Common misunderstandings about digital oscilloscopes

[pascal_sweden]

I found a nice tutorial from Rohde & Schwartz regarding common misunderstandings about digital oscilloscopes. Believe it is very useful for beginners on this forum

1) Your digital oscilloscope's bandwidth

Today's digital scopes use digital processing and employ very sharp, high-order frequency responses that trade minimum sampling rates for maximum bandwidth. The result is high overshoot and ringing when measuring typical digital signals.

http://rohde-schwarz-scopes.com/scope_lie_01.php

2) Your digital oscilloscope's noise specification

Most high sample rate scopes will interleave two, four, eight (or maybe more) digitizers to obtain a sample rate that high. The digitizers need to be accurately interleaved to prevent interleaving spurs in the frequency domain, which manifests itself as noise in the time domain.

http://rohde-schwarz-scopes.com/scope_lie_02.php

3) Your digital oscilloscope's waveform update rate:

Can your oscilloscope really achieve its specified high waveform update rate (acquisition rate) using real signals?

Digital scope manufacturers specify high update rates in excess of 1 million waveforms per second which can be misleading. These update rates are usually stated for certain circumstances or settings, but typically do not include measurements or mask testing which may be turned on when acquiring waveforms.

http://rohde-schwarz-scopes.com/scope_lie_03.php

4) Your digital oscilloscope's analog trigger

Are you really triggering on the waveform you see on the screen?

Most scopes (analog and digital) utilize a separate trigger circuit and a different circuit to acquire the waveform. Since the trigger circuit and acquisition circuit have different bandwidths, different sensitivities and different characteristics, this can cause trigger jitter which appears as jitter on the signal, but is actually jitter coming from the trigger circuit.

http://rohde-schwarz-scopes.com/scope_lie_04.php

[Wuerstchenhund]

I found a nice tutorial from Rohde & Schwartz regarding common misunderstandings about digital oscilloscopes. Believe it is very useful for beginners on this forum

Aside from the fact that there's a beginner's section in this forum where stuff like this should be placed, I hope you're aware that you're citing marketing arguments.

1) Your digital oscilloscope's bandwidth

Today's digital scopes use digital processing and employ very sharp, high-order frequency responses that trade minimum sampling rates for maximum bandwidth. The result is high overshoot and ringing when measuring typical digital signals.

That might be true for some entry-level scopes but most mid-range and high-end scopes do have a slow roll-off beyond the 3dB point, which obviously is news to R&S, as they seem to believe that their scope is exceptional in this area (which it isn't).

I'm not sure what a beginner is supposed to do with that information, or why R&S labels that as "scope lie".

2) Your digital oscilloscope's noise specification

Most high sample rate scopes will interleave two, four, eight (or maybe more) digitizers to obtain a sample rate that high. The digitizers need to be accurately interleaved to prevent interleaving spurs in the frequency domain, which manifests itself as noise in the time domain.

Again, this might be a problem for cheap/low-end scopes, but interleaving spurs aren't a problem for most mid-range and high-end scopes. Aside from the fact that in the price segment of the R&S RTO other manufacturers use much faster single-block ADC hybrids, which again seems to be news to R&S.

Again, I'm not sure where the "lie" is, or what relevance this has for a beginner.

3) Your digital oscilloscope's waveform update rate:

Can your oscilloscope really achieve its specified high waveform update rate (acquisition rate) using real signals?

Digital scope manufacturers specify high update rates in excess of 1 million waveforms per second which can be misleading. These update rates are usually stated for certain circumstances or settings, but typically do not include measurements or mask testing which may be turned on when acquiring waveforms.

Mask testing isn't a problem for any decent somewhat modern DSO except maybe some bottom-of-the-barrel scopes, and these are hardly the class of scopes that compete with an R&S RTO (a $15k+ scope). Math operations are different, but even the RTO slows down when using advanced math.

As before, I can't see where the "scope lie" should be, and as with the topics before it's generally not of much relevance for a beginner who's unlikely to use advanced maths anyways.

4) Your digital oscilloscope's analog trigger

Are you really triggering on the waveform you see on the screen?

Most scopes (analog and digital) utilize a separate trigger circuit and a different circuit to acquire the waveform. Since the trigger circuit and acquisition circuit have different bandwidths, different sensitivities and different characteristics, this can cause trigger jitter which appears as jitter on the signal, but is actually jitter coming from the trigger circuit.

