Modern digital scopes: real-time sampling or equivalent-time sampling

[pascal_sweden]

I was wondering if all modern digital scopes today offer real-time sampling?

For most digital scopes the sampling rate is divided by the number of active channels.
I guess this is because they perform real-time sampling?

What do most digital scopes offer today? Both options?

Are their other differences in the way digital scopes sample the incoming signal, besides actual sample rate, real-time or equivalent-time sampling?

Like to have a better understanding about the different possible implementations, and how they differ between Tektronix, Agilent, Rigol, etc.

Rigol DS1000/2000: Do Rigol scopes offer both options?

Tektronix TLS216: this scope had massive amount of 16 channels, had 2 GS/s per channel, irrespective of how many active channels you were using. How did they do that?
Did that scope use real-time sampling or equivalent time-sampling?

Philips PM3340: This scope offered 2GS/s back at that time. But it was equivalent-time sampling. Does anybody know if it also offered real-time sampling, and what the actual bandwidth was?

[EEVblog]

I've heard the new Rigol 1000Z does not offer equivalent time sampling any more.

[pascal_sweden]

Is it correct that you were going to review the DS1054Z?
Would be cool as well to have a review of the MSO2072A, and specifically the Logic Analyzer features.

[EEVblog]

Is it correct that you were going to review the DS1054Z?

Yes, one is on the courier truck now.
It's a loaner, won't be getting my own for some time, not in stock in Oz yet.

[edavid]

What do most digital scopes offer today? Both options?

Yes.

Rigol DS1000/2000: Do Rigol scopes offer both options?

Dave is correct, the DS1000/DS2000 are real-time sampling only.

Tektronix TLS216: this scope had massive amount of 16 channels, had 2 GS/s per channel, irrespective of how many active channels you were using. How did they do that?
Did that scope use real-time sampling or equivalent time-sampling?

Real-time sampling.  The tradeoff is that the record length is only 2K points.

Philips PM3340: This scope offered 2GS/s back at that time. But it was equivalent-time sampling. Does anybody know if it also offered real-time sampling, and what the actual bandwidth was?

Bandwidth is 2GHz (-3dB).  I think it does have real-time sampling at a sufficiently low sample rate, but I can't figure out what that is from the manual.  I suspect it's quite slow (KSPS).

http://exodus.poly.edu/~kurt/manuals/manuals/Other/PHILIPS%20PM%203340%20Operating.pdf

[nixfu]

As the processing power gets greater and they have much larger sample sizes, scopes will all eventually be real-time where they can capture an entire signal at the full rated bandwidth in a single pass and no longer have to sample multiple cycles "equiv-time-sampling" to get their rated resolution.   

From what I remember, the 1000z have a sample rate of 1Gs/sec which is  a10x nyquist ratio, which is pretty darn good real time capabilities.   

Dave, I know you have discussed this before but you should briefly touch on this in the video when you get this 1000z loaner on the bench.   Maybe talk about directly the individual capabilities of the 1000z relate to its specs compared to the general discussions on scope bandwidth, sample depth etc you have done in previous videos. 

[DanielS]

As the processing power gets greater and they have much larger sample sizes, scopes will all eventually be real-time where they can capture an entire signal at the full rated bandwidth in a single pass and no longer have to sample multiple cycles "equiv-time-sampling" to get their rated resolution.

Processing power is not really an issue since most of the "processing" is simply dumping samples in RAM. Processing power would affect the relationship between points per trigger and waveform update rate though.

The biggest reason for ETS is the cost and availability of high-speed ADCs: you can put together something that can do a couple of GSPS ETS for little more than a hundred bucks but if you want 4GSPS from a single ADC, the ADC alone will cost you a lot more than that.

[Wuerstchenhund]

I was wondering if all modern digital scopes today offer real-time sampling?

Yes, most scopes today are Real-Time Sampling (RTS) scopes. Some also support Equivalent Time Sampling (ETS), but not all of them.

In general, ETS-only scopes are a thing of the past unless you need more than 65GHz bandwidth. Above that there are some older scope designs which go to 100GHz and which are still sold, but these will probably be replaced by RTS scopes soon.

For most digital scopes the sampling rate is divided by the number of active channels.
I guess this is because they perform real-time sampling?

No, not really. Yes, it's RTS, but the reason why on some scopes the max sampling rate is dependent on the number of active channels is because they use interleaving (combining a channel's ADC with the ADC of an inactive channel) to reach higher sampling rates when only one or two channels are active.

What do most digital scopes offer today? Both options?

As I said above, ETS-only scopes are dead outside extremely high bandwidths, and this has been the case for more than 15 years.

