loading

So, you're an electronics hobbyist and want a 1GHz* active probe, or a professional and tired of blowing your 2000$+ active probe in sketchy circuits when all you wanted to know are the whereabouts of some RF-signal?

Then stay with me!!

This Instructable will show you how to build a 1GHz* Fet-based Active Probe, the Fetprobe, for about 10$*, provided you have access to an electronics lab. It is based on an Elektor-magazine article (see the pdf's addendum, section literature in my thesis) beside some other designs. However, as topic of my bachelor-thesis I wanted to find out how good these designs really are and how far one can push them.

If you go down this road, you need access to a lab that is equipped for some RF-fun, a cheap way for ordering RF-components and some rapid prototyping capabilities. Options some DIY-electronic-aficionados may not have. Although, my Bachelor-thesis comes with a lot of measured performance data, you will get pretty similar results if you stick to this tutorial.

In case you want to learn some more and don't mind a more scientific approach than have a look at my thesis included as PDF-file in last step.

Equipment needed to build this probe:

  • PCBs (512µm Rogers RO4003 w. 17µm copper dual-sided or similar) based on my gerber files
  • SMD soldering gear and tweezers!!
  • oscilloscope for debugging and testing
  • DC power supply
  • access to RF rated SMD component-kits (caps, resistors, inductors)

Components you need to get:

  • bf998 dual-gate Mosfet(s) SC-61B 4pin
  • RF-Rated 0603 caps around 1pF
  • 10M 0603 resistor (bias)
  • spring-loaded tips often called pogo pins (RF_in and GND)
  • voltage regulator, i.e. LM317LCDR SOIC 8pin
  • SMA or BNC connector (RF_out)

add. Equipment needed to design & develop your own probe:

  • PCB mill, capable of SMD boards 125µm track-gap or similar etching process
  • VNA (Vector Network Analyzer)
  • swiss-army-knife RF software like AWR Microwave Office from National Instruments (layouting, spice-sim, RF-sim)


*: handles 1MHz-500MHz really well, usable up to 1GHz though (if you can accept a bit of dispersion and amplitude error, as shown in the S21-graphs in the last step

**: means you may need a membership in a laboratory, or acquire some people's affection who do (i.e. via cold beverages). Also, the 20$ price tag suggests you only need to buy the special components, not the standard things which lie around in an RF-lab anyway.

Step 1: Motivation

What is an active probe, anyway?

Assume you got a device you want to repair, and may encounter signals between 1MHz and 1GHz.. just stick in your beloved passive probe that came with the oscilloscope, no worries?

Not that great.

First and foremost, even if your probe is good for the RF-signal's frequency you try to investigate, it still adds some capacitance to that trace you are probing. Some circuits may not like that and this is where the active probe comes to the rescue. Here, the amplifier in the probe's head drives the cable and and other unavoidable capacitances so your circuit doesn't have to.

However there is more to it:

A passive probe, as shown in the diagram, uses a variable RC-element (labeled "comp") in its input path to compensate the transfer function (S21*) for lower frequency applications.

While that works for a certain freq. range, for higher frequencies (~200-800MHz+) the elements cannot be approximated as lumped elements anymore, and the compensation breaks down.

Also, a ground lead, especially a long and curled up one can prove dangerous, because it adds unwanted inductance, and may pick up some RF as well. That's why you should always keep it as short and straight as possible, or remove it at all and add a more direct ground connection, like the RF grounding-spring to your passive probe if you are working on RF-circuits. My active probe has the same kind of spring-loaded tip for the GND connection as for the RF input.

So, what can this DIY active probe do?

  • be cheap & simple
  • relatively easy to build and maintain
  • probably suitable for low-level field or student use

How does it do that?

  • high input impedance S11*
  • stable transfer function
  • good pulse response, little nonlinear distortion (see THD in the pdf for more info)

*: The general term for resistance is impedance, and in the context of how much input impedance a measurement tool has, the input reflection coefficent S11 comes in handy (smaller = better).

S21 is the forward transmission coefficient, being a measure of the output signal's power in relation to the input signal's (here: being mostly stable over frequency = best).

Step 2: Circuit Diagram

How to design such a device?

In essence, it has to be an impedance converter: High impedance (less power flowing away) for the DUT (device under test), controlled impedance (nice strong signal, low distortion) for the measurement instrument.

This can be based around an operational amplifier or a mosfet, but for me the fet-solution turned out to be better (see thesis-pdf in last step).

Above, you can see the schematic* of the Fetprobe, the fet-based Active Probe.

