Keysight announced new 110 GHz scopes

[TheUnnamedNewbie]

I can't seem to find anyone else mentioning this, but it seems like Keysight have pushed LeCroy off the throne when it comes to bandwidth with the recently (July, I think?) announced UXR real-time scopes, going up to 110 GHz. Perhaps even more impressive is the fact that they are claiming 10 bit ADCs (will have to see what the actual ENOB is at 110 GHz...). Also, 4 channels at 110 GHz (I think? I don't seem to find any reference of it being similar to the older 65 GHz scopes where you had 4 channels at 30 GHz or could use 2 at 60?)

Pretty cool, I will take two, thank-you-very-much.

Link to page

[rsjsouza]

Pretty cool. Unless it is a typo, though, I see something strange:

Achieve next-generation technology breakthroughs with 80, 100, and 110GHz or real-time bandwidth

Hopefully it is a typo. Otherwise, this will be much less impressive.

This oscilloscope probably caters to the automotive low-range radar crowd that operates in the 70~80GHz - big business there.

[Faith]

Achieve next-generation technology break throughs with 80, 100, and 110GHz of real-time bandwidth

Says “of” for me...

[rsjsouza]

Haha... I guess they just fixed it. I copied and pasted from their page earlier today, just to avoid any typos.

[Wolfgang]

The first scope I ever saw with a 2.6kW power requirement ! 
It makes nices pictures and heats your lab 
People with a heavy TEA syndrome must have one ! 

[Zucca]

How many of those/year will thay sell?

My guess 10 units a year.

 

[iMo]

I would rather buy a Bugatti Chiron..

[Wolfgang]

Or a decent helicopter ! 
My favourite: Airbus H125

[bd139]

I would rather buy a Bugatti Chiron..

Would be cheaper!

[BrianHG]

A new toy! Yippee... 
Now all we need is for some impossible circumstance where an engineer on this forum somehow ends up with the top 4 channel one and questions us on how to connect and measure the AC mains with it... 

[iMo]

I would rather buy a Bugatti Chiron..

Would be cheaper!

Ok, you are right, the Veyron will be cheaper.. And I'll save some money and buy a DS7054..

[EEVblog]

110 GHz of real-time oscilloscope bandwidth
High definition 10 bit analog-to-digital converter (ADC)
The industry’s highest ENOB at bandwidths up to 110 GHz
Industry’s lowest noise floor with less than 1 mVrms of noise
Industry’s lowest jitter with less than 25 fsrms of intrinsic jitter and less than 35 fsrms of inter-channel jitter
Unrestricted 256 GSa/s per channel sample rate – with a real-time resolution of 3.09 ps per sample

Up to 100x faster performance for some measurements – enabled by a powerful
new measurement acceleration ASIC and memory controller capable of
5 trillion integer operations per second (IOPS)

Straight out of Area 51

[cheeseit]

You think Dave's gonna find one in the dumpster..? That would be an awesome teardown!

[iMo]

Do you think the knocking at the signal input connector will cause a glitch in the trace?

[DaJMasta]

I think knocking at the input may send your front end entirely out of alignment.  Anything short of a gentle touch, kid gloves, in a clean room probably voids the calibration 

[rsjsouza]

110 GHz of real-time oscilloscope bandwidth
High definition 10 bit analog-to-digital converter (ADC)
The industry’s highest ENOB at bandwidths up to 110 GHz
Industry’s lowest noise floor with less than 1 mVrms of noise
Industry’s lowest jitter with less than 25 fsrms of intrinsic jitter and less than 35 fs rms of inter-channel jitter
Unrestricted 256 GSa/s per channel sample rate – with a real-time resolution of 3.09 ps per sample

Up to 100x faster performance for some measurements – enabled by a powerful
new measurement acceleration ASIC and memory controller capable of
5 trillion integer operations per second (IOPS)

Straight out of Area 51

No kidding. At maximum frequency, I wonder how much voltage the inputs can withstand...

You think Dave's gonna find one in the dumpster..? That would be an awesome teardown!

Probably not, but Shahriar on the other hand...

Do you think the knocking at the signal input connector will cause a glitch in the trace?

I suspect that breathing close to the input connector will cause a glitch. 

[Tomorokoshi]

The first scope I ever saw with a 2.6kW power requirement ! 
It makes nices pictures and heats your lab 
People with a heavy TEA syndrome must have one ! 

And it weighs in at 42 kg! Welcome to a new generation of boat anchors!

[mikeselectricstuff]

I think knocking at the input may send your front end entirely out of alignment.  Anything short of a gentle touch, kid gloves, in a clean room probably voids the calibration 

It comes with a torque wrench for tightening the connectors

[bd139]

The torque wrench probably costs more than a DS1054Z as well 

[StillTrying]

Perhaps even more impressive is the fact that they are claiming 10 bit ADCs (will have to see what the actual ENOB is at 110 GHz...).

The preliminarily data sheet says the average ENOB from DC to 110GHz is 4.5, as it's 10 at one end it might be only 2 at the other.

[BrianHG]

110 GHz of real-time oscilloscope bandwidth
High definition 10 bit analog-to-digital converter (ADC)
The industry’s highest ENOB at bandwidths up to 110 GHz
Industry’s lowest noise floor with less than 1 mVrms of noise
Industry’s lowest jitter with less than 25 fsrms of intrinsic jitter and less than 35 fs rms of inter-channel jitter
Unrestricted 256 GSa/s per channel sample rate – with a real-time resolution of 3.09 ps per sample

Up to 100x faster performance for some measurements – enabled by a powerful
new measurement acceleration ASIC and memory controller capable of
5 trillion integer operations per second (IOPS)

Straight out of Area 51

No kidding. At maximum frequency, I wonder how much voltage the inputs can withstand...

You think Dave's gonna find one in the dumpster..? That would be an awesome teardown!

Probably not, but Shahriar on the other hand...

Do you think the knocking at the signal input connector will cause a glitch in the trace?

I suspect that breathing close to the input connector will cause a glitch. 

How do they ship this thing and expect it to maintain calibration?

[DaJMasta]

It comes with a torque wrench for tightening the connectors

Well yeah, they wouldn't be able to sell it outside of an installed lab environment if the requirements for the front panel were that sensitive... and they probably wouldn't bother putting controls for hand use in the same place as any measurement port that sensitive.


