segments and segmented memory. The user can specify how many segments of equal length the memory is divided into. When the first trigger event is met, the oscilloscope will store the sampling points, fill the first segment of the acquisition memory, and then start waiting for the next trigger event. When the condition is met, the oscilloscope will store the waveform in the next segment. This continues until all segments are filled. During this period, the user can also interrupt the oscilloscope to stop waiting for the trigger event and display the captured waveform. The segmented mode is particularly effective for burst signals with a very small duty cycle. Many serial buses, optical fibers, and communication signals belong to this category. By using segmented memory, the oscilloscope can maintain a fast sampling rate and capture waveforms up to several seconds, hours, or even days.

Deep memory oscilloscopes make segmented memory more powerful. First, users can get more segments, so that measurements and analysis have enough samples and sufficient statistical significance to help analyze the root cause of the problem. When the number of segments is fixed, the storage depth of each segment is increased, so that more signal activities around each trigger point can be viewed.

●Maintain faster sampling rate (assuming capture time length remains constant)

Another major benefit of deep memory is its ability to maintain fast sample rates at different timebase settings. Engineers tend to think of oscilloscope specifications as constant values, independent of the oscilloscope function settings, but this is not the case. Let's take a quick look at how memory depth affects the oscilloscope's sample rate and overall bandwidth when the user changes the oscilloscope's horizontal timebase. Most oscilloscope users don't think about memory depth, but in fact, it has a huge impact on ensuring that the oscilloscope maintains other key specifications, including sample rate and effective bandwidth.

For example, an engineer selects a 4GHz bandwidth oscilloscope with 10Mpts of memory per channel and a maximum sampling rate of 10GSa/s as standard. These specifications look good, and the engineer starts using the oscilloscope, and the engineer selects the faster 10ns/div actual setting. The oscilloscope uses only 1kpts of memory to sample at 10GSa/s. The entire 4GHz effective bandwidth is guaranteed as the engineer expected. If the engineer needs to see a longer waveform on the screen, then he tunes the horizontal time base to a slower setting; he selects 200μs/div. In order to sample 10 horizontal divisions at 10GSa/s, the oscilloscope needs to be equipped with 20Mpts of memory, so the oscilloscope automatically reduces the sampling rate to half of the original, or 5GSa/s. Now, the oscilloscope can collect 10Mpts of waveform data to fill 10 horizontal divisions. The engineer needs to see a longer time window, so he changes the oscilloscope time base to 1ms/div. Since the oscilloscope is only equipped with 10Mpts of acquisition memory, in order to capture 10ms waveform activity, the sampling rate is forced to be reduced to 1GSa/s. Therefore, it can be concluded that when the user changes the time base setting, insufficient memory depth will cause the oscilloscope's sampling rate to decrease, which also means that the oscilloscope's effective bandwidth becomes smaller.

By changing the horizontal timebase control at a fixed memory depth setting, the oscilloscope automatically reduces the sampling rate, and at the same time the overall effective bandwidth is also reduced. The front end of the oscilloscope still allows frequency components up to 4GHz to pass through. However, the oscilloscope with a reduced sampling rate is now prone to aliasing problems because its sampling speed is not fast enough to support the high-frequency components passing through the front end. Due to the change in the timebase setting, the sampling rate is reduced from 10GSa/s to 1GSa/s, and the effective bandwidth of the oscilloscope is also reduced from 4GHz to 400MHz. If the maximum memory depth of the oscilloscope is only 10Mpts, this phenomenon will occur, resulting in measurement errors.

What if the oscilloscope is equipped with 1Gpts memory instead of 10Mpts? At a slower timebase setting of 10ns/div, the oscilloscope will maintain a maximum sampling rate of 10GSa/s, so that no waveform distortion will occur in the rated bandwidth of the oscilloscope up to 4GHz. At slower sweep speeds of 1ms/div and even up to 5ms/div, the oscilloscope can still capture at a sampling rate of 10GSa/s, thereby ensuring 4GHz bandwidth at different timebase settings. You may or may not believe that when changing the horizontal timebase setting, the memory depth is indeed closely related to the effective bandwidth of the oscilloscope. Oscilloscopes with deeper memory depth can not only guarantee the sampling rate and effective bandwidth at fast timebase settings, but also maintain a high sampling rate at slow timebase settings to ensure effective bandwidth.

●Get better measurement and analysis results

The third value of deep memory is superior measurement quality. There is a trade-off in how much memory an oscilloscope uses for analysis. If the memory depth is set too low, it will be difficult for the oscilloscope to provide meaningful analysis. For example, when performing eye diagram recovery and jitter analysis, if the memory is set too low, the oscilloscope cannot perform PLL clock recovery because it cannot see enough edges. In addition, shallow memory will hinder statistical analysis, such as FFTs and histograms, because these functions do not have enough points to operate on. If the memory is set too high, the trade-off is a significant increase in analysis processing time, which in turn slows the oscilloscope's response.

What are the disadvantages of using deep memory? Deep memory settings reduce waveform capture rates because the oscilloscope must process information about the data before it can display the results on the screen. Deep memory settings slow down this task and increase the delay between the oscilloscope triggering and capturing waveforms, slowing down the user's operation and requiring the use of dual windows to simultaneously display signal details. Oscilloscopes from different manufacturers have very different waveform capture rates when using deep memory.

Conclusion

When choosing an oscilloscope, it is best to choose one with sufficient bandwidth to meet future needs. Of course, sometimes you may not want to keep deep memory turned on all the time, because that may cause the waveform capture rate to drop, but the value of a deep memory oscilloscope is that when needed, it can be set to deep memory to meet debugging and testing needs.

Deep memory provides the following advantages: at a fixed sampling rate, longer time windows can be captured, while faster sampling rates and effective bandwidth of the oscilloscope can be maintained. In addition, deeper memory can help users' oscilloscopes obtain better measurement and analysis results. Segmented memory allows the oscilloscope to better utilize the acquisition memory and is suitable for capturing burst signals with small duty cycles.

The main disadvantage of deep memory is that it slows down the waveform capture rate. When deep memory is enabled, the waveform capture rates of oscilloscopes from different manufacturers vary greatly. If the user's application is biased towards debugging and real-time fault analysis, deep memory is not important. If the user's application is biased towards post-processing analysis, deep memory can provide more benefits.