Something amazing! Keysight UXR 110GHz, 256GS/s, 10-bit Scope Teardown & Exp.

[Berni]

On these scopes you don't always just use the scope software in fullscreen and nothing else. Sometimes you also run just regular non touchscreen designed windows programs on it and the UI scaling feature in Win 7 are not always perfect.

For example when doing modulation analysis you tend to run this regular oldschool windows looking app on the scope:
https://www.keysight.com/zz/en/software/application-sw/89600-vsa-software.html

EDIT: Tho maybe they started going to Windows 10 now. The UI scaling is known to be much improved there.

  OMG, how is this even a thing.  It's Keysight's own source code.  It they want a higher resolution, even if they were using Windows95 for their OS, in their scope code, they would change the font and button size and ignore Window's scaling BS, or, they have already programmed their software correctly so that even WindowsXP would scale their on screen elements correctly.  Why are we questioning this trivial thing on a 1.6m$ scope?
They actually know what monitor size and res is mounted in the scopes...  It's not like a separate purchased program on a random consumer home computer with a random setup.

Yes but people also use software not made by Keysight on these scopes, there is built in integration with MATLAB and similar tools. Some might use things like LabView or some other programing environment to run automated tests that control other more "dumb" gear like switches or synthesizers. Also the scopes infiniium software can also be installed on a normal PC to use it for opening and analyzing files exported from the scope.

Also even tho the scope might look huge it is still the same 15 inch LCD that all the other Keysights PC based scopes like the 9000, 90000, S-series etc. When sitting from it at a distance that PC monitors tend to be used its not all that big of a screen.

But i do agree that it could have a at least a little bit more resolution. When sitting from these at a normal distance you can see the pixels when you look closely, something like 1600x1200 would be welcome to see but there is no point going above Full HD resolutions. My Agilent MSO6000 scope from 15 years ago is doing 1024x768 on a 6.3 inch display and it looks lovely crisp (They don't even bother to anti-alias the fonts because you couldn't see it anyway)

[David Hess]

According to the specs, It looks like the 2 channel version he looked at "only" requires 1370 VA and can run off 120V.  The 4 channel version only works off 200-240V

Common power outlets in the US are 1800 watts under the best conditions and more typically limited to about 1200 watts.  (1) Even if they had made a version which could run on a 120 volt 20 amp outlet, they are not any more common than a 240 volt outlet.

(1) The biggest "cheap" UPSes are not much larger than 1200 watts even with input power factor correction.

It blows my mind the extent to which the accuracy of something like this relies on both good RF design and some very clever DSP.  The whole time-interleaved sampling seems like it must rely heavily on the (DSP) software to calibrate out the subtle differences in all those signal paths.

Modern integrated pipelined ADCs commonly do self calibration.  DSOs going back decades which use interleaved converters also commonly do self calibration.  But the difference is that 4-way interleaving was the typical maximum.

[srce]

OMG, how is this even a thing.  It's Keysight's own source code.  It they want a higher resolution, even if they were using Windows95 for their OS, in their scope code, they would change the font and button size and ignore Window's scaling BS, or, they have already programmed their software correctly so that even WindowsXP would scale their on screen elements correctly.  Why are we questioning this trivial thing on a 1.6m$ scope?

They actually know what monitor size and res is mounted in the scopes...  It's not like a separate purchased program on a random consumer home computer with a random setup.

While I doubt it is a problem here, often the screen resolution of the graticule is made to be a fixed integer ratio of the digitizer resolution to prevent aliasing or the extra processing necessary to correct aliasing.  So for instance 200 counts from 8-bits might translate into 8 vertical divisions of exactly 25 pixels each and changing screen resolutions would produce a more smeared display.  Preserving this would require specific programming for different screen resolutions.

It looks fine whatever res you run it at - as mentioned, you can move windows on to a different monitor. Also, the same s/w can be run standalone on a regular Windows PC for offline analysis.

[tv84]

What was the scope used in the development phase of this one?  (chicken-egg...)

[mmagin]

What was the scope used in the development phase of this one?  (chicken-egg...)

The really interesting question: how do they calibrate that calibration probe.
They have to characterize the imperfections in it's output to a significantly better accuracy than this scope can measure.  Unless there's something new under the sun in calibration methodology?

[Mr. Scram]

As two minute papers guy says: “What a time to be alive!”

This is a staggering amount of work. I’m honestly surprised that they could make the numbers add up to justify this project. I hope they sell well. It really cements Keysights place at the top of the T&M pile!

Thanks for the video Sharihar. You have no peers on YouTube when it comes to high end electronics educational content.

It may very well be a loss leader. No amount of advertising can buy you the reputation this buys the company. That would be a bit like what Volkswagen was doing with the Bugatti Veyron. Losing a few million on each car is much cheaper than starting a racing team. I understand the Chiron is now making money, mostly due to a lot of knowledge and infrastructure already being present. I suspect Keysight's situation may be similar.

[Fungus]

That would be a bit like what Volkswagen was doing with the Bugatti Veyron. Losing a few million on each car is much cheaper than starting a racing team. I understand the Chiron is now making money, mostly due to a lot of knowledge and infrastructure already being present.

I'm sure that tripling the purchase price for the Chiron didn't hurt, either.

(or whatever the price hike was - a million $ extra per car will sure help recover some R&D)

[markietas]

According to the specs, It looks like the 2 channel version he looked at "only" requires 1370 VA and can run off 120V.  The 4 channel version only works off 200-240V

Common power outlets in the US are 1800 watts under the best conditions and more typically limited to about 1200 watts.  (1) Even if they had made a version which could run on a 120 volt 20 amp outlet, they are not any more common than a 240 volt outlet.

Pretty much every commercial building and most newer homes in the US use 20a 120v circuits for all of their outlets. Though they often are not nema 5-20r, but 5-15r witch doesn't have the sideways blade but is still rated to pass 20a of current.

Although anyone who can afford this scope can afford to get a dedicated outlet installed anyways.

[Mr. Scram]

I'm sure that tripling the purchase price for the Chiron didn't hurt, either.

(or whatever the price hike was - a million $ extra per car will sure help recover some R&D)

I do think that increasing the price on the Chiron helps covering the costs and making them profitable. Although even at that price they'd have lost millions on each Veyron. It's reported each one ended up costing about 8 or 9 million.

I'm curious Keysight is doing something similar, though I suspect they might want to keep that to themselves.

[Hugoneus]

A few other notes:

The scope has to go down to DC. This is often very important in optical applications.

The ADCs which have been fine tuned for this scope will be useful for many many years.

The fact that the same mainframe can be upgraded from 13GHz to 110GHz is very important for long term cost and return on investment.

[Bassman59]

As two minute papers guy says: “What a time to be alive!”

This is a staggering amount of work. I’m honestly surprised that they could make the numbers add up to justify this project. I hope they sell well. It really cements Keysights place at the top of the T&M pile!

Thanks for the video Sharihar. You have no peers on YouTube when it comes to high end electronics educational content.

It may very well be a loss leader. No amount of advertising can buy you the reputation this buys the company. That would be a bit like what Volkswagen was doing with the Bugatti Veyron. Losing a few million on each car is much cheaper than starting a racing team. I understand the Chiron is now making money, mostly due to a lot of knowledge and infrastructure already being present. I suspect Keysight's situation may be similar.

I doubt that Keysight is losing money on each of these oscilloscopes. Face it, the customers who need this product really need it, and they're very aware of what it costs to build something like this. They're happy that they can even buy this oscilloscope.

Remember, the cost to build this product isn't $1.3 million. I'd be surprised if it cost more than $500k. And remember this is just one product in a range, and a lot of the R&D for was used for that range.

[Berni]

Yep a lot of the tech is reused from the S series like the ADC, Aquisition ASIC, PCIe PC backhaul, Windows softwre etc...

They proabably sell a lot of those S series, so investment return is easy. The new tech in this is the front end and samplers along with the digital magic to make it woek together. Even that must have cost a fortune to develop

[srce]

Yep a lot of the tech is reused from the S series like the ADC

Isn't it faster? 64GSa/s vs 40GS/a?

[TheUnnamedNewbie]

Yep a lot of the tech is reused from the S series like the ADC

Isn't it faster? 64GSa/s vs 40GS/a?

It isn't 100% the same, it has been updated a bit - perhaps the process has been further fine-tunes allowing for better performance/faster transistors. At least, that is what I have been told.

[David Hess]

The really interesting question: how do they calibrate that calibration probe.
They have to characterize the imperfections in it's output to a significantly better accuracy than this scope can measure.  Unless there's something new under the sun in calibration methodology?

Keysight sends the pulse generator to NIST which calibrates it using an electro-optic sampling system which is considerably faster than a mere 110 GHz.  Or maybe Keysight has their own electro-optic sampling system.

Older (and slower) sampling oscilloscopes can actually calibrate themselves by measuring their own sampling strobe kickout so no external absolute reference is required but this fails above 10s of GHz where sampling gates become weird.

[mmagin]

The really interesting question: how do they calibrate that calibration probe.
They have to characterize the imperfections in it's output to a significantly better accuracy than this scope can measure.  Unless there's something new under the sun in calibration methodology?