You can be damn sure a little finger oil on the center pin if your 1mm connector will screw with performance at 100GHz, though 

[exe]

I wonder how it compares with, e.g., 1054z or sds1202x-e. Can it do SPI decoding? Bode plot? Does it have touch interface and, if so, can I draw with my finger? Can I control it from my smartphone or from browser? 

[ruffy91]

For 10% extra $ you can get someone designing for you a smartphone which can control it.

[mikeselectricstuff]

It comes with a torque wrench for tightening the connectors

Well yeah, they wouldn't be able to sell it outside of an installed lab environment if the requirements for the front panel were that sensitive... and they probably wouldn't bother putting controls for hand use in the same place as any measurement port that sensitive.


You can be damn sure a little finger oil on the center pin if your 1mm connector will screw with performance at 100GHz, though 

At the price of this I'm surprised they don't have motorised connetcors that automatically clamp to the correct torque.

[mikeselectricstuff]

The first scope I ever saw with a 2.6kW power requirement ! 

They should have put a 3-phase supply on it just for bragging rights

[Muxr]

I wonder how it compares with, e.g., 1054z or sds1202x-e. Can it do SPI decoding? Bode plot? Does it have touch interface and, if so, can I draw with my finger? Can I control it from my smartphone or from browser? 

These top end scopes probably make for terrible day to day electronic shop scopes. The front end is likely also too sensitive and easy to damage while picking up all sorts of wide band noise you don't need to pick up for simple measurements. They are often used in conjunction with other instruments (and probes) which provide perfectly impedance matched test outputs. Not really meant to be used with passive probes.

This is a LeCroy 100Ghz scope doing measurements: https://youtu.be/U3w_EWgGQuk?t=4177

[Dubbie]

I am looking forward to the episode where Dave takes this canyoning. I couldn’t see anything in the data sheet about IP67, but surely nobody would release a new scope these days that wasn’t at LEAST IP54.

[rx8pilot]

Scope Month!

Make a video letting everyone know how you would use 110Ghz of real-time BW!!!

[Keysight DanielBogdanoff]

Perhaps even more impressive is the fact that they are claiming 10 bit ADCs (will have to see what the actual ENOB is at 110 GHz...).

The preliminarily data sheet says the average ENOB from DC to 110GHz is 4.5, as it's 10 at one end it might be only 2 at the other.

It's pretty good across the whole bandwidth.

I wonder how it compares with, e.g., 1054z or sds1202x-e. Can it do SPI decoding? Bode plot? Does it have touch interface and, if so, can I draw with my finger? Can I control it from my smartphone or from browser? 

Let's just say it's not really in the same price bracket by a few orders of magnitude.

And it weighs in at 42 kg! Welcome to a new generation of boat anchors!

Yeah, it's a pain getting them out of transit cases.

How do they ship this thing and expect it to maintain calibration?

Refrigerated, air-cushioned mule train.

You think Dave's gonna find one in the dumpster..? That would be an awesome teardown!

Probably not, but Shahriar on the other hand...

Nothing's final, but we are talking with Shahriar and will explore other channel options down the road .

[Keysight DanielBogdanoff]

Pretty cool, I will take two, thank-you-very-much.

Here you go -> 

[Tomorokoshi]

Pretty cool, I will take two, thank-you-very-much.

Here you go -> 

Daniel, thanks for taking our snark in stride.

[Wolfgang]

Hi,

I am curious if this machine works by the same operating principle as the LeCroy 100GHz.

You probably have seen Shariars video about it. The LeCroy does it by splitting the whole input frequency range in 3 ranges of 33GHz, mixing the upper two ranges down to baseband, digitizing them and then by combining the baseband signals in a huge proprietary FPGA and a ton of software.
This works, but creates huge problems equilibrating group delay at the filter band edges, eliminating artefacts, ...

- does the Keysight work the same way ?
- how did you verify your spec values (risetime, flatness, groupd delay, ...) ? (I asked the same question to LeCroy and they did not know)
- I guess the price range is about 1M€, am I approximately right ?
- whats the real-time bandwidth of a single 10Bit ADC frontend ? My guess: 33GHz

Best regards
  Wolfgang

[Keysight DanielBogdanoff]

Hi Wolfgang,

Great questions!

- does the Keysight work the same way ?
We do NOT do frequency interleaving like LeCroy or Tektronix (or even our Infiniium Z-Series scopes). We have custom ASICs that we've designed specifically for 110 GHz operation. Each channel gets its own front-end module and acquisition board to support 110 GHz on all 4 channels at all times.

- how did you verify your spec values (risetime, flatness, groupd delay, ...)
We're following up with R&D

- I guess the price range is about 1M€, am I approximately right ?
It's not cheap. We are happy to discuss pricing with potential buyers

- whats the real-time bandwidth of a single 10Bit ADC frontend ? My guess: 33GHz
110 GHz! It's the real deal.

[rx8pilot]

Daniel -

What industries, products, or technologies do you suppose will be the most interested in such an exotic achievement?

What type of strategies are involved in getting the signals from the DUT to the front-end of the system? As difficult as it may be to develop the front-end and ADC's - the probing must be something rather special as well.

This may be the missing link to make the Batterizer achieve its goals of creating free energy.

[Lukas]

A google image search found this article http://eetimes.jp/ee/articles/1807/13/news034.html containing some interesting slides of a press conference. Unfortunately I haven't been able to find more slides of that presentation. Maybe Daniel can help us out?

[Wolfgang]

Hi Wolfgang,

Great questions!

- does the Keysight work the same way ?
We do NOT do frequency interleaving like LeCroy or Tektronix (or even our Infiniium Z-Series scopes). We have custom ASICs that we've designed specifically for 110 GHz operation. Each channel gets its own front-end module and acquisition board to support 110 GHz on all 4 channels at all times.

- how did you verify your spec values (risetime, flatness, groupd delay, ...)
We're following up with R&D

- I guess the price range is about 1M€, am I approximately right ?
It's not cheap. We are happy to discuss pricing with potential buyers

- whats the real-time bandwidth of a single 10Bit ADC frontend ? My guess: 33GHz
110 GHz! It's the real deal.

Chapeau - 110 *real* GHz without interleaving is good.

Your answer how you verify performance is a bit too general for me. Of course you do your R&D. If I am to spend about a million, I would like to know in detail why I should trust the specs. Do you have any literature recommendations so I can educate myself ?

