Nothern Utah WebSDR Logo - A skep with a Yagi Northern Utah WebSDR
Receiving equipment


A downconverter for RTL-SDRs:

It is quite common for HF coverage via an RTL-SDR dongle to be achieved through the use of an upconverter  - a device that takes the HF spectrum and shifts it up by 100, 125 or even 200 MHz.  The reason for doing this is easy to understand:  Continuous coverage across the HF spectrum is afforded, which is very convenient - but one must still take care when doing this:
Another method of providing HF coverage is with the use of a frequency downconverter.  This is fundamentally different from the upconverter in that instead of shifting the entire HF range upwards by some amount, one takes the narrow slice of interest and converts it down to a lower frequency.  Doing this solves/minimizes several problems, such as:
If the RTL-SDR can receive HF in its "Direct" mode, anyway, why would you use a downcoverter?  As mentioned previously, the A/D sample rate of the RTL2832 chip is 28.8 MHz, which means that both 20 and 10 meter coverage via this mode has the problem of being too close to the Nyquist limits:  In the case of 20 meters, half of the sample rate (14.4 MHz) is just above the top of the band at 14.35 MHz and images cannot be easily filtered.  In the case of 10 meters the 28.8 MHz sample rate lands right in the middle of 10 meters which means that even if you could build an effective filter to suppress images, you'd only be able to effectively cover a small-ish portion of the band.  At the various bands in-between 20 meters (17, 15 and 12 meters) coverage is possible, but good filtering is still required - and the undersampling means that the sensitivity will be a bit worse than it would be at lower frequencies, presuming that the low-pass filtering in the dongle on the "Direct" branch weren't an issue.  In short, in "direct" mode the RTL-SDR dongle works best from around 1 MHz up to around 12 MHz:  Above this, it becomes increasingly difficult to construct anti-aliasing filters that are both simple and effective.

A converter has the obvious disadvantage that it isn't as convenient:  You probably can't just go out and buy a downconverter like this - and the design considerations require that one pick the frequencies of the local oscillator and down-converted frequency range a bit carefully with respect to the intended coverage.  For example, the local oscillator's harmonic(s) should not land in or very close to the input frequency (the one to be down-converted) as this will result in a very strong signal that could result in intermod products (e.g. birdies.)  The other issue - the down-converted frequency output - should be carefully chosen so that its harmonics are several MHz away from the input frequency coverage as this, too, would result in "birdies" or other undesired responses.

A downcoverter for 15 meters:

As a matter of convenience I chose a 10 MHz local oscillator frequency because inexpensive TCXO devices are readily available, and it adequately meets the design criteria:
A diagram of the as-built downconverter may be seen in Figure 1.

Figure 1:
Top:
 Inside the 15 meter downconverter for RTL-SDR devices, using the "Direct" branch input
Bottom:  Schematic diagram of the converter
Click on an image for a larger version.
Close up of one of the broadband RF amplifier sections
Schematic diagram of the 15 meter downconverter

Circuit Description:

As with the diagram depicted in the bottom frame of Figure 2, on the RF Distribution page (link), the first input stage uses a bandpass filter "borrowed" from the QRP Labs web site, from their "Band Pass" filter products (a link to that page is here).  In the assembly manual - which may be found on that web page - you will find a technical description of the filters (along with some representative band-pass plots) that provide enough information for you to build your own filters.  If you wish, you may buy these modules in kit form (and I can recommend that any of the kits sold by QRP Labs are worth getting!).  This filter does the job of both attenuating the receive image that would otherwise be present at the sum of the local oscillator and the 15 meter band (e.g. 31.00-31.45 MHz) as well as reducing the overall amount of energy impinging on the mixer from signals outside the 15 meter frequency range.

Following this bandpass filter is an amplifier based on the 2N5109 transistor - chosen in this application due to its superior performance over a standard 2N3904 transistor:  The latter will work, but both the gain and noise figure will be somewhat higher.  This amplifier overcomes the losses related to U1, a diode-ring mixer which has an intrinsic loss of about 7 dB.  One might wonder why a diode-ring mixer is used rather than something like the NE602:  The simple truth is that the '602 has far inferior signal-hanling capabilities and is easily overloaded compared to a diode mixer and its associated amplifier.

