BNC 50ohm Feed Thru

[Randy222]

I am just a hobby ham / IoT person, so kHz MHz and GHz are in my toolset, to some extent.

I am evaluating some 50 ohm feed thru items for my scope that only has 1meg Z on inputs.

Looking to see if my testing is valid when evaluating VSWR into the scope.

I opted to not evaluate with scope attached, but put a 1meg terminator on end of the BNC feed thru to simulate being attached to the scope.

I created criteria to mark a usable bandwidth, using VSWR 1.25 as the threshold.

Using my LiteVNA64 I get bandwidth of about 650kHz to 300MHz (VSWR <= 1.25). Beyond those ends the VSWR starts to rise quickly.

However, in observation, with an open ended feed thru the VSWR was fairly flat and low from about 10kHz out to near 3GHz.

Is my terminated method valid for evaluating VSWR between a transmitter (VNA) and receiver (simulated scope 1meg)?

[rf-messkopf]

I did a similar measurement some time ago and posted the results in a German electronics forum. See also the attachment.

The feedthrough termination is model J01006A0013 made by Telegärtner, click here for a datasheet.

The traces in telegaertner.png mean the following:

-red: measured when the scope side of the termination is open.
-green: measured when attached to a vintage HP 54503A scope with vertical scale set to 500 mV/div (front panel says 7 pF input capacitance).
-blue: measured between two ports of the VNA, where the resulting S-matrix was transformed such that port 2 (scope side of the termination) has a virtual impedance of 1 MΩ, with an additional 5Ω series resistor and a 7 pF capacitor from that resistor to ground.

As you can see, things start to become really ugly at frequencies beyond 500 MHz (the specification says more than 15 dB return loss at frequencies between 200 and 500 MHz, with the scope end open). Nevertheless, the termination does meet its specification.

For comparison, I also measured a BNC tee with a 50Ω BNC termination on one end (bnc-tee.png, same meaning of the traces colors as above). Actually, that one turns out to be better than the feedthrough termination. 

[TimFox]

Out of curiosity, what is the measured DC resistance of your 50 ohm terminators?

[rf-messkopf]

Out of curiosity, what is the measured DC resistance of your 50 ohm terminators?

I grabbed two Telegärtner model J01006A0013 feedthrough terminations from my adapters drawer. They measure 49.711 ohms and 49.879 ohms. I can't tell which one I used in the return loss measurements over a year ago.

[joeqsmith]

I tried to modify one that had been damaged.   Being dead, I had no reference but my results were poor.   

https://www.eevblog.com/forum/testgear/pasternack-pe6008-50-thru-term/msg947602/#msg947602

[Randy222]

The DUT's I have are no-name stuff, china made most likely.

I used three of my "best" non calibrated meters, one being a Fluke87

These results are same no matter what end I ohm at.

(tenths resolution)
BNC 1 - 50.4 50.3 50.3
BNC 2 - 50.2 50.1 50.1

[tszaboo]

For comparison, I also measured a BNC tee with a 50Ω BNC termination on one end (bnc-tee.png, same meaning of the traces colors as above). Actually, that one turns out to be better than the feedthrough termination. 

This is the second time I hear this. Why do you think this is happening?

[Bud]

Is my terminated method valid for evaluating VSWR between a transmitter (VNA) and receiver (simulated scope 1meg)?

No, you are missing an equivalent capacitor on the scope end of the adaptor. Refer to your scope what input capacitance it has.

[rf-messkopf]

For comparison, I also measured a BNC tee with a 50Ω BNC termination on one end (bnc-tee.png, same meaning of the traces colors as above). Actually, that one turns out to be better than the feedthrough termination. 

This is the second time I hear this. Why do you think this is happening?

Because the termination on the BNC tee is much better than the termination resistor in the feedthrough device. The manufacturers don't seem to spend much effort on the quality of the resistive element in feedthough terminations (often just a THT resistor, see Joe's posting above) as the resulting return loss is spoiled by the scope input anyway.

I noticed that in my measurements the return loss depends very sensitively on the set virtual scope input capacitance. And this capacitance depends on the individual scope.

These feedthrough devices are good up to a couple of MHz, but do not replace a scope with a dedicated 50 ohms signal path.

