The “problem”

I’ve been wanting to have an high-impedance (Hi-Z) probe for some time now.

Up until now I’ve been searching for something like an HP 85024a which comes up on the second hand market from time to time. It would be overkill for my needs, but still nice… :)

Recently I came across an old Elektor/Elektuur article from April 2004 titled “Poor man’s 1GHz active probe” by David Jewsbury. Some more research turned up the bachelor thesis of Thomas Hirschbuechler titled “Design of a low-cost 1GHz active probe”.

I decided it would be a good learning experience to try and recreate/adapt the design and make some measurements on it.

Design

The board measures 60 by 25mm. As I do not intend to use the probe up to 1 GHz I did not bother to use Rogers or some other high-performance material and opted for regular FR4. Design is done in Diptrace V4 which I can recommend as an easy-to-learn EDA.

The current state of the design is in my Github repository although I need to update the schematic with the revisions I made during prototyping.

v0.1 Prototype board with many changes.

The total cost of the current version is well below €50 where the SMA connector is probably the most expensive part.

Measurements

Bode plot (0-25MHz)

After tweaking and making some changes on the initial design, measurements turned out pretty good (taking in to account thatI do not intend to use the probe much above 200-300MHz). I made a bode plot 0-25MHz with my MSO5074 scope which has a 25MHz signal generator build-in:

Bode plot

S21/S11 from 300kHz to 1GHz

For higher frequencies I turned to my 300kHz-1.3GHz HP vector network analyser:

S21/S11

Frequency(MHz)S21(dB)S11(dB)
0.300-33.06-0.38
5.000-20.140.00
10.000-17.76-0.08
15.000-17.05-0.16
20.000-16.81-0.33
25.000-16.79-0.45
30.000-16.85-0.55
35.000-16.90-0.59
40.000-16.91-0.62
45.000-16.89-0.61
50.000-16.79-0.59
55.000-16.68-0.60
60.000-16.60-0.64
65.000-16.58-0.73
70.000-16.62-0.81
75.000-16.78-0.87
80.000-16.96-0.87
85.000-17.11-0.85
90.000-17.16-0.81
95.000-17.11-0.75
100.000-16.96-0.76
110.000-16.48-0.94
120.000-16.39-1.31
130.000-16.85-1.42
140.000-17.27-1.14
150.000-17.22-0.94
160.000-16.78-1.12
170.000-16.48-1.70
180.000-16.70-2.03
190.000-17.07-1.67
200.000-17.18-1.24
210.000-16.97-1.22
220.000-16.74-1.75
230.000-16.58-2.42
240.000-16.57-2.32
250.000-16.63-1.72
260.000-16.76-1.40
270.000-16.87-1.68
280.000-16.73-2.40
290.000-16.40-2.77
300.000-16.31-2.27
310.000-16.58-1.65
320.000-16.82-1.54
330.000-16.58-2.10
340.000-16.05-3.04
350.000-16.00-3.15
360.000-16.51-2.22
370.000-16.85-1.62
380.000-16.59-1.79
390.000-15.90-2.83
400.000-15.54-3.96
410.000-15.85-3.21
420.000-16.27-1.93
430.000-16.30-1.63
440.000-15.93-2.45
450.000-15.54-3.97
460.000-15.48-4.02
470.000-15.69-2.36
480.000-15.88-1.49
490.000-15.75-1.88
500.000-15.31-3.34
520.000-14.59-2.93
540.000-15.47-1.38
560.000-14.85-3.91
580.000-14.35-1.92
600.000-14.86-1.63
620.000-13.94-4.06
640.000-14.06-1.33
660.000-14.22-1.98
680.000-13.06-3.99
700.000-13.42-0.69
720.000-13.14-2.69
740.000-11.91-2.54
760.000-13.07-0.33
780.000-12.18-3.99
800.000-11.48-0.62
820.000-12.07-1.25
840.000-12.13-5.44
860.000-12.24-0.25
880.000-12.64-2.42
900.000-11.78-2.84
920.000-12.09-0.69
940.000-12.39-4.53
960.000-11.77-2.24
980.000-12.55-2.87
1.000.000-11.21-8.23

The low frequency response (up to, say, 5MHz) isn’t particularly good. Starting around 5MHz, S21 is fairly flat around -17dB up to 400MHz. From there on it starts to increase.

S11 is between 0 and -4 dB up to 800MHz which I think is very good.

Things I’d like to figure out

  • First of all: are my assumptions regarding the measurements correct?
  • How do I improve the low-frequency response to be more linear?
  • What causes the response (both S21 and S11) to ripple rather continuously? Should I suspect my measurement setup?
  • Would I be better to lower the FET supply voltage (or input bias) in order to end up with an even -20dB gain, so that converting the voltage magnitude would be easier?

Plans

I’ll make a v0.2 with the changes and some additional improvements and then a properly shielded enclosure (likely pcb material).

Final thoughts

These results are (much) better than I expected from my first attempt, so I am trying to figure out if I am measuring the correct parameters and whether my interpretation is correct…

While this is mostly an exercise and learning experience, I’m rather happy with the results so far.