DC coupled 2.7 GHz Active Probe Project

[joeqsmith]

Thanks for the update. 

Update time. I have a few things to talk about 
...
I've also been doing some work investigating the usage of stands and the effects on the frequency response of the probe. I've found that this is only really the case when using a VNA to measure the response of the probe. In usage with an oscilloscope, there is no affect on the output of the the probe, whether it's handheld or held using a stand. I'm still not sure the underlying reason why with VNAs, the probe benefits from grounding with a copper strip when using a stand.
...

I doubt there is a difference, outside of what you are measuring.  The last data I collected showed the 3dB loss at 1.5GHz.  I could certainly repeat this test using a 1.5GHz source and scope. 

***
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg5547989/#msg5547989

[lasmux]

Thanks for the update. 

Update time. I have a few things to talk about 
...
I've also been doing some work investigating the usage of stands and the effects on the frequency response of the probe. I've found that this is only really the case when using a VNA to measure the response of the probe. In usage with an oscilloscope, there is no affect on the output of the the probe, whether it's handheld or held using a stand. I'm still not sure the underlying reason why with VNAs, the probe benefits from grounding with a copper strip when using a stand.
...

I doubt there is a difference, outside of what you are measuring.  The last data I collected showed the 3dB loss at 1.5GHz.  I could certainly repeat this test using a 1.5GHz source and scope. 

***
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg5547989/#msg5547989

Okay yes, that would be a more precise measurement. Specifically looking at 1.5GHz, where you saw the change. I have measured signals with harmonics around that frequency band (1.5Ghz-1.8GHz), with which I've not noticed a change, but I can have another look with the fundamental frequency of a measured signal being 1.5GHz-1.7Ghz. Now that I have a stand built it's very straightforwards to test for me

[lasmux]

Thanks for the update. 

Update time. I have a few things to talk about 
...
I've also been doing some work investigating the usage of stands and the effects on the frequency response of the probe. I've found that this is only really the case when using a VNA to measure the response of the probe. In usage with an oscilloscope, there is no affect on the output of the the probe, whether it's handheld or held using a stand. I'm still not sure the underlying reason why with VNAs, the probe benefits from grounding with a copper strip when using a stand.
...

I doubt there is a difference, outside of what you are measuring.  The last data I collected showed the 3dB loss at 1.5GHz.  I could certainly repeat this test using a 1.5GHz source and scope. 

***
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg5547989/#msg5547989

I just did the test quickly at 1.6GHz and 1.5GHz. A 1.6GHz signal went from a 506mV peak-to-peak, to a 525mV peak-to-peak readout, when I grab the probe whilst its in the stand. A 1.5GHz signal went from 645mV to 670mV. So about a -0.33dB difference for both.

Like I said from earlier testing, signal traces didn't change visually when handling or not handling the probe.

[joeqsmith]

I just did the test quickly at 1.6GHz and 1.5GHz. A 1.6GHz signal went from a 506mV peak-to-peak, to a 525mV peak-to-peak readout, when I grab the probe whilst its in the stand. A 1.5GHz signal went from 645mV to 670mV. So about a -0.33dB difference for both.

Like I said from earlier testing, signal traces didn't change visually when handling or not handling the probe.

Your setup even peaks up rather than attenuates the signal.   

***
Reading through the old posts, are you testing with the same hardware revision of what was sent?   You also talk about adding a ground strap.   Can you post a photo of your setup when you ran this test? 

[lasmux]

The signal is attenuated when not holding the probe. I see that my wording before was potentially a bit misleading. This is the same behaviour (but greatly reduced) as what is observed in a VNA I think?

The hardware revision is different, but nothing has changed in the sensor/amplifier part of the probe, only the power supply. With this hardware revision, I still see the same large attenuation effect when using VNAs with a stand (not handheld), and without the ground strap.

Theres a photograph of an example ground strap setup in the datasheet on page 7:
https://www.lasmux.com/wp-content/uploads/2025/04/LD-ASP-2.7_r1_4.pdf
Edit:
Very rough diagram of setup which doesn't show distortion in frequency response.

[joeqsmith]

I don't remember and will need to review that part of the video to know for sure what I was seeing at the time.   

For this test, are you grounding the probe like figure 9, or figure 7 of your document?

***
Also, when you are using the holder in place of your hand, is the holder metal?  Grounded? 
***
Reviewing the video I made,  it does peak up as I place my fingers near the probe's body.  Which is what you are seeing.  Keep in mind that the probe holder I am using is metal. It is also grounded through the base.  Think of it as being a metal rod placed along side of your probe, then I add my fingers to the mix.  I'll set it back up and let you know if I can provide any additional insight.   It should reproduce.

