22 Replies Latest reply on Mar 7, 2021 1:24 AM by friedmule

    2GHz probe from article - noob questions?

    friedmule

      Since I am as noob as you can get do I know that this project may be far over my head but I have read a lot, asked around and got lots of conflicting information.

      The whole project started because I want to be able to measure different places to see how a signal is changing. That does often demand the ability to use more than one ground.

       

      In my search did I find the above article and hope to build something similar and hopefully get it to work with a so wideband as I may be able to.

      While asking on another forum did I got told things I do not understand about the solution in the magazine.

      1) The coaxial cable is mounted with a long lead and will not work correctly.

      2) The ground plane have to be removed underneath the input-resistors.

      3) The input resistors may not be "folded" but have to be straight to work.

       

      Why do I then ask here instead of asking on the other forum, simple, I feel I have used plenty of that persons time and if I now start to "debate" with that helping person, do I appear ungrateful and is just wasting time?

      My reason for being unsure if the help I have got is 100% correct is that the article is made by someone who has measured his solution and that the magazine has chosen to published it.

       

      The images attached to this post is my try to build the same as in the article, but without the power supply part.

        • Re: 2GHz probe from article - noob questions?
          andrewj

          You need to link to the article Robert; it would also be useful to link to the other thread where this has been discussed with you.

           

          I don’t know enough about probes to answer questions but others here will.  2GHz sounds highly specialised though, as must the oscilloscope you will be attaching them to! 

           

          It may be worthwhile saying what you intend to do in a little more detail as there my be a more appropriate solution - for example, there was a recent blog published here on creating a probe which may be of more use.

          1 of 1 people found this helpful
          • Re: 2GHz probe from article - noob questions?
            shabaz

            Hi friedmule

             

            At 2 GHz, and especially for a high impedance probe (which is what that circuit is trying to be) the construction style needs to be different. Every fraction of a pF matters to ground and to other connections, and personally I think only 0402 or lower sized parts help there, with small land patterns, and keep traces separated, or thin, and not near ground plane.

            Here's an idea of what I mean, this was my attempt at a non-differential probe for a different purpose (E-field probe, where you don't actually touch the signal, you just place the probe on top of the insulated trace on the PCB). All Rs and Cs are 0402, look at the difference compared to the 0805 sized resistor mounted at the Q3 position. Notice how to keep capacitance low, the ground is kept extremely thin at the side where the probe goes, so that the capacitance is minimal (the probe signal in an inner layer in this case, you can faintly see it).

            Anyway, even after all that effort, performance sucks, see here:

            In my defence, I think I know what's causing the issue just to the left of the centre of the display, so there's a chance I can fix that sometime, and the rest can be calibrated out on a spectrum analyzer (not on a scope).

             

            The funny thing is, a single resistor gives better results (for sure this isn't a definitive test, but for ballpark examination of a board this could be good enough):

            For that last trace, the single resistor looked like this:

            And then I just placed the end of the wire on the trace (insulated). Of course, this is all very different from what you're making, my point is merely that at these high frequencies, it's best to mitigate issues by following device datasheets as closely as possible (I have not checked for your IC), and follow some practices like keeping capacitance on the input low, use the right bypass capacitors (they look large, so worth checking what the datasheet suggests, or any reference design from the chip manufacturer) and so on.

             

            Generally, if you're working at 2GHz, the circuit under test will be designed for 50 ohm connections, and then you don't need any probe, you can directly use coax into your 'scope.

             

            If you have no probe with your 'scope, even a 100 or 200 MHz probe could be good enough for 90% of use-cases, and then swap to the coax whenever you encounter 50 ohm load connections, and then you'll experience the full bandwidth of the 'scope.

            But I believe the circuit you are working on will function to an extent, just don't expect it to be functioning with no anomalies all the way to 2 GHz.

            4 of 4 people found this helpful
              • Re: 2GHz probe from article - noob questions?
                friedmule

                Thanks a lot, great answer! :-)

                 

                Maybe should I try to explain what I hope to achieve:

                Please ignore this infantile example, it's just to illustrate.

                Let's say you have some components that does something to your signal, but will know how and why. It could be on both sides of a full bridge rectifier. On one side do you have 12 VAC and on the other side, do you have DC. If you tried to see both sides on one scope, could you either use two probes in math mode, or a probe that isolates the ground.

