There is a follow-up articles to this one that describes a circuit that properly matches the source/load over a wide frequency range - See the article: "Revisiting the limited attenuation high-pass filter for the KiwiSDR" link and "Revisiting the limited attenuation High Pass Filter - again".
Since the original posting of this blog entry I was made aware of an 1977 article on this very topic - you can read it HERE. (The article in question begins on page 3 of the PDF.)
* * *
One of the issues common with using a broad-band, direct-sampling SDR (software-defined radio) like the KiwiSDR is that of overload by strong, low-frequency signals, such as those on the AM (mediumwave) broadcast band - but there's another problem that should be considered as well: The high generally-high signal levels at lower HF frequencies. If one looks at an spectrum analyzer connected to a broad-band receive antenna during the evening, one will immediately note that the lower the frequency, the higher the signals seem - particularly the background noise.
This becomes problematic if one is using an antenna with a relatively flat gain across the entire HF spectrum - and one wishes to make the receiver usable at both the top and bottom ends of this range. As an example, I have a KiwiSDR connected to an antenna that is rated to cover from 3 to 30 MHz with roughly constant gain, but I noted that at the top end of the frequency range, around the 10 meter amateur band, the overall system gain was not quite sufficient to "hear" the background iononspheric noise.
The obvious solution to this gain deficit is to install an RF amplifier - which I did - but this had the effect of increasing the already-strong signals below 5-10 MHz even more, resulting in occasional "OV" indications on the KiwiSDR's S-meter signalling to me that the RF levels were high enough to "clip" the A/D converter. While this wasn't too much of a problem during normal conditions, if the lower HF band were particularly noisy - as often occurs in the summer with thunderstorms on the same continent - reception across the entire HF spectrum was compromised when the loud static crashes would occasionally saturate the A/D converter.
It occurred to me that while I had about the right amount of system gain on 10 meters, I had far more than I needed at lower frequencies and could throw some of it away, so I set about designing a filter that would reduce signals at the low end of the HF spectrum, but have minimal effect at the upper end.
A "limited" high-pass filter:
Note: The filter detailed below is NOT recommended as it does not match well to 50 ohms across Ll frequencies - see the follow-up article HERE for one that provides a good match/return loss.
The obvious solution to this would be the addition of a high-pass filter - but there's a problem: Even a minimal high-pass filter would have increasingly-higher attenuation at lower and lower frequencies - potentially in the many 10s of dB - but we don't really want to get rid of the lowest frequencies. What we need is a filter that will "knock down" signals by a significant amount - but not so much that they become inaudible.
In analyzing the signal levels, I determined that the goal of the design would be to leave signal levels above about 10 MHz unaffected, but reduce the signals below 8 MHz or so by 10-15dB. This amount of attenuation (about 2 "S" units) would significantly reduce the amount of RF energy entering the A/D converter at the lower end of the frequency range (about 2 "bits" worth) but analysis of the noise floor and signal levels at these lower frequencies indicated that I would still be able to hear the noise floor.
The diagram of this filter is shown below:
No real attempt was made to make this filter's input and output impedances "flat" across the entire HF spectrum - and to be sure, below about 14 MHz its input impedance a bit high, but this will have little practical effect on its operation - and we really don't need to be too precise, anyway.
As tested on a spectrum analyzer, the insertion loss is 12-13dB from DC to about 4 MHz at which point it gradually drops to about 2dB at 11 MHz and then dropping to less than 1dB by 30 MHz. When doing an "A/B" comparison with and without the filter on the KiwiSDR, the waterfall above 10 MHz looked unchanged, but the signals below about 7 MHz were much less "bright" - and most importantly, the occasional "OV" indications on the S-meter pretty much stopped appearing altogether.
In my opinion, the RF input on the KiwiSDR is slightly deaf, requiring a bit of gain (say, 6-10dB) to be able to reliably hear the background ionospheric noise on the higher HF bands - particularly when they are closed - a problem compounded by normal amplitude roll-off as one nears the Nyquist frequency. To this end, the KiwiSDR at this location is preceded by a low-noise, high dynamic range RF amplifier that is flat from a few 10s of kHz to well over 30 MHz.Update:
After this article was originally written it was determined by several testers using different KiwiSDRs that the absolute sensitivity of a KiwiSDR is on the order of -155dBm/Hz for 0dB S/N at 28 MHz. This sensitivity level is about 6-8 dB below the expected noise at a "quiet" site using a unity gain antenna on the 10 meter band.
In the real world, it is likely that 10-12 dB of overall signal amplification should preceded a KiwiSDR to allow it to be sensitive enough to hear the noise on a "quiet" 10 meter band and the weakest signals. If amplification is used, it should be placed as close to the antenna as possible in the signal path, but after a filter such as that described on this page (the filter will reduce the probability of overload by strong signals below 10 MHz and its loss at 10 meters is low and will have minimal effect), and before any splitter if you plan to feed more than one receiver from that signal path. When the overall amount of amplification is calculated, be sure to include the loss of a splitter is taken into account. For example, a 4-way splitter will incur about 7dB of loss, so if you wish the KiwiSDR to "see" 12dB of additional signal at 10 meters you will need an amplifier with a gain around 20dB.
The components for construction of this filter aren't critical: The capacitors are high-stability NP0 (a.k.a. C0G) ceramic types while L1-L3 are wound using 30 AWG enameled wire with L1 and L2 having 15 turns and L3 having 12 turns on T37-2 toroidal cores, respectively. L4 is a an inexpensive molded inductor and its value can be anything from 2.2 to 3.3 uH, or one could make it by winding 25 turns on the same type of T37-2 toroidal cores as used for L1-L3.
A look at the Kiwi's waterfall with the filter:
Figure 3, below, shows this filter in place on the KiwiSDR at the Northern Utah WebSDR site:
If I build another of these filters I'll push the "knee" up 1-2 MHz higher, starting the roll-off of signals below 11-12 MHz, instead.
This filter seems to be very effective in reducing the total signal power from lower HF frequencies while having minimal effect at higher frequencies. Because the signal+noise levels from a broadband antenna are much higher at the lower end of the spectrum, it is possible to reduce these signals by 2 "S" units or so without dropping the background noise - or the signals themselves - below the noise floor of the receiver.
For information on a filter system that is specifically designed to attenuate AM (MW) broadcast band signals, see the article "Managing HF signal dynamics on the RTL-SDR (and KiwiSDR) receivers", also on this blog.
There is a follow-up article to this one - "Revisiting the limited attenuation high-pass filter for the KiwiSDR" link where a variation of this filter is presented that passes the AM broadcast band and frequencies below it and is recommended for those installation where you wish to receive longwave signals.
This page stolen from ka7oei.blogspot.com