I wonder what you are using for the USB signal generator?  Are you trying to use one of the low cost VNAs in CW mode as an RF source?   If so, which?

[lasmux]

The stand has a large piece of metal inside of it to act as a weight, but this is isolated from the probe. There is no electrical contact with the probe. The metal part is probably 10-15mm from the probe body. So I'm guessing it's not doing anything.

I will try to be more specific on what test setups I've been working on. The signal generator I am using is the USB LTDZ 35-4400M spectrum analyser in CW mode. I use this because it generates a very fast 75ps edge, and is useful for evaluating the bandwidth of the probe. It outputs a fairly sensible waveform normally. I think I have also seen the same lack of effect when using the LibreVNA in CW mode, but I don't exactly remember.

Test setup 1.
No change in frequency response with or without stand/touching the probe.


Test setup 2.
Probably better ground connection between signal generator and scope.
No change in frequency response with or without stand/touching the probe.


Test setup 3.
Using different scope. USB sampling scope from SJL instruments.
No change in frequency response with or without stand/touching the probe.


Test setup 4.
Measuring frequency response on LibreVNA, with ground strap fitted. (figure 7 in datasheet).
No change in frequency response with or without stand/touching the probe.


Test setup 5.
Measuring frequency response on LibreVNA, WITHOUT ground strap fitted.
Notable change in frequency response when not holding the probe by hand.

[joeqsmith]

I first did a sanity check using the LiteVNA.  No attempt was made to calibrate the VNA as we just want a relative measurement.  I used the same holder as before, mounting the probe using the same paper clamp.   I laid everything on top of a section of PCB as before.  I used a homemade coplanar waveguide that is setup as a load standard.  I just want to know if I would see the same notch. 

LiteVNA_mount_hand: showing S21 without my hand touching the probe (RED) and with my hand (YELLOW).   About a 2.5dB difference.   So, at least this setup behaves reasonably close to what I expected. 

I used SJL Gigawave 6400 with a BNC to SMA adapter to attach your probe to channel 1 as shown.   I used my Marconi 2024 RF generator for a source, set to 1.4GHz. 

SJL_mount_hand: Orange showing with my hand attached.  Yellow/Red showing with my hand away from the probe.  I would need to hunt around a bit to find the sweet spot but I can clearly see a difference with both the VNA and scope, as I would expect. 

[lasmux]

Thanks for the testing Joe. I'm going to have to do some more testing to see if I can reproduce your results. Do you have a bit of copper tape handy to test the ground strap?

[joeqsmith]

Thanks for the testing Joe. I'm going to have to do some more testing to see if I can reproduce your results. Do you have a bit of copper tape handy to test the ground strap?

I'll solder up a dedicated test board with a strap and retest.   

[lasmux]

I am wondering whether the metal stand you are using has something to do with this also. As you noted, there is a large grounded metal rod right next to the whole probe, but I see from the photograph that the bottom of that rod extends well beyond the bump grips of the probe. The circuitry beyond the bump grips is particularly sensitive, which is why it's recommended in the manual to only hold the probe from above the bump grips. I will run my own tests as well later.

[hpw]

As you have a LeCroy WaveMaster 7300a.

Just compare with an AP034 differential probe or any none differential how it perform.

I mean CMRR & channel skew wise... while your dual probe use may looks as working but measurement wise 

[joeqsmith]

I am wondering whether the metal stand you are using has something to do with this also. As you noted, there is a large grounded metal rod right next to the whole probe, but I see from the photograph that the bottom of that rod extends well beyond the bump grips of the probe. The circuitry beyond the bump grips is particularly sensitive, which is why it's recommended in the manual to only hold the probe from above the bump grips. I will run my own tests as well later.

I made up two different fixtures, added a few inches of braid to one, and picked some O-rings that will allow me to ditch the paper clip and secure your probe to the holder without it extending beyond the grips.

I'll run it later today.

[lasmux]

Having a piece of metal (grounded or otherwise) near the probe tip does indeed attenuate the output of the probe, so I think that's definitely something to try. I've still not been able to recreate the -3dB change going between handheld and stand held on a scope.

