DC coupled 2.7 GHz Active Probe Project - Now Available!

[joeqsmith]

M2 is no load, C2 is the loaded waveguide.  PP061 resistive probe blue.   Rise time is roughly 500ps for the probe, or around 700MHz.   Edge without the probe attached measures about 200ps.   

With a better scope and lower transitions the probe doesn't  look so impressive.      

[joeqsmith]

But if you think that's bad here is that same homemade resistive probe.      That's s mighty big loop, lots of inductance.  Not too surprised it peaks up.   I would expect your probe would do a much better job with this 200ps edge.   

Scope is an old WM8500A 5GHz BW. 

[joeqsmith]

Using a different driver with my Tektronix P6202A probe.  It makes a decent filter. 

[MathWizard]

Cool project, I'd buy this probe down the road.

How much worse would it be if you used all the same chips, but used all standard size through hole parts ? What if you didn't pay attention to spacing or layout ? I'm just trying to get a feel for how sensitive these circuits really are. If 2GHz is ~15cm waves, thats getting down to circuit size.

1 of the next things I need to buy, is cables, terminators, and splitters, for using between my scopes and signal gen. And I want to try making 1 of those passive probes that's just coax and a 1k resistor, and has very low capacitance. Dave showed 1 in a video on probes. I can't remember their downside, besides low series resistance. But even mucking about with BJT amplifiers lately, I realize how big 15pF of 10x probes can be.

[joeqsmith]

Cool project, I'd buy this probe down the road.

How much worse would it be if you used all the same chips, but used all standard size through hole parts ? What if you didn't pay attention to spacing or layout ? I'm just trying to get a feel for how sensitive these circuits really are. If 2GHz is ~15cm waves, thats getting down to circuit size.

1 of the next things I need to buy, is cables, terminators, and splitters, for using between my scopes and signal gen. And I want to try making 1 of those passive probes that's just coax and a 1k resistor, and has very low capacitance. Dave showed 1 in a video on probes. I can't remember their downside, besides low series resistance. But even mucking about with BJT amplifiers lately, I realize how big 15pF of 10x probes can be.

MathWizard, for a 20:1 ratio, you want to use a 950 ohm resistor, not a 1k.   The scope's input is 50, so the circuit you are probing sees 1000ohms (assuming a perfect world).

The following was taken from this paper:
https://people.ece.ubc.ca/robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf

Analog bandwidth is a measurement specification that simply defines the frequency at which the measured amplitude of a sinewave is 3 dB lower than the actual sinewave amplitude (see IEEE-1057). Figure 1 shows an idealized amplitude roll off error as a sinewave signal approaches the specified bandwidth frequency of a measurement device having a first order or single pole Gaussian response. At the rated bandwidth, the measurement error approaches 30%! If you want to make a measurement on a sinewave that has only 3% error, you would only want to measure sinewaves much lower in frequency than the rated bandwidth of the oscilloscope, about 0.3 times the rated instrument bandwidth. Because most signals are more complex than sinewaves, it is a general rule of thumb is use a measurement device, like an oscilloscope, that has 5 times the bandwidth of the signal you intend to measure (explained later and shown in Figure 5).

***
In case it wasn't clear, the following was also taken from the same paper: 

In fact, every probe manufacturer assumes that at the maximum specified bandwidth, the probe’s frequency response is down 3 dB.

[lasmux]

But if you think that's bad here is that same homemade resistive probe.      That's s mighty big loop, lots of inductance.  Not too surprised it peaks up.   I would expect your probe would do a much better job with this 200ps edge.   

Scope is an old WM8500A 5GHz BW. 

Yeah, that's a lot of ringing! Very neat how little it's loaded the original signal though.

[lasmux]

Cool project, I'd buy this probe down the road.
How much worse would it be if you used all the same chips, but used all standard size through hole parts ? What if you didn't pay attention to spacing or layout ? I'm just trying to get a feel for how sensitive these circuits really are. If 2GHz is ~15cm waves, thats getting down to circuit size.

Thanks, I'll let you know when it's ready

The probe amplifier input and feedback network on the probe is all 0402 passives. Even with that I have to have various snubbing resistors etc as stray inductances/capacitances cause havoc with the frequency response. You might be able to get something working with through hole components, but it would take much more effort to get it working correctly. And you probably wouldn't be able to reach 2GHz bandwidth.
Compared to the DIY 1k resistive probes, it is more general purpose due to the 1M input resistance, and in theory will have a more precisely tuned frequency response, but has lower bandwidth potential.

