The end is near. LeCroy and their 100GHz scope...

[rsjsouza]

EDN article

I really did not know LeCroy was there yet...

If Dave ever has one for evaluation and teardown, I wonder if there will be a sticker inside: "WARNING: continuous exposure to this oscilloscope internals can cause cancer from non-ionizing radiation."

(still remember my astonishment about a 100MSPS oscilloscope in early 90's...)

[CodyShaw]

Boy would it be freaking cool to be on the development team for that bad boy...

Now Dave needs to get his hands on one for a teardown!

[alm]

The novelty is that it's a real-time scope with 100 GHz bandwidth. Lecroy has been offering 100 GHz sampling scopes for a while.

[tinhead]

The novelty is that it's a real-time scope with 100 GHz bandwidth.

right, but they cheating a bit "DBI works by splitting the input signal into multiple frequencies which are downconverted to the bandwidth of the digitizing system. Then it acquires and digitizes each band, using DSP to re-assemble them and compensate for distortion".

[alm]

This has been going on for a while, I think the fast real-time Agilent scopes do the same. They're multiplexing multiple ADCs, somewhat similar to what the cheap 1GS/s scopes (Rigol et al) do, but in the frequency domain instead of the time domain. The main issue is that (automated) calibration takes forever, since everything has to be matched to minimize distortion.

[free_electron]

This has been going on for a while, I think the fast real-time Agilent scopes do the same. They're multiplexing multiple ADCs, somewhat similar to what the cheap 1GS/s scopes (Rigol et al) do, but in the frequency domain instead of the time domain. The main issue is that (automated) calibration takes forever, since everything has to be matched to minimize distortion.

the one i used for the jim williams pulse generator doesn't. the sampler time interleves analog samples ( basically 8 sample/hold amplifiers with differential outputs that are shifted in time) . that feeds into 8 a/d converters.

[alm]

I checked for any sources on Agilent using frequency interleaving, but I couldn't find any. Must have been mistaken on that.

So they poke fun at Lecroy for using frequency domain interleaving while they use time domain interleaving?

[JuanPC]

  LeCroy.

[dr.diesel]

Now Dave needs to get his hands on one for a teardown!

Here is their contact info:

contact.corp@teledynelecroy.com

Let's all email, asking them to send Dave one for a "review", cough, with a screwdriver.

[SArepairman]

they will mail him one thats welded shut if they do send anything at all lol

[xrunner]

How do you couple a 100 GHz signal from a system into a scope without disturbing it (the signal)? I am depending on the more knowledgeable people here to school me.

[EEVblog]

How do you couple a 100 GHz signal from a system into a scope without disturbing it (the signal)? I am depending on the more knowledgeable people here to school me.

You essentially can't. But if you spend a lot of money and have a grey beard, you can get close and will know what the effects and limitations are.

[EEVblog]

Let's all email, asking them to send Dave one for a "review", cough, with a screwdriver.

Surely they have an old prototype front end board lying around? Mailbag please!

[free_electron]

How do you couple a 100 GHz signal from a system into a scope without disturbing it (the signal)? I am depending on the more knowledgeable people here to school me.

You essentially can't. But if you spend a lot of money and have a grey beard, you can get close and will know what the effects and limitations are.

you need one of those bob pease beards... with in-beard-breadboard

[c4757p]

How do you even generate that 100 GHz test signal? Jesus, that's a tenth of a terahertz! DC to daylight indeed...

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

[dfmischler]

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

I don't think so.  Red light (the lowest frequency in the visible spectrum) starts at about 400 THz.  Then again, I'm already older than that.

[xrunner]

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

Well my eyeball can probe daylight and "measure it" - i.e., put it into a visual frame for my brain ... but perhaps what it's seeing is being disturbed by itself and what I see isn't really what's out there - outside my brain. 

OK I'm stopping myself before somebody else does.

[c4757p]

I was kidding!

I still want to know more about how it all works, though. 100 GHz is amazing.

[dr.diesel]

Pretty amazing.  Wonder what type of input connector is on it?  Type N is really only good to about 18Ghz.

