What is the importance of memory depth when selecting a digital storage oscilloscope?

Question

I'm preparing to purchase a digital storage oscilloscope, to complement my 100 MHz analog oscilloscope. I've been demoing units and narrowing down the choices.

The EEVBlog tutorial on DSOs advises that the three most important properties of DSOs are:

1. Analog bandwidth
2. Sample rate
3. Memory depth

I am looking at a couple of two-channel 100 MHz DSOs, the Tektronix TDS2012C and the Rigol DS2102. Some other features aside, these two appear to have identical performance in terms of analog bandwidth and real time sample rate.

However, the Tektronix oscilloscope's memory depth is a paltry 2,500 points compared to the Rigol's generous 14 million points (expandable to 56 million). This seems like a major reason to opt for the Rigol. (Not to mention it is ~$200 less.)

Is there a reason not to weigh memory depth heavily? Perhaps some reason that the Tektronix doesn't need as much?

I know I want to be able to capture multiple digital events/waveforms from microcontrollers and look for things that change over time, etc. Ample memory depth is needed for this. The fact that the Rigol and Tektronix differ by orders of magnitude concerns me. Tektronix has a superb reputation, whereas Rigol is new to me. The vast difference in memory depth seems to be the final word. Should it be?

Answer 1

NOTE: A lot of this is now outdated, as Tektronix have released some interesting scopes lately (2015).

This started out as a comment, but I'm expanding it to an answer:

Basically, Tektronix is not competitive in the digital oscilloscope market any more.

Your comparison is fundamentally flawed too. You are comparing Tektronix's bottom-of-the range scope to Rigol's middle-of-the-range model.

The actual Rigol scope that best matches that Tektronix scope is the DS1102E.

• Both have tiny, crappy, QVGA (320*240) screens
• Neither have intensity grading.
• The Rigol still has a lot more sample memory, 1 Mpoints vs. 2.5 Kpoints.

Note that the Rigol scope listed above is only US$400!

---

Really, if you're shopping for a US$500-US$3000 DSO, the only two players on the market even worth bothering to look at (at least at the current time) are Rigol and Agilent. They are the only two people on the market that offer intensity grading (Rigol call it "Ultravision" and Agilent call it "InfiniiVision").

This is a technique that actually measures the time the input waveform spends at each ADC value per X-axis time-step, and actually varies the intensity of the drawn scope trace to reflect the period of time the input spent at that voltage. This produces a display that actually somewhat resembles a traditional cathode-ray oscilloscope. It is absolutely a excellent feature, and I, at least personally wouldn't even consider a DSO that lacked it at this point.

Basically, Tektronix are just not producing DSOs worth looking at. They did have some good DSOs in the early 2000s: they produced a nice, primitive DSO, garnered a significant chunk of market share, and basically then sat there resting on their laurels and stopped innovating. This is supported by the teardowns I've seen of some of their late-model scopes, which were using rather ancient silicon for their processing. _{^{Note that this is changing, but only for Tektronix's higher end. They're doing some really cool stuff with their MSO devices (mixed-signal oscilloscopes). They basically combine a spectrum analyzer and a DSO, and for RF work, they look excellent. They're also $50K+}}.

Then, Agilent came along and basically completely wiped the floor with them in short order, with their much deeper memory scopes, and introduced intensity grading.

Now, Rigol have subsequently come out with a competitive mid-range scope line that makes them also worth considering, together with Agilent.

---

As far as I can tell, Tektronix's superb reputation should only really be applied to cathode-ray oscilloscopes (I have several, all Tektronix). They really didn't take the transition to digital, and its high innovation rate well at all.

If I were buying a scope now, I would look for:

Absolutely essential at ANY price-point:

• Greater then 100 KPts memory.
• 640*480 or larger screen. _{^{This is why I never bought one of the cheaper Rigol scopes}}

Absolutely essential a >~$1K price-point:

• Intensity grading.

Nice to have:

• High waveforms/second
  • This ranges from merely nice to totally essential, depending on what you are using the scope for. If you're glitch-hunting, you pretty much have to have high waveforms/second rates for decent coverage. The Tektronix scopes are an order of magnitude lower in waveforms/second then the Rigol and Agilent scopes (though the latest ($$$) Agilents are even better).
• Protocol decoding, at least as an option

Answer 2

It depends on what you are using it for.

In general memory depth allows you to capture a larger time slice of the signal at the same sampling rate

Some applications:

Large memory depth allows you to sample a low frequency signal at a much higher sample rate and still catch the entire signal, this allows you to analyze transients and small features in the signal that would be lost to Nyquist at lower sampling rates.

It also allows you to use a lower sampling rate and acquire for longer, which may seem pointless to you, but it's not uncommon in science (speaking from the POV of an atomic physicist) to use a scope as a general purpose ADC/data acquisition device, with 14m points you can store minutes of data at kHz sampling rate without resetting the acquisition system (and dealing with the inevitable gap in data that causes), compared to a few seconds with 12k points. This can be very useful when recording, say, the decay of an radioactive source with timelines on the order of a minute. With a larger memory depth you can also sample at larger frequency for the same time slice, giving you better resolution.

Whether or not this is important to you depends on what you will be using the scope for. Personally, in my field of work, signal frequency is relatively low and it's more useful to have a high memory depth at the cost of slightly lower sampling frequency.

I would take the fact that Tektronix's lack of memory depth is basically indicative of their low-end model compared to the upper-middle-of-the-line Rigol offering. Basically Tektronix is more expensive, you pay for construction quality, precision, calibration, warranty, and SNR of their ADCs. You also pay for the name.

Note: Their DPO series two-channel 100 MHz offering is only slightly pricier for 1M sample points; it was my impression that the TDS is an old design (I still use the TDS540, great scope) that has had a few facelifts and is delegated to the budget category.

Not to speculate on their designs, but it not unfeasible that the older design of the TDS uses discrete ADCs and standard microcontrollers for data acquisition, vs. a custom ASIC in the new models which could be easier to expand the memory for.

Answer 3

If you will be using the oscilloscope to analyze digital signals, specially serial data then memory depth is extremely important.

Other situation where memory depth is important is when you have two spaced event and you need to analyze both in detail. If your oscilloscope does not have enough memory depth you have to choose to see an overview of both events at a low resolution or analyze just one of them.

I have a Rigol DS1052D with 1M points memory, and the main reason I choose it over the other candidates was the memory depth.

Between the Tektronix and the Rigol you suggest I would not hesitate in buying the Rigol. Their instruments are well built with a very professional look and feel, not to mention the set of features you can not find on other brands on the same price range.

Rigol has a video explaining the importance of memory depth on an oscilloscope here.

Edit: the video is no longer available