Poor Man 1GHz
description
Transcript of Poor Man 1GHz
-
When it is required to make a measurement at a node of anRF circuit, connecting to the circuit using a normal oscilloscopeprobe, even on the x10 setting can change the behaviour ofthe circuit. For those difficult cases, you need a special probelike the one described here.
Poor Mans 1-GHz
DIY saves a pretty penny
elektuur - 4/200436
David Jewsbury
Any probe presents extra impedancefor the circuit to drive, usually consist-ing of some resistance and straycapacitance, resulting in reduction ingain, or in extreme cases, causinginstability.
The loading effect of the resistanceand stray capacitance can be largelyremoved by using an active probe. Themajor manufacturers in the oscillo-scope market all offer suitable models(see also Scope for Scopes elsewhere
in this issue), but costing over a 1000they are too expensive for amateur use.This article describes a probe that canbe constructed at home, for very littlemoney and has useful performance.
SpecificationsThis probe has some compromises inperformance, as you would expect. InTable 1 it is compared to a commonlyavailable commercial probe, the type85024A from Agilent.
Admittedly the commercial probe, with0 dB loss, is more convenient to workwith, but for most applications a home-brew probe is no disadvantage.
Circuit DescriptionThe circuit is shown in Figure 1. It ishard to imagine anything simpler.A dual gate MOSFET, T1, is used ina source-follower configuration. Thisprovides a low output impedance todrive the coax cable and test equip-
-
ment. The signal at the probe tip isapplied to gate 1. The impedance atgate 1 is a very high resistanceshunted by a few picofarads ofcapacitance. The choice of MOSFETused in the circuit is not critical, anyone of the types listed in Table 2and housed in a SOT143 case can beused with impunity. Be sure how-ever to steer clear of -R suffixdevices because they have a differ-
ent pinout and will not work on theproposed PCB.
Capacitor C1 has a value of about0.5 pF, and is made by patches of cop-per on each side of the board. The gainof the buffer itself is a little less thanone, but because of the voltage divideraction of C1 and the input capacitanceof T1, the overall loss of the probe isapproximately 20 dB, or the input volt-
age is divided by 10.IC1 regulates the supply voltage to astable 5 volts. D1 protects the probe inthe event of the supply leads beingreversed.
ConstructionThe PCB artwork is shown in Fig-ure 2. The board has been designedto allow fitting in a metal tube. All
Active Probe
4/2004 - elektuur 37
78L05IC1
1AC6
470n
C5
470n
D1
SMD
T1
BF998
R4
47
R1
10M
R2
4k7
R36k8
C2
1n
C3100n
C4
RG178
1n
C1
see text
*
*
+8V...+30V
040108 - 11
S
D
G1
G2BF998
Figure 1. Circuit diagram of the DIY active probe. A dual-gate MOSFET guarantees light, uniform loading of RF signals over afrequency range extending well beyond the 1 GHz mark.
Table 2. MOSFET selection guide
TypeCiG1(pF)
Noise figure(dB)
BF990 2.6 2
BF991 2.1 1
BF992 4 1.2
BF998 2.1 1
Table 1. Commercial / homebrew comparison
Agilent 85024A Homebrew probe
Input Impedance 0.75 pF // 1 M 0.75 pF // 10 M
Bandwidth 300 kHz to 1 GHz ( 1.5 dB),or 1 GHz to 3 GHz ( 2.5 dB) 100 kHz to 1.5 GHz ( 2.5 dB)
Gain 0 dB nominal 20 dB nominal
1-dB compression point 0.3 V RMS not measured
-
the components are surface mounted(SMD), but assembly by hand is rea-sonably easy with a fine tipped sol-dering iron and tweezers. The partsare assembled on one side of a dou-ble sided 1.6-mm thick PCB. Connec-tions are made between the groundplane on each side of the board withsolid wire soldered on each side. Fly-ing leads take the power to the probeand a length of coax ending with aBNC plug take the output to the testinstrument. Heat shrink coax is usedto strain relieve the leads. The RF andGround probe are made from steelpins filed to a point. Pins borrowedfrom the familys clothes repair kit areexcellent.
Testing and use ofthe probe
After connecting the probe leads toa suitable power supply, the probeshould draw between 10 and 30 mA.If all is in order, connect the probe toa spectrum analyser. Applying an RFsignal to the probe should result inan output seen on the spectrumanalyser. To get accurate results it isimportant that the ground probecontacts an RF ground close to theprobed point on the circuit. It is alsoimportant to hold the board by theedges to prevent stray effects fromfingers on the circuit. If the imped-ance at the probed point is 50 ,
then the peak on the spectrumanalyser should be about 20 dB lessthan the power at that point in thecircuit. Commercial probes werenotoriously sensitive to electrostaticdischarge, but seem more robustthese days. Although T1 has internaldiodes to protect against ESD it iswise to take normal precautionsagainst unwanted static, whileusing the probe, as for any sensitiveelectronics.
(040108-1)
elektuur - 4/200438
C2
C3C4
C5C6
D1IC1R1
R2 R3
R4
T1
040108-1
Figure 2. The PCB has been designed with compactness and low input capacitancein mind hence the use of SMD parts.
COMPONENTSLISTAll resistors and capacitors: SMD, 0805 case
Resistors:R1 = 10MR2 = 4k7R3 = 6k8
R4 = 47
Capacitors:C1 = PCB capacitorC2,C4 = 1nFC3 = 100nFC5,C6 = 470nF
Semiconductors:D1 = 1A diode, SMDT1 = BF998 in SOT143 case
(see Table 2)IC1 = 78L05 in SO-8 case
MasterclassIt is important to realise that the probe is measuring RF voltage, but the dis-played quantity is usually the power that the probe is delivering to thespectrum or network analyser. The voltage at the probe tip is given by:
Where P is the displayed power in dBm, and L is the loss in the probe in dB.If the probe is being used for faultfinding purposes or only an approximatemeasurement is needed, L can be taken as 20 dB. For accurate measure-ments the probe can be calibrated over its frequency range, using thesetup shown here.The 50- load can be a 51- 0805-style surface mounted metal film resis-tor, soldered at the end of a piece of semi-rigid coax. The resistor shouldbe reasonably non-reactive up to 1 GHz.The loss of the probe is designed to be slightly less than 20 dB so than ifneeded it can be set to exactly 20 dB by trimming small amounts of copperfrom C1 with a scalpel. After calibration, very accurate measurements areavailable in 50- systems. With other impedances there is an additionalsmall error due to the unavoidable residual loading effects of the probe.
V
P L
=
( )1020
10
Spectrum Analyser
Signal Generator
RFprobe
50load
040108 - 12