About buying one - I am a poor radio amateur (DL1DWG), suffering from a partial TEA syndrome (see my website). This machine is fascinating,
but a bit above my price range, I am afraid.

It seems that the high resolution of 10Bits does not help too much at the high end. 4.5Bits of ENOB may be a good value compared to competitors, but it is not really much in absolute terms. I presume that the ADCs use a time-staggered interleaved architecture with several slower ADCs, am I right ? This method works, but has jitter and clock sync issues, as usual.
Is this really better than a single ADC with less bits and higher speed ?

Best regards
  Wolfgang

[Keysight DanielBogdanoff]

@rx8pilot,

We've seen a lot of interest for this in research around optical and communications technologies.

Honestly, a noise floor this low and a BW this high is brand new, so we expect to find new opportunities we didn't even know about. I know we experienced that when we released the 63 GHz scope.

Probing-wise, I believe it will be 99% direct cable connections. Boards will be designed to handle this connection appropriately. This scope will be compatible with existing Infiniium-compatible probes, too.

I think the Batterizer people should buy one .

[Keysight DanielBogdanoff]

Your answer how you verify performance is a bit too general for me. Of course you do your R&D. If I am to spend about a million, I would like to know in detail why I should trust the specs. Do you have any literature recommendations so I can educate myself ?

When we have a warranted spec we're legally obligated to meet it. If I had to guess, we test it with super fancy test gear. For example, bandwidth is conceptually an easy one. Simply input a sweep frequency with a known amplitude and measure the response. Something like noise can require a bit more statistics/confidence calculations, but is basically the same. Input a signal with known characteristics, then measure the output. Sometimes you just need really high-end gear to do that.

It seems that the high resolution of 10Bits does not help too much at the high end. 4.5Bits of ENOB may be a good value compared to competitors, but it is not really much in absolute terms. I presume that the ADCs use a time-staggered interleaved architecture with several slower ADCs, am I right ? This method works, but has jitter and clock sync issues, as usual.
Is this really better than a single ADC with less bits and higher speed ?

Find me a good 256 GSa/s ADC and we'll go that route!  We do use time interleaving. I'm not super well versed in the high end scopes, but I know that the UXR's noise floor is exceptionally low. I'm not sure what discussions were had around putting that number in the data sheet, but this is a custom ADC setup that performs really well. You might be interested in the podcast interview I did a while back with one of our ADC R&D folks: https://eestalktech.com/fast-adc/

At the end of the day, the requirements for an oscilloscope ADC tend to be much more strict than for other ADC applications.

[rx8pilot]

Find me a good 256 GSa/s ADC and we'll go that route! 

Checking DigiKey now to see what they have in stock.........

[Wolfgang]

Hi,

another application the LeCroy people told me about are anti-collision and short-range surveillance pulsed radars in the 70-100GHz range.
Maybe thats a market for Keysight, too.

I would gess that at 110GHz there are *no* probes.  Everything must be via very short, very good, very expensive, still very lossy cables.

If you sell this to the Batteriser people, you could suggest that they use a waveguide input.

[mikeselectricstuff]

@rx8pilot,

We've seen a lot of interest for this in research around optical and communications technologies.

Honestly, a noise floor this low and a BW this high is brand new, so we expect to find new opportunities we didn't even know about. I know we experienced that when we released the 63 GHz scope.

Probing-wise, I believe it will be 99% direct cable connections. Boards will be designed to handle this connection appropriately. This scope will be compatible with existing Infiniium-compatible probes, too.

I think the Batterizer people should buy one .

Maybe hit up Energeous - they've struck out so far finding a frequency they can legally transmit enough RF to charge a phone, maybe they should be looking upwards...

[oPossum]

.

[amyk]

The input connector alone on one of those costs several times more than an entire Rigol DS1054...

[Andrey_irk]

The input connector alone on one of those costs several times more than an entire Rigol DS1054...

I suspect they are hand picking them as it is hard to get decent yield while maintaining low geometrical tolerances.
Daniel,
Do you have your own facility that manufactures these 1mm connectors or you're buying them somewhere?

[Dubbie]

I think that is the highest tech board I have ever seen in my life.

I doubt things get more high tech than this unless you are behind a stack of NDAs as high as your knees.

[bson]

It's actually doing real-time acquisition instead of reconstructing the signal from a frequency domain acquisition?  That's... astonishing.  Wow.  Just wow.  ASIC or not.  Just wow.

[rx8pilot]

The numbers are so ridiculous that they look like typos.....

'320 lanes - 8Gb/s each'

[Berni]

I wonder how it compares with, e.g., 1054z or sds1202x-e. Can it do SPI decoding? Bode plot? Does it have touch interface and, if so, can I draw with my finger? Can I control it from my smartphone or from browser? 

Actually you can do all of that since its a Windows PC based scope and they use the same PC side software on all infiniium scopes.

[Berni]

Probably not, but Shahriar on the other hand...

Nothing's final, but we are talking with Shahriar and will explore other channel options down the road .

Heck yes, he is certainly the right youtuber to demonstrate such a high end piece of kit. Hope we also get to see some of the insides like in the previous ludicrous GHz scope.

[JPortici]

@daniel

I wonder if they guys at Texas Instruments would be interested in making a couple of videos?
They've put out incredible integrated chips for ADAS in the past two years so they certainly have the goods available for the public to demonstrate, and they recently started increasing their youtube presence.

Also because they replaced Lecroys with Tektronix. Time for another change 

[TheUnnamedNewbie]

- does the Keysight work the same way ?
We do NOT do frequency interleaving like LeCroy or Tektronix (or even our Infiniium Z-Series scopes). We have custom ASICs that we've designed specifically for 110 GHz operation. Each channel gets its own front-end module and acquisition board to support 110 GHz on all 4 channels at all times.

I had to just sit and stare at that for a bit. This is incredible. And I imagine as you don't need to deal with frequency interleaving things, there will be less issues with noise due to analog front-end prepossessing and also better stability and repeatability?
I believe we have 1 or 2 of the 63 GHz series scopes in our labs and they really are workhorses for what we do.. I can already imagine the anxiety this scope brings to our lab manager, who will now have a barrage of researches asking him to get one of these. Will these be shown at the 'Millimeter-wave measurement insights' hotspots in Europe?