Trigger jitter isn't a problem for most mid-range and high-end scopes, which also can trigger on any point in the signal. R&S seems to believe that other scopes work like an old Hameg CRO, which they don't.

Seriously, I'm not sure why you're posting such marketing drivel which R&S uses to promote their $15k+ high end scope, something that isn't of relevance for a beginner who'll very likely start with a slow and limited bottom-of-the-barrel scope like a Rigol DS1054z, or in some cases even with an old analog scope.

I'd also caution you to not blindly buy into any marketing BS, as you seem to do. Most of such stuff where one manufacturer compares his product with their competitors isn't worth the paper/the bytes its written on/with.

[pascal_sweden]

Thanks for pointing out that marketing buzz should be taken with a pinch of salt .. even if it comes from a big company

I thought that a reputable company like R&S would tell mostly the truth.

So, all manufacturers use this approach? Agilent/Tektronix? Even your favorite top-of-the-barrel brand LeCroy? =)

Maybe the Chinese are more honest then in the end of the day. They advertise their low end scopes as bottom-of-the-barrel scopes, to use your definition

[Lukas]

2) Your digital oscilloscope's noise specification

Most high sample rate scopes will interleave two, four, eight (or maybe more) digitizers to obtain a sample rate that high. The digitizers need to be accurately interleaved to prevent interleaving spurs in the frequency domain, which manifests itself as noise in the time domain.

Again, this might be a problem for cheap/low-end scopes, but interleaving spurs aren't a problem for most mid-range and high-end scopes. Aside from the fact that in the price segment of the R&S RTO other manufacturers use much faster single-block ADC hybrids, which again seems to be news to R&S.

Again, I'm not sure where the "lie" is, or what relevance this has for a beginner.

Rohde&Schwarz rolled their own monolithic 10GSa/s folding/interpolation ADC when they got into the oscilloscope market [1]. Others [2] interleave like crazy.  Looks like R&S are very proud of their monolithic ADC, so marketing tries to make use of that.

[1] https://www.mikrocontroller.net/topic/182804#1801224 and http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6226133
[2] http://poulton.net/papers.html

[pascal_sweden]

Note that even beginners, who can not afford high-end scopes, might want to get a good understanding about the differences between low-end scopes and high-end scopes.

So from that perspective, I think information like this can not just be classified as the "non beginners" section in the library

Some beginners are very curious, and question the underlying implementation and design decisions, which other industry veterans might just take for granted, because it has been drilled in their mindset.

Maybe beginners want to use the knowledge for having a better view on the limitations in their bottom-of-the-barrel scope, and want to explore ways on how they can work around these limitations. Or maybe they just want to learn more out of theoretical interest. Maybe they want to challenge their technical knowledge which they learned at university.

Even beginners with a limited budget and limited practical experience, might have a good theoretical understanding of signal sampling theorems and digital signal processing. I am sure that there are many "practical beginners" here on this forum, who have an engineering degree in electronics. A practical beginner is not per definition a high school student.

The practical stuff is simply not taught at universities, and sometimes you just get overwhelmed with so many mathematics during classes, that you forget basic principles and common sense. This is very unfortunate, but it is a reality among engineering students.

[Lukas]

For beginners interested in 'how it's done/has been done in the industry' I can strongly recommend the HP journals http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/hpjindex.html . Sure, many of them are pretty dated now, but they still a good job at explaining the design process that lead to the final product.

[Wuerstchenhund]

Thanks for pointing out that marketing buzz should be taken with a pinch of salt .. even if it comes from a big company

I thought that a reputable company like R&S would tell mostly the truth.

So, all manufacturers use this approach? Agilent/Tektronix? Even your favorite top-of-the-barrel brand LeCroy? =)

Pretty much, except (yes, you guessed it  ) LeCroy (which btw is one of the reason I do have a lot of respect for them). Agilent/Keysight and Tek are the worst offenders, they pretty much go the full mile including twisting reality if necessary. They also publish tons of "ours-vs-theirs" compare sheets and videos where they compare their scope against a (usually smaller model) competitor's scope in dubious setups, and massage the outcome so that their product looks better.

