DC coupled 2.7 GHz Active Probe Project - Now Available!

[lasmux]

Looks cool. Those socket connectors at the end, are they the same size as the Agilent active probes?
For me it's a bit strange that you choose to have the 1M to be in series with the signal, those probes have a relatively small series resistance, and the 1M is used to shunt the 1pF. Any reason for that? Not judging, I'm genuinely curious, as I haven't built high speed scope probe yet, only high gain ones for shunt resistors.

I'm not sure the size of the keysight/agilent probes. I think they might be 25mil (0.635mm), the ones I selected are 1mm, so a bit larger maybe. Would that be a problem for you? I will include solderable adapters that you could maybe connect up to those probes, but that's not as seamless.
I could theoretically replace the socket with this which I think would be compatible if you think it would be a deal breaker, although I'd have to change the enclosure which would be a bit of a faff
https://www.mill-max.com/products/pin-receptacle/wire-termination-receptacle-soldercup-type/1134


The 1M is to make the probe a bit more general purpose for low speed signals. Above a few MHz the input resistor divider between R2 and R3 has almost no bearing on the signal attenuation. It's mostly down to the capacitor divider between the parasitic capacitance across the 1M, and C1. To get a linear frequency response from low frequencies to high frequencies, you need to closely match the resistor divider and the capacitor divider attenuation. So in that sense it doesn't matter whether it's 1M or 10k, it just pushes the frequency out further at which point the capacitor divider starts to dominate. But the capacitor divider still needs to be correct due to the high bandwidth of the measurement.

[tszaboo]

Looks cool. Those socket connectors at the end, are they the same size as the Agilent active probes?
For me it's a bit strange that you choose to have the 1M to be in series with the signal, those probes have a relatively small series resistance, and the 1M is used to shunt the 1pF. Any reason for that? Not judging, I'm genuinely curious, as I haven't built high speed scope probe yet, only high gain ones for shunt resistors.

I'm not sure the size of the keysight/agilent probes. I think they might be 25mil (0.635mm), the ones I selected are 1mm, so a bit larger maybe. Would that be a problem for you? I will include solderable adapters that you could maybe connect up to those probes, but that's not as seamless.
I could theoretically replace the socket with this which I think would be compatible if you think it would be a deal breaker, although I'd have to change the enclosure which would be a bit of a faff
https://www.mill-max.com/products/pin-receptacle/wire-termination-receptacle-soldercup-type/1134


The 1M is to make the probe a bit more general purpose for low speed signals. Above a few MHz the input resistor divider between R2 and R3 has almost no bearing on the signal attenuation. It's mostly down to the capacitor divider between the parasitic capacitance across the 1M, and C1. To get a linear frequency response from low frequencies to high frequencies, you need to closely match the resistor divider and the capacitor divider attenuation. So in that sense it doesn't matter whether it's 1M or 10k, it just pushes the frequency out further at which point the capacitor divider starts to dominate. But the capacitor divider still needs to be correct due to the high bandwidth of the measurement.

Problem? No, I have those 1GHz active probes at work (N2795), and saw how versatile  the accessory package is. According to my caliper, the pins are 0.7mm. I don't know if they sell the probe pins separate.
The equivalent input on these are 120Ohm series, 1pF 1Mohm parallel to ground.

[JohnG]

Thanks!
The slew rate of the output is limited to 2V/ns, so for example a 10V rising edge input would be divided by 10x due to the probe attenuation to 1V (before 50 ohm termination), so due to the slew rate limitation would have an additional 500ps rise time (I think) (edit. around a 600ps rise, see post below). You'd have to have quite the unusual signal to be generating this kind of signal though. In my spice simulations this kind of input starts to make step responses/square waves look a little trapezoid. I don't have a high amplitidue pulse generator (or fast enough oscilloscope) to test this properly. In general, I don't know what the rise time is for more sensible signals as my oscilloscope just isn't fast enough (500MHz).

Note though that other active probes such as the Keysight N2796A 2GHz probe also limit their dynamic range at higher frequencies. This is a screenshot from their datasheet:

I measure such signals on a routine basis, so it doesn't seem so unusual to me (maybe I am an outlier, though).