Most of the better scopes also offer an ETS mode as well, but not all do. Especially cheaper low end scopes often lack that feature.

Are their other differences in the way digital scopes sample the incoming signal, besides actual sample rate, real-time or equivalent-time sampling?

Like to have a better understanding about the different possible implementations, and how they differ between Tektronix, Agilent, Rigol, etc.

Not really. The basic principle behind RTS scopes is the same. Some use technologies like DBI (Digital Bandwidth Interleaving) where signal components are downmixed, but this is only used in certain highend scopes (i.e. LeCroy's 100GHz RTS scope demonstrator uses DBI).

However, not all ETS implementations are the same. For example LeCroy uses something called RIS which works different than ETS modes of of other manufacturers. This pdf explains RIS and other ETS modes in more detail:
http://cdn.teledynelecroy.com/files/whitepapers/wp_ris_102203.pdf

Tektronix TLS216: this scope had massive amount of 16 channels, had 2 GS/s per channel, irrespective of how many active channels you were using. How did they do that?

Probably by using a 2GSa/s ADC per channel. Not all scopes use ADC interleaving, as this comes with its own issues when designing a scope.

Philips PM3340: This scope offered 2GS/s back at that time. But it was equivalent-time sampling. Does anybody know if it also offered real-time sampling, and what the actual bandwidth was?

The PM3340 was a 2GHz scope with 250MSa/s real-time sampling rate and 14bit 10bit vertical resolution which also offered 2GSa/s in ETS mode. It was a good scope at it's time (1989; I had a PM3343 PM3320A back then which was the 200MHz version of the PM3340) but by today's standards it's a boat anchor. It's also difficult to fix with lots of unobtainium parts. The high vertical resolution made (and still makes) especially the PM3343 PM3320A sought after for audio work, though.

Edit: The scope was a PM3320A not PM3343, and had 10bit vertical resolution not 14bit.

[alex.forencich]

And when you get into super high bandwidths, then you start to see really interesting solutions like optical-domain time stretching.  In this case, the signal is modulated onto a chirped laser pulse and then sent through a long piece of dispersion compensation fiber.  As it passes through the fiber, the ends of the pulse travel at different speeds (since they are different wavelengths) and the modulated pulse is stretched in time.  At the end, it is detected and sampled by much lower rate ADCs.  It is possible to get into the terasample per second range with this technique. 

[tggzzz]

Philips PM3340: This scope offered 2GS/s back at that time. But it was equivalent-time sampling. Does anybody know if it also offered real-time sampling, and what the actual bandwidth was?

A quarter of a century ago the HP54100 series offered upto 50GHz bandwidth with, IIRC, 40MS/s real-time.

In such discussions it is always worth defining exactly why a fast RTS is necessary. Sometimes it is, but in many important applications ETS is sufficient. Signal integrity springs to mind, particularly when using eye diagrams which are inherently ETS!

[Wuerstchenhund]

In such discussions it is always worth defining exactly why a fast RTS is necessary. Sometimes it is, but in many important applications ETS is sufficient. Signal integrity springs to mind, particularly when using eye diagrams which are inherently ETS!

Eye Diagrams are certainly *not* "inherently ETS"! Some Eye Diagrams *can* be done with an ETS only scope but there's a risk that important signal components are missed.

HP has written a nice document about RTS vs ETS for Eye Diagrams:
http://www.hpl.hp.com/hpjournal/96dec/dec96a1a.pdf
And this was written at a time when RTS scopes didn't have the high sample rates as today's scopes.

Decades ago ETS was a crutch to overcome the insufficient real-time sampling rates of old days ADCs. But it's not 1989 any more!

RTS at sufficient sample rates is better in any ways over ETS. The question should not be if/why RTS is necessary but why bother with all the drawbacks of ETS at all when today scopes have sufficient real time sampling rates up to 160GS/s.

[pascal_sweden]

I wanted to point out a company based in Sweden, SP Devices, who have patented technology to increase the sample rate by using several ADCs in parallel.

They provide other technology as well:
http://spdevices.com/index.php/technology

Maybe one day we will see a Rigol scope with SP Devices technology in side

It is my understanding that Dave has good connections with Rigol Engineering. Maybe he can pass on the company details of SP Devices to Rigol. Of course if this would lead to any business, I would appreciate some credits =)

[nctnico]

I wanted to point out a company based in Sweden, SP Devices, who have patented technology to increase the sample rate by using several ADCs in parallel.

That 'technology' is common knowledge. I doubt their patents are really new. It could be a nice off-the-shelf solution for a manufacturer of low volume specialistic data acquisition hardware.