[Randy222]

Is my terminated method valid for evaluating VSWR between a transmitter (VNA) and receiver (simulated scope 1meg)?

No, you are missing an equivalent capacitor on the scope end of the adaptor. Refer to your scope what input capacitance it has.

Thanks for pointing that out, so I went and tested.
Scope is rated 15pF +- 3
Using my LCR meter my 1meg BNC terminator shows 13.2pF, although I suspect a few pF due to the proximity of the test leads used, I did try to separate them as much as I could, but the lead ports are spaced pretty close to each other.

Meter is not a calibrated unit, but it has given me matching results to OEM specs of parts I checked over the years.

I'll build another 1meg terminator using a different method to see if my VSWR results change much, or I buy one already made.
I'll also need to buy a name-brand feed thru, ThorLabs T4119 or the like.

[Randy222]

Won't need to hash out on this topic, there's plenty here on EEVb for me to digest.

This one is good
https://www.eevblog.com/forum/testgear/inexpenive-50-ohm-feed-through-terminations/msg2548299/#msg2548299

[joeqsmith]

I have the opposite problem in that my two higher speed scopes don't offer high impedance inputs.   They are 50 ohms only.  If you want to use a 10X probe, you must use some sort of buffer or I have higher speed active probes for them.

I assume your scope has much higher BW than what you have for 10X probes and are wanting to use resistive probes with it.  This is the reason for your interest in through terminators.   I normally use the 10X probes unless the loading it too much.   My normal use of the terminators is to provide a simple load for my signal generator before connecting it to a high impedance circuit.    Then again, we can say the scope is a high impedance... Which it is, until it's not. Like everything, those parasitics get in the way...   


**
Tried to make clearer but failed...  Easiest way to put it, the faster you go, the more you need to consider....

https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/

[joeqsmith]

I am thinking if the goal is to get a decent return loss with a short, or um scope attached, it could maybe be done if you wanted to give up some of the dynamic range.   

Using your criteria,  say 20dB return loss with a 3dB cutoff at 350MHz, DC coupled and 6dB loss with a 12-18pF load, uni-directional made with all passive parts.  If this seems reasonable,  I could try and model something up.  May get lucky....

[rf-messkopf]

You could build it like a X10 scope probe, i.e., a compensated voltage divider, with a 50 ohms input resistor. See the attachment for a quick simulation (the portion in the dashed box is the actual feedthrough device). That gives you 20 dB return loss up to 1 GHz if you are able to pull it off in practice. The 10 megohms resistor and the compensating capacitance (which must be variable in practice) will be a challenge at GHz frequencies. You could think of using other division ratios to end up with more reasonable component values.

No idea if such a device would be useful. Maybe it's better to just add a 3 dB attenuator to the 50 ohms side.

[joeqsmith]

I'm thinking something similar but using common values that could be realized in hardware.  With OPs scope being from 12-18pF, thinking I need at least one trimmer to compensate for it.   

A 3dB attenuator to a 50 ohm terminator attached to the scope will improve the return loss but the response will not be very flat.  18pF @ 300MHz is 30 ohms, in parallel with the 50.  I think your other idea is heading down the right path.

Still, if you are wanting to work at GHz with a scope and your scope doesn't support have internal termination, seems like it is time for a new scope.. 

[rf-messkopf]

Another run with a 2 megohns divider resistor and 18 pf scope input capacitance. Compensating capacitance is about 9 pF with 22 ohms in series, selected to give a good compromise between transient response and flatness. Flatness is okay up to 1 GHz, but the return loss is mediocre: 20 dB only up to about 100 MHz. It seems that some tradeoffs must be made with a purely passive device.

[tszaboo]

For comparison, I also measured a BNC tee with a 50Ω BNC termination on one end (bnc-tee.png, same meaning of the traces colors as above). Actually, that one turns out to be better than the feedthrough termination. 

This is the second time I hear this. Why do you think this is happening?

Because the termination on the BNC tee is much better than the termination resistor in the feedthrough device. The manufacturers don't seem to spend much effort on the quality of the resistive element in feedthough terminations (often just a THT resistor, see Joe's posting above) as the resulting return loss is spoiled by the scope input anyway.

I noticed that in my measurements the return loss depends very sensitively on the set virtual scope input capacitance. And this capacitance depends on the individual scope.