A very small input capacitor C_in in the input path guarantees that only veeery little current gets away from the DUT, and next to it a 10M bias-resistor ensures that the mosfet operates in its ideal DC-working-point (linear area of input-output relation. signal now is AC+DC). Now, the small signal itches the double-gated fet, which in turn creates a stronger signal. Followed up by the resistors R1 and R2 which determine the fet's gain and help to better transmit waves over the coax (output impedance matching), especially on a 50Ohm system (typically used for RF-measurement equipment). Then, the DC-part of the signal (introduced for the happiness of the fet) gets chopped off by the capacitor C_out (not part of original signal).

At last, a filter dampens a resonance around 800MHz, with values around C1=C2=1.8pF, L=22nH.

It has to be noted that these values most likely have to be tuned separately for each Active Probe replica, as the resonance and the gain seem to vary due to manufacturing (your soldering, as well as the chip itself**).

*: Showing only the functional parts, no parasitics, they are described in the attached pdf.

**: Something like this could be automatically regulated by other fets/transistors, but the design goal "simplicity" forbids doing that (as well as the notion that every additional component, especially active ones, often introduce additional problems, and it's very hard to get that all under control, pricey components inclusive. That's part of why comparable professional tools may cost a hundredfold).

attached pdf's: how the schematics look like in NI's AWR

Step 3: PCB Design

How to turn the schematic into hardware?

Well, it's considered good practice to first build a prototype and demonstrate the concept (first picture).

Only then, as you know that your circuit actually works, the full-blown design (i.e. with its own voltage reg. instead of having a labPSU dedicated to it) makes sense to be implemented. Don't trust any spice-simulation over a real PCB too much, especially if you cascade many functional instances!

The other pictures show the 3D-View in NI's AWR, as well as the real PCB.

Note that the RF-path is implemented as a 50Ohms line (Wave impedance, not resistance!) and kept clear of other traces for about one trace width or three times the substrate thickness, whatever results in the higher separation. The input-path kept as short as possible, the ground-plane is removed below it and the FET to get rid of parasitic capacities. Note that this first section leading up the FET is not impedance controlled, we want it to be as high impedance as possible to not disturb the circuit we are probing.

At the end of the day, you want to hold it in your hands or mount it to a stand, so two copper clad FR4-circuit boards (w. Kapton-tape finish) where attached as housing. They not only prevent your hands from touching the circuit, the standoffs provide some extra clearance. Therefore, prohibit interfering with the measurements since the input-path's 0.5pF and 10MOhm are in the same ballpark of what your fingers might do! Using FR4 circuit material as enclosure was just an easy way of utilizing the PCB mill to build the housing. If you don't have one, make something similar from sheet metal or even put it in a plastic case. However, be sure everything stays right clear of the business end to avoid unnecessary stray capacitance.

To supply power, the Fetprobe can use everything from 12 to 30 Volts (that is, its voltage regulator can handle). Note that the hardware implementation also features a few extra bypass capacitors (bias, supply of fet) and a few zero-Ohm-links, where protection diodes could go, depending how careful you are about the polarity when you are hooking it up. When choosing caps for the supply be sure to always team up different kinds in parallel. E.q. 2.3µF electrolytic and 100nF ceramic. This helps with overcoming the potentially poor RF performance of the electrolytic caps.

Step 4: Building & Debugging

Now, let's get our hands dirty!

Get the PCB*, put the components on, solder them.

Piece a' cake? Let me put it this way: some of the components are as big as a flea, literally. Unless you have eagle-vision or much experience, a microscope or a nice desktop magnifying glass is mandatory. Some small clamps, a watchmakers vise or other random tools of the right size (i.e. pliers) can help to hold stuff down, like the PCB.

Static Electricity or ESD!!

The transistor you are handling is a very delicate thing made for detecting minute signals, no wonder it will take damage from some 1000V worth of static discharge. Make sure your microscope, vise and most important you are grounded! Ideally, connect yourself with a 1-10Meg resistor in series to ground (If you ask ebay for grounding straps they will happily supply one for a few bucks). This contraption will absorb the power of any discharge (preventing its dissipation inside your electronic devices). Always touch these "grounded" objects before touching the PCB, or something may blow without you noticing at first.