[joeqsmith]

But if you think that's bad here is that same homemade resistive probe.      That's s mighty big loop, lots of inductance.  Not too surprised it peaks up.   I would expect your probe would do a much better job with this 200ps edge.   

Scope is an old WM8500A 5GHz BW. 

Yeah, that's a lot of ringing! Very neat how little it's loaded the original signal though.

To give you some idea of the effects of lowering the inductance, these are two poor man's 10X probes.  The resistors are butted up against the coax center pin with a much shorter probe tip.  All to reduce that big loop.   

As before, M2 is the unloaded waveguide. C2 is loaded waveguide with probe #2 attached.  C3 is probe #2.  M3 is probe #1.    You mentioned the amount of variance seen with homemade probes but these two seem fairly well matched.   

[joeqsmith]

Just for completeness, I reattached the homemade 20X probe to compare it with the 10X probe #1.   Construction is everything and as you pointed out, I did try a few ways to improve it but they all failed to exceed what I could achieve with the simple techniques shown. 

[joeqsmith]

I have thought about trying to add a blade for the ground rather than the wire to help reduce the inductance.   One thing I have done in the past was when I wrap the wire around the coax to form the ground, I use both ends rather than one.  I normally solder this coiled wire.  The problem is getting them attached to the board.   Picotech offers a probe like that today.   Looks like a 20X 6GHz probe would cost about $1000 USD.   

https://www.picotech.com/accessories/gigabit-digital-passive-test-probes

[tggzzz]

I have thought about trying to add a blade for the ground rather than the wire to help reduce the inductance. 

Like this HPAK N2878A?

[Mechatrommer]

I've been working on an active probe design for around a year. The goal started off as creating a DC-coupled active probe to support a photon counting sensor I am also working on, but it was a very fun project and I spent so much time on it that now the plan is to sell it. Could I have some feedback on the probe/performance, and on the contents of the datasheet, before I start buying the first batch of parts... which will be quite expensive.

I have another post that I'm putting together where I'll go into the development process a bit more.

I'm making two versions, a 1GHz version, and a 2GHz version.

The datasheet is here: https://www.lasmux.com/wp-content/uploads/2023/07/LD-ASP-1G_2G.pdf



Quick specs:
Bandwidth: DC-1GHz, DC-2GHz
Input capacitance (measured at 1GHz): 0.7pF
Attenuation: 20x
DC input resistance: 1Mohm

1GHz version frequency response (linear and log axis):


2GHz version frequency response (linear and log axis):


Tip input impedance of both probes, depending on which ground lead is fitted:


The resistive ground lead can be used to stop a resonance developing on the ground connection, which reduced the input impedance above 1.5GHz. I talk about this a bit more in the datasheet.



In terms of step response for the system, I've 'only' got a 500MHz oscilloscope, so can't properly test the rising edge speed unfortunately. This is the probe measuring a 50 ohm terminated 100MHz signal, with a <100ps rise time. This greatly exceeds the bandwidth of the scope so there's some ringing. The trace looks basically identical if I measure the signal directly by the oscilloscope.

Currently I'm aiming for around £150 for the 1GHz version, and £185 for the 2GHz version.

good work!

[joeqsmith]

I have thought about trying to add a blade for the ground rather than the wire to help reduce the inductance. 

Like this HPAK N2878A?

Yes, but with a smaller loop. 

[joeqsmith]

I used some copper spring material to make an "L" shape for the blade.  The tab was formed into a cylinder and soldered to the coax. 

Again, M2 is the unloaded waveguide (probe not attached).  C2 is with the probe loading the waveguide.   C3 is the measured signal from the blade probe.   Of course there are some errors with gain and I can compensate for the phase.  A bit too much gain in the plot but works fairly well compared with the others.   

***
The blob used to mechanically stabilize the resistor is Devcon 5 minute epoxy.   It doesn't appear to hurt the performance of the probe at least at these frequencies.   

[joeqsmith]

Making similar measurement with an old LeCroy 7200.  The only option for saving data is with a floppy drive which I can read with my USB floppy.  For plotting, I use CERN's viewer.   

Shown looking at the 10X blade probe on channel B2.   Most of the error is due to the how the scope works.   It has a step generator built-in and can loopback that signal to the display.   It's is not the same as probing at the end of the waveguide like I have been showing.   I need one more channel for that.   

