HIGH PASS FILTER FOR OPTIMUM HF RECEPTION

[glovisol]

HIGH PASS FILTER FOR OPTIMUM HF RECEPTION

Presence of powerful signals in the MW band, due to public broadcasting activity, as described here for LF:

https://www.sdrplay.com/community/viewt ... f=5&t=3943

can affect HF bands as well.

Continued in next post for text amendment purposes

[glovisol]

HIGH PASS FILTER FOR OPTIMUM HF RECEPTION

Front end overload can produce spurious signals, desensitization and cross modulation at HF frequencies above 2 MHz. A sharp High Pass filter placed between antenna and RSP class receiver input, can protect the front end circuitry by effectively reducing or eliminating the unwanted Medium Wave band signals, extremely strong at night time.

High Pass filters are designed from tabulated Low Pass prototypes through a transformation process set up on a single Excel spreadsheet. Since the HF band above 2 MHz covers several octaves, a Cauer type prototype is not suitable because these filter types, having attenuation poles near the passband, can produce parasitic and unwanted signal transmission at higher frequencies, caused by imperfections in the filter components. On the contrary, a Low Pass Cauer type filter, as used for LF reception in the previous thread, is suitable because of the small size of the required passband below the bandstop, where component imperfections are not significant.

The main requirement for the High Pass filter is a passband from 2 MHz up and a stopband from 2 MHz down, affording at least >20 dB attenuation as near cutoff as 1.6 MHz. For instance a Chebyshev prototype: TH 0.5 dB/N=9 provides 0 dB attenuation @ 2MHz with 0.5 db ripple in the passband and 39 dB attenuation @ 1.6 MHz. Unfortunately the return loss of this filter in the passband is too low, in the order of 10 dB, which is not suitable (Figures 1 & 2).

The chosen Chebyshev prototype is a TH 0.01 dB/N=9, which has zero dB attenuation in the full passband (virtually no ripple) and provides more than 21 dB attenuation @ 1.6 MHz. Attenuation then monotonically increases at lower frequencies and exceeds 70 dB @ 1 MHz, thereby wiping off the entire LF & MW spectrum (Figure 3) while the return loss is 28 dB or better over the full passband. In Figures 2 & 3 the graph frequency limit has been set at 9 MHz, just for convenience: the filters' passband really extends, as will be shown, well over the targeted 30 MHz.

In this thread complete data will be presented (a) for a 500 Ω in / 500 Ω out filter plus 500 Ω to 1 KΩ isolation transformer for RSP Spectrum Processors equipped with balanced Hi Z input and (b) for a 50 Ω in / 50 Ω out filter plus 50 Ω to 50 Ω isolation transformer accommodating 50 Ω inputs.

[glovisol]

FILTER SCHEMATIC, COMPONENT VALUES & INSERTION LOSS PREDICTIONS

The chosen Chebyshev 0.01 dB ripple, N=9 lowpass prototype is the series first, e.g. the first component is a series inductor that becomes a series capacitor as a consequence of the lowpass to highpass transformation. The resulting highpass prototype has 5 capacitors and 4 coils. If we used the shunt first prototype, we would end up with 4 capacitors and 5 coils. Since in general coil Q is significantly lower than capacitor Q (e.g. capacitors at low MHz frequencies are less lossy than inductors) the series configuration is to be preferred for minimizing insertion loss.

The filter schematic with all de-normalised component values for 50 Ω in/out @ 2 MHz cutoff frequency, is shown in Figure 4. The performance shown in previous figures 2 & 3 is based on the hypothesis of “perfect” e.g. lossless components. In the real world capacitors have a parallel loss resistance and inductors have a series loss resistance and the quality factor Q of these components is the ratio between component reactance and component loss resistance.

It is essential to predict insertion loss at the cutoff frequency. From experience good capacitors will exhibit a minimum Q = 800, while coils will have an average Q = 150. The Excel spreadsheet in Figure 5, calculates the component values from the original prototype and the filter insertion loss for the Q values mentioned above. From Figure 5 the predicted insertion loss at 2 MHz is 0.4 dB.

For instance with very high Q capacitors and coil Q = 200, Figure 6 shows a predicted filter insertion loss @ 2 MHz of only 0.27 dB. Since capacitors introduce lower losses than inductors, a filter with fewer inductors will have lower insertion loss, all other conditions being equal.

In the next post we shall examine inductor design and will optimize inductor Q factor by manipulating coil parameters.