Following the diode-ring mixer is another bandpass filter - another one modified from the 30 meter QRP Labs design by reducing the inductance of the "larger" windings by removing a turn or two:  This filter will be a bit narrower than the input and further-reduce off-frequency signals - and it also eliminates the image response from the output.  Ideally, a "diplexer" would be included on the output of a mixer to assure that it was terminated at the image frequency, but this was omitted because of the relatively narrow passband filter on the input that reduces the number and breadth of signals and because this is a "non high-performance" application.

Providing the 10 MHz local oscillator is an inexpensive (approx. $2.50 in single quantities) TCXO.  This device provides justenough drive for the diode ring mixer and it will be stable to within a few Hertz at 10 MHz, assuring frequency stability that is around an order of magnitude better than what would be experienced if one were to do "upconverting" to >=100 MHz as is commonly done.  In our application - in an unheated, not air-conditioned building - the temperature will very wildly - from below freezing to as high as 140F (approx. 60C) so this degree of stability is important.  The down-side of this devices is that it is very tiny - about 2.5x3.5mm - so it is "super glued" to the board bottom-up and tiny (30 AWG) wires are soldered to the surrounding connections.  This device requires a 3.3 volt supply so a standard 5 volt regulator was used with a normal "dim" red LED in series to provide a 1.6-1.8 volt drop.

Following the output bandpass filter is another amplifier - this time, using the generic 2N3904 - which boosts the signal a bit more to overcome the intrinsic (relative) insensitivity of the RTL-SDR dongle:  At this lower frequency, the gain will be higher and the noise figure - set by previous stages - is less important, allowing the use of a general-purpose device.  Hanging on the output of this amplifier is a series L/C network tuned to the 10 MHz local oscillator frequency to reduce its energy by 15-20dB:  This signal is otherwise a bit strong and as we know, it's a good idea to keep the extraneous signals entering an RTL-SDR dongle to a minimum!

The final stage is a simple potentiometer-type attenuator - nothing more complicated being required as constant impedance is not very important for the input of an RTL dongle, particularly if a short cable is used.  As we already know, when dealing with RTL-SDRs on HF, it's best to start out with a bit of extra signal - and then tweak the levels downwards as necessary to find the "sweet spot" between being able to hear weak signals and prevent overload from strong signals.  In the case of 15 meters, at this time of the sunspot cycle when band openings are a bit rare and signals are on the weak side it's probably best to adjust the overall system gain to just be able to hear the background ionospheric noise - and no more!

Testing:

When tested using the "SDR Sharp" program, the combined sensitivity (with R10 set for maximum signal) of the downconverter and RTL-SDR dongle running in "direct" mode was on the order of -130dBm in an SSB bandwidth - less than 0.1 microvolts, and more than enough sensitivity to "hear" anything that would fall on any decent HF antenna in an RF-quiet environment when the band was dead.  When installed on site, R10 was adjusted so that, with a "dead", quiet band the background ionospheric noise was be just registered on the S-meter and A/D converter by several dB to maximize its signal handling capability.

A downconverter for 10 meters:

A similar downconverter was constructed to provide full 10 meter coverage, also using an RTL-SDR dongle.  This converter is nearly identical to the 15 meter converter, except as follows:
The RTL-SDR dongle is configured to run at 2.048 MSPS with the center frequency being 28.676 MHz, providing coverage from 27.652-29.700 MHz.  Because of the RTL-SDR's internal 28.8 MHz clock, there is a moderately strong carrier at this frequency that cannot be avoided.

Addendum:

It was later found to be necessary to add yet another RF amplifier after the one on the downconverter's output to bring the signal level up enough to make the RTL-SDR dongle capable of hearing the background (thermal) noise on the 10 and 15 meter amateur bands on the respective downconverters.  This amplifier was placed after the output level adjustment potentiometer (R10 in Figure 1) to minimize the amount of signal that this amplifier could see.  Because of the comparative deafness of the RTL-SDR in "direct" mode, the apparent increase in noise figure by placing the gain control at that point is irrelevant.

Because of the limited dynamic range of these dongles (no matter whether you use an upconverter, downconverter or "direct") having such a level control on the input is absolutely necessary

Pages about other receive gear at the Northern Utah WebSDR:


Additional information:
 Back to the Northern Utah WebSDR