Oscilloscopes,connectors,VSWR and signal fidelity

[fonograph]

Oscilloscopes like Rohde & Schwarz RTO,Keysight 6000 & S,Lecroy Wavepro HD and Tek MSO 6 bandwidths that go up to 6 - 8 GHz.That is seriously deep in the microwave range,impedance matching becomes extremly important and signal degradation due to impedance mismatch becomes very common.

Here are two important things to keep in mind : 1 All of the above mentioned scopes use BNC connectors,they might use fancy BNCs,they name it "precision BNC" but at the end of the day,its a BNC in 8 GHz rated instrument that is supposed to analyze signals in time domain.

2. VSWR... most scopes I named above dont  provide specs in datasheet that shows VSWR vs frequency.One exception I found is RTO,the spec says under 2 GHz VSWR is 1.25 : 1 and above 4 GHz its 2.0 : 1.

I am not microwave wizard nor a milimeterwave warlock but...  2.0 : 1 VSWR!? Holy schmitt trigger that aint good! Thats catastrophically bad,it might not be such a big deal in spectrum analyser or antenna,but this is oscilloscope! We use them to look at time domain,we need decent time domain signal fidelity,strong reflections like that must wreak havoc on waveform shape if its anything other than steady sinewave.

Even in realistic 5 ENOB scenario,you can see the echoes down to 36 dbc,with 2.0 : 1 VSWR you have 11% of the signal reflected,it will the travel back toward DUT and cause all kinds of nasty schmitt inside there,get reflected again and arrive as delayed echo.This to me sounds like big big problem,maybe I am just noob and there is something that prevents this from being issue but I cant think of anything.

2.0 : 1 VSWR in oscilloscope... just think about it.And thats the 6 GHz RTO,imagine how its like in the 8 GHz BNC scopes like Keysight S,Wavepro HD or Tek MSO 6,they dont give any VSWR information in their datasheets.My point is,isnt a 8 GHz scope with BNC bullschmitt? Isnt 8 GHz BNC scope inherently faulty concept,isnt it stupid device that cant do its job properly at those high frequencies despite being advertised as 8 GHz? Isnt it waste of money? Isnt it practicaly essential to get non BNC scope,with proper microwave connectors to look at 6 - 8 GHz signals with good fidelity?

[mc172]

It's unfortunately how it goes. The people that run the show at the companies that design these things probably think that any scope with a "non standard "scope" connector" will never sell. The scary thing is that they might be right.

I see this all the time. I spend my life having to design 3+ GHz product with shite connectors that were designed almost a century ago for operation up to 30 MHz. I also spend most of my days telling people about this, and hearing "but we did xyz product up to 800 MHz so it's easy to go to 1800".
The people with the money (both the customer and the sales people dealing with the customer) just don't have a clue, nor do they give a shit to get a clue.

The real question is, why the hell would you ever need an 8 GHz time domain device, even if it had N or SMA connectors on it?

[xrunner]

It's unfortunately how it goes. The people that run the show at the companies that design these things probably think that any scope with a "non standard "scope" connector" will never sell. The scary thing is that they might be right.

Sad but true. We have scopes that go into the 100's of MHz and beyond, RF signal generators and spectrum analyzers. The RF gens and SpecAns have N connectors, but the scopes don't. I have a closet full of patch cables with BNCs so to connect up things I adapt the N connectors to a BNC. I guess if scopes had N connectors (or whatever the designers deemed necessary) you could use an adapter as needed ... Why they didn't already all settle on N connectors I do not know. It's just the way it all worked out.

[Andrey_irk]

The real question is, why the hell would you ever need an 8 GHz time domain device, even if it had N or SMA connectors on it?

I think, mostly people use them for constellation analysis and such. How else can you take a look at the signal that has a few GHz of bandwidth? There are real-time SA's but their bandwidth is not nearly as wide.
TDR analysis, eye diagrams are also some of the common use cases of such scopes.
I think, someone like Hugoneus can give you some examples where you can do the job only with a 20+ GHz scope.

As for the BNCs in such devices - I think you are correct! I've measured the performance of some BNCs with a VNA and the VSWR beyond 1 GHz was all over the place. Even a slide touch can ruin everything.
Although, I know that the connectors they use are not the regular ones. For instance, they use only small dielectric support beads instead of filling the connector with it(which increases the bandwidth). Maybe there are some trics to make them more stable as well.
Nevertheless, I don't see what stops them from using "type N" or SMA as the price of adapters (only if you need them!) is negligible given the price of the scopes.

[David Hess]

Some BNCs are better than others.  Back when Tektronix introduced the 7A29 1GHz vertical amplifier for their 7104 1GHz mainframe, they discovered that common BNCs do not match the impedance of common 50 ohm coaxial cable so they ended up making their own.