Regarding the Agilent/Keysight probe example, usually such a graph shows probe derating due to some reliability or safety limitation. The same rolloff is typical for completely passive probes which will not exhibit slew rate limiting. The data sheet was not very clear on this, but from the impedance plot, an 8V AC amplitude signal at 1 GHz will give you about 50ish mA AC RMS of probe tip current, which is not insignificant. These probes have an adjustable offset of +/-12V, which is how they arrive at the 20V max input. This is not slew rate limiting, this is a way to get around the AC voltage limitations of the probe tip amplifier. It's useful for looking at noise on a voltage bus, for example, by adjusting the offset to the bus voltage so you have the full AC dynamic range centered around the bus voltage.

As I mentioned, I'd be happy to test the probe. I have a Bodnar pulser and an HP8131 pulse generator, and an old, but very functional, 6 GHz scope. I should warn you ahead of time that I'm not real quick about getting to these things, so if you find someone locally, you are likely to get a quicker turnaround. Also, if it is really slew rate limited to 1V/ns at the output, it's not something I can use.

John

[lasmux]

Looks cool. Those socket connectors at the end, are they the same size as the Agilent active probes?
For me it's a bit strange that you choose to have the 1M to be in series with the signal, those probes have a relatively small series resistance, and the 1M is used to shunt the 1pF. Any reason for that? Not judging, I'm genuinely curious, as I haven't built high speed scope probe yet, only high gain ones for shunt resistors.

I'm not sure the size of the keysight/agilent probes. I think they might be 25mil (0.635mm), the ones I selected are 1mm, so a bit larger maybe. Would that be a problem for you? I will include solderable adapters that you could maybe connect up to those probes, but that's not as seamless.
I could theoretically replace the socket with this which I think would be compatible if you think it would be a deal breaker, although I'd have to change the enclosure which would be a bit of a faff
https://www.mill-max.com/products/pin-receptacle/wire-termination-receptacle-soldercup-type/1134


The 1M is to make the probe a bit more general purpose for low speed signals. Above a few MHz the input resistor divider between R2 and R3 has almost no bearing on the signal attenuation. It's mostly down to the capacitor divider between the parasitic capacitance across the 1M, and C1. To get a linear frequency response from low frequencies to high frequencies, you need to closely match the resistor divider and the capacitor divider attenuation. So in that sense it doesn't matter whether it's 1M or 10k, it just pushes the frequency out further at which point the capacitor divider starts to dominate. But the capacitor divider still needs to be correct due to the high bandwidth of the measurement.

Problem? No, I have those 1GHz active probes at work (N2795), and saw how versatile  the accessory package is. According to my caliper, the pins are 0.7mm. I don't know if they sell the probe pins separate.
The equivalent input on these are 120Ohm series, 1pF 1Mohm parallel to ground.

My probe divides the signal prior to the amplifier, and then re-amplifies it. The N2795 doesn't attenuate at all to begin with, and then I guess has the amplifier gain configured to divide by 10 perhaps. I'd guess that that could also be why their ESD damage tolerance is so low also, the input is only protected by the 120R resistor.

[lasmux]

Thanks!
The slew rate of the output is limited to 2V/ns, so for example a 10V rising edge input would be divided by 10x due to the probe attenuation to 1V (before 50 ohm termination), so due to the slew rate limitation would have an additional 500ps rise time (I think) (edit. around a 600ps rise, see post below). You'd have to have quite the unusual signal to be generating this kind of signal though. In my spice simulations this kind of input starts to make step responses/square waves look a little trapezoid. I don't have a high amplitidue pulse generator (or fast enough oscilloscope) to test this properly. In general, I don't know what the rise time is for more sensible signals as my oscilloscope just isn't fast enough (500MHz).

Note though that other active probes such as the Keysight N2796A 2GHz probe also limit their dynamic range at higher frequencies. This is a screenshot from their datasheet:

I measure such signals on a routine basis, so it doesn't seem so unusual to me (maybe I am an outlier, though).