R&S isn't half as bad as Tek and Keysight, however either R&S has no clue what their competitors are up to (well, they aren't in the scope business for that long) or believes their customers are morons, because none of the points they highlighted on their "scope lies" page is actually an issue for any scope that competes with their very expensive RTO Series. Yes, it's marketing, but it doesn't really give me a lot of incentive to consider their products when the manufacturer has to resent to deception.

LeCroy doesn't seem to do ours-vs-their-comparison, at least so far. They're also not constantly harking on the competition, but let their products speak for themselves (but then, LeCroy is still very much lead by engineers, even most of their sales team comes from engineering). In fact, the only document they published which I'm aware off where competitors' scopes were used is this one talking about interpolation:

http://cdn.teledynelecroy.com/files/whitepapers/wp_interpolation_102203.pdf

And even there the test parameters are clearly defined so anyone should be able to reproduce the results.

Maybe the Chinese are more honest then in the end of the day. They advertise their low end scopes as bottom-of-the-barrel scopes, to use your definition

Maybe you're right. Both Rigol and Siglent have pretty much lied in the past about the capabilities of their products, so I'm not holding my breath, though.

[Wuerstchenhund]

Rohde&Schwarz rolled their own monolithic 10GSa/s folding/interpolation ADC when they got into the oscilloscope market [1].

Yes, in 2010. That's 9 years after LeCroy rolled out their 10GSa/s monolithic ADC in the WavePro 7k/WaveMaster 8k. Agilent came out with theirs not too long after. If I remember right even Tek had their own 10GSa/s ADC at that time.

Others [2] interleave like crazy.

Not really, the big competitors only interleave to allow even higher sample rates in 2ch config. Just to give you an idea, when R&S came out with their 10GSa/s ADC, others were already at 20GSa/s in a single ADC, and up to 80GSa/s in interleaved configuration.

Looks like R&S are very proud of their monolithic ADC, so marketing tries to make use of that.

They can be, but not because it's such an advance over the competition (which it wasn't, in the scope market R&S is mostly following, not leading). They can be proud of having produced the first 10GSa/s 8Bit ADC that didn't originate in the US and at that time would have fallen under export limitations due to its dual use in military applications.

[Wuerstchenhund]

Note that even beginners, who can not afford high-end scopes, might want to get a good understanding about the differences between low-end scopes and high-end scopes.

So from that perspective, I think information like this can not just be classified as the "non beginners" section in the library

With that argument we should also discuss cooking of shellfish, as this might be something that a EE beginner could also be interested after all 

In all seriousness, if it's something that mostly a beginner would ask then it should be in the beginner section.

Some beginners are very curious, and question the underlying implementation and design decisions, which other industry veterans might just take for granted, because it has been drilled in their mindset.

You have a pretty crude imagination about experienced engineers. I'm pretty sure that the majority doesn't just take things for granted because it has been "drilled in their mindset" (although that does happen, even in this forum we have a certain group of backwarders that regularly appear and tell people how much better analog scopes are and what hogwash all this new-fangled digital stuff is; pretty much the EE equivalent of HiFi vodoo), but simply because they know and are aware how stuff works and how to deal with it.

Maybe beginners want to use the knowledge for having a better view on the limitations in their bottom-of-the-barrel scope, and want to explore ways on how they can work around these limitations. Or maybe they just want to learn more out of theoretical interest. Maybe they want to challenge their technical knowledge which they learned at university.

And in your opinion you believe the marketing pages of a test instrument manufacturer is the right source for that? Seriously?

Even beginners with a limited budget and limited practical experience, might have a good theoretical understanding of signal sampling theorems and digital signal processing. I am sure that there are many "practical beginners" here on this forum, who have an engineering degree in electronics. A practical beginner is not per definition a high school student.

The practical stuff is simply not taught at universities, and sometimes you just get overwhelmed with so many mathematics during classes, that you forget basic principles and common sense. This is very unfortunate, but it is a reality among engineering students.

Well, please enlighten us then what you have learned from these R&S webpages that give you a better understanding of practical electronics engineering? Or are you already off to order a R&S scope?   

[Wuerstchenhund]

For beginners interested in 'how it's done/has been done in the industry' I can strongly recommend the HP journals http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/hpjindex.html . Sure, many of them are pretty dated now, but they still a good job at explaining the design process that lead to the final product.