These feedthrough devices are good up to a couple of MHz, but do not replace a scope with a dedicated 50 ohms signal path.

Right, so the through terminations, that are designed to be used for oscilloscopes, they are low quality, and stop below 1GHz, while the regular terminators will go to GHz regions. Actually, we could compare a system if we use a simple T junction with a through termination.
I'm still thinking there is a bit more to this.

[joeqsmith]

I think with the OPs 300MHz, I assume 3dB flatness, you may be able to pull it off.   I tried using 6dB of attenuation and a network of RLCs.  It would need to be trimmed, just like a probe.   OP may not want to give up 6dB.  I figured I would wait for the constraints before diving in too deep. 

This would make for a fun little contest circuit.  It's something that someone may actually have a use for, unlike an oscillator on a breadboard.   

Right, so the through terminations, that are designed to be used for oscilloscopes, they are low quality, and stop below 1GHz, while the regular terminators will go to GHz regions. Actually, we could compare a system if we use a simple T junction with a through termination.
I'm still thinking there is a bit more to this.

Again, at least for the OP wanting 300MHz, their scopes input capacitance is 18p max, or 30 ohms at 300.   At one point, directly measuring the scope's input, I bet you get a decent match at 50ohms.   Adding R isn't going to give you a flat impedance, well, unless we can say use a 1ohm source rather than 50....

[Randy222]

My noted 650kHz to 300MHz was just width of VSWR <= 1.25
My hobby scope is a rigol dho804 opened up some to be an 814 where -3db is around 180MHz.

It's noted in the EEVb link I posted that the feed thru's act better when you front them with an attenuator.

[joeqsmith]

Another run with a 2 megohns divider resistor and 18 pf scope input capacitance. Compensating capacitance is about 9 pF with 22 ohms in series, selected to give a good compromise between transient response and flatness. Flatness is okay up to 1 GHz, but the return loss is mediocre: 20 dB only up to about 100 MHz. It seems that some tradeoffs must be made with a purely passive device.

Didn't want you to think I was a slacker.   I modified one that I had simulated in SPICE, without adding any attenuation.   From testing I could see the original Pasternak part out performed the two I had originally modified, so  I chose a simple approach by adding a bit of inductance.   Some small gains but really, the idea of working at high speeds with this sort of setup is flawed from the start.   

https://www.eevblog.com/forum/testgear/inexpenive-50-ohm-feed-through-terminations/msg5227260/#msg5227260

[tszaboo]

I think with the OPs 300MHz, I assume 3dB flatness, you may be able to pull it off.   I tried using 6dB of attenuation and a network of RLCs.  It would need to be trimmed, just like a probe.   OP may not want to give up 6dB.  I figured I would wait for the constraints before diving in too deep. 

This would make for a fun little contest circuit.  It's something that someone may actually have a use for, unlike an oscillator on a breadboard.   

Right, so the through terminations, that are designed to be used for oscilloscopes, they are low quality, and stop below 1GHz, while the regular terminators will go to GHz regions. Actually, we could compare a system if we use a simple T junction with a through termination.
I'm still thinking there is a bit more to this.

Again, at least for the OP wanting 300MHz, their scopes input capacitance is 18p max, or 30 ohms at 300.   At one point, directly measuring the scope's input, I bet you get a decent match at 50ohms.   Adding R isn't going to give you a flat impedance, well, unless we can say use a 1ohm source rather than 50....

I've just watched your video about the topic. So I actually wonder if there is a solution to matching a 18pF input to 50 Ohm.
If I approach it as an antenna matching problem, then adding L or C would only match it at one frequency. So I'm thinking a wideband matching with several networks could work better. For example 200 Ohm resistors with 4 different impedances, to compensate for the different frequency components?
But then again, are we looking for flatness in the frequency band, are we looking for keeping the shape of the signal intact? Are these two goals different? DC accuracy?

[joeqsmith]

If OP doesn't respond,  I would say a free for all on the requirements, goals...     rf-messkopf's simulation used values that may have made it difficult to construct.  I wanted to show something in physical hardware.  Adding a small bit of inductance seemed like something I could handle building...