The scope doesn't have a lot of storage and I don't remember it having a way to adjust the trigger independently for each channel.  Even at 1nS/div, I can't align the two waveforms.    Consider we are using 30+ year old scope (the 7200 was released in 1989),  results seem fair.   

[lasmux]

I used some copper spring material to make an "L" shape for the blade.  The tab was formed into a cylinder and soldered to the coax. 

Again, M2 is the unloaded waveguide (probe not attached).  C2 is with the probe loading the waveguide.   C3 is the measured signal from the blade probe.   Of course there are some errors with gain and I can compensate for the phase.  A bit too much gain in the plot but works fairly well compared with the others.   

***
The blob used to mechanically stabilize the resistor is Devcon 5 minute epoxy.   It doesn't appear to hurt the performance of the probe at least at these frequencies.

This is really nice. I love the short L ground blade design. The probe response matches the source waveform really well! Even the ripple is very close.

Making similar measurement with an old LeCroy 7200.  The only option for saving data is with a floppy drive which I can read with my USB floppy.  For plotting, I use CERN's viewer.   

Shown looking at the 10X blade probe on channel B2.   Most of the error is due to the how the scope works.   It has a step generator built-in and can loopback that signal to the display.   It's is not the same as probing at the end of the waveguide like I have been showing.   I need one more channel for that.   

The scope doesn't have a lot of storage and I don't remember it having a way to adjust the trigger independently for each channel.  Even at 1nS/div, I can't align the two waveforms.    Consider we are using 30+ year old scope (the 7200 was released in 1989),  results seem fair.   

Can you save waveforms on the screen? This is how I sometimes do it. I save and display a reference waveform, and then use the delay on the trigger to overlay the live view of another signal.

[joeqsmith]

I believe it has the freedom to do what you suggest, if the scope was 100% working.  It throws the BRAM (Battery RAM) fault on start as the internal battery is dead.  I'm not sure if that is the problem but when I select the internal memory, the scope resets.  The scope shows the memory is garbage and selecting clear will also cause a reset.   I am using a version of firmware that is known to have problems but it supports the plug-ins that I am using.  It's possible that the memory problem could be tied to firmware.  30 years old, maybe even a hardware failure.  This thing is an old VME chassis.    


***
This was a kickstarter resistive probe.  While a try to reduce the inductance, if you read their comments,  "the Vishay FC0402 resistors I use in the attenuator are rated for 30V each. I have a series string of 100-75-75-75-75-50 ohms so with 50V at the tip you shouldn't be over 30 on any single resistor."   Hard to believe they would get any level of performance out of it.

Note how they terminated the test board so we can't see the loading effects.   

https://www.kickstarter.com/projects/azonenberg/akl-pt1-2-ghz-passive-oscilloscope-probe

[joeqsmith]

While we wait for your probes,  attempted to low format the drive, which failed.  The drive I have been using is an old Seagate ST125.  Note the jumper wire is not factory but a mod I sorted out that makes the drive act like the old Mini-scribe drives that came from the factory.  These scopes only supported a few different drives. The KALOK KL320 was another one it supported.  I inserted my backup drive and again, fails out.  That drive has been in storage for maybe 15 years or so.   

I had converted this scope to use rechargeable batteries for the backup.  I had forgot that the low format would fail if they were dead.  So after a full charge,  low format, reinstall, the scope now works.  I went through the same process with the second drive and it works as well.     Putting it back together,  it no longer resets when the internal memory is selected and does not show garbage when you try and display the contents.     

So I tried your suggestion of using the storage and delayed trigger using about a foot or so of coax for a delay.   Works like a charm, well, for now.    

[joeqsmith]

For fun, I took the ECL output into a DC block, into a Mini-Circuits 8,1GHz tripler and then into a wideand amplifier.   The hope was to get a faster edge.  Loaded waveguide is C2 and the bladed probe, C3.   The WM8500A has a BW of 5GHz, or roughly 70ps.   Signal level is a bit low but the homemade probe seems to do a fair job tracking it. 

Granted, you can't use it for high voltages and at low frequencies, even a basic 10X probe would present less loading but if you work with high speed digital, it may be difficult to match the price performance of a resistor and some coax.  Well, and apparently a really good ground.   

[lasmux]

I've seen that probe from Andrew Zonenberg before. He has some really nice other projects also. I think the main issue with that probe is the tip capacitance is actually quite high for a resistive probe, and the bandwidth stops at 2GHz. Not sure why it was necessary to use those Vishay FC0402 resistors. Under 2GHz I don't think a normal 0402 resistor would have been an issue, and they are miles cheaper.