[xrunner]

How do you couple a 100 GHz signal from a system into a scope without disturbing it (the signal)? I am depending on the more knowledgeable people here to school me.

You essentially can't. But if you spend a lot of money and have a grey beard, you can get close and will know what the effects and limitations are.

In all seriousness - I can't imagine using it @100GHz unless a circuit board that used 100GHz class signals had dedicated test points built-in, these being some sort of coax connector, that allowed such a piece of test equipment to be connected. Outside of those test points, it seems to me that it would near impossible to "clip on" to a part of a 100GHz circuit and have it operate in any way, shape, or form, like it was intended. But I am not an expert at 100GHz. 

[free_electron]

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

Well my eyeball can probe daylight and "measure it" - i.e., put it into a visual frame for my brain ... but perhaps what it's seeing is being disturbed by itself and what I see isn't really what's out there - outside my brain. 

OK I'm stopping myself before somebody else does.

you mean we carry two ultra high frequency spectrum analyzers around in our head in the form of eyeballs  ? and two low frequency spectrum analyzers in the form of ears ? and our skin can sense infrared ? and we even have a built in signal generator that can do arbitrary signals. it's called a mouth. and its all connected via electrical interfaces to a massive computer and data storage system called a brain.

dammit. why do we need all that complicated testgear then !

[EEVblog]

In all seriousness - I can't imagine using it @100GHz unless a circuit board that used 100GHz class signals had dedicated test points built-in, these being some sort of coax connector, that allowed such a piece of test equipment to be connected. Outside of those test points, it seems to me that it would near impossible to "clip on" to a part of a 100GHz circuit and have it operate in any way, shape, or form, like it was intended. But I am not an expert at 100GHz. 

Yes, it's not this everyday debugging tool where you go around probing your 100GHz circuit to figure out why it's not working. They are designed for characterisation of new systems and technology, or as you say, measurement of circuits with dedicated build in test point for this very purpose. In either case, a lot of thought and engineering would have gone into the on-board probing test points.

[xrunner]

you mean we carry two ultra high frequency spectrum analyzers around in our head in the form of eyeballs  ? and two low frequency spectrum analyzers in the form of ears ? and our skin can sense infrared ? and we even have a built in signal generator that can do arbitrary signals. it's called a mouth. and its all connected via electrical interfaces to a massive computer and data storage system called a brain.

dammit. why do we need all that complicated testgear then !

'cause some people will either not believe what they see, or they will fabricate a story about what they think they see. 

[Bored@Work]

dammit. why do we need all that complicated testgear then !

Because our build-in stuff only handles some small parts of the spectrum. Esp. spectrum mankind has become very interested in, even addicted to.

[nctnico]

How do you couple a 100 GHz signal from a system into a scope without disturbing it (the signal)? I am depending on the more knowledgeable people here to school me.

The prototype usually is a PCB with test connectors (and passive.active on-board signal pick-up) to connect a probe to. When the design has been verified a new PCB is made without the test connectors. Even an on-board passive probe (divider with two resistors and an SMA socket) is useful to several GHz.

[rsjsouza]

They are designed for characterisation of new systems and technology, or as you say, measurement of circuits with dedicated build in test point for this very purpose.

That is usually the case of development boards used with the initial silicon samples of new devices: the board itself is a sea of SMA connectors tied to the multiple pins of the device for proper characterization but have absolutely no practical/everyday use. Not surprisingly these boards never see the light of the day, but are beautifully constructed (some are almost 1cm thick depending on the device complexity!)

[amyk]

How do you even generate that 100 GHz test signal? Jesus, that's a tenth of a terahertz! DC to daylight indeed...

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

192GHz in 2006: http://www.dtic.mil/dtic/tr/fulltext/u2/a464948.pdf
300GHz in 2011: http://www.ee.ucla.edu/~razavi/papers/Journals/BRApr11.pdf

Generating the signal doesn't seem particularly difficult, compared to being able to measure it...

[w2aew]

I checked for any sources on Agilent using frequency interleaving, but I couldn't find any. Must have been mistaken on that.

So they poke fun at Lecroy for using frequency domain interleaving while they use time domain interleaving?