[tszaboo]

If OP doesn't respond,  I would say a free for all on the requirements, goals...     rf-messkopf's simulation used values that may have made it difficult to construct.  I wanted to show something in physical hardware.  Adding a small bit of inductance seemed like something I could handle building...

I think for something more complicated, a small PCB would be better. What would be even better is ceramic boards.
I did a lot of search for housings for feedthrough terminations, or attenuators. The issue I see is the cost is around 40-50 USD for such a housing, or DIY solutions that show. I kind of feel like 80% of RF design or probe design is mechanical, and the last 20% is the electronics. Most connectors don't come with good dimensional drawing, and the ones that do are some specific imperial thread pitches that drive me crazy.

[G0HZU]

I've just watched your video about the topic. So I actually wonder if there is a solution to matching a 18pF input to 50 Ohm.
If I approach it as an antenna matching problem, then adding L or C would only match it at one frequency. So I'm thinking a wideband matching with several networks could work better. For example 200 Ohm resistors with 4 different impedances, to compensate for the different frequency components?
But then again, are we looking for flatness in the frequency band, are we looking for keeping the shape of the signal intact? Are these two goals different? DC accuracy?
Report to moderator    Logged

The (close to) ideal matching component would be a parallel negative capacitor with a capacitance of -18pF. As this component doesn't exist in a passive form, the next best thing is to add a small amount of series inductance inline with the 50R termination resistor.

A series RL can mimic a 50R resistor in parallel with -18pF over quite a large bandwidth.  See the second image below. You can see it produces a -18pF capacitance over a large bandwidth. However, 46nH is a lot of inductance to add, and so in practice, it can be risky to do this as it can cause a spike to appear on the leading edge of a fast pulse if a really fast pulse is then fed to the scope. This is because (when this network is added to the scope input) the input impedance of the scope will then have an upwards bump up at 100MHz-200MHz or so. So it's usually best to keep the series inductance value fairly low if you want to look at fast pulse edges.

Note that a typical scope input isn't as simple as 1Meg in parallel with 18pF. There will be some series resistance right at the scope input and this is typically in the range 20R to about 50R. This resistance often varies a bit depending on the attenuation setting. A typical scope input can be modelled reasonably well with the network given in the first image below, as long as the correct value for R1 and C1 are used.

[joeqsmith]

I think for something more complicated, a small PCB would be better. What would be even better is ceramic boards.
I did a lot of search for housings for feedthrough terminations, or attenuators. The issue I see is the cost is around 40-50 USD for such a housing, or DIY solutions that show. I kind of feel like 80% of RF design or probe design is mechanical, and the last 20% is the electronics. Most connectors don't come with good dimensional drawing, and the ones that do are some specific imperial thread pitches that drive me crazy.

Feel free to use any construction techniques and burn as much cash as you wish for your demonstration.   Personally, I don't see much value in it as the whole idea is flawed.   

I had ran some other tests that were not shown in the video.  Attached photo showing the scope simulator and terminator on FR4 using 2X1206 100.   The graph compares the custom thru terminator attached to the LeCroy 64xi waveblunder, compared with the simulated 64xi with both the custom terminator and the prototype built on FR4.   Sure, some slight differences but good enough.   

That 10x probe I show on this scope will out perform it to I think we said 60MHz, where the my resistive probe presents a higher impedance load. 
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg4988716/#msg4988716

We can achieve higher BW using a homemade resistive probe and better scope.   
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg5006290/#msg5006290

[Randy222]

I ran VSWR test again, this time fronting the 50 ohm feedthru with a 6db 50 ohm attentuator.

Very flat and below 1.1 from 10kHz to 300MHz (no 1M terminator)
Adding my homebrew 1M terminator the VSWR improved very slightly, like 1.082 to 1.079

Getting in 500MHz+ areas the VSWR curve starts to look like full rectified AC (humps) and in some areas of bandwidth the peaks reach up near 2. I suspect this from the feedthru items since they are only rated to 1GHz, whereas my attenuators are ratsed to 6GHz.

So at least for now, I think I am good with terminating 50 ohm on my 1M scope.

[joeqsmith]

I ran VSWR test again, this time fronting the 50 ohm feedthru with a 6db 50 ohm attentuator.