As you have a LeCroy WaveMaster 7300a.
Just compare with an AP034 differential probe or any none differential how it perform.
I mean CMRR & channel skew wise... while your dual probe use may looks as working but measurement wise 

Under around 400MHz it's pretty well behaved, but yeah, I just ran some tests and the CMRR goes below 10dB at around 600MHz. It was around 30dB at 100MHz. And yes, you do have to tweak the skew on one of the channels to account for differences in cable length etc. But that's quite easy to do. Just measure the same signal with both probes and adjust the skew to minimise the subtracted signal.
In general I've not put a lot of effort into this so far, so these are only early unrepeated results. And also, the performance will depend on how well the probes are matched. In general for measuring differential signals though, I think it is still very useful as most common mode noise on a measurement will be at relatively low frequency where the CMRR of this setup seems to be reasonable, and so a differential measurement will improve the noise on any measurement by a factor of 2x. Also, given there is no ground connection, potentially the signal loading is improved also. I need to test this also.

[joeqsmith]

Having a piece of metal (grounded or otherwise) near the probe tip does indeed attenuate the output of the probe, so I think that's definitely something to try. I've still not been able to recreate the -3dB change going between handheld and stand held on a scope.

Setup1-3
Showing new fixtures.  Cut down wire wrap pins were soldered to the coplanar waveguides to allow a more repeatable connection to probe.  Also shown is ground strap attached to the load board.  The alligator clamp's teeth were cut off and braid was soldered in to get a better surface connection to probe.

Setup4
With new O-rings, I can attach the probe to the holder without the holder extending beyond the grip area of probe.  Note how clamp attaches to the SMA body. 

The LiteVNA was again used.  Yellow with my hand off the probe, Red with it attached.  It certainly had an effect on the overall flatness.  Fingers were placed higher up on the probes body as shown in Setup 5.

The probe was then attached to the SJL Gigawave 6400 and the test fixture stimulated with my Marconi 2024.   Shown with my hand off the probe Yellow/Orange and fingers just below the O-ring as before, Red.   Note, that at higher frequencies we would see it flip (signal decrease with my finger touching probe). 

[lasmux]

That's a very serious ground strap!!! A bit more significant than my simple copper tape 

Thanks for the extensive testing there. It looks to me like the effect has indeed been reduced with the movement of the metal stand away from the probe front end and the addition of the ground strap. Again, for me I only need to use the ground strap when taking VNA measurements whilst also using the stand (not handheld).

In terms of what we have in common between our setups, we have the same probe, the same scope, a fairly similar signal generator, a fairly similar test fixture. Despite that we are getting some differences in results still, although closer now. Given the improvements by reducing the proximity of your metal stand, my feeling these anomalous readings are due to the large metal objects very near or directly pressed against the probe, especially if they are beyond the bump grips, which I don't have because of the plastic/isolated stand I'm using.

In terms of the probe itself. The only difference between my probes and yours is that the revision I am using has a different grounding setup at the battery input. I don't think that would affect this though.

I have done some more testing and confirmed that placing metal objects beyond the bump grips, within maybe 5mm of the area beyond the bump grips can modify the response of the probe.

[joeqsmith]

One thing I show in the photo, but may not have made clear, my wrist is laying on the PCB, which makes a difference. 

Also, keep in mind that someone could very well be trying to probe a board near metal. 

[lasmux]

I have also tried to recreate your results with my hand being 'grounded' to nearby metal etc, without success. I think there may be a couple of effects going on. Firstly, grounded metal increasing the capacitive coupling between the amplifier inputs and ground, which will affect the frequency response. Secondly an affect on the general grounding quality of the probe. I suspect the first one is the dominant effect due to the reduction in the effect when you moved the metal stand away from beyond the bump grips. The way to resolve this would be shielding which we've already discussed as not feasible at this price point, or increasing the capacitive coupling between the amplifier signal input and the signal being measured. This would have the effect of increasing the capacitive loading of the probe, so isn't something I want to be attempting as that is probably the parameter that the probe is most optimised for. It would also require retuning the frequency response of the probe. I will expand the current direction within the manual to 'not hold the probe beyond the bump grips' to include a comment about metallic objects beyond the bump grips somewhat affecting the frequency response.

On the topic of capacitive loading on measured signals. I've done some more testing with the twin probe differential setup, and the S11 return loss on a probed line drops even further from a worst case -15dB to around -16.5dB worst case. I can't find much data about other probes that note this. Is anyone aware of what the comparable value would be for say a Teledyne/Keysight/Tektronix/etc? I will post some more details later on. I need to do a bit more tweaking of the ground connection between the two probes. Currently it is quite long which I don't think is ideal.

[joeqsmith]

On the topic of capacitive loading on measured signals. I've done some more testing with the twin probe differential setup, and the S11 return loss on a probed line drops even further from a worst case -15dB to around -16.5dB worst case. I can't find much data about other probes that note this. Is anyone aware of what the comparable value would be for say a Teledyne/Keysight/Tektronix/etc? I will post some more details later on. I need to do a bit more tweaking of the ground connection between the two probes. Currently it is quite long which I don't think is ideal.

