RF current sampler

This article describes an RF current sampler for HF.

The sampler could be used to monitor a transmission with an oscilloscope or spectrum analyser.

The current sampler is chosen rather than a voltage sampler, as it is much easier to get a flat and predictable frequency response with minimal insertion VSWR.

Fundamental concepts

Lets set out some fundamental concepts that guide the design.

Characteristics of ideal voltmeters and ammeters:

Ideally, voltmeters should be very high impedance devices and ammeters should be very low impedance devices, otherwise they disturb the thing being measured.

The RF current sampler is like an ammeter, it should drop very little voltage so that when it is inserted in the line, current and voltage is essentially the same as without.


The design is to suit up to 2kW in 50Ω continuous. Current at 2kW in 50Ω is (2000/50)^0.5=6.325A, so the probe is rated for 6.5A.

For a maximum of 1W lost in the current transformer, Vp=1/6.325=0.158V.

Choosing a 1:30 turns ratio for the current transformer, Rs=0.158/6.325*30^2=22.48Ω.

The secondary load Rs is 50 || R', so R'=1/(1/22.48-1/50)=40.85Ω which will be supplied by two 82Ω 1W carbon resistors in parallel. Dissipation in the resistors at Ip=6.5A and with the monitor port o/c is (6.5/30)^2*41=1.9W.

Power in the 50Ω monitor device at Ip=6.5A is (6.5/30*22.48)^2/50=0.47W (26.7dBm).

The secondary voltage to primary current calibration factor for the probe loaded by a 50Ω load is 0.158*30/6.325=0.75V/A. Alternatively, for 50W through (1A), the power in the 50Ω monitor is 0.75^2/50=11.25mW which is 10*log(11.25/50)=-36.5dB.

The impedance inserted in the through line is 22.48/30^2=0.025Ω. This design delivers a very low series impedance to the line under test.

The primary will be one turn shielded using a piece of RG58 transmission line, braid grounded at one end only.

Fig 1:

Fig 1 shows the currents that flow in primary circuit of the system. In this instance, the electrical length from one side of the box to the other is less than 1.5° at 30MHz, and less at lower frequencies, so for the purpose of this explanation, it is assumed that there is zero phase shift within the box.

 In a coaxial cable where skin effect is fully developed in the shield conductor, the inside and outside surfaces of the shield are isolated by the skin effect.

 In TEM mode, for a current I (as shown on the outer surface of the inner conductor) there is an equal current I in the opposite direction on the inner surface of the outer conductor. At the right hand end of the shield, that leftwards current must flow around the end of the shield and a rightwards current flows on the outer surface of the shield. At the left end of the shield there is a node with three current paths: the inner surface of the outer conductor, the outer surface of the outer conductor and the tie to the box. In the even, the current into the node from the inner surface of the outer conductor is equal to the current out of the node to the outer surface of the outer conductor, and the current on the tie is zero (near zero in practice).

So, the effect of this structure is that the outer surface of the outer conductor carries the same current as the inner conductor, but it is at near zero potential due to the tie to the box, and so less current capacitively coupled to the secondary. Some might suggest the tie should go to the point where the secondary circuit is 'grounded', the BNC connector on the lower edge of the box (not shown in Fig 1), and that it is the centre of the braid that should be 'grounded'. The difference is negligible in this application, and some configurations are less practical, impractical even.

The current transformer will use an inexpensive (A$1) Fair-rite 2643625002 ferrite core which will easily slip over the RG58 transmission line with the secondary winding in place.

The primary magnetising impedance of the core that appears in shunt with and needs to be much greater than the transformed secondary impedance, approximately 41/30^2=0.045Ω.

Fig 2:

Fig 2 shows the primary magnetising impedance of the core calculated using Calculate ferrite cored inductor. Z (0.781+j9.76) is greater above 1MHz. 

Fig 3:

Fig 3 shows from the manufacturer's data, the primary magnetising impedance of the core that appears in shunt with the transformed secondary impedance (0.045Ω). The magnetising impedance is much greater over the whole HF range, and so magnetising current is insignificant, the core is most suitable.