[Lukas]

I guess that there's still a small market for high-bandwidth low-samplingrate ETS scopes. When you're doing analog work, you often don't need single-shot capture.
Yes, technology is advancing but a 1GHz ETS scope still would be way cheaper than a 1GHz real time scope.

[David Hess]

Are their other differences in the way digital scopes sample the incoming signal, besides actual sample rate, real-time or equivalent-time sampling?

Like to have a better understanding about the different possible implementations, and how they differ between Tektronix, Agilent, Rigol, etc.

Not really. The basic principle behind RTS scopes is the same. Some use technologies like DBI (Digital Bandwidth Interleaving) where signal components are downmixed, but this is only used in certain highend scopes (i.e. LeCroy's 100GHz RTS scope demonstrator uses DBI).

However, not all ETS implementations are the same. For example LeCroy uses something called RIS which works different than ETS modes of of other manufacturers. This pdf explains RIS and other ETS modes in more detail:
http://cdn.teledynelecroy.com/files/whitepapers/wp_ris_102203.pdf

Most equivalent time sampling implementations work the way described in this LeCroy white paper.  Modern implementation tend to replace the analog time to digital converter with some variation of a transition midpoint timing time to digital converter.

The oldest DSO that I am aware of which supports random equivalent time sampling which LeCroy calls random interleaved sampling is the Tektronix 7D20 (70 MHz) with a 40 MS/s real time sampling rate and 2 GS/s equivalent time sampling rate released in 1983.  It is the ancestor of the 2440 and TDS600 series DSOs.  The contemporary HP 19860A used sequential equivalent time sampling.

The 7D20 was followed by the the Tektronix 2220 (60 MHz) and 2230 (100 MHz) with 20 MS/s real time sampling rates and 2 GS/s equivalent time sampling rates which were both released in 1986 by which time HP already had the 54100A (1 GHz) with 40 MS/s a real time sampling rate and a 100 GS/s equivalent time sampling rate released in 1985.

[Wuerstchenhund]

I guess that there's still a small market for high-bandwidth low-samplingrate ETS scopes.

Based on the very limited availability of ETS scopes I guess this market is probably tiny. If it wasn't then I'm pretty sure that the major brand like Agilent/Keysight, Tek and LeCroy wouldn't all just offer a single ETS scope which has been on the market between ~10 to ~15 years and only serves for extremely high bandwidths above their RTS scopes.

When you're doing analog work, you often don't need single-shot capture.
Yes, technology is advancing but a 1GHz ETS scope still would be way cheaper than a 1GHz real time scope.

Well, if you need that 1GHz scope for a very limited set of applications then maybe, but then you'd probably also need a RTS scope for everything else so away goes a large part of your price advantage.

Of course aside from the fact that you would already pay a premium just for the fact that ETS only scopes are very low volume products while RTS scopes are produced en masse.

[pascal_sweden]

That 'technology' is common knowledge. I doubt their patents are really new. It could be a nice off-the-shelf solution for a manufacturer of low volume specialistic data acquisition hardware.

The interleaving technique as such is indeed common knowledge. The challenge with interleaving however is to correct for the manufacturing variations of the characteristics of the individual ADC, in order to obtain the optimal resolution.

The patent of SP Devices relates to the estimation of the mismatch error and the reconstruction of the signal with all mismatch errors suppressed. The proprietary technology provides a background estimate of the gain, offset and time-skew errors of the ADCs without the need for any special calibration signal or post-production trimming. The signal is recreated with minimal latency.

[tggzzz]

In such discussions it is always worth defining exactly why a fast RTS is necessary. Sometimes it is, but in many important applications ETS is sufficient. Signal integrity springs to mind, particularly when using eye diagrams which are inherently ETS!

Eye Diagrams are certainly *not* "inherently ETS"!

Grrr. Annoyingly, you are correct

Some Eye Diagrams *can* be done with an ETS only scope but there's a risk that important signal components are missed.

HP has written a nice document about RTS vs ETS for Eye Diagrams:
http://www.hpl.hp.com/hpjournal/96dec/dec96a1a.pdf
And this was written at a time when RTS scopes didn't have the high sample rates as today's scopes.

I'm understand why they say that, but based on a recent experience, my preference is still for displaying isolated dots without linear (or other) interpolation. That way I can see what's being measured. I haven't tried a scope with the eyeline technique, but I'm sure that if it is implemented well it will be valuable in some limited circumstances. I wonder if other manufacturers implement it?

Decades ago ETS was a crutch to overcome the insufficient real-time sampling rates of old days ADCs. But it's not 1989 any more!