Daniel, what kind of cable calibration options do these have? Even with just a few tens of centimeters of cable, the attenuation at 110 GHz will be quite significant - Any 'easy' compensation solutions?
Part of me was actually expecting the next generation of scopes to have solutions using external measurement heads, similar to the PNA MM-wave network analyzer systems - doing some kind of time-interleave sampling head that outputs a number of time-interleaved channels into the output?

How many of those/year will thay sell?

My guess 10 units a year.

 

Now, I can't speak for Keysight, but I think you would be surprised at how 'much' of these they sell. Sure, this wont be something mass-manufactured, but I think they will sell quite a few more than 10 a year. People say the same about the 63 GHz scopes or the top-of-the-line VNAs, and I already know about a hand-full of those just in my proximity (ofcourse, the fact that I am working in a research park that has about 1500 EE researchers helps...)

[Zucca]

few years ago at KS RnD:

- "Let's design a new scope, something cool"
- "Oh yes boss can't wait"
- "Bandwidth?"
- "Let's do more than 100GHz!"
- "Yeahhh" clapping and cheering, girl dancing.
- (smart EE guy, fat, smells but not too much, big coke in the right hand, thick glasses with flux on the lens, dirty hair, long beard, tshirt with ketchup stain I am an engeneer.. I am good with math)
   "We can do easly get 163,9GHz at -3dB, I know how"
- (slim guy, tie, smells good, PRADA titanium glasses, Iphone, Ipad, Iwatch, Macpro, rolex on the right hand, hair wax, perfect simmetric sideburns)
   "Then we release a 110GHz machine to get the market, then we could sell the upgrade to 150GHz"
- (smart EE guy, thinks:"Here we go again..., why not 105GHz, and sell the 110GHz and then the 120.. 130 ..140 .150 Ghz... bullss.."
- "Any other questions?"
- People sucking pens and looking to the room corners.
- "Have a great day everyone!"

Why 110GHz ?

[TheUnnamedNewbie]

Why 110GHz ?

Because that is the upper limit of the 1 mm connector they use. To go higher, they would have to go for different connectors, and currently the only coaxial connector available is Anritsu's .8 mm connector which goes to I think 145 GHz. Any more than that you would need to go to waveguides, but no waveguide can go down to DC, and metal waveguides are quite narrow-band (compared to what you need from a scope).

[cowasaki]

I'm having trouble with an Arduino blinking an LED, will this show me the waveform?

[Zucca]

Yes, if you are driving the LED at 110GHz

[cowasaki]

Yes, if you are driving the LED at 110GHz

I'll have to check the sketch to make sure it isn't higher than 110GHz

[iMo]

Red LED is 475THz. No luck with that gear.

[Berni]

few years ago at KS RnD:

- "Let's design a new scope, something cool"
- "Oh yes boss can't wait"
- "Bandwidth?"
- "Let's do more than 100GHz!"
- "Yeahhh" clapping and cheering, girl dancing.
- (smart EE guy, fat, smells but not too much, big coke in the right hand, thick glasses with flux on the lens, dirty hair, long beard, tshirt with ketchup stain I am an engeneer.. I am good with math)
   "We can do easly get 163,9GHz at -3dB, I know how"
- (slim guy, tie, smells good, PRADA titanium glasses, Iphone, Ipad, Iwatch, Macpro, rolex on the right hand, hair wax, perfect simmetric sideburns)
   "Then we release a 110GHz machine to get the market, then we could sell the upgrade to 150GHz"
- (smart EE guy, thinks:"Here we go again..., why not 105GHz, and sell the 110GHz and then the 120.. 130 ..140 .150 Ghz... bullss.."
- "Any other questions?"
- People sucking pens and looking to the room corners.
- "Have a great day everyone!"

Why 110GHz ?

Why?

Because LeCroy made a 100GHz scope so 110 > 100. It is a competitive market after all. He did actually use his Mac Pro to do some market research.

[EEVblog]

- I guess the price range is about 1M€, am I approximately right ?
It's not cheap. We are happy to discuss pricing with potential buyers

Use the EEVblog discount code "Crikey" for $100k off 

[Kalvin]

- I guess the price range is about 1M€, am I approximately right ?
It's not cheap. We are happy to discuss pricing with potential buyers

Use the EEVblog discount code "Crikey" for $100k off 

Do not forget to mention that buying it from Amazon will give you a nice 7% commission!

[EEVblog]

Details on how it works:
https://www.mwrf.com/analog-semiconductors/inp-ic-technology-powers-instruments-past-100-ghz

[EEVblog]

Nothing's final, but we are talking with Shahriar and will explore other channel options down the road .

Shahriar (by way of his RF voodoo group at Bell Labs) is literally one of the few hundred people in the world who would actually need to buy it

[TheAmmoniacal]

Shahriar (Hugoneus) don't like mixing work and YouTube, but it's time to get one of your research boards/fixtures and bring it over to Keysight for probing and demonstration!

[iMo]

..Now, I can't speak for Keysight, but I think you would be surprised at how 'much' of these they sell..

Imagine - they sell 50pcs and they break even

[EEVblog]

..Now, I can't speak for Keysight, but I think you would be surprised at how 'much' of these they sell..

Imagine - they sell 50pcs and they break even

That's nothing.
I worked in the military market were 2pcs was the break even and the remaining 3pcs was profit 

[BravoV]

The real question is, when this will ended up at ebay like other boat anchor scope I got. 

[EEVblog]

Shahriar (Hugoneus) don't like mixing work and YouTube, but it's time to get one of your research boards/fixtures and bring it over to Keysight for probing and demonstration!

I thought he once bought home a scope from work that was on demo loan to work, and the scope cost more than his house?

[iMo]

..Now, I can't speak for Keysight, but I think you would be surprised at how 'much' of these they sell..

Imagine - they sell 50pcs and they break even

That's nothing.
I worked in the military market were 2pcs was the break even and the remaining 3pcs was profit 

And all 5 pcs then destroyed during a day at Woomera

[TheUnnamedNewbie]

Details on how it works:
https://www.mwrf.com/analog-semiconductors/inp-ic-technology-powers-instruments-past-100-ghz

To be honest, this doesn't say that much about what they are doing to make it work. The main reason for people to go to InP is power and noise. Even CMOS does 110 GHz and above, but the issue is the noise and the low breakdown voltages, which means high currents and thus matching is hard and needs band-limited networks or other wizardry. When designing this scope, I'm quite confident the hard part is not designing an amplifier at 110 GHz. It is designing one that works at 110 GHz all the way down to DC. Oh, and that phase noise on the SAH will involve a lot of magic.