I can second the recommendation, the HP Journal was from a time before they engaged in 'optimistic marketing'.

For those that are looking for stuff about more modern kit, some interesting documents covering the basics can also be found in LeCroy's (pretty badly organized) app note and whitepaper portfolio:

http://teledynelecroy.com/resources/details.aspx?doctypeid=1&mseries=352

http://teledynelecroy.com/resources/details.aspx?doctypeid=5&mseries=352

Beginners might be interested in particular stuff like that:

Understanding scope probes:
http://teledynelecroy.com/doc/probes-probing

How to calculate ENOB and SNR:
http://cdn.teledynelecroy.com/files/appnotes/computation_of_effective_no_bits.pdf

Equivalent Time Sampling:
http://cdn.teledynelecroy.com/files/whitepapers/wp_ris_102203.pdf

X/Y plots:
http://teledynelecroy.com/doc/calculating-area-in-xy-displays-waveform-math-finds-area-enclosed-by-xy-display

Troubleshooting high speed serial signals with a DSO:
http://teledynelecroy.com/doc/troubleshooting-high-speed-digital-signals-with-dsos

FFT:
http://teledynelecroy.com/doc/more-about-the-fft

Averaging:
http://teledynelecroy.com/doc/eres-vs-boxcar-averaging

[Lukas]

Rohde&Schwarz rolled their own monolithic 10GSa/s folding/interpolation ADC when they got into the oscilloscope market [1].

Yes, in 2010. That's 9 years after LeCroy rolled out their 10GSa/s monolithic ADC in the WavePro 7k/WaveMaster 8k. Agilent came out with theirs not too long after. If I remember right even Tek had their own 10GSa/s ADC at that time.

As in my [2] reference, most of the Agilent ADCs are based on interleaving a massive amount of ADC slices on a single die. You still have to accurately align their individual sampling phases.

Others [2] interleave like crazy.

Not really, the big competitors only interleave to allow even higher sample rates in 2ch config. Just to give you an idea, when R&S came out with their 10GSa/s ADC, others were already at 20GSa/s in a single ADC, and up to 80GSa/s in interleaved configuration.

See above. I guess R&S marketing is targeted against on-chip interleaving. That's how I at least understood it.

Looks like R&S are very proud of their monolithic ADC, so marketing tries to make use of that.

They can be, but not because it's such an advance over the competition (which it wasn't, in the scope market R&S is mostly following, not leading). They can be proud of having produced the first 10GSa/s 8Bit ADC that didn't originate in the US and at that time would have fallen under export limitations due to its dual use in military applications.

Sample and 'number of bits' rate isn't everything.  R&S specify an ENOB of ?7bit for their 8bit ADC, Keysight specify the ENOB for their 10bit scopes. Lecroy doesn't even specify an ENOB for their 12bit scopes. If the ENOB would be noteworthy, they'd specify it. In a 'competitive comparison' [1] Agilent states an ENOB of 6.4bits@1GHz for LeCroy's 12bit scope, whereas their 10bit scope has an ENOB of 8 bits. Go figure.

[1] http://cp.literature.agilent.com/litweb/pdf/5991-4437EN.pdf

[commie]

I believe Teledyne(pls correct me if I am wrong) in 1975 introduced the 4558 dual opamp, fantastic symmetric opamp. They now cost as little as £8.00/100, why? because everyone  is using them. 

[T3sl4co1l]

Oft-repeated lie:

"You need a sample rate significantly faster than the signal bandwidth, to see anything at all"

 

[Wuerstchenhund]

As in my [2] reference, most of the Agilent ADCs are based on interleaving a massive amount of ADC slices on a single die. You still have to accurately align their individual sampling phases.

I'm sorry but your "reference" just goes to a website containing a list of other documents (of which many aren't even oriented correctly). You'd have to be more specific as just off-loading a list of documents and expecting me to find the one that supports your statements.

Sample and 'number of bits' rate isn't everything.  R&S specify an ENOB of ?7bit for their 8bit ADC, Keysight specify the ENOB for their 10bit scopes. Lecroy doesn't even specify an ENOB for their 12bit scopes. If the ENOB would be noteworthy, they'd specify it. In a 'competitive comparison' [1] Agilent states an ENOB of 6.4bits@1GHz for LeCroy's 12bit scope, whereas their 10bit scope has an ENOB of 8 bits. Go figure.