[glovisol]

INDUCTOR DESIGN AND SELECTION

For this frequency range and impedance levels air core inductors can be used. The "COIL 32" software, readily available, can be used to optimise inductor parameter selection. In general larger diameter coils, wound with larger diameter wire and with a pitch that keeps turns spaced, exhibit a higher Q than smaller diameter close wound coils with thinner wire. Inductor parameters related to physical dimensions, calculated with COIL 32, are shown in Figure 7, where it is shown that close wound coils are not suitable, having Q's below 100, or barely above.

A good choice for both the 2.79 uH and the 2.32 uH inductors is coil diameter of 30 mm, with 1mm dia. enamelled wire wound with 2 mm pitch: predicted inductor Q is above 210. With the higher Q the predicted filter insertion loss drops to 0.299 dB, as shown in the fresh calculation of Figure 8.

In the next post we shall examine design and data for the 500 Ohm filter version.

[glovisol]

500 OHM FILTER DESIGN & DATA

Following the same procedures, schematic, component data and insertion loss for the High Pass filter @ 500 Ohm in/out impedance are shown in Figures 9, 10 & 11 below. Using 30 mm diameter coils the Q factor is over 270, hence lower losses than those of the 50 Ohm filter can be predicted.

Filter topologies in relation to the RSP Processor used (RSP1/RSP1A/RSP2/RSP2PRO/RSPduo) will be examined next.

[glovisol]

HIGH PASS FILTER TOPOLOGIES

The proposed High Pass filter has been designed for two different impedance levels: 50 Ohm and 500 Ohm. The lower impedance version not only accommodates RSP Processors with only one coaxial 50 Ohm input, like the RSP1A, but simplifies prototype filter testing as well.

Three different topologies are presented, but infinite combinations are possible (depending on antenna type and Processor input arrangements) by altering the impedance ratios of the transformers. In any case, according to my experience, the use of a transformer at the input or the output of the filter is always advisable, because it keeps the Spectrum Processor balanced with respect to ground (normally beneficial with local noise problems) and affords input protection against transients that could damage the RSP input circuitry. Furthermore one can choose to place the ground connection, if required, where it suits him most.

Topology #1 accepts a standard coaxial unbalanced 50 Ohm input and the ground connection is connected to terminal B. This topology, so to speak, provides a balanced input to RSP Processors which are not equipped with one.

Topology #2 extends the fully balanced & symmetrical configuration to the filter itself. Inductors are the same as for #1, but capacitance values must be doubled. T1 is a 1:1 ratio toroid transformer to be described later on along with T2 for Topology 3.

Topology #3 accepts a long wire antenna, such as a Beverage, or a wide loop antenna (as the triangular loop VK7JJ style) with a nominal output impedance at, or near to, 500 Ohm and exploits the excellent balanced, high impedance, 1 KOhm input facility of the RSP2 and/or the RSPduo.

The table below summarises the proposed topologies.

TOPOLOGY-------RSP1------------------RSP1A------------------RSP2/RSP2PRO----------------------RSPduo

# 1..................50/50 Umbal...........50/50 Unbal.............50/50 Unbal...................50/50 Unbal.

# 2..................50/50 Bal...............50/50 Bal.................50/50 Bal......................50/50 Bal.

# 3..........................................................................500/1000 Bal................. 50/1000 Bal.

Figure 12 shows the three different schematics.

[glovisol]

Back from Christmas holidays, work in progress. I am posting picture of the 4 filter coils. I have used the no pitch version after all, it is easy to make them using Kodak film roll containers (30 mm dia. exactly) and the no pitch means Q's of only 150, but I can assemble & test the filter prototype in a short time. With these Kodak containers I can stick a screw in to do fine tuning of the inductance value.....will report complete filter soon.

Cheers,

glovisol

[glovisol]

The filter is beginning to work, but still some way to go. In order to simplify testing I dropped from N=9 to N=7 and this is the prototype result with the uploaded screen below: cutoff frequency is still too high, symptom that coil inductances are lower than calculated. I did changes to the design and obtained a big advantage in topology, but I shall report further on.

[glovisol]

Here below is illustrated the effect of substituting two of the four coils used with higher Q counterparts, while also centering the cutoff frequency. Improvement in the sharpness of attenuation at cutoff frequency is clearly visible.

[sdrom33]

Hi govisol can you explain the measurement setup for filter? What spectrum scope? When you giving build data?