Very flat and below 1.1 from 10kHz to 300MHz (no 1M terminator)
Adding my homebrew 1M terminator the VSWR improved very slightly, like 1.082 to 1.079

Guessing you did not watch my video where I show that the capacitor is the dominant factor.  You need to measure with your scope, or improve your load to show anything meaningful.   

Attached simulation using your 6dB attenuator attached to a 50 ohm thru terminator, then to your scope.  For the scope, I kept the 1M constant and then swept the capacitor from 0 to 20pF.    With 0 capacitance, the VSWR is perfect (which is what you measure with your 1M load).   At 20pF, it's 1.4 at 300MHz.   

[joeqsmith]

If we fix the scope's capacitance at 16pf then sweep the inductance,  VSWR is below 1.5 to 300MHz for all inductance values.   But then look at what the scope sees.   

[joeqsmith]

What I was hinting at with the 6dB loss was not to add an attenuator like previously shown.  Rather I was thinking to make a uni-directional divider with a bit of inductance added to grounded leg.   Note the VSWR is a wash but what the scope sees can now be fairly flat.

Obvious question may be, why didn't I demo this.   Problem was to cut the pin to add the series R and then somehow add the other parts without unsoldering the first, then get the thing back together without damaging it....   Seemed like a lot of effort for something that I considered a bad idea to start with.  So I took the lazy way out and showed a simple inductor.   

[G0HZU]

I think that is the way some low cost scopes may do the internal 50R termination. There's usually already something like a 20R series resistor just inside the BNC input. This is there all the time.

Then, to provide the 50R termination mode they exploit the series 20R and switch in a 30R resistor in shunt. This sets up a 50 ohm potential divider and it masks some of the input capacitance from the input. The penalty is some loss in the divider but the scope will apply a fudge factor for this to keep the calibration correct.

Earlier this year, I mentioned this stuff about adding some series inductance and the resistive divider in this thread here in posts #4 and #14:

https://www.eevblog.com/forum/beginners/whats-the-input-capacitance-of-an-oscilloscopes-50-ohm-input/

[Randy222]

I ran VSWR test again, this time fronting the 50 ohm feedthru with a 6db 50 ohm attentuator.

Very flat and below 1.1 from 10kHz to 300MHz (no 1M terminator)
Adding my homebrew 1M terminator the VSWR improved very slightly, like 1.082 to 1.079

Guessing you did not watch my video where I show that the capacitor is the dominant factor.  You need to measure with your scope, or improve your load to show anything meaningful.   

Attached simulation using your 6dB attenuator attached to a 50 ohm thru terminator, then to your scope.  For the scope, I kept the 1M constant and then swept the capacitor from 0 to 20pF.    With 0 capacitance, the VSWR is perfect (which is what you measure with your 1M load).   At 20pF, it's 1.4 at 300MHz.

But my 1meg terminator itself does show it has some pF, it does not appear to be 0F. In general, poorly constructed terminators, thus I do expect some pF to be there.

I guess my next step is, connect VNA through attentuator+feed-thru to scope.

[joeqsmith]

....But my 1meg terminator itself does show it has some pF, it does not appear to be 0F. ...

If you do watch that video, you may notice that the two terminators I made back in 2016 exhibited a worse return loss than the OEM part when connected  to my scope.  That was a big hint to what was coming.  I posted how the OEM part was nothing more than an axial resistor.  That added parasitic inductance improved the performance.   Swapping that axial part for the 4, then the 2 X 1206's lowered that inductance.   

[joeqsmith]

All scopes are different.  The two GHz scopes I have are fixed at 50 only.   That LeCroy 7200 I showed has SMA connectors on the ICs themselves.  I wouldn't be surprised if the waveblunder doesn't have a different front for the two modes.     

[joeqsmith]

Looks like KeySight has a whole series on oscilloscope front end designs.  Here showing two separate paths:

https://youtu.be/1sVEWzMjkBo?t=258

[tszaboo]

We can achieve higher BW using a homemade resistive probe and better scope.   
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg5006290/#msg5006290

I've been experimenting with Z0 probes myself. I have a design with multilayer PCB, and pogo pin to pick up signals, or 100 mil pin header at the end. I think I got quite good results, below is a comparison to a N2795A active probe. This is probably the sharpest square wave I could do without designing something, it's for sure is limited by the source. It shows faster rise time than the N2795A so I'm hoping it will be above a GHz, but I'm yet to make a test setup to measure it.
This is at work. The reason I'm interested in the through terminators is because I cannot really afford a new scope with 50 Ohm inputs at the home lab.