Just so, but see below...

RTS at sufficient sample rates is better in any ways over ETS. The question should not be if/why RTS is necessary but why bother with all the drawbacks of ETS at all when today scopes have sufficient real time sampling rates up to 160GS/s.

The issue is, always has been, and always will be, cost and size. Of course the decision points change with time

[David Hess]

RTS at sufficient sample rates is better in any ways over ETS. The question should not be if/why RTS is necessary but why bother with all the drawbacks of ETS at all when today scopes have sufficient real time sampling rates up to 160GS/s.

I agree but not all DSOs made today have sufficient real time sampling rates to support their input bandwidth without aliasing even with input signals which are completely below their Nyquist frequency.

I wanted to point out a company based in Sweden, SP Devices, who have patented technology to increase the sample rate by using several ADCs in parallel.

There is nothing new here.  Interleaved sampling is at least 3 decades old and linearity correction for interleaved digitizers is almost as old.

I guess that there's still a small market for high-bandwidth low-samplingrate ETS scopes. When you're doing analog work, you often don't need single-shot capture.
Yes, technology is advancing but a 1GHz ETS scope still would be way cheaper than a 1GHz real time scope.

I would say that most of the time it is not required for analog design work including digital signal integrity.  I am actually disappointed in the current crop of 100 MHz 250 MS/s DSOs because of their lack of ETS.

[Wuerstchenhund]

However, not all ETS implementations are the same. For example LeCroy uses something called RIS which works different than ETS modes of of other manufacturers. This pdf explains RIS and other ETS modes in more detail:
http://cdn.teledynelecroy.com/files/whitepapers/wp_ris_102203.pdf

The oldest DSO that I am aware of which supports random equivalent time sampling which LeCroy calls random interleaved sampling is the Tektronix 7D20

Tek's Random Equivalent Time Sampling is *NOT* the same as Random Interleaved Sampling. Read the paper again. There are some important differences!

(70 MHz) with a 40 MS/s real time sampling rate and 2 GS/s equivalent time sampling rate released in 1983.  It is the ancestor of the 2440 and TDS600 series DSOs.  The contemporary HP 19860A used sequential equivalent time sampling.

The 7D20 was followed by the the Tektronix 2220 (60 MHz) and 2230 (100 MHz) with 20 MS/s real time sampling rates and 2 GS/s equivalent time sampling rates which were both released in 1986 by which time HP already had the 54100A (1 GHz) with 40 MS/s a real time sampling rate and a 100 GS/s equivalent time sampling rate released in 1985.

That is all well and nice but that was 30 years ago and bears not much relevance for modern day scopes.

[EEVblog]

I agree but not all DSOs made today have sufficient real time sampling rates to support their input bandwidth without aliasing even with input signals which are completely below their Nyquist frequency.

I'm not really aware of any mainstream one that isn't?
Tek set the benchmark 15+ years ago with the TDS200 series RTS scopes.

I would say that most of the time it is not required for analog design work including digital signal integrity.  I am actually disappointed in the current crop of 100 MHz 250 MS/s DSOs because of their lack of ETS.

But 250MSPS is just enough to reconstruct the waveform if you use Sin X/x interpolation. In theory I believe that in most cases you can't get a huge amount more real information out of a 100MHz bandwidth limited signal once you pass the 2.4x mark, i.e. 240MSPS for 100MHz bandwidth. But that does depend on the filter type used etc.
Either way, it think it's pretty marginal.
That is likely why Rigol have simply dropped ETS on the 1000Z.

[tggzzz]

Well, if you need that 1GHz scope for a very limited set of applications then maybe, but then you'd probably also need a RTS scope for everything else so away goes a large part of your price advantage.

IMNSHO digital problems should be captured and debugged in the digital domain, and analogue problems in the analogue domain. The interesting case comes where analogue signals are being interpreted as digital signals. The classic example of that is often referred to as the "signal integrity" of digital signals.

For such cases, it is possible and sufficient to:

• use an analogue scope to look at the eye diagram, and ensure the signal integrity is sufficient
• something will have "interpreted" that analogue waveform as a digital signal; use a logic analyser (in whatever form) to look at that the result of that "interpretation"
• debug that digitised signal

Yes, of course there are cases where that isn't possible, but I contend it would be practical for the majority of cases in which people think they need a fast real-time scope with extremely long capture buffers.

[David Hess]

However, not all ETS implementations are the same. For example LeCroy uses something called RIS which works different than ETS modes of of other manufacturers. This pdf explains RIS and other ETS modes in more detail:
http://cdn.teledynelecroy.com/files/whitepapers/wp_ris_102203.pdf

The oldest DSO that I am aware of which supports random equivalent time sampling which LeCroy calls random interleaved sampling is the Tektronix 7D20

Tek's Random Equivalent Time Sampling is *NOT* the same as Random Interleaved Sampling. Read the paper again. There are some important differences!

Great!  Maybe you can point them out because I read the paper in detail and the only difference they identify is filtering after interleaving which is not unique to LeCroy and they do not even include that difference in their summary.

[Wuerstchenhund]

Some Eye Diagrams *can* be done with an ETS only scope but there's a risk that important signal components are missed.

HP has written a nice document about RTS vs ETS for Eye Diagrams:
http://www.hpl.hp.com/hpjournal/96dec/dec96a1a.pdf
And this was written at a time when RTS scopes didn't have the high sample rates as today's scopes.

I'm understand why they say that, but based on a recent experience, my preference is still for displaying isolated dots without linear (or other) interpolation. That way I can see what's being measured. I haven't tried a scope with the eyeline technique, but I'm sure that if it is implemented well it will be valuable in some limited circumstances. I wonder if other manufacturers implement it?

Ahem, this document is from 1996! It's almost 20 years old! The technology that HP has offered back then is no longer relevant!

And the question if you see an interpolated signal on the screen or not has absolutely nothing to do with ETS or RTS. It simply depends on your scope, i.e. can you switch off interpolation or is it forced enabled. For example, on LeCroy scopes interpolation is optional that has to be enabled by the user, as their philosophy is that a scope shall always display a signal unaltered by default. Other vendors have different mindsets.

But again, this has nothing to do with ETS vs RTS.

RTS at sufficient sample rates is better in any ways over ETS. The question should not be if/why RTS is necessary but why bother with all the drawbacks of ETS at all when today scopes have sufficient real time sampling rates up to 160GS/s.

The issue is, always has been, and always will be, cost and size. Of course the decision points change with time

Cost, well, a 100GHz ETS scope is slightly cheaper than a 65GHz RTS scope, but we're talking about prices in the region of $50k to $100k and more. Chances are that if you're working on such high complexity projects then you can afford a proper scope as well.

For everything in the lower bandwidth ranges (i.e. below 5GHz) ETS isn't a sensible options. There are no ETS only scopes available in that bandwidth range, so you'd either have to settle for one of the few USB ETS scopes or hunt for a museum piece on ebay, which will be old, big, loud, noisy, power hungry, and chances are good spare parts are no longer available.

I appreciate that we have many older EE's in this forum who have grown up with analog scopes and who never got really into DSOs, but really, it's not 1986 any more, and scope technology and prices have changed a lot since then. And there's a reason why scopes these days up to 65Ghz (or 33GHz if you're Tek) are RTS scopes.

[Wuerstchenhund]

Tek's Random Equivalent Time Sampling is *NOT* the same as Random Interleaved Sampling. Read the paper again. There are some important differences!

Great!  Maybe you can point them out 

From the document:

"...How RIS Implementations Vary Among Scopes
RIS is a word coined by LeCroy. Usually, on other vendor's scopes, this mode goes by names like Equivalent Time (ET), Repetitive Equivalent Time (RET) or Sequential Sampling. Some vendors have stopped offering this mode in their high-end scopes.

The main implementation difference between LeCroy and other vendors is the filtering of the final interleaved RIS trace. In other words, some vendors simply acquire the waveform segments and interleave them together. There are pros and cons to both methods: On one hand, the noise that is filtered out of the RIS waveform can only come from one source - the non-repetitive nature of the waveform or trigger and serves to degrade the rendition of the true analog waveform. On the other hand, the removal of this noise hides any non-repetitiveness that can lead to confusion unless the operation is understood.

LeCroy believes that the removal of the noise and resulting higher performance and better signal fidelity is more important than the preservation of the non-repetitive variations. Confusion is avoided by understanding the scope operation. LeCroy believes this because the resulting RIS waveform is confusing even if this filtering is not applied if the waveform is not repetitive. The only solution to confusion in RIS mode is an understanding of how the mode works regardless of the scope vendor.

because I read the paper in detail and the only difference they identify is filtering after interleaving which is not unique to LeCroy

Who else does it exactly the same then? Tek not, according to the documents I've seen and what I remember from the Tek DSOs I've used. So if you have some reference please share it.

and they do not even include that difference in their summary.

No, they don't, because unlike say Agilent's or Tek's our-vs-theirs comparisons where the main aim is to make the own product look good, the aim of this document is to describe what RIS does, what its limits are and how the implementation varioes between vendors.