Regarding the Agilent/Keysight probe example, usually such a graph shows probe derating due to some reliability or safety limitation. The same rolloff is typical for completely passive probes which will not exhibit slew rate limiting. The data sheet was not very clear on this, but from the impedance plot, an 8V AC amplitude signal at 1 GHz will give you about 50ish mA AC RMS of probe tip current, which is not insignificant. These probes have an adjustable offset of +/-12V, which is how they arrive at the 20V max input. This is not slew rate limiting, this is a way to get around the AC voltage limitations of the probe tip amplifier. It's useful for looking at noise on a voltage bus, for example, by adjusting the offset to the bus voltage so you have the full AC dynamic range centered around the bus voltage.

As I mentioned, I'd be happy to test the probe. I have a Bodnar pulser and an HP8131 pulse generator, and an old, but very functional, 6 GHz scope. I should warn you ahead of time that I'm not real quick about getting to these things, so if you find someone locally, you are likely to get a quicker turnaround. Also, if it is really slew rate limited to 1V/ns at the output, it's not something I can use.
John

I think because of how I divide the voltage prior to the amplifier, I am less succeptible to damage induced by high amplitude/high frequency signals. Last time I checked (edit. in spice) for a 28V p-p signal at 2GHz into the probe, there would be less than 1mA RMS going into the op amp, with most of the power dissipated within the input resistor network, which can handle it. I do need to run some more simulations on this though. But thanks for the clarification, I had misunderstood that the dynamic range line on that graph was not referring to the probe dynamic range in general, but more focusing on the voltage derating.

The slew rate is limited to 2V/ns at the output (remember the 10x output attenuation). For example, your HP8131 has a maximum p-p voltage of 5V, with a 200ps transition time. The slew rate limited rise time on that signal would be just slew rate limited to 250ps on my probe (theoretically). So only slightly longer, maybe here the bandwidth is reduced to around 1.3-1.4GHz for such signals. For the Bodnar pulser, the maximum p-p voltage is only 1.2V, so the slew rate limitation of the probe wouldn't affect the reading, it would be bandwidth limited here and wouldnt achieve the 40ps rise time.

If no-one else volunteers, I would be very grateful if you'd have a look at the probe. It'll be really interesting to see how it performs on a fast oscilloscope!

[joeqsmith]

Looks like you have done a nice job with it.   Any idea on the selling cost yet?   With the 50 ohm drive, I can see using it for applications outside of a scope probe.   You mentioned using it with your VNA.  There was a thread some time back about someone wanting to make a buffer to drive their 50 ohm inputs.  I had posted a commercial one and some of my own but they are capped at a bit over 100MHz.   

If you are looking for people to evaluate them, I may be able to help as well.

Thanks for posting and keep us updated.   

[lasmux]

Hi Joe,
I've appreciated a lot of your forum posts on VNAs this past year I've been working on this project!

Currently I'm aiming for around £150 for the 1GHz version, and £180 for the 2GHz version on ebay. I've not gone through my BOM thoroughly yet though, so things may shift, potentially down. I can do it a bit cheaper for the forumites as 12% ebay fees wouldn't apply, I just need to figure out how to set up a store correctly on my website  .

And yeah, I think some kind of sensor preamp, or an input buffer for a VNA are very valid use cases.

[joeqsmith]

Very reasonable.   Just looking at used prices for the LeCroy 2.5GHz single ended HFP2500, about $400 - $500 USD.   Even then, you don't know if they work or how they were treated.  It's a total crap shoot.   

For the most part, when looking at single ended low voltage signals, I tend to stay with resistive dividers.  Cheap and minimalish loading.    I have a few LeCroy 4ish GHz diff probes for my scope.   I thought about trying to build a DC -  >>GHz differential probe for logic signals but at the time, I wasn't able to source popcorn parts that would do it.   I haven't checked in several years and would guess there are much higher performance parts available today.   

Having that DC-100MHz buffer around has been helpful.  I've used it to drive my spectrum analyzers as well.  Nice thing I can use what ever 1X // 10X  probe with it.  Downside, 100MHz.   

Thanks.  Yes, I've had a lot of fun playing with these low cost VNAs.  I'm very impressed with what they have been able to achieve in such a small package.  Especially with that LiteVNA and Dislords firmware.   Hopefully your scope probes will be as successful.   