The Agilent scopes that are >33GHz BW do appear to using a form of Digital Bandwidth Interleaving, similar to LeCroy in their Real-Edge mode. They don't say that this is what their doing, but detailed analysis of various things like noise floor, fine resolution frequency and phase response, etc. all point to the use of this technique.

[rsjsouza]

192GHz in 2006: http://www.dtic.mil/dtic/tr/fulltext/u2/a464948.pdf
300GHz in 2011: Generating the signal doesn't seem particularly difficult, compared to being able to measure it...

Back in 1996 I recall a presentation by HP (before it became Agilent) showcasing their EESOF program and the presenter mentioned the Patriot missile (used thoroughly in the first gulf war) had a close range radar system operating at 93GHz on its guidance module...

[c4757p]

192GHz in 2006: http://www.dtic.mil/dtic/tr/fulltext/u2/a464948.pdf
300GHz in 2011: http://www.ee.ucla.edu/~razavi/papers/Journals/BRApr11.pdf

Generating the signal doesn't seem particularly difficult, compared to being able to measure it...

It's still damn impressive, just slightly less. Thanks for the links, though - good reading!

otherwise, its like a 5 years old kid messing with DS1052E.

Five-year-old me would have loved a DS1052E. Though I probably would have put my foot through the screen, or found some equally stupid way to break it...

[Marco]

The novelty is that it's a real-time scope with 100 GHz bandwidth.

right, but they cheating a bit "DBI works by splitting the input signal into multiple frequencies which are downconverted to the bandwidth of the digitizing system. Then it acquires and digitizes each band, using DSP to re-assemble them and compensate for distortion".

How are frequency domain interleaved ADCs more cheating than time domain interleaved ADCs.

[c4757p]

Hmm. On one hand, the technology involved in splitting up signals into manageable chunks like that is fascinating. On the other hand, I do love the simplicity of a traditional scope: BNC - analog crap - CRT deflection plates...

Of course, even most basic DSOs split the signal into AC and near-DC.

[Marco]

Of course, even most basic DSOs split the signal into AC and near-DC.

Hmm, how do you mean? I thought ADCs in DSOs were generally DC coupled.

[c4757p]

They are, but the signal is often split before a lot of the signal conditioning circuitry and recombined right before the ADC driver amplifier. We've got schematics floating around here somewhere for the DS1052E frontend, I'll see if I can find them.

Here. Not as dramatic as I may have made it sound, but the AC does skip across the first amplifier through that 330pF capacitor.

I guess this is how the majority of scope frontends work; I was more or less just remarking that even though the naive description is often that the signal drives the CRT "directly" through the amplifiers, that's not really true. Or rambling because I just woke up. You decide.

[Marco]

Ah okay, yeah I read about that in some books on oscilloscope front end design ... basically the ideal HF amplifier is a source/emitter follower, but they have lousy 1/f noise and DC offset stability, whereas opamp circuits have great 1/f noise and DC offset stability.

[c4757p]

There are whole books on oscilloscope front end design?? Any you can recommend? I'd love to see if I can find one...

[synapsis]

There are whole books on oscilloscope front end design?? Any you can recommend? I'd love to see if I can find one...

I second this.

So far all I've read about high speed ADCs is this paper, which was pretty informative for an amateur like me. (PDF)

https://www.ece.cmu.edu/research/publications/2009/CMU-ECE-2009-010.pdf

[alm]

There is the 'slightly dated' Tektronix Circuit Concepts series. It is probably by far the most detailed information you'll find, though. Back when they were showing to the world how awesome their designs were; before the lawyers and MBAs came in and started calling it property. Pretty sure it does not discuss ADCs, however .

[c4757p]

Oh, that's awesome, thank you!

Is is possible to feel nostalgia for a time you never experienced?

[JuanPC]

How do you couple a 100 GHz signal from a system into a scope without disturbing it (the signal)? I am depending on the more knowledgeable people here to school me.

Roswell means anything to you? 

[JuanPC]

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

Well my eyeball can probe daylight and "measure it" - i.e., put it into a visual frame for my brain ... but perhaps what it's seeing is being disturbed by itself and what I see isn't really what's out there - outside my brain. 