Suggestions on SMD-Soldering:

Of course, for proper instructions, look for an in-depth Instructable or ask colleagues, but here are some things I struggled with / remember doing:

  • prepare the solderpads via scratching them a little, i.e. with a glass-fibre brush (SMD gear)
  • use the thinnest solder possible (i.e. 0.3 mm)**
  • put additional flux on the pcb, it helps (even if your solder has a flux-core)
  • too much solder is too much, especially in the RF-path: If a blob forms, remove it partially with solder-wick, or unwanted capacities will give you headaches (you only want a smooth slope from the component's lead to the pad, no mountain of solder)
  • too little may fool you: since too little solder in the RF-path may result in an open circuit which equals a capacitance, it'll probably still somewhat work, but give you variable results!
  • don't be fooled into assembling everything at once, i.e. test the power supply section BEFORE you put the main device (i.e. the fet) on the board, just in case the v-regulator's resistors got swapped and now it produces an inadequate voltage, or there's now power at all, or something (see notes on picture with green vise)
  • don't stress the tiny traces too much: They will rip, and better pray its not the RF-path you just killed (tough, it won't help if you did, so better be safe than sorry). Use a tweezers-style desoltering-tool or two solder-irons at the same time in case you need to remove a component.
  • play the touchy-game, if the graphs look strange: By putting your fingers anywhere on the circuit, you may be able to detect faulty solder joints, as well as how a little additional parasitic capacity influences the setup (here: low voltages & very low currents = little power => no danger).

*: always use a gerber-viewer like ViewMate to check whether all holes and components got exported just fine, also check whether it got loaded properly into your milling-machine software.

**: note that unlike on a pcb with a soldermask, here you have to be very precise with solder usage and can't just spam some on there and it'll only stick to the pads where it belongs and nowhere else. This project doesn't have a soldermask (additional manufacturing process, probably additional capacity), so too much solder will stick to the traces, and you don't want it to be there (may add capacity, inductance) so remove it with solder-wick.

Step 5: Measurements & Thesis

After all the building, measurements are to be made!

The first two pictures show the setup:

A 50-Ohm test-trace is being probed in the middle using the Fetprobe. Here a VNA (Vector Network Analyzer), which has been calibrated to the specific cables in use, is used both as signal source as well as signal analyzer. All the following graphs are generated this way measuring different parameters and trying out different active probes.

First, a comparison of the Fetprobe to a commercial HP probe is presented. Secondly graphs on how the Fetprobe behaves on different power levels are shown. Broadly speaking everything works almost perfectly from 1MHz up to 500MHz and for voltages up to 2V peak and if you can tolerate minor amplitude errors and distortion it is well useable up to 1GHz.

dBm to voltages table: www.minicircuits.com/pages/pdfs/dg03-110.pdf

If you want to know more about the process, and how the Fetprobe behaves in a pulse-response scenario, or on THD (Total Harmonic distortion) measurements, be sure to check out the pdf of my BSc-thesis.

When you made it here, all that's left to say is: Happy soldering :)