[KungFuJosh]

I just need to figure out how to set up a store correctly on my website 

You already have woocommerce installed, you just need to make sure it's setup correctly, and setup the products, payments, and shipping options. I can help in exchange for a couple 2G probes. 😉

[lasmux]

Very reasonable.   Just looking at used prices for the LeCroy 2.5GHz single ended HFP2500, about $400 - $500 USD.   Even then, you don't know if they work or how they were treated.  It's a total crap shoot.   
For the most part, when looking at single ended low voltage signals, I tend to stay with resistive dividers.  Cheap and minimalish loading.    I have a few LeCroy 4ish GHz diff probes for my scope.   I thought about trying to build a DC -  >>GHz differential probe for logic signals but at the time, I wasn't able to source popcorn parts that would do it.   I haven't checked in several years and would guess there are much higher performance parts available today.   
Having that DC-100MHz buffer around has been helpful.  I've used it to drive my spectrum analyzers as well.  Nice thing I can use what ever 1X // 10X  probe with it.  Downside, 100MHz.   
Thanks.  Yes, I've had a lot of fun playing with these low cost VNAs.  I'm very impressed with what they have been able to achieve in such a small package.  Especially with that LiteVNA and Dislords firmware.   Hopefully your scope probes will be as successful.   

I'm glad you think the pricing is reasonable. In general, probes are in a weird space in terms of pricing. They don't feel like they should be so expensive, but they are quite difficult to get right, and contain expensive and low volume custom components. That said, I feel commercial active probes are unreasonably expensive. High speed passive probes aren't much better.

Yeah, DIY Zo probes are definitely a reasonable (and dirt cheap) alternative. That said, their linearity can be a bit questionable, and I think if you measured the input impedance on one of them (I've not done this), I suspect it may be a bit variable. The inductance of the ground connection really screws up everything, which is why the expensive active probes often provide ground blade/leaf connections. In my probe I can counter the issue with a resistive ground connection. I don't think that would work in a Zo probe.

I'll send you a PM tomorrow regarding sending you a probe for you to have a play with. Thanks for offering to help out

I just need to figure out how to set up a store correctly on my website 

You already have woocommerce installed, you just need to make sure it's setup correctly, and setup the products, payments, and shipping options. I can help in exchange for a couple 2G probes. 😉

I really hated setting up that website lol. The woocommerce plugin is just rammed full of features that make figuring out how to do something simple, very difficult. I really just need to watch a few tutorials on youtube and I'm sure it'll all be easy.

[KungFuJosh]

I really hated setting up that website lol. The woocommerce plugin is just rammed full of features that make figuring out how to do something simple, very difficult. I really just need to watch a few tutorials on youtube and I'm sure it'll all be easy.

If you're only selling 2 products, you can skip woocommerce and just use PayPal buttons. If you plan on expanding later, that's a different story.

[lasmux]

I had another look around, there's a little bit of testing of Zo probes done by... you!
https://www.eevblog.com/forum/testgear/fifty-ohm-probes/msg607659/#msg607659
And some others:
https://www.eevblog.com/forum/projects/lo-z-probe/msg801477/#msg801477
There's also this one:
http://jahonen.kapsi.fi/Electronics/DIY%201k%20probe/

Everyone seems to get very different results. Not quite sure what's going on. I think my active probes are probably better performing up to their stated bandwidth, but lower bandwidth potential than a well built Zo probe. Although getting a linear response across the bandwidth seems quite tricky on a Zo probe.

I really hated setting up that website lol. The woocommerce plugin is just rammed full of features that make figuring out how to do something simple, very difficult. I really just need to watch a few tutorials on youtube and I'm sure it'll all be easy.

If you're only selling 2 products, you can skip woocommerce and just use PayPal buttons. If you plan on expanding later, that's a different story.

I might go with this. Thanks for the suggestion.

[joeqsmith]

I had another look around, there's a little bit of testing of Zo probes done by... you!
..
Everyone seems to get very different results. Not quite sure what's going on. I think my active probes are probably better performing up to their stated bandwidth, but lower bandwidth potential than a well built Zo probe. Although getting a linear response across the bandwidth seems quite tricky on a Zo probe.

Indeed, I have tried various configurations when making resistive probes.  Some better than others.   You can image that component selection and construction is going to have a major effect as you move beyond a GHz. 