Nothing's final, but we are talking with Shahriar and will explore other channel options down the road .

Shahriar (by way of his RF voodoo group at Bell Labs) is literally one of the few hundred people in the world who would actually need to buy it

I think the guys at the optical groups are going to be more in need of such a device. Ofcourse, only Shahriar himself can comment on this, but I think most of what his group does involves band-limited stuff, and thus a lot of VNAs and Specans, but no scopes.

[borjam]

Find me a good 256 GSa/s ADC and we'll go that route! 

Checking DigiKey now to see what they have in stock.........

I've got a bunch of SN74LS256000000012N here. Want to borrow a couple_ 

[iMo]

..To be honest, this doesn't say that much about what they are doing to make it work. The main reason for people to go to InP is power and noise. Even CMOS does 110 GHz and above, but the issue is the noise and the low breakdown voltages, which means high currents and thus matching is hard and needs band-limited networks or other wizardry..

My guess those InP stuff work at 150-200degC at least in that front-ends. I doubt a Si CMOS would be happy with that..

[Wolfgang]

There is a markeiting reason, too: You need to top LeCroy which has 100GHz.

[filssavi]

My guess those InP stuff work at 150-200degC at least in that front-ends. I doubt a Si CMOS would be happy with that..

I  don't think the InP was chosen for high temperature operation (or better still I think that wasn't even in the minds of the designers), for a couple of reasons, first and foremost the bandgap of InP isn't that much higher than silicon's one (1.34eV vs 1.12eV) so I wound't expect it to be much better (SiC and GaN with a bandgap of between 3 and 4 eV are) also at high temperature semiconductors properties degrade significantly so even if you could operate there you definitely wouldn't do it (noise increases a lot with temperature, electron mobility decreases, depending on the structure of the devices losses might go up)

[TheUnnamedNewbie]

..To be honest, this doesn't say that much about what they are doing to make it work. The main reason for people to go to InP is power and noise. Even CMOS does 110 GHz and above, but the issue is the noise and the low breakdown voltages, which means high currents and thus matching is hard and needs band-limited networks or other wizardry..

My guess those InP stuff work at 150-200degC at least in that front-ends. I doubt a Si CMOS would be happy with that..

Why would they run at that temperature? The main reason is as I stated noise, and the fact that the higher mobility in InP makes it easier to design high-gain stages at high frequencies.

[Wolfgang]

Its not that they want to run at such high temperatures, they have to because all the amp stages need to be run at high currents to achieve the desired bandwidth. Its the same reason why high-speed op-amps run hot. You need  lot of current to charge and discharge your parasitic capacitances (e.g., Miller) in a very short time, and that means gas guzzling.

[Zucca]

There is a markeiting reason, too: You need to top LeCroy which has 100GHz.

I wish LeCrap will come out with a 120GHz tomorrow.

[Wolfgang]

I think that 120GHz is less needed than a few bits more of ENOB. 4.5 Bits is about 20 discernible levels of voltage. Maybe enough for an eye diagram but not what you would normally expect from a scope for a million bucks.

One more thought about DUT connections - Even very short 1mm 110GHz cables have attenuations of a few dB. And there phase is not really immune to shifting them around on a table. Phase and attenuation are heaviliy dependent on frequency, on top of that. So, pulse fidelity is not really what you want.

Is there a "VNA mode" to normalize these effects out of your measurements ?

[Eric_S]

I'm having trouble with an Arduino blinking an LED, will this show me the waveform?

Just remember to plug your old passive BNCs straight in, and you'll be employee of the year in no time.

[iMo]

Its not that they want to run at such high temperatures, they have to because all the amp stages need to be run at high currents to achieve the desired bandwidth..

That is the reason, sure (mind these circuits use the bipolar transistors which heat usually much more than the cmos do). When one carefully reads through Dave's link above (I did), you may find the picture N.2 showing with 150degC (they mention 275degC too) your new oscilloscope will work for some time without a need for replacing the burned parts. Also mind the few kilowatts of power you need to pay for, in order to see those nice pictures on the screen

[Berni]

I think that 120GHz is less needed than a few bits more of ENOB. 4.5 Bits is about 20 discernible levels of voltage. Maybe enough for an eye diagram but not what you would normally expect from a scope for a million bucks.

One more thought about DUT connections - Even very short 1mm 110GHz cables have attenuations of a few dB. And there phase is not really immune to shifting them around on a table. Phase and attenuation are heaviliy dependent on frequency, on top of that. So, pulse fidelity is not really what you want.

Is there a "VNA mode" to normalize these effects out of your measurements ?

The software for there PC scopes does have what they call PrecisionProbe and InfiniiSim that can help you measure the response of your probing setup and compensate it out, or it can be fed in existing S parameters or RLC equivalent models to compensate out. On some scopes this is also hardware accelerated to reduce the speed impact of the extra computational work.

I'm pretty sure these are expensive optional features (as its usual for these high end scopes)

[Wolfgang]

... or they hope you will buy an additional 110GHz VNA in order to calibrate your setup ...

[TiN]

Now, let's go other end of the spectrum and get that Keysight 3459B SuperDMM (C) for a refresh! 5 Hz bandwidth in max resolution mode is enough 

[Berni]

Now, let's go other end of the spectrum and get that Keysight 3459B SuperDMM (C) for a refresh! 5 Hz bandwidth in max resolution mode is enough 

So have you collected enough those 3458A and need a new challenge?

[Keysight DanielBogdanoff]

To answer some questions:

Daniel, what kind of cable calibration options do these have? Even with just a few tens of centimeters of cable, the attenuation at 110 GHz will be quite significant - Any 'easy' compensation solutions?

I believe there is a custom calibration board that helps with the cable/probe cal.

..Now, I can't speak for Keysight, but I think you would be surprised at how 'much' of these they sell..

Imagine - they sell 50pcs and they break even

That's nothing.
I worked in the military market were 2pcs was the break even and the remaining 3pcs was profit 

This was an astoundingly long-term and expensive R&D project, so it'll take a bit to break even. You probably would be surprised by how many 63 GHz scopes we sell. It's not a huge number, but it's not small, either.

I'm having trouble with an Arduino blinking an LED, will this show me the waveform?

In exquisite detail. You'll probably want to implement a bandwidth filter to knock down the extra 100 GHz of 1/f noise, though.

few years ago at KS RnD:

Just a few years!?!

Why 110GHz ?

The 1mm connnector does max out at 110 or so. But, there's also some art and uncertainty to this level of technology. You shoot for targets and sometimes land above/below them. The ASICs were obviously designed to handle these speeds, but 110 GHz is where we landed .

[Eric_S]

Speaking of things with silly frequencies. I guess one these could be useful for 5g research. IIRC the FCC has opened up a spectrum in the 64-71GHz band.

[gkeeth]

Daniel,

Out of curiosity, can you explain the reason for the altitude spec, in particular the "non-operating" spec? (I'm assuming the operating spec is for cooling reasons, especially with the temperature derating).

[Tomorokoshi]

The 1mm connnector does max out at 110 or so. But, there's also some art and uncertainty to this level of technology. You shoot for targets and sometimes land above/below them. The ASICs were obviously designed to handle these speeds, but 110 GHz is where we landed .

As Mr. Tufnel would say, "This one goes to 11".

[Marco]

How do these super high frequency ADCs work in general? Are they just scaled down variants of the LLNL sampler at this point, two buffered transmission lines with a bunch of sampling gates, ping ponging between triggering the gates on one line and the other? Or maybe buffered sampling gates, each with a CCD like bucket brigade to queue up some samples for slower ADCs?

Anyone want to link some patents, I'm too lazy to search

[Wolfgang]

When it comes to top-end machinery like this, the air gets very thin.

Your market is:
1) Communication tech (ultra high speed networking, electro-optical stuff, ultra high speed busses ...)
2) Short range radar and object recognition
3) Military (they alwas have some awkward use for top gear, whatever it may be)
4) R&D, e.g. high speed physics, ..., some prestigeous universities, ...
5) The people that would have bought a Bugatti or Helicopter otherwise
My guess is around 1000pcs in 3 years, maybe thats optimistic.

I cannot even think of fabricating the DUTs you need such a machine for, why buy it ? You would probably leave the rocket science to
some key manufacturers and buy their final products or ready-to-use modules from China after a year or two. Why make what you can
buy, especially if all the testing is so expensive. You need a very good reason to do that.

Another guess:
- In the first two years, software and hardware will be a buggy desaster
- Competition will follow up within a year or two, maybe based on band-splitting technology
- Next big step is 145GHz (limit of 0.8mm connector) by LeCroy, Tek or Anritsu
- Prices of 1mm equipment (cables, connectors, adapters, ...) will come down a lot in the next few years due to larger production volume
- Batteroo will ask for a 1mm to banana plug adapter

[Keysight DanielBogdanoff]

Altitude seems to be a consumer-safety spec. Basically, this means it won't blow up unexpectedly if you take it up a mountain or onto a plane. My guess on the temp derating is that it's a heat transfer concern with thinner air, but I'm digging into it.

[Wolfgang]

Hi Dave, maybe a watercooled oscilloscope would be a solution ?! It would be an industry first, am I right ?
And their would be no altitude limit. 

[BrianHG]

Hi Dave, maybe a watercooled oscilloscope would be a solution ?! It would be an industry first, am I right ?
And their would be no altitude limit. 

The radiator will still need to work effectively in the thinner atmosphere.  So, there would still be an altitude limit.

[Keysight DanielBogdanoff]

When it comes to top-end machinery like this, the air gets very thin.

Your market is:
1) Communication tech (ultra high speed networking, electro-optical stuff, ultra high speed busses ...)
2) Short range radar and object recognition
3) Military (they alwas have some awkward use for top gear, whatever it may be)
4) R&D, e.g. high speed physics, ..., some prestigeous universities, ...
5) The people that would have bought a Bugatti or Helicopter otherwise
My guess is around 1000pcs in 3 years, maybe thats optimistic.

I cannot even think of fabricating the DUTs you need such a machine for, why buy it ? You would probably leave the rocket science to
some key manufacturers and buy their final products or ready-to-use modules from China after a year or two. Why make what you can
buy, especially if all the testing is so expensive. You need a very good reason to do that.

I haven't seen the forecast, but I'd expect 1000 in 3 years is on the high side. Buyers will definitely trend heavily towards the R in R&D.

Another guess:
- In the first two years, software and hardware will be a buggy desaster
- Competition will follow up within a year or two, maybe based on band-splitting technology
- Next big step is 145GHz (limit of 0.8mm connector) by LeCroy, Tek or Anritsu
- Prices of 1mm equipment (cables, connectors, adapters, ...) will come down a lot in the next few years due to larger production volume
- Batteroo will ask for a 1mm to banana plug adapter

The software is the same as the other Infiniium oscilloscopes, so it's pretty darn good. There's always a chance of hardware bugs, but we've been designing and testing this with all the rigor you'd expect from us. With something this valuable and high tech, you probably wan't the warranty anyways . The 63 GHz scopes have something like 10,000 ACTIVE components, btw.

I doubt we'll see something reactionary. This isn't a project you can just spin up in a couple years. The custom ASICs alone will take years for most teams even if they implement band splitting.

145 GHz seems like a logical step to me, it'll be interesting to see what the future holds!

I agree that the 1mm space will see some growth, it's largely been a corner case tech for a while.

[Keysight DanielBogdanoff]

Just chatted with our mechanical engineers about the derating, it seems to be mostly related to the power supply. As you'd expect, you need extremely clean power for something like this. If the supply we use gets altitude derated, so does the scope.

[Keysight DanielBogdanoff]

Hi Dave, maybe a watercooled oscilloscope would be a solution ?! It would be an industry first, am I right ?
And their would be no altitude limit. 

Yes, this occurred to me, too. You also wouldn't get fan noise! I'm not sure how it would affect performance, though. The whole "warm up" period is important for precision measurements. It would be a really cool hack/upgrade, though. 

[Wolfgang]

Hi Dave,

I hope its really as bug free as you say. Given the extreme hardware complexity and the timeframe, plus the added problems regarding calibration, deembedding, deskewing, jitter removal, ..., all more severe due to the higher bandwidth I am not so optimistic about software quality.

BTW, I have an Infiniium scope (only a humble DSOS0604), and I like it, but this is a completely new frontier. I also have some counterexamples to the Keysight claim for excellent software - e.g. BenchVue.

