DC coupled 2.7 GHz Active Probe Project - Now Available!

[lasmux]

I'm organising an initial manufacturing run of a batch of 25 PCB assemblies now. Going through some normal back-and-forth with the assembly house over some of the details.

[joeqsmith]

Nice.  Just let me know when you are ready. 

[points2]

I'm organising an initial manufacturing run of a batch of 25 PCB assemblies now. Going through some normal back-and-forth with the assembly house over some of the details.

Hi lasmux,
any info about when the probes will be available ?

[lasmux]

I'm organising an initial manufacturing run of a batch of 25 PCB assemblies now. Going through some normal back-and-forth with the assembly house over some of the details.

Hi lasmux,
any info about when the probes will be available ?

Soon. Did I say that last time though 

[lasmux]

In other news, I have the first batch of 25 PCBs here (at long last) and the first couple have tested fine!!


I've also spent what felt like an eternity soldering together a mountain of probe tips. Each signal tip is in three parts. There's a pin to go into the probe, a series resistor, and then a sharp tip, all cut down and then soldered together. This is then carefully glued for rigidity, and finally tidied up with some semi-rigid heatshrink. 10 minutes assembly per tip. 50 tips. I've managed to burn through every decent podcast I know of. Fortunately the ground tip sockets are a bit simpler to assemble.

[lasmux]

I'm super proud to say that at long last, these are available at the site:
http://www.lasmux.com/product/single-ended-active-probes/

The datasheet has also been extensively updated, so worth checking that out!

It's surprising how much time and effort it is to get something like this up and running. Production issues, delays, customs and shipping, fighting insurance companies... the list just goes on and on. My hourly wage for this first batch, even if I exclude the 1.5 year long design work, is very, very low. Hopefully the next batch will be smoother.

[mawyatt]

Nice effort & probe, well done 

If/When we need an Active Probe we'll order one

Best,

[lasmux]

I think one of the biggest things I've learnt from this project is that manufacturing (more than one-offs) things which aren't just straight PCBs is really hard, despite having some R&D experience in many of these aspects. It's also been surprising how long seemingly trivial tasks take. I had a bunch of estimates on how long various tasks would take for each probe. Pin assembly, testing, customs paperwork, etc. For almost everything I've underestimated the time requirements, despite having some experience in all of these things in the past.

I have been thinking on ways to improve my workflow for subsequent batches, including test jigs, and assembly jigs for the pins. It's definitely tricky as these kinds of tooling jobs take time to set up, and the time-saving payoff will take a while to reach.

[joeqsmith]

Lasmux has provided me with one of his new probes to review.   I gave it a quick check and everything appears to work out of the box.  The bug rug was not supplied with the probe.  I added it for a bit of insurance.   

As always, if there is something specific you would like to see, feel free to ask. 

[joeqsmith]

Making some progress on the review today.  The probe came with baby papers (data from the mfg) which included S-parameters.  I had asked Lasmux to provide more details about their setup and have asked that they make this public.   Shown is the setup I am using.   

[joeqsmith]

I'm not a fan of the placement of the probe tips having the offset as it makes it a challenge to attach to the board.   Also, the shape of the probe body makes it difficult to clamp to.  Note my use of a paper clip around the cables to hold the probe in place.   

[joeqsmith]

My review.

[dmderev]

Could you share schematics, please?

[2N3055]

Could you share schematics, please?

No.
It is not open source..
And schematic is minor part of the story. Most of the work is in physical layout...

[joeqsmith]

Could you share schematics, please?

You would need to contact the manufacture for that level of documentation.  They did provide some details about the design in the start of the thread: 
https://www.eevblog.com/forum/testgear/12-ghz-active-probe-project/msg4974892/#msg4974892

For what they are charging, I can't see a reason why anyone would want to try and make their own anyway.

I thought about showing the prototype on camera but they had posted photos already so it seemed redundant.  And with the prototype being hand built, I didn't want to give viewers the impression that this was the same quality of the final product.   

IMO, if you are needing a higher BW, low capacitance probe with high DCR and you are working with low voltages, this thing is a steal at the price.   

[lasmux]

Thanks for the review joeqsmith!! Excellent overview of it!

I'm glad your experience was overall positive, albeit with a couple of niggles. I love the array of probes you have to compare it against! And of course, good to see the new resistive probes in action also.



One thing that struck me was how well the Lecroy PP002A passive probe performed when you were loading the 1GHz clock signal later in the video. Despite it having a 14pF (!!!!) tip capacitance. It actually performed about the same as the reviewed 0.7pF active probe. I think this is because the clock is such a narrow band signal, and when I've done testing on passive probes, I've seen that their applied load is very frequency dependent. Maybe at 1GHz, the Lecroy load is particularly gentle on the probed signal? For example, the below graph was the S11 reflection coefficient (blue line) of a load I measured last year, with an HP 6pF 500MHz passive probe attached across the load. See at 1.5GHz, there's a pretty good reflection coefficient. At 600MHz, not good.




As you noted, I forgot to post a more detailed method for measuring the probe frequency response. Here it is:

There are three graphs provided with the datasheet (two of these measured for each individual probe in the test report). In the datasheet these are figures 1 top and bottom, and figure 2.

Datasheet figure 1 is captured on a libreVNA, S21 and S11, with a custom coplanar waveguide connected to port 1, see photograph. The VNA has been calibrated with the coplanar waveguide in place, so the reference plane is at the end of the coplanar waveguide (where the load is fitted in the photograph). For the through calibration, the load and SMA-BNC adapter are removed to allow the blue SMA cable (port 2) to connect to the coplanar waveguide SMA. I've found that SMA-BNC connectors are often not very good at higher frequencies. It took me a few attempts to find one which didn't lead to nasty reflections.



For the effective impedance measurement (datasheet figure 2), a custom SMA 50Ohm load was built so I could place the probe directly at the termination point, so the probe acts as a perfect(ish) parallel load to the termination/VNA reference plane. This was constructed from four 200Ohm 0402 resistors soldered to a trimmed SMA PCB edge connector. The custom load has a very similar return loss compared to a commercial load, measured up to 4GHz. The globs of solder on the load are there to allow the probe tips a larger targets to contact when taking impedance measurements.

The VNA is calibrated with the custom load, and then the probe is placed across the load.

The VNA is then set to output the impedance magnitude. The effective impedance of the probe is then seen as a load directly in parallel with the 50 ohm VNA load, so the probe effective impedance is calculated as 1 / ( (1 / measured impedance) - (1 / 50)).

[joeqsmith]

Thanks for letting me have a look at it.

One thing that struck me was how well the Lecroy PP002A passive probe performed when you were loading the 1GHz clock signal later in the video. Despite it having a 14pF (!!!!) tip capacitance. It actually performed about the same as the reviewed 0.7pF active probe. I think this is because the clock is such a narrow band signal, and when I've done testing on passive probes, I've seen that their applied load is very frequency dependent.

For those who did not watch just some background:
The oscillator was in series with the waveguide and then attenuator before connecting to the scope.  What was shown was what difference the scope sees when the test probes were attached to the waveguide.   For the PP002A, I used the provided ground spring.   During the video, I did not terminate the probe to an actual scope.

Here I show it attached to the scope simulator used during the review as well as the actual scope.  It's basically a wash but best to prove it.  When looking at the return loss of the PP002A using the same technique used to measure your probe,  you can see how it peaks at 500 then starts to drop.   I suspect some of this is due to the long spring. 

Had I used a 500MHz clock, I would imagine your probe would have far out performed it.   Similar, had I used a higher source impedance driver,  I could have made your probe look much better than the home made resistive probe.   This is why I started out by using that book to cover some of the basics about loading.   

Also lets be very clear and state the obvious.  Had I shown what the scope measures when the using your probe vs the LeCroy probe on that 1GHz clock, there would be a MAJOR difference!!   The only way I could have done this test is to use a homemade 1Meg/50ohm buffer.   That would have been meaningless.   The only other option would have been to use that LeCroy buffer with the WaveMaster along with the PP005 as that is a combination someone could possibly still purchase.   

Again, a bit difficult to know what to show. 

[lasmux]

Yeah I totally understand that it's difficult to show all the context details. Not critisizing, I was just originally puzzled.

It was good to see a lot of the performance specifications get some external validation. I really appreciate all the work you put into the review!

Would the following be a fair summary of your conclusions?
Pros
Pretty good probe frequency response, with similar bandwidth to what's specified.
Low tip capacitance leads to low impact on DUT signal, with performance in line with specification.
Really good price/performance ratio. Multiple times cheaper than comparable used probes from ebay.
9 hour battery life.
Not a proprietry connection to test equipment.

Cons
Noise level is higher than other probes, so less useful for analogue work. This is unfortunately unavoidable due to the input design.
Needs to be handheld to achieve measured probe response linearity. This is interesting as all of my testing has been handheld so far, which is maybe why I didn't observe this. I will look into this further for the next production run, although no promises of any improvement. If I can't fix this I'll stick a note in the datasheet about this.
A bit tricky to clamp when not using handheld. Is this a good thing given the above issue
Offset ground pin can be a bit fiddly. I agree it can be a bit tricky. It works best when you pop the ground pin in a ground via nearby the signal to be measured.

Conclusion.
General mark of approval, especially at price point. Best for high frequency digital measurements. Less useful for analogue measurements due to noise.

