Video Teardown & Experiements with LeCroy's 100GHz, 240GS/s Scope

[jimon]

Hugoneus, w2aew,

Thanks for explanation

[miguelvp]

Very nice and concise for all the information presented. Even if it was a long video it was well organized and to the point.

[Radio Tech]

I had to sit up late to watch tis one. Great video   

[KF5OBS]

BTW, Chris met Mr Lecroy the other week, and didn't even think to ask him onto the Amp Hour! 

Do you want me to ask him if he wants to be on the Amp Hour?

[HighVoltage]

I really enjoyed the factory tour at the end of your video.
It was also interesting to see that LeCroy is using lots of Agilent gear.

You should do the same with Keysight and show us a really good factory tour of the Keysight production facilities.

[Wuerstchenhund]

It was also interesting to see that LeCroy is using lots of Agilent gear.

What else should they use? LeCroy doesn't do RF generators and other such kit themselves, and for that stuff Agilent/Keysight gear is probably the best you can get.

[mikeselectricstuff]

Would have been interesting to see what they used to cal the 100GHz version

[janoc]

vaualbus, jimon,

Let me clarify a few things which you kindly brought up.

The Keysight 62GHz scope uses essentially the same technique for conversion. Their ReadEdge system is a diplexer, filter and dual-band down-converter. This can be inferred from their frequency planning and block diagrams, even though they don't directly mention it.

The two laser experiment is a very important test. (I was not going to let them get away with a cheesy test!)  O0 When you beat two lasers together in a PD (in non-linear detection) you get a beat frequency which is the difference between the two tones (You also get the sum, but that is huge). So in this case we are first beating a 194THz laser with a 194.05THz laser which would produce an electrical tone at 50GHz. By increasing the second laser to 194.1THz, we produce an electrical tone at 100GHz! The scope then digitizes this electrical tone which is at the edge of its bandwidth. I hope this is more clear now. 

Mindblowing stuff, thanks for this video! 

[Hugoneus]

Would have been interesting to see what they used to cal the 100GHz version

Unfortunately I could not film that. Even a small conversation we had about it had to be cut out of the video. They wish to keep it confidential.

[andrija]

That board with those heatsinks and memory modules is one of the most beautiful pieces of electronics I have ever seen. It's beautiful. It looks like a work of art.

[rx8pilot]

That board with those heatsinks and memory modules is one of the most beautiful pieces of electronics I have ever seen. It's beautiful. It looks like a work of art.

Yes, I was stunned by the component count. I think he said 15,000 components! Wonder what the reject rate is on the PCB's? ....and the extreme cost of the silicon would make the build process the most nerving moment I could think of.

It doesn't just have to work - it has to work perfect. It seems that the slightest oops would disable a PCB worth $100k+ Kinda nuts.

[bktemp]

The acquisition board is relly a great piece of art!
I am a bit surprised they have used 5 independent power supplies on the board. A lot of point of load converters make sense, but different power supplies make a design normally much more difficult, because if one power supply fails the unpowered parts can easily get damaged when input signals from the powered sections are still active.
On this board it seems the 4 channels are completely seperate. The 5th power supply is probably for the trigger and control circuit.

[Dongulus]

After watching the video, I wanted to find a little more information about the Teledyne LeCroy's DBI processing technique. I found this paper which offers more detail about how the DBI system works, a bit the design of the input amplifier for the MSH (monolithic sample and hold), and how the digital processing works.

[chicken]

It doesn't just have to work - it has to work perfect. It seems that the slightest oops would disable a PCB worth $100k+ Kinda nuts.

100% agree, but not perfect... I wonder how it feels to put a soldering iron onto that PCB

[IanB]

100% agree, but not perfect... I wonder how it feels to put a soldering iron onto that PCB

I hope that's a bodge wire and not a botch wire 

(A bodge is a fix, while a botch is a screw up.)

[EEVblog]

It doesn't just have to work - it has to work perfect. It seems that the slightest oops would disable a PCB worth $100k+ Kinda nuts.

I've worked on boards like that, and it is a very time consuming and expensive process. Things like first doing a trial place/reflow solder run with identical packages but with much cheaper parts, just to test there are no reflow issues, and then you have a cheaper physical sample to work from as a mechanical sample.
And then once assembled with real parts (triple checked and signed off at each stage) you'll do anything to salvage a board that has issues. So hand reworks are not uncommon.
Wasting countless multi-thousand dollar blank boards for all sorts of tasks and checks is commonplace.
If you find a design flaw, it's actually a cheaper and safer option to mod the boards than to go re-spin and re-qualify a board, so quite common to find mod wires and bodged chips and parts.
Extensive built in testing is half of the design process.
It can take a year or more to get one of these boards into production.

[rx8pilot]

Management has nerves of steel to commit to projects like this. To have that much time and effort packed into a single PCB (which is only part of the system). I would imagine that just defining the PCB qualification parameters is a massive job in itself. My only reference is in aerospace mechanics where the designs are certainly complicated, but the process of determining the tolerances allowed can be even more complicated. They have to balance performance specs with physical manufacturing limitations. For the work I did with NASA, we would see that when a part is impossible to manufacture, they would invent an all new machine, material, or process to make it. $$ values were as out of this world as the spacecraft they make.

LeCroy had to learn some special tricks to pull this off. So impressed with that level of dedication.

[Hugoneus]

Dave knows a lot about PCB manufacturing and qualification so I'll just echo his words. I also suspect that they have a way of testing the boards before the main ADCs are attached to it, to make sure all is good before the 4 THA and 8 ADCs are added.

[rx8pilot]

Dave knows a lot about PCB manufacturing and qualification so I'll just echo his words. I also suspect that they have a way of testing the boards before the main ADCs are attached to it, to make sure all is good before the 4 THA and 8 ADCs are added.

The custom chips are escorted by an elite security team and transported in an armored car. 

[electr_peter]

How many layers does ADC board have? I would guess at least 8-10.
Also, how did they produced that metal PCB frame? Casting or CNC only? Seems to be very big, but narrow part.

[EEVblog]

I also suspect that they have a way of testing the boards before the main ADCs are attached to it, to make sure all is good before the 4 THA and 8 ADCs are added.

Quite likely.
Wouldn't be hard (in the scheme of things) to build in a board level data connection (via pads or connector). Keeping signal integrity is the tricky part.
Alternatively dummy data (not real time of course) could be fed in via JTAG boundary scan.
In fact this board could have one kick-arse boundry scan system in place.

[electr_peter]

One thing I really like about such scope is that it is pushing the limits. It gets to 100 GHz, but just barely (almost all aspects are pushed to the limit, only extensive DSP corrections helps to keep it in one piece). I am sure a small amount of tweaking could bring another GHz or so for current model and future versions will be even better.

It is certainly not the case of having 200 GHz BW scope and configuring it via some jumpers to "entry range" 100 GHz BW version.

[Len]

I hope you took that optical receiver with you when you left! I'd love to see a teardown and explanation of that. How does it mix the lasers? How does it convert the resulting optical signal to an electrical signal?

[CosPhi]

Yes, wonder also a bit about how this laser stuff works

[Hugoneus]

The heterodyne conversion of the two lasers does not happen in the optical domain. It happens in the electrical domain. There are ways to mix light in a non-linear fashion directly in the optical domain, but that is not what is happening here.

The photo detector is only sensitive to the light signal intensity and it operates in square law detection mode. The two lasers illuminate the photo detector at the same time and the beat frequency is generated already in the electrical domain (constructive and destructive addition of light on the surface of the detector).

There are limitations however, for example the two lasers can't be orthogonal in polarization.