List of my probes:
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg4988716/#msg4988716

You are using the coplanar waveguide, terminated to 50 ohms, measuring S11, then attaching the probe?   You would like to see that same measurement using my diff probe?  They do provide the capacitance, noted in that list if that helps.

[lasmux]

On the topic of capacitive loading on measured signals. I've done some more testing with the twin probe differential setup, and the S11 return loss on a probed line drops even further from a worst case -15dB to around -16.5dB worst case. I can't find much data about other probes that note this. Is anyone aware of what the comparable value would be for say a Teledyne/Keysight/Tektronix/etc? I will post some more details later on. I need to do a bit more tweaking of the ground connection between the two probes. Currently it is quite long which I don't think is ideal.

List of my probes:
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg4988716/#msg4988716

You are using the coplanar waveguide, terminated to 50 ohms, measuring S11, then attaching the probe?   You would like to see that same measurement using my diff probe?  They do provide the capacitance, noted in that list if that helps.

I am calibrating the VNA with a 50ohm terminated coplanar waveguide in place, and then measuring the return loss by attaching the probe to the waveguide. So there is -60dB S11 worst case return loss before the probe is attached, and then -15dB when it is attached.

I would love to see the return loss measurements on those probes if you don't mind! Maybe the HFP3500 which is the same as the Lasmux probe capacitance. Also the WL300 would be a glorious measurement with a 0.1pF tip capacitance!!

[joeqsmith]

I am calibrating the VNA with a 50ohm terminated coplanar waveguide in place, and then measuring the return loss by attaching the probe to the waveguide. So there is -60dB S11 worst case return loss before the probe is attached, and then -15dB when it is attached.

I would love to see the return loss measurements on those probes if you don't mind! Maybe the HFP3500 which is the same as the Lasmux probe capacitance. Also the WL300 would be a glorious measurement with a 0.1pF tip capacitance!!

You attach the WG to the test cable, then cal at the opposite end of the WG, then replace with terminator? 

Or

Short / open at and of cable, then attach WG with load?

Or

SOL at end of cable, then attache WG with load?

[lasmux]

You attach the WG to the test cable, then cal at the opposite end of the WG, then replace with terminator? 
Or
Short / open at and of cable, then attach WG with load?
Or
SOL at end of cable, then attache WG with load?

The first one. I attach the waveguide to the test cable, then SOL calibrate on the opposite side of the waveguide to the test cable.
Port1 || ---- test cable --- || --- waveguide --- || SOL

[joeqsmith]

You attach the WG to the test cable, then cal at the opposite end of the WG, then replace with terminator? 
Or
Short / open at and of cable, then attach WG with load?
Or
SOL at end of cable, then attache WG with load?

The first one. I attach the waveguide to the test cable, then SOL calibrate on the opposite side of the waveguide to the test cable.
Port1 || ---- test cable --- || --- waveguide --- || SOL

I used the LeCroy WR64x rather than the WaveMaster as we just need to power up the probes.  I used the Agilent E8357A (hacked to 9GHz) with ecal.  Used unknown thru to cal the VNA and left port 2 attached allowing us to collect all four S-parameters if you want them.  I used some very short pins in the HFP3500.    The pins on the 300 are fixed.   

While I am confident in the data, be aware that Keysight no longer serviced these units when I bought mine.  I wrote some custom software to allow me to align it.  The equipment I used to align it were checked against equipment from where I work that are under calibration.   The ecals and adapters were characterized using Keysight standards that I loaned out.   

[lasmux]

Arg, sorry Joe I thought I had replied to this.

Thanks for that test data. Very interesting results.

That 0.1pF tip capacitance of the WL300/D300 probe is just marvellous! No idea how they manage that.

Interesting how the return loss of the HFP3500 is a bit worse than the lasmux probe, despite being a similar tip capacitance.

[lasmux]

Alongside the new stand which I'm going to start shipping with the probe soon, I also have a new EMI shield to go over the probe. This is still being developed, but is essentially a semi-conductive (carbon filled plastic) shoud that covers the sensitive parts of the probe. I'm currently evaluating different materials that are better and worse at shielding. Not quite ready to share the shielding noise attenuation level just yet, but it's going in the right direction.

The stand went through a last series of design edits to make the legs fold so it fit better in the case. I'm building these now. They have a large steel weight in them to weigh down the probe and keep it in place when measuring circuits. It's a bit of a game-changer for usability IMO.

I've also got a much improved case it's now shipping with.

See attached images