                Of course does a differential probe not galvanic isolate but may still be usable?

                 

                In short would I like to be able to place a probe wherever I want to, without having to think about common ground. I would love the probe to go from DC (know it's hard) to the highest possible below 2-ish GHz.

                 

                Why did I then choose the above solution? Because it's about the only circuit I could find, that in some way maybe could fulfill my needs. But if any of you have another idea will I be very happy!! :-)

                  • Re: 2GHz probe from article - noob questions?
                    shabaz

                    All probe methods are compromises, each method has strengths and weaknesses. The standard probes with a 'scope (i.e. passive 10x or 100x probes) are all-round useful and general-purpose, but as you say, they have a common connection that is also grounded on most 'scopes.

                    It's not normally an issue, because 90% of the time you're measuring with respect to 0V, so the probe ground can be clipped to 0V (provided 0V is connected to ground, or isolated, say a battery powered circuit).

                    If you wish to do what you're suggesting (place two or more probes wherever you want and not having the same common connection) then you're limited to the more specialist probes but they all have their weaknesses too. Add on top of that your requirement for DC-2GHz, and I'm not sure there's any reasonable option for everyday use. You could try the circuit, I'm just not sure it will be effective for you, and I have no alternative circuit.

                    If you just wish to use a single probe and not worry about where the ground clip of the probe is connected, you can already do that with normal passive probes, provided your circuit is running from batteries (which it often is for a beginner).

                    Also, 2 GHz is sometimes a curse, you can get more noisy waveforms (and end up having to set the 'scope's input bandwidth to lower settings if what you're interesting in seeing is the main signal present).

                    3 of 3 people found this helpful
                      • Re: 2GHz probe from article - noob questions?
                        friedmule

                        Okay, that taught me much.

                        My needs will luckily properly never be above 30V-50V so x100 would be nearly useless, do I think.

                        1x and 10X for better bandwidth maybe what I need.

                         

                        The probe I have above, what is that normally mainly used for? (willy-nilly probing, low/high voltage, other)

                          • Re: 2GHz probe from article - noob questions?
                            shabaz

                            It sounds it, but actually the reverse is true, the 10x and 100x will load the circuit under test less, allowing you to have better bandwidth into the 'scope. If 100x results in too much noise on the display for the smaller signals then 10x can be used. The 1x is generally not used due to extremely low bandwidth not suitable for observing many circuits. For instance I have some probes with optional 1X setting, but I've never slid that switch there so far.

                            The probe above is a low-voltage differential probe for differential signals, if it works correctly (but I think it may have anomalies, but I've not built it so speculating). It's not for general use, the general use probe is the normal passive probe. The other probe you mention has weaknesses too, it is only 60 kHz bandwidth, so that won't serve as a general probe.

                            1 of 1 people found this helpful
                    • Re: 2GHz probe from article - noob questions?
                      friedmule

                      I have read a lot and tried to understand the original article better but it has given me more questions instead:-)

                       

                      As I understand has the original author tested his design and fund it somewhat fine. But I do also read from you that his design is in fact not very good.

                      The size should be 0402 instead of 1206, the front component do I understand have to be in a straight line and not "folded" like in the article. The ground plane (and the other?) may not be under the signal resistors due to capacitance.

                       

                      In short, how, why and ??? Did he design and measure like he did if it is all a bad-ish idea? :-)

                      • Re: 2GHz probe from article - noob questions?
                        jc2048

                        While asking on another forum did I got told things I do not understand about the solution in the magazine.

                         

                        1) The coaxial cable is mounted with a long lead and will not work correctly.

                        2) The ground plane have to be removed underneath the input-resistors.

                        3) The input resistors may not be "folded" but have to be straight to work.

                         

                        1. The person responding would have been concerned about the matching of the cable. It's not a <work>/<doesn't work> thing. If you have a discontinuity, then it will show on the resulting waveform. The original author is going from a stripline trace (hopefully) to the transmission-line properties of a length of coax with something much less well-defined in between. The person responding in the thread was flagging it as a potential problem area that could possibly be done better. In practice, it's right at the end of the line, near the series termination, where things are a bit messy anyhow, so it may not make too much difference. What the original author did could be adapted quite easily [shorter cable tails and a scapel to cut the trace down to just the length needed], so it's not really a make-or-break issue on a prototype layout. A better arrangement might be to take the output stripline trace to the kind of connector that fits to the edge of a board, which is the kind of thing you would do on a commercial product.