[1] http://cp.literature.agilent.com/litweb/pdf/5991-4437EN.pdf

That's a good example of someone falling into the trap of marketing BS (and using an Agilent "infomercial" isn't really helping your case). LeCroy doesn't specify ENOB because as a figure it is worthless without a description of the environment (i.e. frequency) it was calculated at, which of course you'll rarely find in Agilents or even R&S's marketing blah.

You might want to read this paper from Analog Devices which should explain what ENOB really does:
http://www.analog.com/media/en/technical-documentation/technical-articles/MS-2124.pdf

Let me quote:

"Though often quoted, the ENOB is insufficient to describe a high speed converter’s performance. High speed converters are famously multiparametric, and no single number can hope to capture what takes an entire specification table to describe. ENOB does make a reasonable starting point for comparing candidate converters,so long as you do not depend excessively on the number’s significance."

and

"More valuable are the SINAD vs. frequency characteristic curves, which many high speed converters present in their data sheets (Figure 2). These allow you to identify at least typical performance at the frequencies of interest to your application instead of at the spot frequencies the converter manufacturer chose for the data sheet’s  specification table."

If you really think when R&S says their ADC has an ENOB of 7bit and Agilent that their ADC has an ENOB of 8bit that you know which ADC is better then you've been taken for a ride.

LeCroy doesn't publish ENOB numbers because as a single figure they are useless. However, in an example in one of their app notes they measure and calculate the ENOB for the WaveMaster 820zi 20GHz scope, which for their test parameters (1.4GHz sine) comes out at 5.33bit, which isn't bad for a 20GHz 40GSa/s scope. But thankfully aside from that well documented example they leave the ENOB BSing to their competitors and focus on the relevant parameters.

[tautech]

I'd also caution you to not blindly buy into any marketing BS, as you seem to do. Most of such stuff where one manufacturer compares his product with their competitors isn't worth the paper/the bytes its written on/with.

+1
Like any bias comparative analysis, they are always  to make themselves look/spec better.

This marketing "oneupmanship" has been going on for decades. 

[vk6zgo]

I'm a bit dubious about the triggering one:-

http://rohde-schwarz-scopes.com/scope_lie_04.php

If their argument here is justified,then my old BWD which obtains triggering from a 220k Ohm resistor hanging off the output of the vertical deflection amplifier should have great triggering.

I can't remember any radical problems with triggering in the  many good quality analog 'scopes ,or in the few DSOs I have used.

"Straw man"?

[Rupunzell]

DSO, digitizing systems are going to slide, dice and chop up what once was a constantly varying natural event into bits and pieces.

These bits and pieces will be reconstituted into a form of visual information.

Each step of this process contains and results in distortions of what once was the original constantly varying event. Limitations and advantages of this process must be kept in mind by digitizing systems to avoid being fooled by the information presented.

Core difference between digitizing systems (DSO) and Analog time domain instruments, Analog O'scopes process the signal as is with great fidelity if they are any good. This difference is why high fidelity, high quality analog O'scopes do not lie while DSO can lie and deceive it's users that are un-aware of limitations baked into their design.

If would be ignorant and claim one technology is always better than the other as both technologies have their advantages and disadvantages. Best and most accurate results can occur when the measurement need determines instrument and measurement choice.

All else is marketing and taking advantage of user-market ignorance.


Bernice

[Wuerstchenhund]

Oh boy, that didn't take long, did it? 

Core difference between digitizing systems (DSO) and Analog time domain instruments, Analog O'scopes process the signal as is with great fidelity if they are any good. This difference is why high fidelity, high quality analog O'scopes do not lie while DSO can lie and deceive it's users that are un-aware of limitations baked into their design.

I'm sure it's complete news to you, but FYI, analog scopes *do* lie, they lie a lot (Dave did even a video about that some time  ago). In fact, *every* test instrument lies to an extend, and any engineer worth it's merits should be very well aware of that.

