I bought a Mini-whip active antenna (fig. 1) from a China online seller because of its low cost, ~USD2. Tracing the PCB layout, it appears to be a clone of PA0NHC ver. 2013 [1], although with some minor changes (fig. 2). Subsequent googling reveals that the manufacturer is SZ-plaza [2].
Fig. 1: Photo & dimensions of the mini-whip & its biastee
Fig. 2: PA0NHC version 2013 (top) is likely the inspiration for the SZ-plaza clone (bottom)
Unfortunately, the item was plagued with two catastrophic issues. Firstly, upon powering up, the tantalum caps C4 & C5 (22 uF) heated up rapidly (red circled in fig. 3 photo and circuit). Additionally, the PCB drew ~100 mA, when the expected current is ~40mA. Prying off the two caps, the current returned to normal. Tantalum caps are notorious for this failure mode - heating due to internal leakage. This failure is unsurprising because the caps are rated 16V and therefore, are cavalierly used in a 9V-15V circuit, i.e. with almost no de-rating. Whereas, tantalum cap guidelines call for a voltage rating of >2x the max. operating voltage [3].
Fig. 3: Photo of Mini-whip (top) and circuit diagram obtained by tracing the PCB layout (bottom). The hot and leaky tantalum capacitors are red circled in both photo and circuit
The second surprise was the abnormally low gain over HF (fig. 4).
Fig. 4: Abnormally low gain which was subsequently traced to Q1's Drain and Source being mistakenly swapped in the PCB layout!
The Mini-whip's gain was measured with 50R generator impedance; i.e. directly connected to a nanoVna (fig. 5), although this is incorrect because an antenna simulation network (dummy antenna) is required to simulate the very short whip. However, this is a quick and dirty way to ascertain whether the Mini-whip is working or not. The cause of the low gain was ascertained after checking the Q1 (BF998) datasheet, specifically its lead identification. The PCB layout has mixed up Q1's Drain and Source! This was further confirmed by buyers' comments on the product advertisement (fig. 6).
Fig. 5: gain measurement setup
Fig. 6: buyers commenting on Q1's reversed connections on the product advertisement page. The leaky cap is also mentioned
The remedy was to flip Q1 and then followed by bending down the upward pointing legs. Following the repair, the gain increased by >25 dB over the original (before) condition (fig. 7).
Fig. 7: The gain increases by >25 dB after correcting the erroneously swapped Source and Drain of Q1
Additionally, the voltages around Q1 also returned to correct values, whereas previously, they indicated that the lower transistor was biased to pinch-off (fig. 8).
Fig. 8: Voltages around Q1 before and after flipping the device indicate a return to correct operation
After further testing, I discovered another deadly fault - the output reflection coefficient |S22| > 1 over part of the HF range! This fault is depicted by the area enclosed by the red dashed box in the Smith chart (fig. 9)! Its position outside the chart circle means any signal reflecting from the output port will be stronger than the original - a sure recipe for oscillation!
fig. 9: Part of the output reflection coefficient |S22| > 1. This is represented by the trace outside the Smith chart (red dashed box) - an ominous warning of instability!
As it is impossible to determine the unstable region's frequency range by reading the Smith chart, the required information has to be obtained from a rectangular plot (fig. 10). The peak above the dotted red line is the unstable region which spans 20-24 MHz.
fig. 10: Whereas the unstable frequency range cannot be read off the Smith chart, a Cartesian plot can show |S22| > 1 occurs over 20-24 MHz (peak above the red dashed line)
What is the cause of this instability? Q2 is a common collector / emitter follower (fig. 11) - a configuration that is notorious for instability [4]. To counter the inherent instability, a resistor (R6) is customarily added to the input. In the PA0NHC original, R6 was a 100R resistor (red circled in fig. 11), but the SZplaza clone reduced it to 33R, presumably to increase gain.
fig. 11: R6 (red circled) at the Q2 input is intended to prevent oscillation but its value has been detrimentally reduced from the original design's 100R to 33R in the physical sample
Before committing to repair, simulation was used to investigate R6's influence on stability. Simulation predicts that increasing R6 value can beneficially move the 'loop' inside the Smith chart (fig. 12).
fig. 12: Simulation predicts that increasing R6 value can beneficially shift the 'loop' inside the Smith chart to stabilize the mini-whip. When R6 = 50 Ω or 130 Ω (blue & pink traces), the resultant S22 is totally enclosed by the chart's unity circle
Finally, the hypotheses was experimentally validated by changing R6 to 120R on the physical board. Following this repair, measurement shows that the S22 is fully enclosed inside the Smith chart (blue trace in fig. 13).
fig. 13: After increasing R6 to 120 Ω, the measured S22 is now totally enclosed inside the Smith chart; i.e. stable operation (blue trace). The original S22 (black trace) is shown for comparison
Nevertheless, the new S22 is still blemished - it is far away from the Smith chart centre, hence indicating a poor match to the 50Ω coax connecting to the receiver. At the risk of sounding sacrilegious, perhaps the entire Mini-whip architecture of Common-Drain followed by Common-Collector is flawed from the onset because it reduces the output Z too much!
In a nutshell, the mini-whip may be a simple circuit, but it still requires careful investigation to tease out all the faults.
The task of repairing the mini-whip was aided greatly by circuit simulation (fig. 14). The modeling of the second stage Q2 was compounded by the unknown part number and the manufacturer did not reply to our email. So, we arbitrarily elected to use BCX56 because it was used in ON1BES's mini-whip fork. Cstray represents the parasitic capacitance of the input protection diodes D1-2 (1N4148). To replicate the whip antenna, a dummy aerial X2 is inserted between source and the amplifier input. The dummy aerial's circuit values are copied from PA0NHC. As the values of capacitors C1, 4, 7 & 11 are unknown, they are arbitrarily assumed to be 100 nF in the simulation.
Fig. 14: SZ-plaz mini-whip circuit model with a PA0NHC dummy aerial X2 inserted between source and the amplifier input.
The experimental gain exceeds -16 dB over HF (fig. 15). The gain peaked at 22 MHz due to the series resonance of the dummy antenna's 6.8 pF and L2. Total gain variation is 23 dB - this is a trade-off from using L2 to boost the gain. The modeled gain (blue dashes) exhibits the same shape as the experimental (solid black) and has a maximum error of 5 dB.
Fig. 15: The experimental gain exceeds -16 dB over HF. The gain peaked at 22 MHz due to the series resonance of the dummy antenna's 6.8 pF and L2. The modelled gain (blue dashes) exhibits the same shape as the experimental (solid black) and has a maximum error of 5 dB.
Although mini-whips promise high input impedance, in reality they only meet that condition at the lower frequency range. As the frequency increases, the input impedance gradually drops until it is almost 50R at the input series resonance frequency (fig. 16). The varying Zin is a trade-off from adding the input inductor L2 to increase the gain at ~ 22 MHz. The original PA0RDT without L2, has lower gain, but its Zin is constantly high,. The modeled S11 (blue dashes) has the same shape as the experimental (solid black).
Fig. 16: Although mini-whips are predicated on high input impedance, in reality, they are such, only at the lower frequency end. As the frequency increases, the input impedance gradually drops until ~50R (chart centre) at the input series resonance frequency. The modelled S11 (blue dashes) has similar shape to the experimental (solid black).
After increasing R6, the measured S22 is fully enclosed inside the Smith chart (fig. 17). Quantitatively, the modeled S22 (blue dashes) is poorly matched to the experimental (solid black), but has the same general shape.
Fig. 17: Measured S22 (solid black) is fully enclosed inside the Smith chart. Quantitatively, the modelled S22 (blue dashes) is poorly matched to the experimental, but has the same general shape.
Fig. 18: Modeled OIP3 exceeds 20 dBm
Fig. 19: Modeled P1dB exceeds 8.5 dBm
[1] "Pa0nhc miniwhip active receiving antenna". http://yo3kxl.netxpert.ro/diverse/antene/miniWhip/Pa0nhc%20miniwhip%20version/20121025miniwhip.htm
[2] SZ-plaza advertisement, http://www.sz-plaza.com/pd.jsp?id=51
[3] J. Gill, "Surge in solid tantalum capacitors", Avx, https://www.technonet.co.kr/data/bbsData/tantal1.pdf
[4] M. Chessman & N. Sokal, "Prevent emitter-follower oscillation", Electronic Design, Jun. 1976. www.hifisystemcomponents.com/downloads/articles/Prevent-Emitter-Follower-Oscillation.pdf