Another from 2017 where I was playing with a cheap demo board.  The resistive elements were placed onto a board. 

https://www.eevblog.com/forum/microcontrollers/typical-speed-of-fpgas/msg1279898/#msg1279898

Even my LeCroy PP061 probes I mentioned are not good for these higher frequencies.  The diff probes I have are good for about 4GHz.  I also have the PP066 which are a 7.5GHz resistive probe.     

https://cdn.teledynelecroy.com/files/manuals/pp066-rp4030-user-manual.pdf

[joeqsmith]

It would be nice if probes did not load down the circuit you are wanting to test.  I started out building one of Bob Pease's active FET probes.  From memory, maybe 100MHz BW.   It saved me a few times as a young engineer when the 10X probes we had were presenting too much load for my circuits.   For home, I eventually bought a Tektronix P6202 active probe.  This served me well until I started playing with faster circuits.   

Capture1 lists the higher BW probes I have for my hobby use along with the basic specs.  I included yours as a reference. 

I have a LeCroy coplanar waveguide test board and used this as the thru for the LiteVNA.   I then attached both a LeCroy PP061 and PP005 probe to a scope and used them to probe the board.   I was never able to locate a manual or data for the PP061.

Capture2 is looking at S21.  The old 10x probe presents less of a load than the resistive probe until the cross over point at 70MHz. 

Capture3  is looking at S11.  The impedance the VNA sees with the probe attached.  This is not the probe, but the probe in parallel with port 2.   Typically I would have some driver and load and want to probe between them. 

I think to run this sort of test I would want to use my old PNA.  Some of my probes are limited to about -4dBm.  And I would want to use the unknown thru model for that waveguide.   

***
Added photo of my old Tektronix active probe along with with the probes for my old WaveMaster.   

[lasmux]

You have a lot of probes!

Thanks for this data. Interesting. I guess with a lot of the active probes, their manufacturer specific interface makes it very difficult to actually measure their response. Obviously you're able to measure the impedance. Regarding capture 2... it looks like it's true that passive probes don't help signal integrity  The same is also true of certain Ukrainian active probes that can be found on Ebay these days. I bought one because it was cheap, and the signal loading was not pretty. Response linearity was apalling too. Not bad for $25, but not that useful either.

I personally have used three methods to measure the impedance.
1. I use a SMA T piece on the VNA port 1, with a 50 ohm termination. I then place the probe in the other side of the T piece. This is better than using the port 2 as termination as I think there will be worse reflection artefacts by not measuring near the termination point.
2. I use a custom built 50 ohm load which has had it's resistors exposed, and measure directly across those.
3. A custom 'open' SMA termination which I can measure directly across.

1 and 2 tend to give fairly close results. 3 tends to give lower impedance values at lower frequencies for some reason. I'm not sure if this is because it's trying to measure impedances much much higher than 50 ohms. Maybe it gets less accurate. I'm not sure.

For bandwidth measurements, I use the same setup method 1/2, but then have the probe output measured by the VNA port 2.

[joeqsmith]

LeCroy does or did offer an accessory that allows their probes to be used with a standard scope.  Keysight and Tektronix also offers this for some of their probes.   I have yet to find a used one for mine. 

I would think to measure a probes impedance you want to have a fixture that you can remove all the errors right to where the probe attaches.   No added 50 ohm termination, no Ts, no extra length...   Your option 3 with an open that had the reference plane matched to the SOL standards, then just measure S11.   For low frequencies (GHz ish) I would just use those coplanar waveguide boards.  Then attach the probe in place of where the two 100 ohm resistors attach.   Beyond about 1.5GHz, I wouldn't trust it.   

Where my low cost VNAs don't have good return loss for port 2, the old PNA is another story.   Still just to see the loading effects between those two probes, good enough to use the Lite. 

***
I should add that when you measure the impedance, you need to choose the proper method for best accuracy.  Depending on the range, you would use the shunt, shunt thru, or series method.    In your case, you may want to use two different methods to cover the range.   