Note that a typical scope input isn't as simple as 1Meg in parallel with 18pF. There will be some series resistance right at the scope input and this is typically in the range 20R to about 50R. This resistance often varies a bit depending on the attenuation setting. A typical scope input can be modelled reasonably well with the network given in the first image below, as long as the correct value for R1 and C1 are used.

That's a very good point. I'm going to guess it depends on the attenuator they have in front of the input buffer.
For realistic measurements with a 50Ohm input, we can probably guess the input signal is going to be in the 5VRMS range though.

So what can we do? Connect a VNA on the scope input, and see whatever it measures?

[G0HZU]

Here's an old plot showing the simple scope model against a VNA measurement of an old Tek 465 scope and a Tek TDS2012 input. I can't remember what values I used for R1 and C1 for the TDS2012 but it was probably about 40 ohms and 20pF.

The agreement for Rp and Cp out to about 150MHz was very good as you can see in the plot below. This model is much better than the basic 1Meg in parallel with 20pF model that many people continue to use.

There's also a plot showing a classic old Tek 465 input being compared to the basic model. Again, the agreement is very good. In this case, the values for the model are given next to the plot.

[G0HZU]

I've got quite a few scopes here, all made by Tek or HP or Hameg, and I've measured the inputs of all of them with a VNA at some point.

For example, I have an old HP Infinium scope (500MHz) and this simply switches in a shunt 50R resistor at the input when 50 ohm mode is selected. There is some obvious printed inductance in series with the 50R load and HP will have added this deliberately to help improve the input VSWR at higher frequencies. However, the VSWR of the Infinium scope is only low up to about 250MHz. It reaches 1.6:1 up at 500MHz for example. This is despite the extra inductance in series with the 50 ohm resistor.

The most modern scope I've got here is a Tek MSO4104 and this has a dedicated 50 ohm signal path when the 50 ohm input is selected. A relay is used to swap across to the 50 ohm path. In the 50 ohm mode, the input VSWR of this scope is very good to well past 500MHz. It's still quite an old scope though...

[Randy222]

....But my 1meg terminator itself does show it has some pF, it does not appear to be 0F. ...

If you do watch that video, you may notice that the two terminators I made back in 2016 exhibited a worse return loss than the OEM part when connected  to my scope.  That was a big hint to what was coming.  I posted how the OEM part was nothing more than an axial resistor.  That added parasitic inductance improved the performance.   Swapping that axial part for the 4, then the 2 X 1206's lowered that inductance.

I did not watch vid yet, but will.

My two items are just axial metal film 1% 1/4w.

One is a single 1meg.
The 2nd is two 2meg in parallel, but they loop away from each other.

Both poorly made items have the resistors soldered to center pin with about zero space between center pin and resistor, then the other axial lead bends around on about 1/16" radius on 180deg turn to reach back to ground.

Thanks for all the info.

[G0HZU]

Here's a VSWR plot of the Tek MSO4104 when set to the internal 50R path. I've measured it out to 1GHz.

The VSWR is quite low up to about 850MHz. If I switch to the 1Meg input and try and fit an external 50R through termination (a commercially made BNC model) the VSWR is then <1.3:1 up to about 100MHz. Above this frequency, the VSWR rises quite fast. By 200MHz it hits 1.9:1 for example.

I can swap this through termination for an inductively compensated 50R termination and this does improve the VSWR, but it will cause some rising and falling edge issues when trying to look at a very fast pulse.

[joeqsmith]

A better model to me would as a minimum include the scopes transfer function for the different ranges.   

Looking at the manual for my LeCroy 7200, they spec the VSWR at 1.2, BW is 4GHz with a 125pSec transition for 50mV and up.  Otherwise, 3GHz / 150ps.  Isolation is 60dB min at 1GHz. 