It is plausible that others dont have an InP front-end process as advanced as you have. So my guess would be that they enhance bandwidth by expanding their bandsplitting approach.

[gkeeth]

Just chatted with our mechanical engineers about the derating, it seems to be mostly related to the power supply. As you'd expect, you need extremely clean power for something like this. If the supply we use gets altitude derated, so does the scope.

Altitude seems to be a consumer-safety spec. Basically, this means it won't blow up unexpectedly if you take it up a mountain or onto a plane. My guess on the temp derating is that it's a heat transfer concern with thinner air, but I'm digging into it.

Thanks for looking into it! I'm mostly curious about the "Non-operating" spec, though. Does that refer to when the scope is unplugged/turned off? If so, what would go wrong with the scope, something pressurized internally? Or does it refer to when the scope is plugged in & turned on, but not being actively used for measurements? In that case, does it just mean that if you go above 15,000 feet the scope will be damaged by overheating?

[Wolfgang]

Using heatpipes would add the advantages of reaching thermal equilibrium faster than just by fans, it would be quieter, and maybe consume less heat in total than all the air cooling approach. When looking at the fotos of your front-end modules you could use even more current for more bandwidth if cooling ist further improoved. All the heatpipes should end in a central heat exchanger block that could be either air or water cooled.

[mikeselectricstuff]

Hi Dave, maybe a watercooled oscilloscope would be a solution ?! It would be an industry first, am I right ?
And their would be no altitude limit. 

Yes, this occurred to me, too. You also wouldn't get fan noise! I'm not sure how it would affect performance, though. The whole "warm up" period is important for precision measurements. It would be a really cool hack/upgrade, though. 

I'd think it would be easier to regulate temp and get more uniform temp distribution.
Less vibration maybe.
Leaks could get real expensive though.

[Jr460]

I can see some other uses for scope this fast.   The newer Fiber Channel specs. 

Right now 16G is what most people are buying, 32G and 64G are due soon.  By due I mean that you by the HBAs and switches fro vendors without having to wait for special order. 

If you are playing with digital signals 64G, you need more than that see what is going on and build and test the line interfaces.

[Marco]

That's PAM4, so doesn't need frequencies quite that high.

[coppice]

Hi Dave, maybe a watercooled oscilloscope would be a solution ?! It would be an industry first, am I right ?

Water cooled electronics is usually a source of pain and misery. I've never worked with anything water cooled that didn't have at least a little staining from whatever colour antifreeze was used.

[Wolfgang]

A 2.6kW stove withe the noise level of  landing jet in your lab is a nuisance, too.
IBM ECL watercooled mainframes ruled the world for 2 decades 
And what if power density leaves you no choice ?

[mikeselectricstuff]

There's always Fuorinert - a lot less messy.

[bson]

In exquisite detail. You'll probably want to implement a bandwidth filter to knock down the extra 100 GHz of 1/f noise, though.

It doesn't have a 2MHz filter??? Pffft.

[KE5FX]

Hi Dave,

I hope its really as bug free as you say. Given the extreme hardware complexity and the timeframe, plus the added problems regarding calibration, deembedding, deskewing, jitter removal, ..., all more severe due to the higher bandwidth I am not so optimistic about software quality.

The software won't be substantially different than it has been for several years.  There is no reason to expect any quality regressions there.

BTW, I have an Infiniium scope (only a humble DSOS0604), and I like it, but this is a completely new frontier. I also have some counterexamples to the Keysight claim for excellent software - e.g. BenchVue.

Most of the time, scope firmware bugs manifest themselves through interactions between complex features.  "When I enable SPI decoding on channel 3 while monitoring a CAN bus on the MSO inputs, it fails to trigger on pattern xxxx-xxxx-xxxx," stuff like that.  These scopes are not engineered to see how many features they can cram into them, or to compete with whatever Rigol just announced.  They are used to trigger on fast one- or two-channel signals to look at eye diagrams, and possibly to capture the corresponding data records for offline analysis.  As long as the core signal processing is done correctly -- which it will be -- the users will be happy.

[Wolfgang]

Hi,

if you say that the software for this scope is old, I like it.  It will have fewer bugs then.
Owning quite some Keysight equipment (see my profile) I know that software quality is not too bad in general, but there are exceptions (BenchVue).
When I get you right the software on my DSOS0604 is the same as on the new scope with 110GHz ?

I dont know what RIGOL announcement you are referring to, but the UXR and all the RIGOLs have no common customer base 

[Berni]

Yep given that they most likley use  the same software means there should be very little in the way of bugs. So far all infiniium scopes have used the same software. They just drop support for old scopes in certain versions of it once it becomes too annoying to maintain while adding support for newly released scopes. So basically if your infiniium scope is not 20 years old and runs Win 7 and up then you can use the latest and greatest software.

With how many advanced features are crammed into this software it makes sense to just use it everywhere, rewrtiing all of this would take ages even for a large team of programmers. But most of these features are locked away under a optional license. On your particular DSO S-Series even the bandwidth is simply a license key (But obviously not unlockable up to 110GHz ).

[glarsson]

Does it have the classic 20MHz bandwidth limiting function that "all" oscilloscopes have? :-)

[srce]

Yep given that they most likley use  the same software means there should be very little in the way of bugs. So far all infiniium scopes have used the same software.

I've got an S series - there's plenty of s/w bugs.

[Keysight DanielBogdanoff]

Yep given that they most likley use  the same software means there should be very little in the way of bugs. So far all infiniium scopes have used the same software.

I've got an S series - there's plenty of s/w bugs.

If you have bugs, we'd love to hear about them! You can shoot the contact centers a quick e-mail or phone call (see my signature).

Does it have the classic 20MHz bandwidth limiting function that "all" oscilloscopes have? :-)

In exquisite detail. You'll probably want to implement a bandwidth filter to knock down the extra 100 GHz of 1/f noise, though.

It doesn't have a 2MHz filter??? Pffft.

The Infiniium Oscilloscopes employ some DSP techniques, so you get user-definable bandwidth filters! I believe there are also some hardware filters as well.

@Wolfgang, yes, it's the same GUI for the UXR and the S-Series.

[egonotto]

Hello,

my respect, this is great scope.

If a customer buy such a scope would an employee bring the scope and train the customer?

Best regards

egonotto

[Wolfgang]

I would say at prices like this a lengthy presales period with ample application support is the minimum.