For BW,  I would still stick with the terminated coplanar waveguide, but in this case, I have one from LeCroy.   The T adds error.  Same with the custom terminator.  Consider cal'ing at the end of the waveguide, then leaving the cal standard attached when making the measurement.  You just want to know the probe.  Hope this makes some sense.   

The other test would be to just measure some basic digital signals and compare results.

I had told a friend about your project and projected selling price.  I looked up the current cost of the PP005 on Digikey and even that probe would cost more.  So yes, I think your target price is still very fair, assuming the probe works.   Let me know once you have these first set built up.   

***
Reminds me of my attempt to replicate someones experiment where they attempted to measure the capacitance of some resistors.  Far away from 50 ohms....
 
https://www.eevblog.com/forum/rf-microwave/shunt-capacitance-of-1206-smd-resistors-jeroen-belleman-december-2010/

[joeqsmith]

The PP005A attached to the home made coplanar waveguide with the terminators not populated.   The probe wasn't terminated.  Tip capacitance seems to measure around 6p rather than 11.  Note too that rather than 10M, it measures 100k (300 shown).   If I have time tonight, I will try and measure it in series with my low frequency network analyzer to give you some idea of the difference.   That NA can go up to 150MHz. 

[joeqsmith]

Using a PECL buffer that feeds a splitter.  One leg which goes to one of the scopes inputs (50ohms).  The other goes to the LeCroy coplanar waveguide test board which was then terminated to 50 ohms.   Scope probe when then attached to second channel on scope and then used to probe the waveguide. 

No load:  No second probe attached.  This scope is only spec'ed for 600MHz (approx 600ps), and that PECL driver is about 350ps.  Cable adds a fair bit of loading.  Idea is to show the effects of the probe loading.

PP002:  This is a 350MHz 10X probe.   Causes a fair bit of loading.  Think about that PECL signal going off to some other part of the circuit rather than the scope.   Also, with the splitter, things are somewhat isolated.   

PP061:  The resistive probe still has some loading (not a great probe)  but much better than the 10X probe.   

Once you get a bit closer,  I'll bust out my faster scope and active probes.     

[joeqsmith]

Next, I removed the splitter and connected the LeCroy waveguide directly to the PECL driver.  The opposite end of the waveguide connected to the scopes 50 ohm input.  Basically, how you would probe a typical signal on a board but the scope is again the target load.     

145:  The PP061 resistive probe pink.  M1 is the unloaded waveguide (nothing attached).  C1 is the loaded waveguide.  Again, some loading.  Scopes BW is limiting the edge to about 600ps. 

146:  The PP005 500MHz 10X probe.   It really messes with the signal.  I did not clear the traces so ignore the averaged data.

147: The P6202A 500MHz active probe.   Has very little loading effects compared with the 10X passive probe.   According to the link below, the probe cost about $1300 in 1998.  The cables are thick and heavy making the probe difficult to use.   You can start to see why the resistive probes were a better choice in most cases.   

http://www.barrytech.com/tektronix/probes/tekp6202a.html

[lasmux]

I think your second set of measurements will be better? With the splitter you will have two 50 ohm terminations (to ground, not VCC-2V) on the PECL driver, which it might not like? I've not used PECL before though to be honest, only a bit in simulation.

Very interesting how the PP005 really does a lot of damage to the signal. The PP061 lowers the PECL voltage levels slightly with it's loading? But the shape of the loaded waveform is more similar. The P6202A waveform has almost no loading (at these frequencies), but the output waveform maybe has slightly more peaking on the rising edge? It's still qualitatively very similar.

[joeqsmith]

I think your second set of measurements will be better? With the splitter you will have two 50 ohm terminations (to ground, not VCC-2V) on the PECL driver, which it might not like? I've not used PECL before though to be honest, only a bit in simulation.

I had used a resistive splitter, so with the two ports loaded with 50 ohms, the PECL driver sees 50 ohms.  The signal the scope sees is 6dB lower than the original.  The reason I thought about running this test was to somewhat isolate the two ports.   

https://www.microwaves101.com/encyclopedias/resistive-power-splitters

Very interesting how the PP005 really does a lot of damage to the signal. The PP061 lowers the PECL voltage levels slightly with it's loading? But the shape of the loaded waveform is more similar. The P6202A waveform has almost no loading (at these frequencies), but the output waveform maybe has slightly more peaking on the rising edge? It's still qualitatively very similar.