<p>After spending like two hours or so reading this ible, your paper and researching a bit about dual gate mosfet I'm very impressed about how this all turned out. Despite of the detailed documentation I still have a ton of questions:</p><p>1. How can a maker use this? Can it's output be directly connected to an oscilloscope or is an output termination impedance required?</p><p>2. How significant is the advantage of this probe compared to a passive probe with a hobby-level 100/200/300MHz scope? A typical example might be probing a 100MHz clock line or similiar.</p><p>3. If I understand correcly the gate 2 of the FET is used to set a DC operation point, the signal on gate1 is then superimposed onto it. Why is a seperate bias voltage needed? How was the value of R1 choosen?</p><p>4. What parameters limit the maximum input amplitude before clipping/distortion starts?</p><p>5. Do you think it could be possible to use a standard FR4 PCB with maybe about 5% tolerance in impedance? I'm askin because right now the board is by far the most expensive part in the BOM and limits the availability for makers.</p><p>I've to admit that most of the questions above are due to my poor knowledge of RF/HF ciruits in general, please excuse if any of them is extraordinary dumb.</p><p>Nevertheless I do love this kind of do it yourself high-spec/low-cost personal lab equipment as it allows anyone to get his hands on more advanced electrnics. That said, thank you very much for sharing your bachelor thesis with us, and welcome to the community, Thomas!</p>
<p>Thanks<br>for asking! I can tell you straight away (Pages refer to the thesis):</p><p><br>1)The<br>Termination can be omitted in principle if the cable-length L is<br>smaller than the wavelength/2 (c=lambda * f, c=2/3 * speed-of-light<br>(because of coaxcable)), since no standing waves can form (due to<br>boundary conditions) and wavepropagation can be neglected. (i.e. 1ft<br>cable = 0,3m &rarr; max. ~250MHz).</p><p><br>3.1)<br>No, the working point gets applied via R_bias, the second gate is<br>merely short-circuited for gain stabilisation. The 9V at the drain is<br>the supply, not the bias. I changed the setup-schematic to reflect<br>that (&quot;ext. DC&quot; referred to the thesis' experimental bias,<br>oops..)</p><p><br>3.2)<br>see Page 7: &quot;This [additional bias voltage] is required as the<br>Fet needs to work in the positive input voltage range, and without<br>the DC-offset it would handle negative voltage transitions<br>significantly worse.&quot; (Any transistor rated for AC has to have a<br>bias-voltage, as it can only open and close like a valve so to speak,<br>but not provide negative voltage directly. So you need to superimpose<br>some DC, let the transistor do its thing, then chop off the DC again)</p><p><br>3.3)<br>R1 was chosen by playing around with adjacent values of 47Ohm to get<br>to 1/20 signal &bdquo;amplification&ldquo; (&rarr;x20 probe). You can estimate<br>the gain via formula 2.1 of page 7 (having calculated &quot;S&quot;<br>after a measurement with R1=47Ohm), but that's only an idealization<br>;)</p>
<p>1) That's a great rule of thumb! It'll come in useful one day, thats for sure.</p><p>3) Ok, I think I get it know: I assume you drive the transistor to it's limits, so I_DS&asymp;30mA. The voltage across the resistor is therfore &asymp;1.4V and the remaining U_G1-S&asymp;1V will turn the transistor on. The gain of &asymp;20 correlates with the figure 6 from the <a href="http://www.nxp.com/documents/data_sheet/BF998.pdf">datasheet</a> at high voltages of U_G2-S (I somehow first assumed that there is no gain, just like when a npn's collector is connected to the supply). Is that correct so far?</p><p>This also gives a hint about question 4, as the input amplitude must be Uin &le; ∆Uout,max/gain &asymp; 1.4V/20 = 70mV. This also explains why you've measuered a pcb trace in step 5, the voltage drop across the trace offers a great low impedance source!</p><p>One last question, why has the BF998 be choosen for this project? After all the second gate isn't used beside forcing a linear gain, assuming there are single gate mosfets with a linear gain.</p><p>In the current semester I have a course where the basics of transistor circuits/amplifiers are explained and it's exiting to see actual ciruits based on some of the ideas (we haven't discussed FETs yet, though). Today I leared something, thanks!</p>
<p>2) If your 100MHz clock-signal is sinusoidial and relatively strong, your average passive probe probably will do fine. However, if you've got a non-sinusoidial (and/or weak signal), e.g. rectangular, your scope may show a worse representation due to relatively lowpass behaviour of the passive probe, meaning some of the harmonic components needed to form the signal may be cut off, as well as the measurement influencing the (weak) signal (by weakening it further).</p><p>5) Probably ~400-500MHz limitation with FR4: The problem with FR4 is its relatively &quot;Epsilon_r&quot; (dielectric constant), so my guess is you may be able to work with that, but have to deal with more variation from board to board, as well as locally if you're unlucky, presumably especially at the output filter. Nevertheless, I'd like to see a FR4 implementation turns out!</p><p>(Why the BF998?): That's in the thesis, page 4-7 ;)</p><p>(x20 gain): there is no x20 voltage &quot;gain&quot;, please see my reply to (3.3: 1/20 ampl. = x20 attenuation), but ID=30mA is correct. The reason that VG1, the bias voltage (on the RF_in) is 2.4V (figure 3.16: thesis page 25) is that negative parts of the input signal can't pull it out of the linear area and towards the left in bf998's diagram6, so to speak.</p><p>4)(input rating): no, please see the last graphs of the instructable and calculate dBm to V via the provided linked table. The bias-voltage is probably the first limit, so you had to drive the mosfet with an unsafe high bias (uncertain life expectancy) to get a higher Vpp_in,max. In that respect the Vpp_in,max=2V, as given on the last page of the instructable, seems quite reasonable. Also refer to the THD figures (e.g. page 30).</p>
<p>This seems so interesting.Need to try this out!</p>
Wow! Very nice. I appreciate your sharing of the entire design process and the results!

About This Instructable

9,455views

134favorites

License:

More by thirschbuechler:Ubuntu recent files quicklists like on windows (in quicklaunch / taskbar) DIY 1GHz* Active Probe For Under 20$** another 0$ bike bottle holder, pipe based 
Add instructable to: