The rtl-sdr can be told to run in a mode called “direct sampling mode”, which with a small hardware mod allows the dongle to tune to the HF frequencies where ham radio and many other interesting signals are found. This means that no upconverter circuit is required.
However, the difficulty with direct sampling is that a hardware modification to the dongle is required. Also, the performance can not be expected to be as good as an upconverter.
The direct sampling mode was originally discovered and discussed in this Google groups thread.
YouTube user Superphish was able to receive HF AM broadcast radio, and a decode a HF weather fax signal at 5.8MHz using the direct sampling mod.
YouTube user 907h9879070g9790 was able to use the direct sampling mod to receive and decode a RTTY signal at 147KHz.
The hardware mod is quite simple for someone with good soldering skills. It involves opening the dongle casing, and soldering a “random wire” or “long wire antenna” to pin one or pin two of the RTL2832U chip. This allows RF signals to directly enter the RTL2832, bypassing the tuner chip. Even a short wire a few meters long will be sufficient for picking up broadcast AM stations. For ham signals, a longer wire is probably required. Care must be taken with the mod as this bypasses all electrostatic discharge (ESD) protection.
Pin one and two are located below the circle indentation on the RTL2832U chip. It might be easier to solder the wire to one of the two capacitors which are connected directly to pin one and two.
Some more useful images of pin layouts can be found at superkuh’s rtl-sdr blog post.
Activating direct sampling mode requires a software driver adjustment. Luckily, SDRSharp has this adjustment as a built in option. In SDRSharp, go to configure, and change the sampling mode to “Direct sampling (I branch)”, or ”Direct sampling (Q branch)”.
The I branch corresponds to pin one/pin two, and the Q branch corresponds to pin four/pin five which can also be used. Pin three is not connected.
If a wire antenna is connected to one of these pins, and the correct direct sampling branch is selected, you will be able to receive signals between 0-14.4 MHz. Frequencies between 14.4 and 28.8 MHz can also be received with a band pass filter.
Since the RTL2832U uses two pins to create a differential input, a balun can be used to connect both pins one and two to an antenna. Essentially this mod involves finding or winding your own balun, and connecting the balun inputs to both pin one and two, and the other end of the balun to your antenna. Here is a pdf file by mikikg showing schematics for this modification.
Here is a blog post by Dekar who used a transformer from an old ADSL modem as the balun.
Another site in Italian with a lot of useful images is here. He used a 4:1 self wound toroid as the balun and got what looks to be good results.
Google translate can be used to translate the pages, but the images are the most important things here.
Most experimenters of this mod find that FM interference is a problem, and thus low pass filters are necessary. The Italian mod page shows schematics and images of a 60 MHz low pass filter combined with the direct sampling mod. His results show significant improvements in out of band signal rejection with the low pass filter.
Some users also report that adding in a low noise amplifier (LNA) can help improve reception. A good Reddit thread discussing improvements to this mod can be found here.
As the RTL2832U input impedance is unknown, it is difficult to match impedance with the antenna. In Dekars mod, he assumed a 1:2 impedance conversion would help, but it actually made the signal worse.
Direct Sampling with No Hardware Mod Discovery
There is also a discovery for the E4000 tuner found by Reddit user Anonofish. Anonofish found that when tuning to frequencies between 3686.6MHz – 3730MHz, the rtl-sdr dongle acted as if it was in direct sampling mode, passing the radio signal directly past the tuner and into the RTL2832U chip. This allows the lower end of the HF spectrum to be received without any hardware modifications.