From my previous posts where I provided some basic metrics for my probes, you can see the PP061 is a 10X probe, so the DCR was 500 ohms (which includes the scope's input).   The waveguide was loaded with 50 ohms by the scopes other channel.  Now we added another 500 ohms.   Your comment about the P6202A "at these frequencies" is key.   It's not a perfect 500 ohms and the response is not going to be flat as we start to look at faster signals.   

Currently the scope is doing a good job of masking what the actual signal looks like.       

***
Thinking about your previous comment, I am guessing you thought I was using a "T" which is not the same thing as a splitter.

[joeqsmith]

They claim 1GHz, for $250 and it is not DC coupled.  I would have no use for something like this.   
https://www.ebay.com/itm/254162548019?hash=item3b2d44b533

$29 and again, not DC coupled. 
https://www.ebay.com/itm/175751864133?hash=item28eba09f45

There doesn't appear to be a low cost 2GHz product that would compete with you.  The only choice for the hobbyist is the used market...

[joeqsmith]

I had another look around, there's a little bit of testing of Zo probes done by... you!
....
Everyone seems to get very different results. Not quite sure what's going on.

Because much of what I had shown here by attempting to embed resistors in the cable and such, isn't really how I typically would make a probe, I wanted to show what something a bit more common.   This is a 20X (953 ohm) probe.   Zoomed in 1ns/div.  Again, fairly low frequencies and the scope is hiding the details but hopefully it helps show that the results shouldn't vary by much.    The cable (Teflon) and connector came from Pasternak.   Surface mount resistor is stabilized with a bit of shrink tube.  Not a whole lot invested.  Maybe a half hour labor and $15.   

Maybe we can run some of these on the VNA and my faster scope later on to help paint a clearer picture.   

[lasmux]

Currently the scope is doing a good job of masking what the actual signal looks like.       
[...]
Thinking about your previous comment, I am guessing you thought I was using a "T" which is not the same thing as a splitter.

Ah thanks about clarifying about the splitter. I had indeed misunderstood, and yes, I have the same problem with my oscilloscope. The true signal form is hidden behind the insufficient bandwidth.

With regards to those other probes, I have seen those. They're not great tbh. The first ones construction looks quite good, but the input capacitance is 3.5pF. This is really quite bad, not far off a good passive probe. As you said, not DC coupled either. The second one looks very similar to other active probes based on the BF-998 that are on ebay. I have a similar version from Ukraine. The frequency response linearity is awful (swings +/- 8db or so across the bandwidth), and the input capacitance is in reality much higher than the quoted 0.5pF. I measured the tip loading drop to 20 ohms at around 500MHz, which again, is really bad.

I had another look around, there's a little bit of testing of Zo probes done by... you!
....
Everyone seems to get very different results. Not quite sure what's going on.

Because much of what I had shown here by attempting to embed resistors in the cable and such, isn't really how I typically would make a probe, I wanted to show what something a bit more common.   This is a 20X (953 ohm) probe.   Zoomed in 1ns/div.  Again, fairly low frequencies and the scope is hiding the details but hopefully it helps show that the results shouldn't vary by much.    The cable (Teflon) and connector came from Pasternak.   Surface mount resistor is stabilized with a bit of shrink tube.  Not a whole lot invested.  Maybe a half hour labor and $15.   
Maybe we can run some of these on the VNA and my faster scope later on to help paint a clearer picture.   

Indeed, there is a very low loading here! Would definitely be interesting to see how it performs on the VNA/faster scope.

[joeqsmith]

Before doing this, another option may be to make a new one that uses a high grade SMA rather than the BNC.  I should have some cable and connectors on-hand.   The tip loop is also fairly large on this one.   I wouldn't want to give you the wrong impression about these  homemade probes once we start speeding things up.   

To look at the BW with the VNA, I am thinking we stay with the same waveguide.  Terminate the backside to 50.  Then attach the probe to the second port.  Using my homemade probe, we should see 26dB.   I would need to use my old Agilent for this.   Maybe then check the tip capacitance as well.  I could run your probe as well as my resistive probes but none of my active or standard 10X probes.