Normally, I'm not using the scopes inputs directly.  More useful for me is how the probe+scope effect the circuit being measured.  See attached:

[joeqsmith]

...
I've been experimenting with Z0 probes myself. I have a design with multilayer PCB, and pogo pin to pick up signals, or 100 mil pin header at the end. I think I got quite good results, below is a comparison to a N2795A active probe. This is probably the sharpest square wave I could do without designing something, it's for sure is limited by the source. It shows faster rise time than the N2795A so I'm hoping it will be above a GHz, but I'm yet to make a test setup to measure it.
This is at work. The reason I'm interested in the through terminators is because I cannot really afford a new scope with 50 Ohm inputs at the home lab.
...

2.2ns edge rates are fairly slow.  As a comparison, back in 2015 a member was posting some data for a ring oscillator they had made using an LS part.  I still had several old 7400 DIP parts and joined the fun.   Shown is an F with fall times around 600ps.

https://www.eevblog.com/forum/projects/waveforms-in-a-74ls04-ring-oscillator/msg645166/#msg645166
I had made a cooler for this test and tried a National military F but could not achieve those edge rates.
https://www.eevblog.com/forum/projects/waveforms-in-a-74ls04-ring-oscillator/msg1033429/#msg1033429

[G0HZU]

I've got a little pulse gen board here that uses AC logic gates and it is also quite fast for rise time. I'll dig it out tomorrow.

Usually, the basic scope model does quite well on all attenuator ranges. Often the changes in input Rs are quite minor, especially with the newer scopes that use cheaper technology in the front end.

By contrast, my old 500MHz HP Infinium scope does show a fairly significant change in the Rs series resistance according to the attenuation selected. This scope uses old school relays and a very exotic looking front end PCB with printed components. There are different sized printed resistors ahead of each attenuator so I assume this corresponds to the differences I see on the VNA.

Normally, it will be this series Rs that will get hot if the scope is driven with large AC waveforms up at V/UHF. Some scopes have to be used with caution in this respect or they can easily be damaged. The 1Meg input might only be rated at about 3Vrms up at UHF for example.

[joeqsmith]

I have that PECL driver I show in the video that I put together but even that is fairly slow.    That early 1970's Tektronix tunnel diode I showed is the fasted pulser I have.   That old WaveMaster isn't near fast enough to show anything.  I did recently have it and smacked a filter with it for fun to look at the resonance.  Signals are far beyond the limits of that old scope. 

https://www.eevblog.com/forum/testgear/show-us-your-square-wave/475/

[tszaboo]

...
I've been experimenting with Z0 probes myself. I have a design with multilayer PCB, and pogo pin to pick up signals, or 100 mil pin header at the end. I think I got quite good results, below is a comparison to a N2795A active probe. This is probably the sharpest square wave I could do without designing something, it's for sure is limited by the source. It shows faster rise time than the N2795A so I'm hoping it will be above a GHz, but I'm yet to make a test setup to measure it.
This is at work. The reason I'm interested in the through terminators is because I cannot really afford a new scope with 50 Ohm inputs at the home lab.
...

2.2ns edge rates are fairly slow.  As a comparison, back in 2015 a member was posting some data for a ring oscillator they had made using an LS part.  I still had several old 7400 DIP parts and joined the fun.   Shown is an F with fall times around 600ps.

https://www.eevblog.com/forum/projects/waveforms-in-a-74ls04-ring-oscillator/msg645166/#msg645166
I had made a cooler for this test and tried a National military F but could not achieve those edge rates.
https://www.eevblog.com/forum/projects/waveforms-in-a-74ls04-ring-oscillator/msg1033429/#msg1033429

I know 2.2ns is nothing to write home about, It was one of the whatever MCU board I had lying around. I just wanted to see if it's comparable to a GHz active probe. It was already outperforming a 500MHz passive probe that Keysight packs with the scope, so I was about to make a test setup with a high speed clock buffer, but then it's holidays time now.

[G0HZU]

I managed to find my old logic gate boards. These have HCT and AC gates and I drive them from an old Datapulse pulse generator. By passing the pulse through several gates, it does achieve a reasonably fast rise time at the output. According to the 4104 scope, the rise time is about 600ps. The original goal for these boards was to try and find the source impedance of the gates, or at least to find the optimal series resistor to put at the output when driving 50R coax.

The plot below shows a straight 50R measurement via a series matching resistor and some 50R coax into the 50R input of the 4104.