[joeqsmith]

Yeah I totally understand that it's difficult to show all the context details. Not critisizing, I was just originally puzzled.
measurements due to noise.

No problem.  I just wanted to be clear about what was shown and why.  IMO, the two primary criteria are 1) don't mess with the circuit I am probing. 2) let me see what is actually happening.  Then comes all the other things like tip selection, body styles...  Of course, no probe meets those first to constraints, so it's a compromise depending on what I am trying to measure.   

Just to drive that point,  shown with the PP002A 350MHz 10X probe connected to the AP-1M buffer, then to the WaveMaster.  During the review I showed how the loading for this probe was on par with your probe when attached to the 1GHz NEL oscillator.  But when use the scope and probe to measure the signal, we get a flat line with a bit of offset (M3).   Not at all useful.   

Compare that to your probe (M2).   The signal is high enough that the noise isn't a factor and with that higher BW, we get some details about the clock. 

Now compare that with the HFP3500 (M4).   What's the right answer?  I would say I need a better scope and probes to answer that!   

It was good to see a lot of the performance specifications get some external validation. I really appreciate all the work you put into the review!

Would the following be a fair summary of your conclusions?
Pros
Pretty good probe frequency response, with similar bandwidth to what's specified.
Low tip capacitance leads to low impact on DUT signal, with performance in line with specification.
Really good price/performance ratio. Multiple times cheaper than comparable used probes from ebay.
9 hour battery life.
Not a proprietry connection to test equipment.

Beyond these, I also felt the overall quality of the probe is good.   The size of the body fits the hand well.  It was provided with a good variety of tips.   The hard carry case is a bonus.       

One thing you may want to consider is a way to guard the input from ESD when the probe is being handled.  Hence my use of the bug rug. 

***
Also, with there being no auto power off and having left the probe on a few times already,  if it wasn't clear, I have been using the probe with a lithium ion rechargeable battery.   These only output 7.4V when charged and it seems to be fine.   

***

Cons
Noise level is higher than other probes, so less useful for analogue work. This is unfortunately unavoidable due to the input design.
Needs to be handheld to achieve measured probe response linearity. This is interesting as all of my testing has been handheld so far, which is maybe why I didn't observe this. I will look into this further for the next production run, although no promises of any improvement. If I can't fix this I'll stick a note in the datasheet about this.
A bit tricky to clamp when not using handheld. Is this a good thing given the above issue
Offset ground pin can be a bit fiddly. I agree it can be a bit tricky. It works best when you pop the ground pin in a ground via nearby the signal to be measured.

Conclusion.
General mark of approval, especially at price point. Best for high frequency digital measurements. Less useful for analogue measurements due to noise.

One thing about SJL's scope, they advertise it as being 6GHz.  The unit they provided not only is easy to verify, they surpass it with a wide margin.   IMO, you appear to want to advertise the product right at the limits of what you could measure using a $1000 VNA and what ever standards it came with.  Personally, I wouldn't do this.  I would be much happier buying a product rated to 2GHz but seeing it perform at 2.5GHz. 

The whole hand placement on the body of the probe IMO effecting the measurements needs to be addressed.   Then again, maybe not.  It's similar to the noise level.  I can tell you that my first dive into active probes, I would have been working on analog designs with higher voltage levels.  This probe would have been a very welcome addition.  It far out performs my first commercial active probe.  I would say it has a place, even as it is today.   

At the selling price, I see it as a win.  Hobbyist are not typically doing cutting edge designs at home and may just need something better than their 10X probe.  IMO this probe could very well fit that need.   Buyers just need to be aware of the limitations.

[joeqsmith]

A similar setup using a PECL driver but with a 10MHz fundamental.   C1/M1 is directly looking at the output of the PECL driver through the waveguide and about 6" of coax.   Consider this is still not what the actual signal looks like as we just added a fair bit of capacitance which is dampening that fast edge.     

C2/M2 is the Lasmux probe   
C3/M3 is the PP002A/AP-1M probe/buffer
C4/M4 is the HFP3500

I took the data zoomed out a ways to better show the noise and settling times.   Then zoomed in, aligning the edges. 

[joeqsmith]

Tektronix P6202 vintage active probe attached to C4. 

[lasmux]

Beyond these, I also felt the overall quality of the probe is good.   The size of the body fits the hand well.  It was provided with a good variety of tips.   The hard carry case is a bonus.       

That's good to hear!

One thing you may want to consider is a way to guard the input from ESD when the probe is being handled.  Hence my use of the bug rug.

Also, with there being no auto power off and having left the probe on a few times already,  if it wasn't clear, I have been using the probe with a lithium ion rechargeable battery.   These only output 7.4V when charged and it seems to be fine. 

