Opamp-based active antenna

Pieter-Tjerk de Boer, PA3FWM web@pa3fwm.nl

(This is an adapted version of part of an article I wrote for the Dutch amateur radio magazine Electron, May 2025.)

Since September 2024, the Twente WebSDR uses a new active antenna. Like the previous one (see [6]) it's an E-field probe, also known as MiniWhip. But there are major differences: it uses the probe's current rather than voltage, and the active component is an opamp rather than a few discrete transistors. In this article we study the principle used, also known as the 'Nordholt antenna' in some amateur radio circles, and how an opamp behaves in it, particularly w.r.t. intermodulation.

The 'Nordholt antenna'

In 1980, Ernst Nordholt and Durk van Willigen (ex PA0DVW), who both were working at Delft University of technology at the time, came up with a new principle for active antennas [1,2], in which the amplifier input has a low input impedance. As Van Willigen recently told me, he was building an active antenna at that time based on the usual principle: a whip connected to an amplifier with a high input impedance. Nordholt was doing research for his PhD on feedback amplifiers, and realised that a low input impedance might be a better choice.

A short whip acts as a voltage source in series with a small capacitance. An amplifier with a high input impedance does not load this whip, and 'measures' the unloaded voltage, which at given field strength is independent of the frequency. In contrast, an amplifier with a low input impedance short-circuits the whip, and 'measures' the short-circuit current. Because of the series capacitance, this short-circuit current increases proportionally to the frequency. So if one desires a frequency-independent response, one has to compensate for this in the amplifier. Figure (a) shows how Nordholt and Van Willigen approached this. Their circuit is effectively an opamp of which the non-inverting input is grounded, and of which the inverting input is connected to the whip and to a feedback capacitor. The opamp with feedback 'strives' to have no voltage between its inputs, thus making the inverting input a virtual ground which short-circuits the whip. This is just the standard circuit for using an opamp as an inverting amplifier, but using capacitors instead of resistors. As a result, the total amplification (from voltage source to output) is frequency-indepedent, with the ratio of C1 (the whip's capacitance) and C2 (the feedback capacitor) determining the amplification factor.

An advantage of such a low-impedance input is that one can easily apply overvoltage-protection at the input using a few diodes, without distorting the desired signals: Nordholt's design could withstand 7 mm long sparks. Another advantage, particularly for military applications, was that if e.g. seawater splashes near the antenna, its capacitance (in parallel to the amplifier input) does not reduce the sensitivity. They patented the idea and sold it to Philips, but it has not seen much use.

Practical schematic

Nordholt built his 'opamp' from seven discrete transistors and a pile of passive components, because back then opamps suitable for sufficiently high frequencies were not yet available as ICs. Nowadays that's different. I've built the WebSDR's new antenna using an OPA818, as shown in the schematic. (The Wuerth and Bourns chokes were advised somewhere on the internet as being very suitable because of their high self-resonant frequencies, but I can't find anymore who gave this advice and where. Different inductors can also be used.)

Figure (b) shows only the essentials. Compared to the earlier figure (a), the feedback circuit became more complicated. The most important is the 2.5 pF capacitor, taking the role of C2 in the first figure. Together with the antenna element's capacitance C1, which is 6 pF, this gives a voltage amplification of 6 / 2.5 = 2.4.

In parallel this, there's a series circuit of 22 pF and 4.7 kΩ. This reduces the amplification at lower frequencies. Because background noise is stronger at lower frequencies, a lower amplification is enough while still ensuring the desired signals stay above the receiver's noise level. And reducing the amplification reduces the risk of intermodulation.

Finally, there's a 33 MΩ resistor in parallel. This is needed to give the opamp correct feedback for DC. But this resistor does reduce the amplification further at lower frequencies, so it should be chosen very large; Nordholt even used 1 GΩ (gigaohm!). On the other hand, it shouldn't be made too large. There's always a bit of DC current in an opamp input, and the only path for this DC current is via this resistor. If that current were e.g. 1 µA, there would be a 33 V voltage drop over the resistor, which far exceeds the opamp output's range. The OPA818 has a FET input and consequently a very low input current, specified as 700 pA maximum. That is one of the reasons this opamp is very suitable for this circuit.