The channel 2 trace is my old Zo probe. This probe has lots of bandwidth, probably flat to about 3GHz. It's a 30dB probe that I normally use with a VNA or a spectrum analyser.

You can see that the probe did OK here. It does add a few artefacts to the pulse shape, but I'm not sure what to expect from this setup in terms of the 'perfect' pulse shape.

It really is quite important to have a scope with a low VSWR over a huge bandwidth in order to get a result like this when using a Zo probe. In the past, I've used an attenuator at the scope input with the 500MHz Infinium scope. The plot below is with the Zo probe directly fed into the 50R input of the Tek 4104.

[tszaboo]

I managed to find my old logic gate boards. These have HCT and AC gates and I drive them from an old Datapulse pulse generator. By passing the pulse through several gates, it does achieve a reasonably fast rise time at the output. According to the 4104 scope, the rise time is about 600ps. The original goal for these boards was to try and find the source impedance of the gates, or at least to find the optimal series resistor to put at the output when driving 50R coax.

You got this square wave from a bunch of generic inverters? Damn, I haven't even considered to try that, I was expecting much worse results. I have to try this when I get back to the lab.

[G0HZU]

Looking at the board, it is just a couple of AC gates in series and a 39R series output resistor. The HCT version used four gates each with a series resistor and the outputs are then summed together into a single output. The rise-time of the HCT version is a lot slower.

Both chips are SMD and are mounted upside down on some bare copper board. I think the HCT chip is a 74HCT244 and the AC gate version is probably just a hex inverter. I can just about make out 74AC04 on the bottom side of the board where I once labelled it with a marker pen. The input to the AC series chip is to pin 1 and pins 2 and 3 are connected together and the output is from pin 4. It's tempting to try going through three gates in series to see if there is any improvement.

The pulse gen I use to drive it is nothing special. It's an old Datapulse 101 with a ~5ns risetime. The AC gates speed this up to a sub 1ns rise time.

The -30dB Zo probe is a simple homebrew design that works to about 3GHz. I've got several homebrew Zo probes here and one of them is good to about 6GHz.

[G0HZU]

Here's what happens if I put the Zo probe on channel 3 set to 1Meg input and I fit a commercial 50R BNC through termination at the input of channel 3. You can see lots of artefacts appear now and these are caused by (re)reflections in the cable of the Zo probe. It's possible to estimate the cable length by looking at the delay time of the artefacts as they appear on the waveform.

[Randy222]

Here's what happens if I put the Zo probe on channel 3 set to 1Meg input and I fit a commercial 50R BNC through termination at the input of channel 3. You can see lots of artefacts appear now and these are caused by (re)reflections in the cable of the Zo probe. It's possible to estimate the cable length by looking at the delay time of the artefacts as they appear on the waveform.

Now put a ~3db 50ohm attenuator on front side of feed thru. Does the trace look better?

[joeqsmith]

I know 2.2ns is nothing to write home about, It was one of the whatever MCU board I had lying around. I just wanted to see if it's comparable to a GHz active probe. It was already outperforming a 500MHz passive probe that Keysight packs with the scope, so I was about to make a test setup with a high speed clock buffer, but then it's holidays time now.

Point being for lower speed signals like this, it may not make enough difference. 

Shown is the fast edge from my old Tektronix scope demo board.  C1 is a standard 10x probe, M1 is a LeCroy resistive probe attached using that Pasternak thru terminator.  M2 is the same resistive probe, using the scopes internal 50 ohm termination.    Sorry, I didn't comp the 10x probe but point being there is little difference when working with these slower signals.   

***
Just in case there is some confusion, I have attached a link showing the board.  There is no terminator.  The resistive probe presents more of a load than the 10X at DC.  Again, crossover for these two probes was around 60-70MHz where the 10X probe starts to load things down.   Best to know what you want to measure and buy the equipment for the job. 

https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg5164857/#msg5164857

[G0HZU]

Here's what happens if I put the Zo probe on channel 3 set to 1Meg input and I fit a commercial 50R BNC through termination at the input of channel 3. You can see lots of artefacts appear now and these are caused by (re)reflections in the cable of the Zo probe. It's possible to estimate the cable length by looking at the delay time of the artefacts as they appear on the waveform.

Now put a ~3db 50ohm attenuator on front side of feed thru. Does the trace look better?

See below