OK I'm stopping myself before somebody else does.

[Marco]

There are whole books on oscilloscope front end design?? Any you can recommend? I'd love to see if I can find one...

Well not front end design specifically, but with chapters devoted to it ... can't vouch for the quality, but this is what I found :

Art and Science of Analog Design by Steve Roach
Wideband Amplifiers by P. Staric

Both concerning relatively low (<GHz) bandwidth designs BTW ... old tech, which can be build with discretes.

[echen1024]

192GHz in 2006: http://www.dtic.mil/dtic/tr/fulltext/u2/a464948.pdf
300GHz in 2011: http://www.ee.ucla.edu/~razavi/papers/Journals/BRApr11.pdf

Generating the signal doesn't seem particularly difficult, compared to being able to measure it...

It's still damn impressive, just slightly less. Thanks for the links, though - good reading!

otherwise, its like a 5 years old kid messing with DS1052E.

Five-year-old me would have loved a DS1052E. Though I probably would have put my foot through the screen, or found some equally stupid way to break it...

I got my first Multimeter when I was 5. It was a cruddy little POS, but provided hours of fun. I graduated to a Tek 2205 at 10 years old, and have been using that scope for 13 continuous (I'm in High school), and recently bought a Siglent SDS1102CNL.

[JuanPC]

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

Well my eyeball can probe daylight and "measure it" - i.e., put it into a visual frame for my brain ... but perhaps what it's seeing is being disturbed by itself and what I see isn't really what's out there - outside my brain. 

OK I'm stopping myself before somebody else does.

you mean we carry two ultra high frequency spectrum analyzers around in our head in the form of eyeballs  ? and two low frequency spectrum analyzers in the form of ears ? and our skin can sense infrared ? and we even have a built in signal generator that can do arbitrary signals. it's called a mouth. and its all connected via electrical interfaces to a massive computer and data storage system called a brain.

dammit. why do we need all that complicated testgear then !

one day, i took a small dose of LSD-25, half a dose,
the walls were breathing in and out...
like this, but faster:

like this fast, but real:


i wanted to know why i see that...
i stared at the mirror in my bathroom,  firmly looked into my eyes,
and noticed that both eyes were moving synchronized with a Horizontal LFO near 1Hz-5Hz,
like a Synth,
and i guessed that was causing the Walls to Breath,

was impressive to think how "the machine works."

like this, but less crazy 

[FrankBuss]

How do you even generate that 100 GHz test signal? Jesus, that's a tenth of a terahertz! DC to daylight indeed...

There is a newer video where they explain how they generate a test signal, and some more background information:

He says that there is no test equipment that can generate a 100 GHz sine wave, so they are using a 16.6 GHz signal generator and a six times multiplier, feeded to the scope through a wave guide (see video 1:40). Doesn't look like a standard BNC

They don't say how much the scope costs, but of course it is too expensive if you have to ask.

[dannyf]

I think the fast real-time Agilent scopes do the same.

Yes. And the el cheapo scopes do that as well.

[Rufus]

I think the fast real-time Agilent scopes do the same.

Yes. And the el cheapo scopes do that as well.

Resurrect a year old thread to quote with no attribution a sentence devoid of any meaning without context so you can add another equally meaningless sentence. Nice job.

[Dave Turner]

Astronomers have been working with Terahertz and above for years.  Optical processors have supposedly been 'around the corner' for some time. Hmm maybe 100Ghz and above should be looked at from another angle.

[dannyf]

The technology is there: there are optical amplifiers for fiber transmissions.

[Noise Floor]

I'd like to see them move that signal off center and see how stable it is.  Ever since the whole "12 bit" scope marketing exaggeration they did, I have my doubts on how useful this really is. 

I have little doubt it works in a very small band, but unless you are targeting a specific application in the band, you are likely served better by a more overall well rounded and engineered scope.  People get hung up on some top level specs and fail to account for real world use in many cases is my perspective.

[dannyf]

they are using a 16.6 GHz signal generator and a six times multiplier,

What about using a sub-pico-second pulse generator and filter out the lower harmonics?