[innkeeper]

so maybe i am gonna say something here that might be obvious to some but maybe not to others reading this thread,

VSWR spec on a scope is only viable if the scope is providing a known termination impedance such as 50 ohms,

in most test cases though, an oscilloscope is a high impedance device and as such no swr spec would be given as the input impedance is not a termination impedance.

[Andrey_irk]

so maybe i am gonna say something here that might be obvious to some but maybe not to others reading this thread,

VSWR spec on a scope is only viable if the scope is providing a known termination impedance such as 50 ohms,

in most test cases though, an oscilloscope is a high impedance device and as such no swr spec would be given as the input impedance is not a termination impedance.

Agree. But then again, who uses high impedance inputs in GHz region?

[TheSteve]

so maybe i am gonna say something here that might be obvious to some but maybe not to others reading this thread,

VSWR spec on a scope is only viable if the scope is providing a known termination impedance such as 50 ohms,

in most test cases though, an oscilloscope is a high impedance device and as such no swr spec would be given as the input impedance is not a termination impedance.

Anything above 500 MHz is most commonly terminated with 50 ohms internal to the scope. Most scopes I have seen with 50 ohm terminations give a spec of an SWR of 1.5:1 over its rated input frequency. All of the high frequency active probes use the 50 ohm termination in the scope.
I haven't seen any issues with the BNC's on my Keysight @ 1 GHz when using good quality patch cables but going to 6 GHz + is a whole different game.

[fonograph]

Rohde & Schwarz just released a 50 ohm only scope and it have BNCs.It have better VSWR than the RTO,its 1.25 : 1 under 4 GHz and 1.4 : 1 from 4 to 8 GHz.Thats better but still,1.4 is 2.78% energy reflected.

In my opinion,bandwidth,connector type and ADC bit depth should be such that at the highest frequency the scope is rated for,the percentage of reflected energy due to connector type should be less than half LSB.

[mc172]

The real question is, why the hell would you ever need an 8 GHz time domain device, even if it had N or SMA connectors on it?

I think, mostly people use them for constellation analysis and such. How else can you take a look at the signal that has a few GHz of bandwidth? There are real-time SA's but their bandwidth is not nearly as wide.
TDR analysis, eye diagrams are also some of the common use cases of such scopes.
I think, someone like Hugoneus can give you some examples where you can do the job only with a 20+ GHz scope.

If you are serious about doing so, you use a signal analyser, not an oscilloscope. These have proper connectors on them.

[fonograph]

WTF are you writting about? There is plenty of scopes with proper microwave,even milimeter wave connectors.Please stop hijacking and derailing my thread into "hurr durr who needs microwave scopes,just use SA derp"

Even uber expensive spectrum analysers have like 500 MHz real time bandwidth,UXR scope have 110 GHz real time bandwidth.Also the UXR have second best milimeter wave connector type that is currently in production,it uses the 1mm type,only higher performance connector is the Anritsu 0.8mm.

[mc172]

I'm not derailing your thread, you just did that yourself.

[dmills]

I would note that a common trick to improve VSWR is to place an accurate 50 ohm pad in line close to the scope input, as long as you can get it really close to the mismatch this will soak up much of the reflection (And if you are using a scope you are probably not short of signal level).
 
High speed scopes lie (Well, actually, **ALL** scopes lie) and you have to do a fair bit of interpretation of what you are seeing, but this is true even of the picture on a 100MHz instrument once you get above wiggly DC. Probe technique matters and in RF the probing is usually not an irrelevant load on the circuit.

My usual use for a fast scope is more of the eye diagram then actual specifics of a waveform use case, equivalent time sampling is good enough for this, and more often then not this is what you really care about at the physical layer.
Not to say that there are not good use cases for the quick scopes, but for me, in my use case I care about eye and jitter at the physical layer and for higher level things, well I have other tools once in the digital domain.

Regards, Dan.

[tggzzz]

I see this all the time. I spend my life having to design 3+ GHz product with shite connectors that were designed almost a century ago for operation up to 30 MHz. I also spend most of my days telling people about this, and hearing "but we did xyz product up to 800 MHz so it's easy to go to 1800".

Nothing changes.

40 years ago the project on the next bench to mine was to enable 2Mb/s to be transmitted between telephone exchanges down bundles of twisted pairs that were specified at 1.6kHz (i.e. for speech). The telco did not want to ask the cable manufacturers to do any measurement/qualifications/changes, since that would have allowed then to increase prices.

I always thought that project was, um, limited since my project was for the first generation of practical optical fibres

[fonograph]

So I was right that 8 GHz BNC scopes are not good idea?