Table 1: Design parameters
Frequency range 1-30MHz
Maximum power in 50Ω 2kW continuous with 50Ω load on the monitor port   
Maximum current 6.5A  
Coupling factor Vs/Ip=0.75V/A with 50Ω load on the monitor port
-36.5dB with 50Ω load on the monitor port

Table 1 sets out a summary of the design parameters.


Fig 4:

Fig 4 shows the key parts. The case has been drilled and coax connectors fitted, the ferrite core secondary wound and the through coax line prepared.

Note the detail of the coax. It is a piece of RG58C/U, shield to be grounded at one end only so that it acts as an electrostatic shield to reduce sensitivity of the unbalanced secondary to the through line voltage.

Fig 5:

Fig 5 shows the interior of the assembled unit. The ferrite core is held in place with a generous application of hot melt glue to the coax jacket before working the wound core over the glued jacket.

Fig 6:

Fig 6 shows the completed current sampler.

Fig 7:

Fig 7 shows the measured amplitude response of the sampler response is within ±0.1dB over HF (±1% voltage variation). It is a few tenths of a dB lower than predicted, possibly the result of some flux leakage, resistor tolerance, with contribution from the accuracy of the 50Ω termination and other measurement errors. Coupling factor averages -38.5dB or 0.72V/A.

The measurements validate the choice of a medium µ ferrite toroid. Though #43 material is a lossy material, in this application, flux leakage is very low (due to the medium permeability) and the magnetising impedance is very high so the effect of core loss is negligible, leading to a predictable and flat response.

Fig 8:

It must be stressed that this calibration, and the flat response require termination of the sampler port with a 50Ω load. If the port is connected to an oscilloscope, it must have a 50Ω input impedance or a suitable inline termination must be used, any standing waves on the connecting cable will disrupt the accuracy. Fig 8 shows a suitable inline termination to be used at the CRO input jack (TE Connectivity 1-1478207-0).


If a -40dB coupling factor is preferred, then if P=50W, Ip=1A, Vs^2/50=50*10^(-40/10), Vs=0.5V and so Rs=0.5/1*30=15Ω and R'=1/(1/15-1/50)=21.4Ω which could be approximated with two parallel 43Ω 1W carbon resistors.

The current transformer and load resistance element of this circuit could be combined with the detector circuit of Measuring common mode current for a direct reading RF Ammeter. RF Ammeters were once used widely, but since the source of thermocouple ammeters from disposals channels dried up, hams seem to think that RF Ammeters are no longer available (Straw 2007).

A correspondent asked what would be needed for a similar sampler for the MF band.  This design works for MF, measured response does not roll off until frequency is less than 300kHz. Current measurements are often used with MF antennas of known R to infer the power input to the antenna system.


(Danzer 2009) describes a sampler with a high impedance secondary circuit. If it has a low voltage drop in the primary circuit, it will only be by virtue of considerable flux leakage from the low permeability powdered iron core and the low magnetising impedance.

(Hallas 2012) quotes Danzer's design, but gives a diagram of his own with no discussion of the importance of a low impedance secondary circuit.

(W2AEW 2011) gives a voltage probe which does not discuss the loading effect of the oscilloscope and connecting cable, the effect of standing waves and bandwidth.

A new fashion with the evolution of AM in ham use is a sampler box that includes both a raw RF sample and a envelope detector output for the purposes of a trapezoidal pattern on an oscilloscope to monitor modulation. Of course, whilst the pattern shows the depth of modulation correctly, any linearity inferences apply ONLY to the detector circuit and NOT to the transmitter itself. For an indication of end to end modulation linearity, the audio sample must be from the audio input. Unfortunately, that doesn't work well with modern DSP based modulators which have substantial latency, and the modulation might be many ms delayed which causes an apparent phase shift dependent on modulating frequency and a resultant elliptical shape to the upper and lower edges of the trapezoidal pattern.

If projects do not give expected or measured performance characteristics, performance is unknown!

References / links



Version Date Description
1.01 31/07/12 Initial.

© Copyright: Owen Duffy 1995, 2017. All rights reserved. Disclaimer.