But thank you for proving my point:

(although that does happen, even in this forum we have a certain group of backwarders that regularly appear and tell people how much better analog scopes are and what hogwash all this new-fangled digital stuff is; pretty much the EE equivalent of HiFi vodoo)

[Rupunzell]

Yep, this point will never end until YOU get some real world experience with serious Analog design and why high quality, high fidelity analog O'scopes DO NOT LIE if the user knows instrument limitations and how to apply said instrumentation properly.

Can you prove DSO's don't lie? If yes, post your results. I don't care about what Dave or others have posted. Present the facts, test conditions and results completely for all to examine, duplicate and prove as fact.

Then prove high quality, high fidelity analog O'scopes lie.

Nuff said.

Bernice

Oh boy, that didn't take long, did it? 

I'm sure it's complete news to you, but FYI, analog scopes *do* lie, they lie a lot (Dave did even a video about that some time  ago). In fact, *every* test instrument lies to an extend, and any engineer worth it's merits should be very well aware of that.

(although that does happen, even in this forum we have a certain group of backwarders that regularly appear and tell people how much better analog scopes are and what hogwash all this new-fangled digital stuff is; pretty much the EE equivalent of HiFi vodoo)

But thank you for proving my point

[dadler]

Ahh ad hominem, condescension....not long until we have a nazi reference.

Oh Internet, I love thee.

[Rupunzell]

Indeed, trolling with baiting often brings up NAZI at some point. That is pretty much when any meaningful discussion of any topic has flat-lined.


Bernice

Ahh ad hominem, condescension....not long until we have a nazi reference.

Oh Internet, I love thee.

[Rupunzell]

Issues with DSO signal fidelity and information presentation has been discussed once before.
https://www.eevblog.com/forum/testgear/my-first-oscilloscope/75/


Bernice

[ale500]

The video from the signal path about the 100 GHz Teledyne-Lecroy Oscilloscope have a nice but not very in deep explanation of a modern interleaving system. In the end, they use digital signal processing to reconstruct the original signal . If I understood correctly one could use the same method at lower frequency, just as a proof-of-concept right ?
Agilent/HP also published a paper on how their 8 bit ADCs (several GS/s) worked a couple of years back, and they talked about ENOB quite lengthly... but where is that document again...

[Rupunzell]

LeCroy short on what they are doing..
http://cdn.teledynelecroy.com/files/whitepapers/interleaving_process_in_dbi_scopes.pdf

Note the use of a mixer and how the bits and pieces of information is to be sorted out and presented to the user.


Alternatively, Hypress did a superconducting Josephson junction based 50 Ghz sampling scope in 1987.
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1487395&url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F9953%2F31997%2F01487395


And a bit of history:
http://www.kahrs.us/~mark/pdf/papers/MTT2003.pdf


Bernice

The video from the signal path about the 100 GHz Teledyne-Lecroy Oscilloscope have a nice but not very in deep explanation of a modern interleaving system. In the end, they use digital signal processing to reconstruct the original signal . If I understood correctly one could use the same method at lower frequency, just as a proof-of-concept right ?
Agilent/HP also published a paper on how their 8 bit ADCs (several GS/s) worked a couple of years back, and they talked about ENOB quite lengthly... but where is that document again...

[Lukas]

As in my [2] reference, most of the Agilent ADCs are based on interleaving a massive amount of ADC slices on a single die. You still have to accurately align their individual sampling phases.

I'm sorry but your "reference" just goes to a website containing a list of other documents (of which many aren't even oriented correctly). You'd have to be more specific as just off-loading a list of documents and expecting me to find the one that supports your statements.

http://poulton.net/papers.public/2010_cicc_GHz_ADCs.pdf Slide 27. Their 20GSa/s ADC is made up of no less than 80 ADC slices.

Regarding ENOB, etc: Still wonders me why I get more specs for a $50 ADC than for a $50k 'scope.

The video from the signal path about the 100 GHz Teledyne-Lecroy Oscilloscope have a nice but not very in deep explanation of a modern interleaving system. In the end, they use digital signal processing to reconstruct the original signal . If I understood correctly one could use the same method at lower frequency, just as a proof-of-concept right ?
Agilent/HP also published a paper on how their 8 bit ADCs (several GS/s) worked a couple of years back, and they talked about ENOB quite lengthly... but where is that document again...

Very likely to be there, since this guy seems to be one of the ADC gurus at HP/Agilent/Keysight: http://poulton.net/papers.html