                         

                        2. If you look at the Analog Devices datasheet, they show their test board layout and the way that they remove the planes away from the input and output pins. The reasons are different in the two cases.

                         

                        For the input pins, it degrades the margins for the feedback loop. That would result in ringing on sharp edges and, if you were unlucky, oscillation of the part. In theory, current-feedback devices like this shouldn't be as susceptible to capacitance at the input pins as voltage-feedback devices, but it's still a potential problem.

                         

                        For the output, the problem is with the emitter-followers inside the chip's output stage that are fussy about driving capacitance directly at the output. [The emitter-follower response makes it look somewhat inductive with frequency, so you're getting a resonance with the capacitance.]

                         

                        3. This relates to parasitic capacitance: to ground if you have a ground plane; end-to-end along the resistors; and [maybe] from the input to the feedback output pin. If you don't have the ground plane, it also relates to the differential impedance between the two inputs, which is further complicated by the two inputs of the amplifier not matching in the way that is the case with a voltage-feedback amplifier. So it's very involved to tease apart and explain [and the person on that forum may have been answering quickly and not thought through all of that anyway - if they were used to doing rf work, it would just have 'looked wrong' to them and flagged-up warning signs].

                         

                        The original [Elektor] author made something that worked well enough for his own use. Perhaps you should buy one of his boards and try it out to see if it works well enough for what you want [he comments on one of those EEV threads with contact details - I can't find it again, but you could probably manage that better than me]. Then maybe see if you can improve on it.

                         

                        Be aware that the voltage that the chip will tolerate between the two inputs to the device is low, so you will either need to be aware of that and take care when using the probe [I would suggest that you work out what the maximum differential voltage at the probe tips is before you use it and how far you can move each away from the common ground] or consider what you might do to protect the chip.

                        3 of 3 people found this helpful
                          • Re: 2GHz probe from article - noob questions?
                            friedmule

                            Thank you for your very clear and detailed answer! :-)

                             

                            To me does it sound like it may be a better idea to start out selecting components for what i need instead of relying on an old build and then maybe use a one layer board?

                            If possible, would a +/- 36V be more like what I could ever need.

                            The above design has a max 2GHz but the datasheet talks about140KHz bandwidth, not sure what that means in that design?

                              • Re: 2GHz probe from article - noob questions?
                                jc2048

                                You've got the wrong part number on your schematic: it was ADA4927-1, not ADA7927-1.

                                 

                                These are the pulse responses from the datasheet. In some ways, they give you a better idea of how your probe might perform than a raw bandwidth figure [I think they've made a mistake with the vertical scale on the two rightmost plots - you'd think that would be V rather than mV, wouldn't you - and the bottom left plot where you'd think it was the CM output rather than the differential one].

                                 

                                 

                                The wave shape is reasonably good with a truly differential output [the part is intended as a front-end to the differential inputs of a D/A converter], less so if you just grab one of the outputs relative to GND [though that may be good enough for what you want].

                                 

                                If you want to draw up a reasonable spec, how about going to the website of a major manufacturer like Tektronix, Le Croy, or Agilent, download the datasheets for the kind of probes you're interested in, and have a look at what they are able to achieve. That will then tell you the very best that you might manage and you can draw up a less demanding spec that's closer to what you'll actually manage with your own level of experience and skills [remember that some of these companies have over 50 years of accumulated experience and all have some of the best designers going - I wouldn't even pretend to be able to do the kind of stuff that they do].

                                  • Re: 2GHz probe from article - noob questions?
                                    friedmule

                                    Great explanation and idea to look at the main developers site. I had done that before trying to make my own and i found 36GHz probes and 1:1 6GHz versions.

                                     

                                    My dream goal would be a "perfect" floating 1:1 — (0 to 1.5GHz) — probe +/- 36V.

                                     

                                    I have some idea that if i find the right chip, keep the components small and close, then should it be possible to make something usable?

                              • Re: 2GHz probe from article - noob questions?
                                friedmule

                                There are of course also the possibility to isolate by using a light-bridge.

                                The signal does then have to be converted in some fashion but does maybe best remain an analog signal?

                                Please ignore C1 + C4, it was an oops in my strain of thoughts:-)