When I bought my humble S-Series I got a unit on loan from them plus application support on phone and an engineer visiting my place. 

Good service is a sales argument, especially with expensive equipment. 

[srce]

Yep given that they most likley use  the same software means there should be very little in the way of bugs. So far all infiniium scopes have used the same software.

I've got an S series - there's plenty of s/w bugs.

If you have bugs, we'd love to hear about them! You can shoot the contact centers a quick e-mail or phone call (see my signature).

Thanks - but they've been reported months ago.

[iMo]

I would say at prices like this a lengthy presales period with ample application support is the minimum.

This type of equipment is not acquired such an enthusiastic EE calls KS and after a confirmation the gear is on stock he sends an order via email
Except some private hobbyist all serious organizations or companies (who usually need such an equipment) recall a tendering process organized by their Procurement department. And those procurement guys know pretty well how to squeeze KS or Lecroy in, be sure..
 

[Rich@RohdeScopesUSA]

Daniel - congrats to you and the Keysight team on the UXR!  Super impressive achievement 

-Rich

[cpuerror]

I'm going to wait for Hantek to come out with their 110Ghz scope I think it will be much more affordable.

[Scratch.HTF]

1+ terasample/second ADCs, anyone?

[VK5RC]

Thanks for the post UNN and comments esp Daniel from K-S, would have been a fascinating development, would love to have been a 'fly on the wall'.
That is a LOT of data coming at you. 250GSa/S x4(channels) x10bits 

[srce]

Thanks for the post UNN and comments esp Daniel from K-S, would have been a fascinating development, would love to have been a 'fly on the wall'.
That is a LOT of data coming at you. 250GSa/S x4(channels) x10bits 

It's very impressive - 10Tb/s. Is it DDR or something more specialised? Lots of pins on the ASIC to drive that many memory banks in parallel. 

[exe]

Thanks for the post UNN and comments esp Daniel from K-S, would have been a fascinating development, would love to have been a 'fly on the wall'.
That is a LOT of data coming at you. 250GSa/S x4(channels) x10bits 

Uhm, at full speed it has much lower ENOB, I wonder if this affects required memory bandwidth...

[srce]

Thanks for the post UNN and comments esp Daniel from K-S, would have been a fascinating development, would love to have been a 'fly on the wall'.
That is a LOT of data coming at you. 250GSa/S x4(channels) x10bits 

Uhm, at full speed it has much lower ENOB, I wonder if this affects required memory bandwidth...

It's 5 bits ENOB at max sample rate, so that would only halve it. That's still a lot of mem b/w. I suspect they'd still probably need to store at least 8 bits anyway.

[nfmax]

It doesn't matter: you still have to get the bits into memory somehow, whether they are effective or not!

[Wolfgang]

Agreed. Hardware standards propose 8Bits parallel anyway, I guess.

[Keysight DanielBogdanoff]

Hello,

my respect, this is great scope.

If a customer buy such a scope would an employee bring the scope and train the customer?

Best regards

egonotto

Thanks! You'd have to talk to your local application engineer (AE) about on-site support and setup time.

Daniel - congrats to you and the Keysight team on the UXR!  Super impressive achievement 

-Rich

Thanks, Rich!

Thanks for the post UNN and comments esp Daniel from K-S, would have been a fascinating development, would love to have been a 'fly on the wall'.
That is a LOT of data coming at you. 250GSa/S x4(channels) x10bits 

It's very impressive - 10Tb/s. Is it DDR or something more specialised? Lots of pins on the ASIC to drive that many memory banks in parallel. 

I think I'm allowed to say... it's a hybrid memory cube. 

[srce]

Thanks for the post UNN and comments esp Daniel from K-S, would have been a fascinating development, would love to have been a 'fly on the wall'.
That is a LOT of data coming at you. 250GSa/S x4(channels) x10bits 

It's very impressive - 10Tb/s. Is it DDR or something more specialised? Lots of pins on the ASIC to drive that many memory banks in parallel. 

I think I'm allowed to say... it's a hybrid memory cube. 

Interesting - not come across that before.

Here's a micron part
datasheet for anyone interested.

£400 for 2GB - 896 pin - 1.28Tb/s bandwidth. 4 x 16 SERDES links at 15Gb/s each. So you'd probably just need to have two of those per scope channel.

Nice 

[Berni]

Holy crap that is a speedy memory chip

But the chip is actually pretty cheap at such a price tag when you go find a FPGA that is callable of running this monster at full wack. The interface to it is 4 links, each having 8 to 16 pairs and TX and RX is separate. So that gives us 4 x 16 x 2 = 128 serdes channels! Even FPGAs with 4 serdes channels at these speeds cost an arm and a leg.

But seeing this has those 4 links... does that mean that this is also technically quad-port memory with all ports capable of individual write+read operation?

[Muxr]

If they are rolling their own ASICs they could have leveraged something like HBM2 as well with 256 GB/s per package.

[srce]

If they are rolling their own ASICs they could have leveraged something like HBM2 as well with 256 GB/s per package.

From a quick google - that looks harder to integrate though.

For HMC it looks like you have 128 differential data pins @ 15Gb/s (512 IO in total) in a BGA package - whereas HMB2 looks like 1024 data pins @ 2.4Gb/s (1696 IO in total as there's lots more control pins per channel) and the memory has to be integrated in package on an interposer. If you need two, I'd guess you're going to struggle with needing >3000 IO

[Brumby]

I'm still catching up with this....

It's actually doing real-time acquisition instead of reconstructing the signal from a frequency domain acquisition?  That's... astonishing.  Wow.  Just wow.  ASIC or not.  Just wow.

[vaualbus]

I'm just curious  to the overall system architecture.
So we know  that we have 4 boards each handling one channal,
Than there has to be another boeard for handling each channel 1.2Tb/s data.
I beat it is full of asic/fpga/DSP that converts that signal into I guess PCI 3.0 (or thunderboalt?) interface that go to the computer side.
I will beat that you use almost all of the pci lanes avaible from the processor (i7 or i9 by the way?).
Than I will beat there is yet another board handling reference clock for the four channel.
And a monster power supply of course.

[FrankBuss]

In case someone missed it:

https://www.eevblog.com/forum/testgear/something-amazing!-keysight-uxr-110ghz-256gss-10-bit-scope-teardown-exp/