Graph (c) shows the final frequency characteristic. Along the curve the dominant feedback components are indicated, and also the cut-off frequencies are marked. When the feedback is mostly capacitive (like above 13.5 MHz), the amplification is constant. But where a resistor dominates in the feedback (such as between 1.5 and 13.5 MHz), the amplification decreases with decreasing frequency.

Intermodulation

A major point of concern in wideband active antennas is intermodulation, as by definition, such an antenna deals with all signals from the entire spectrum.

Jochen Jirmann, DB1NV, already published several times [3] about active antennas using opamps, and tells in those publications that such an opamp circuit behaves totally different than a normal transistor circuit. If the signals become too large, the opamp output cannot keep up: either because it saturates at the supply voltage, or because its output voltage needs to change faster than the opamp's 'slew-rate' allows. Either case gives serious distortion with lots of intermodulation products. If one then lowers the signal levels, so the opamp can handle them, then this distortion quite suddenly disappears: the feedback then makes the opamp cancel its own distortion, and as a consequence, the circuit would be almost intermodulation-free.

Unfortunately, practice turns out to be different. The figure shows the level of the intermodulation products measured at the output of the opamp circuit shown above, when two 'desired' signals of 6 and 6.1 MHz are fed in and their amplitude is varied. The horizontal axis shows the level of the desired signals, measured at the output; and the vertical axis shows the level of the intermodulation products. The most important lines are the red and blue ones, measured at 0.1 and 5.9 MHz: this is where we would expect intermodulation products of second and third order.

All the way at the right, so at the highest signal levels, the red and blue line indeed show the behaviour discussed above: when the opamp is overloaded the intermodulation products are strong, and when we make the signals weaker, the intermodulation levels decrease rapidly. At output levels of +8 and +2 dBm, respectively, they even disappear almost entirely. The temptation is to stop measuring there, as the intermodulation is gone.

However, if we do continue with lower input levels, we see the intermodulation products return! Only when we attenuate the signals much further, to around -10 dBm, the intermoduation products get weaker again, albeit much more slowly now.

How is this possible? I found an explanation in [4]. The graph line labeled 'IM2' (second order intermodulation) represents the strength of the intermodulation product measured at 6.1 - 6 = 0.1 MHz. But on that same frequency, also a product of e.g. fourth order could appear, because 6.1 + 6.1 - 6 - 6.1 also equals 0.1 MHz. So, what we measure here is the sum of various intermodulation products of second, fourth and even higher orders. Usually those products of fourth and higher order are (much) weaker than those of second order, so we can pretend to measure only second-order intermodulation. But if the opamp is overloaded and produces much distortion, the higher-order products are no longer negligible. Furthermore, it may be (and apparently is) that such a fourth-order product is in opposite phase to the second-order product. Then there may be a signal level at which they are both equally strong but opposite, and thus cancel. That explains the sharp 'notch' in the red line. A similar consideration applies to the blue line, measured at 5.9 MHz: not only the expected third-order intermodulation (6 + 6 - 6.1 = 5.9 MHz) lands there, but also fifth and higher.

At much lower signal levels, around -10 dBm, the contributions of higher order do become negligible, so at 0.1 and 5.9 MHz we only measure second and third-order intermodulation. Those are expected to become weaker by 2 or 3 dB (respectively) per dB of attenuation of the desired signals. Slopes corresponding to 2 or 3 dB-per-dB are indicated with dotted lines.

At first glance, the cancellation of the second/third-order intermodulation by the fourth/fifth (and higher) order may seem nice: less intermodulation is better. But we must not forget that the fourth and higher order intermodulation also produces new intermodulation products at other frequencies. So the total amount of intermodulation does not decrease; it just ends up on different frequencies than where we usually measure.

In fact, this phenomenon does not just occur in opamps. DB1NV pointed me to an application note [5] of the RF power fet RD100HHF1, which contains a graph showing this same phenomenon.