[SeanB]

Pulse repetition rate is too low..............

[Dave Turner]

The technology is there: there are optical amplifiers for fiber transmissions.

Dannyf - that was sort of my point.

To my knowledge serious research has been going on to create optical devices analogous to semiconductors at least as far back as the 70's, if not further. I'm not counting holographics. Where said research has got to I don't know, however given that the research is successful, becomes commonplace and miniaturised we should see an increase of several orders of magnitude in high speed computation. transmission etc. Hopefully in what's left of my lifetime.

[Alex Eisenhut]

How do you even generate that 100 GHz test signal? Jesus, that's a tenth of a terahertz! DC to daylight indeed...

I swear, at the rate we're going, we'll have scopes that can probe literal daylight by the time I'm 40...

Probably RTD based. BWO's been doing that for decades. They go to THz.

As for a "scope" with a light input, why did you wait so long?

http://w140.com/tekwiki/wiki/7J20

[Alex Eisenhut]

How do you even generate that 100 GHz test signal? Jesus, that's a tenth of a terahertz! DC to daylight indeed...

There is a newer video where they explain how they generate a test signal, and some more background information:

He says that there is no test equipment that can generate a 100 GHz sine wave, so they are using a 16.6 GHz signal generator and a six times multiplier, feeded to the scope through a wave guide (see video 1:40). Doesn't look like a standard BNC

They don't say how much the scope costs, but of course it is too expensive if you have to ask.

"no test equipment that can generate a 100 GHz sine wave"? Really?

"Doesn't look like a standard BNC"

1mm connector goes to 110GHz. Wonder why they make 'em if no one can generate 100GHz....

[Alex Eisenhut]

The technology is there: there are optical amplifiers for fiber transmissions.

Doped fiber amplifier. It's all done in the physics of light. Not "electronics" in the traditional sense, although it's all done with electrons one way or another...

http://en.wikipedia.org/wiki/EDFA

[Wuerstchenhund]

Ever since the whole "12 bit" scope marketing exaggeration they did

What '"12 bit" scope marketing exaggeration'?

[Noise Floor]

Ever since the whole "12 bit" scope marketing exaggeration they did

What '"12 bit" scope marketing exaggeration'?

They pushed very hard the fact that they used a 12 ADC compared to other market competitors using 8 bit ADCs.  The problem I found is what I spoke to above, the extra ADC resolution was only useful in a very narrow band.  I believe this was a result of the scope not being engineered for 12 bits throughout as well as the ADC itself having lots of noise and distortion at many input frequencies. 

Our lab at work bought an HDO early on I learned to not trust it pretty quickly relative to other scopes I have access to, for the applications I was working on.  I bold that part because I do think the extra resolution could be used in certain settings, but people should know the limitations of all their equipment and not buy into the marketing of 12 bit always being better.  That said, I did really like the UI and footprint of the scope and I think Lecroy innovated well in that space.  I just don't care for over the top exaggerated marketing, which this 100GHz demo smells like to me, although I can't claim that for certain.

[Wuerstchenhund]

Ever since the whole "12 bit" scope marketing exaggeration they did

What '"12 bit" scope marketing exaggeration'?

They pushed very hard the fact that they used a 12 ADC compared to other market competitors using 8 bit ADCs. 

Well, that's what marketing is about, highlighting where your product is different/better than the competition. At the end of the day, marketing *is* exaggeration. It's not just LeCroy doing it for their scopes, everyone does it.

This aside, LeCroy did have a point: Agilent's pseudo 12bit scope was indeed using 8bit ADCs with oversampling. Nothing wrong with highlighting that.

And if you really believe that LeCroy was exaggerating then I would really like to hear what you think about Agilent quite often in their marketing brochures ('whitepapers') twists reality 'til it screams.

The problem I found is what I spoke to above, the extra ADC resolution was only useful in a very narrow band.  I believe this was a result of the scope not being engineered for 12 bits throughout as well as the ADC itself having lots of noise and distortion at many input frequencies. 