Keysight sends the pulse generator to NIST which calibrates it using an electro-optic sampling system which is considerably faster than a mere 110 GHz.  Or maybe Keysight has their own electro-optic sampling system.

Huh, wow, thanks for the link.

[Hugoneus]

The really interesting question: how do they calibrate that calibration probe.
They have to characterize the imperfections in it's output to a significantly better accuracy than this scope can measure.  Unless there's something new under the sun in calibration methodology?

Keysight sends the pulse generator to NIST which calibrates it using an electro-optic sampling system which is considerably faster than a mere 110 GHz.  Or maybe Keysight has their own electro-optic sampling system.

Huh, wow, thanks for the link.

Also note that for characterizing things like ENOB, all you need is a clean synthesizer.

[snoopy]

If the last 25 years are anything to go by, 25 years from now this will be an entry level spec scope.

Yeh probably being made by Rigol

[Berni]

Don't think the sort of progress we seen recently can keep up for long. Eventually things get harder and harder to make.

A similar thing can be seen with PC CPUs. Things seamed pretty steady until we get past the first Pentium. At that point CPU frequencies started really going up and they kept iterating and improving on the process to squeeze more and more speed out of it. Consumer chips quickly broke the 1GHz mark and kept going while the power consumption of these new fast CPUs skyrocketed to 100W and beyond. They improved only a little bit MIPS per MHz but kept rapidly getting faster just cause they could crank the clock speed higher and higher. But then we hit the 3GHz mark and things suddenly slowed down a lot. At that point multicore started becoming a consumer thing because it was now easier to add a extra core than to make it faster.

Go forwards to today we now have consumer chips that get past 4,5 GHz but that's about it. They did improve the MIPS per MHz quite a bit as that was the new easiest way forward, but even that route slowing down lately. So again they are resorting to putting even more cores into CPUs. Similar thing is happening with how there FLOPS performance is pushed forward by simply adding more parallel computational units with the new AVX512 instructions.

So once a way to drastically improve performance is found it generally does not last forever, at some point the idea becomes unpractical and so you need to find a new trick to keep going.

That being said i am curious about what scopes we will get 25 years from now. Do we get a cheap under 1000 USD Rigol with 1GHz of bandwidth and a 12bit ADC?

[TheUnnamedNewbie]

That being said i am curious about what scopes we will get 25 years from now. Do we get a cheap under 1000 USD Rigol with 1GHz of bandwidth and a 12bit ADC?

I do think that is likely. From what I have seen, there has been a massive push towards high speed ADCs, with the need for higher bandwidths and newer applications such as SDR. Techniques for low-power, wide-band and low-cost amplifiers will also improve, giving us lower cost at those points too.

[Fungus]

That being said i am curious about what scopes we will get 25 years from now. Do we get a cheap under 1000 USD Rigol with 1GHz of bandwidth and a 12bit ADC?

Easily. Probably for $500.

The price for low-volume goods like these is mostly (R&D cost / projected sales volume)+markup.

In 25 years the "DSO ASIC" and the "analog input ASIC" will be done and dusted, R&D will be close to zero.

Edit: Having said that, the price of the probes fro 1GHz might be more than the 'scope. 250Mhz might be the entry level with a price jump above that for active probes.

[David Hess]

That being said i am curious about what scopes we will get 25 years from now. Do we get a cheap under 1000 USD Rigol with 1GHz of bandwidth and a 12bit ADC?

I doubt it because there is more to increasing resolution past 8 bits than the ADC.  Most 12 bit DSOs do not have 12 bit performance in their analog signal conditioning.  Try measuring their linearity across their bandwidth and their 12 bit settling time.

This is reflected in the ENOB (effective number of bits) specification given at different frequencies in the oscilloscope's specifications.  If this specification is not given in detail, then you can know that the DSO performs poorly in this regard.

Now *sampling* oscilloscopes hit 10 and 12 bit performance or even better decades ago but they have no analog signal conditioning.

[Fungus]

I doubt it because there is more to increasing resolution past 8 bits than the ADC.  Most 12 bit DSOs do not have 12 bit performance in their analog signal conditioning.  Try measuring their linearity across their bandwidth and their 12 bit settling time.

That can all be calibrated out in a DSP.

[Scratch.HTF]

I've read that power companies in North America are required to provide split phase 240V.

[Bassman59]

I've read that power companies in North America are required to provide split phase 240V.

That’s what is brought from the utility pole to a residence, two 120 V hots 180º apart and a neutral. The panel will have a two-pole main breaker feeding the two phases in the panel. Both 120 V and 240 V circuits can be taken from the panel.