For a fair comparison to other circuits that do not act so strangely, we could use the dotted lines, which extrapolate the pure second and third order intermodulation levels. That results in an IP2 of +68 dBm and an IP3 of +35 dBm. That is not particularly good; a good (but simple) two-transistor circuit also achieves this.

B.t.w., the graph only serves as illustration of the phenomenon. The exact numbers depend on many factors, such as the properties of the opamp being used, but also the amplification factor and the frequency.

Noise

Besides intermodulation, there is another worry, namely whether the amplifier's own noise does not swamp the desired signals. Let's chart the noise sources in the given circuit, and find their noise levels at 30 MHz, calculated back to an equivalent noise voltage at the whip.

First, there is the opamp's input voltage noise. According to the datasheet, it is 2.2 nV/√Hz. This can be seen as an extra voltage at the non-inverting input, which in this circuit will be amplified by a factor (1/6+1/2.5)/(1/6) = 3.4. The signal from the whip is amplifed by a factor 2.4, so effectively this noise corresponds to 2.2 × 3.4 / 2.4 = 3.1 nV/√Hz at the whip.

Next, there's the opamp's input current noise: this increases with frequency and is 1 pA/√Hz at 30 MHz. This current flows through 6 pF and 2.5 pF in parallel, which at 30 MHz together have about 600 ohms of reactance, resulting in a noise voltage of 0.6 nV/√Hz. This is substantially less than the opamp's voltage noise.

Then there's the noise from the resistors. The 33 MΩ resistor has a high noise voltage, but at high-frequencies most of this noise is short-circuited by the 2.5 pF capacitor, as I showed in [7]. The 4k7 resistor is more disturbing: it's unloaded noise voltage is 8.8 nV/√Hz, with 2.5 pF in parallel, 3.6 nV/√Hz remains; referred back to the input that is 1.5 nV/√Hz. Again, this is less than the opamp's own noise.

I must say I only calculated these noise levels, I didn't measure them. In the end, we only care whether the circuit's own noise is substantially lower than the noise received by the whip anyway (from the atmosphere, deep space, and nowadays also often 'man-made'). At a 'rural' location, ITU R.P372 says the ambient noise level is about 10 nV/√Hz at 30 MHz on a vertical antenna with an effective height of 1 m. That is already substantially more than the calculated noise of the circuit, while our location on a university campus undoubtedly is noisier than 'rural'.

Practical experience

Although a-priori it was not clear whether the new circuit would really be better than the old one, in practice it does turn out to be. In the upper part of the HF range, the receiver has clearly become more sensitive, thanks to the extra amplification at high frequencies; we need that due to the rather long antenna cable and the by itself somewhat insensitive SDR. And there clearly are fewer intermodulation products visible, particularly in the long-wave range, thanks to the intentionally reduced amplification at low frequencies, where there are many strong signals. The opamp circuit may not really be better on its own, but the extra flexibility to optimize the amplification at different frequencies has brought an improvement.

B.t.w., Nordholt's own circuit [1,2] is substantially more complicated, and hard to reproduce nowadays because some transistors are no longer in production. But its performance is better: its IP2 and IP3 are +80 and +53 dBm, and that's even measured at a higher frequency then where I tested my circuit.

References

[1] Ernst H. Nordholt, Durk van Willigen: A new approach to active antenna design. IEEE Transactions on antennas and propagation, 1980.
[2] E.H. Nordholt: Breedbandige actieve antennes voor 5 kHz tot 30 MHz. Elektronica, 7/1981.
[3] Jochen Jirmann, DB1NV: Rauscharme HF-Operationsverstärker, auch für Aktivantennen. UKW-Tagung Weinheim, 2023.
[4] A tutorial on applying op amps to RF applications. Texas Instruments, SNOA390B, 2013.
[5] RD100HHF1 PushPull Amplifier application note. Mitubishi RF semiconductors, 2005.
[6] Technische notities van PA3FWM, Electron 7/2020. Or online in English here.
[7] Technische notities van PA3FWM, Electron 10/2022. Online in English here.

---

Text and images (except the last schematic) on this page are copyright 2025, P.T. de Boer, web@pa3fwm.nl .
Republication is only allowed with my explicit permission.