Our lab at work bought an HDO early on I learned to not trust it pretty quickly relative to other scopes I have access to, for the applications I was working on.  I bold that part because I do think the extra resolution could be used in certain settings, but people should know the limitations of all their equipment and not buy into the marketing of 12 bit always being better. 

I agree, but that is no different with other scope properties, every engineer should know how important certain features are for his specific applications, that's part of the job.

To me this sounds like your place simply bought the wrong tool for the job, which does happen (this is why most scope manufacturers do offer lending a scope for a short time for evaluation). And quite frankly, any person that blindly buys a new and very expensive piece of test equipment without making sure it's beneficial for the application or making use of getting a loaner first should be 'promoted' to cleaning the lab floor.

That said, I did really like the UI and footprint of the scope and I think Lecroy innovated well in that space.  I just don't care for over the top exaggerated marketing, which this 100GHz demo smells like to me, although I can't claim that for certain.

Again, if you think LeCroy is exaggerating (which is a bit difficult in such a live demo, especially if there is a certain reputation to protect) then go and have a read through some Agilent whitepapers.

[Noise Floor]

Hey Wuerstchenhund,

You won't much disagreement from me with what you said.  I'm not familiar with the white papers you are referring to, perhaps we should create a thread for "marketing wall of shame".  I was only speaking to my experience in this case and have strong memory of being hitting over the head with "12-bit!" marketing emails and ads for a while.

[Wuerstchenhund]

perhaps we should create a thread for "marketing wall of shame". 

That would be futile, as the list would certainly be endless. It's much easier to only list the positive examples, as this makes for a rather short list 

I was only speaking to my experience in this case and have strong memory of being hitting over the head with "12-bit!" marketing emails and ads for a while.

Don't forget that marketing has a purpose, not only for the manufacturer that wants to sell its goods. Simply because for expensive kit the person who wants it is often not the person who pays for it. An engineer may want a certain scope but if he can't convince his finance controller that this is the best thing since sliced bread then he simply won't get it. Sad but true in many workplaces. Marketing is the bait engineers use to get what they want 

[Noise Floor]

Don't forget that marketing has a purpose, not only for the manufacturer that wants to sell its goods. Simply because for expensive kit the person who wants it is often not the person who pays for it. An engineer may want a certain scope but if he can't convince his finance controller that this is the best thing since sliced bread then he simply won't get it. Sad but true in many workplaces. Marketing is the bait engineers use to get what they want 

I never thought of it that way, very interesting.

[David Hess]

There are whole books on oscilloscope front end design?? Any you can recommend? I'd love to see if I can find one...

Well not front end design specifically, but with chapters devoted to it ... can't vouch for the quality, but this is what I found :

Art and Science of Analog Design by Steve Roach
Wideband Amplifiers by P. Staric

Both concerning relatively low (<GHz) bandwidth designs BTW ... old tech, which can be build with discretes.

I thought of that article as well.  It is included in "The Art and Science of Analog Circuit Design (EDN Series for Design Engineers)" edited by Jim Williams:

http://www.amazon.com/Science-Analog-Circuit-Design-Engineers/dp/0750670622

"Analog Circuit Design: Art, Science and Personalities (EDN Series for Design Engineers)" which was also edited by Jim Williams discusses oscilloscope vertical amplifier design among other things.

http://www.amazon.com/Analog-Circuit-Design-Personalities-Engineers/dp/0750696400

[David Hess]

Ah okay, yeah I read about that in some books on oscilloscope front end design ... basically the ideal HF amplifier is a source/emitter follower, but they have lousy 1/f noise and DC offset stability, whereas opamp circuits have great 1/f noise and DC offset stability.

Unfortunately oscilloscopes are stuck using source followers for high and controlled input impedance and good transient response.  The DC stability in operational amplifiers comes from a fully differential input and it is worth noting that analog oscilloscopes do a single ended to differential conversion early in their signal chain to take advantage of this.  Some (most?) DSOs do as well.

The 1/f and other noise is not helped by using small area devices to limit input capacitance for good high frequency performance.

[Sailor]

The Tektronix  book 'Vertical Amplifier Circuits' is great reading.