Glad you managed to fix your scope I've never been good at fixing old oscilloscopes. Last time I took apart my HP 54111D, I managed to achieve nothing but waste time and stress out about killing myself on the CRT.
Nice tuning of that delay haha. Very nice match between the source/measured signal. But yeah, damn dude, that resistive probe is working a charm. Really nice. As always, I think there is a place for active probes above resistive probes. Where you want reduced signal loading, probably sub 2GHz, and for frequency response stability.
My original application was my actively quenched APD as a photon counting module, where I wanted to measure across a 1k ballast resistor restricting current through an APD. A passive probe had way too much capacitance and would have slowed down the response of the signal, a resistive probe would have made the ballast resistor ineffective and probably damaged the APD. An active probe was the only way that I could think of.

On another topic, I was considering a differential version of the probe. I've been playing around in LTSPICE, and I could probably get over 1GHz of bandwidth, with the same 0.7pF tip capacitance, +/- 30V input dynamic range, but I suspect my CMRR wouldn't be great without a bit of work on matching the two inputs closely.

[joeqsmith]

I agree, active probes certainly have their place.  I would like to see your diff probe if you pull it off.  When I investigated it, I could not locate parts to do the job.  That has been a long time ago and I suspect based on your goals, you will be able to pull it off.   That would be a handy probe to have on hand.  Much of the high speed digital has been differential for some time.   

Searching ebay, even junk 1GHz diff probes have an asking price of over $200.   Not suggesting anyone would pay that.  Looking for my D300, this one was about the cheapest I could find which at least appears complete and they offer returns. 

https://www.ebay.com/itm/175614100115

Of course, that's about 4X the BW you are targeting but still, I think there is money to be made if you can keep the costs down.

When my wife and I were first married, I was needing a higher speed scope than we had at work.  I rented a scope similar to that and had it shipped to our house so I could sort out how to run it before taking it to work.  My wife was worried that something would happen.  I think the cost was more than our house at the time.    

[hpw]

On another topic, I was considering a differential version of the probe. I've been playing around in LTSPICE, and I could probably get over 1GHz of bandwidth, with the same 0.7pF tip capacitance, +/- 30V input dynamic range, but I suspect my CMRR wouldn't be great without a bit of work on matching the two inputs closely.

Yeah, as I currently use the AP034 diff probe (see picture) as nice using 0.1" pin headers for fixed or tips connections over the active HFP2500 or the crappy active Cal Test CT4121. Even the CT4121 as shielded front header still picks up RF waves.

Hp

[tszaboo]

I've seen that probe from Andrew Zonenberg before. He has some really nice other projects also. I think the main issue with that probe is the tip capacitance is actually quite high for a resistive probe, and the bandwidth stops at 2GHz. Not sure why it was necessary to use those Vishay FC0402 resistors. Under 2GHz I don't think a normal 0402 resistor would have been an issue, and they are miles cheaper.

I just looked at those FC0402 resistors and they are 5 USD in single pieces.
I really don't get it. 0402 SMD thin film inductor is going to set you back 2c, but if you want a resistor with the same technology, it's going to cost 300 times more.

[joeqsmith]

S11 & S21 for the blade probe using the PNA.  My homemade PCB waveguides are going to be useless.    Starts out at 500 ohms as and drops off rather quickly reaching 100 ohms at 6 GHz.   Mag isn't very stable  (when holding all the parts in my hands, moving them around...).  Then there is de-embedding the interconnect.   Consider it all just a gross measurement to give us some idea how it behaves.

[JohnG]

My original application was my actively quenched APD as a photon counting module, where I wanted to measure across a 1k ballast resistor restricting current through an APD. A passive probe had way too much capacitance and would have slowed down the response of the signal, a resistive probe would have made the ballast resistor ineffective and probably damaged the APD. An active probe was the only way that I could think of.

There is another way that might work. You make the 1k ballast part of your probe by putting a 953 ohm resistor in series with 50 ohm microstrip or similar transmission line right on the same PCB. Just don't forget to terminate the other end of the line if it is not connected to your scope. You get your intended loading and measurement at the same time. Note that this does not work as well if the ballast is connected to the positive bus, but sometimes it is possible to do something clever and make it work anyway, especially if you can put in a series dc blocking cap (if you don't need a dc response).

John