The tips (and power inputs) are actually already ESD protected. The signal socket has a series 1MOhm resistor going into the op amp, which itslef has inbuilt ESD protection.
Interesting thought about an auto power off function. I'd need to see what is available. Maybe a danger with that is that it powers off whilst someone is taking a measurement, without them realising. Of course the alternative is using a 9V battery every time you forget to switch it off for a few hours. I like that you've used a rechargable 2S lithium battery

One thing about SJL's scope, they advertise it as being 6GHz.  The unit they provided not only is easy to verify, they surpass it with a wide margin.   IMO, you appear to want to advertise the product right at the limits of what you could measure using a $1000 VNA and what ever standards it came with.  Personally, I wouldn't do this.  I would be much happier buying a product rated to 2GHz but seeing it perform at 2.5GHz. 

I do agree with the sentiment. Almost all of the probes in the first production batch were over 2.7GHz by my measurements. I guess the issue is that different peoples test kit will measure different bandwidths, which may leave them dissapointed. I'll consider that.

The whole hand placement on the body of the probe IMO effecting the measurements needs to be addressed.   Then again, maybe not.  It's similar to the noise level.  I can tell you that my first dive into active probes, I would have been working on analog designs with higher voltage levels.  This probe would have been a very welcome addition.  It far out performs my first commercial active probe.  I would say it has a place, even as it is today.

I've done a bit of testing on this today, and unfortunately there's not an obvious fix. I tend to get somewhat consistent frequency response results when holding the probe in various different positions by hand, with different levels of contact, but letting go of it completely does seem to affect the response negatively for some reason. I think it'll have to stay as it is for the time being at least, and I'll need to update the documentation/website to make it clear that it's performance is characterised during handheld use, and using a clamp/stand will have an effect on the frequency response. I do agree with you however, that it is still a very useful tool irrespective of this. I'll keep looking and maybe figure out a fix at some point.

At the selling price, I see it as a win. [...] Buyers just need to be aware of the limitations.

Yes for sure.

[lasmux]

A similar setup using a PECL driver but with a 10MHz fundamental.   C1/M1 is directly looking at the output of the PECL driver through the waveguide and about 6" of coax.   Consider this is still not what the actual signal looks like as we just added a fair bit of capacitance which is dampening that fast edge.     

C2/M2 is the Lasmux probe   
C3/M3 is the PP002A/AP-1M probe/buffer
C4/M4 is the HFP3500

I took the data zoomed out a ways to better show the noise and settling times.   Then zoomed in, aligning the edges.

Nice plots.

A couple of observations:
Noise is definitely apparent there on the lasmux probe. The probe noise combined with the relatively low PECL voltage transition amplitude (400mV) makes for a noisy signal when zoomed out.
On the zoomed in signal, the rise time of the original signal is around 1ns and the HFP3500 and the lasmux probe seem to track it fairly well, albeit with a small DC offset on the lasmux probe. Not surprising as it's well within the bandwidth of both probes. The P6202 has a bit of overshoot, and the PP002A probe doesn't seem to have the bandwidth to keep up with the edge.

[joeqsmith]

The P6202 has a bit of overshoot, and the PP002A probe doesn't seem to have the bandwidth to keep up with the edge.

I have the OEM tools for the P6202 to compensate it but you have to remove the plastic cover from the tip and expose the delicate circuitry (mechanically not very sturdy).     

I dropped the fundamental down low enough so at least we could see something beyond a flat line with that 10X probe.  It does make for a nice filter. 

Attached showing the updated spreadsheet for some of my probes.  Note the dynamic range of my active probes are all less than half of yours.  Maybe between you and SJL, you may want to give up dynamic range to improve the noise.   

[lasmux]

From the very simplified schematic I posted much earlier in the thread (see below), that would involve reducing the ratio R2/R3. R3 is 20k, so there's an attenuation of 50x on the input. The signal is then amplified by the (not-unity-gain stable) op amp back up to a 10x attenuation factor. This becomes 20x a net attenuation by the 50 ohm output impedance/oscilloscope impedance. The issue with reducing the attenuation on the input is that the capacitive divider of the parasitic capacitance across R2, and the C1 + parasitic capacitance across R3 has to maintain the same ratio for the AC gain. The parasitic capacitance across R3 is great enough that C1 would have to be a negative capacitance to maintain the correct ratio, which obviously isn't an option. Reducing the ground around that node to reduce the parasitic capacitance just makes external noise pickup a lot worse.

Another option is to reduce the impedance of the op amp input node so external noise pickup is less of an issue. So say reduce R2 to say 100k, and reduce R3 to 2k. It makes it a 100k DC resistance probe though, not a 1M, although that's probably a reasonable tradeoff if it actually helps. I'd have to try it to see what the change is.