For accurate frequency tuning even amongst large temperature in an SDR, a Temperature Compensated Crystal Oscillator (TCXO) should be used as the main oscillator. Standard RTL-SDR dongles used a frequency of 28.8 MHz and do not come with a TCXO, but for some time now we have been selling our own branded dongles that come with a TCXO built in (out of stock at the moment sorry – back in the first half of April!). If you have an older or other dongle that does not have a TCXO it can be an interesting exercise to hack one in yourself. The biggest problem though, is that 28.8MHz TCXO oscillators are not commonly found for sale in low quantities.
Over on YouTube user devttys0 (Craig) has uploaded a video that thoroughly explains the theory behind creating a home brew 28.8 MHz TCXO out of a standard non-temperature controlled 19.2 MHz oscillator. The build involves halving the frequency, and then filtering and using the third harmonic as the clock signal (19.2/2 * 3 = 28.8 MHz), as well as creating the temperature compensation circuitry.
We have just released a new and improved RTL-SDR unit in our store, which we are currently pricing at $19.95 USD, or $24.95 USD including 2x telescopic antennas. The unit comes with the following improvements:
1 PPM temperature compensated oscillator (TCXO) – Accurate tuning and almost zero temperature drift (2 PPM initial offset, 1 PPM temperature drift)
SMA female antenna port – Most dongles use the less common MCX or PAL antenna ports. Ours use SMA which is much more common so more adapters and antennas are available for it. It is also more durable and has lower insertion losses.
R820T2 tuner – More sensitive/lower noise floor than the older R820T tuner. 100% compatible with software for the older R820T.
Improved component tolerances – Allows the RTL-SDR to work more optimally over all frequencies.
Experimental: 4.5V USB powered bias tee – Can be enabled by soldering two pads on the PCB together. This allows the RTL-SDR to power LNA’s (like the LNA4ALL and HABAMP) and active antennas through the coax cable.
Experimental: Break out pads for direct sampling – Allows easier soldering to pins 4 & 5 on the RTL2832U for enabling the direct sampling mod.
For US customers we highly recommend that you buy from our Amazon store as if you spend over $35 you will receive free shipping from a local Amazon warehouse. This usually takes less than 1 week for delivery. Prime subscribers can also get free 2 day shipping if bought on Amazon. If you like you can also use our international cart to buy from our Chinese warehouse with free shipping.
International customers can get free shipping from our warehouse in China. We will always try to use the fastest tracked air mail shipping method available to us, which will be ePacket, EMS air mail or similar if possible. This should get the parcel at your door within 2 weeks, but please note that this time is heavily dependant on the customs and postal agencies within the destination country which we have no control over. Countries such as Italy, Canada, Brazil, Russia and middle eastern countries are known to have extremely slow customs agencies. If you prefer you can also pay more for express shipping and we will use DHL, UPS, FEDEX or EMS Courier. Just use our cart to select the shipping method you prefer. We also kindly remind customers that with international shipping you are responsible for any customs duties or taxes incurred by the shipment.
Shipping status meanings: ‘In process’ means that your order information has been sent to the warehouse and the parcel is being packed. A tracking number will follow usually by the next business day.
Please note that tracking updates may take a few days to show up.
We will provide 6 months warranty on manufacturing defects. Please note that if you try the direct sampling or bias tee mods then any warranty will be voided, so please ensure your dongle is working before trying these.
If you suspect a manufacturing fault please email us at [email protected] and include your order number and name. Please include details of the fault and a picture of the fault if it is physical damage. If the unit is faulty we will issue either a refund or send a new unit out depending on your preference.
Temperature Compensated Oscillator (TCXO)
The 28.8 MHz oscillator used in most RTL-SDRs is passive and not frequency accurate. This means that when you tune to a known frequency, it will likely be offset by a few kHz. Usually the PPM offset on a normal RTL-SDR is in the range of 30 – 150 PPM. Furthermore, as the dongle warms up, the frequency will drift up to ~20+ PPM until the temperature stabilizes.
The 1 PPM Temperature Compensated Oscillator (TCXO) in our units provides accurate tuning with an initial offset of 2 PPM and a 1 PPM temperature drift over time. This means that a known signal will appear where it should on the frequency spectrum and will not significantly drift in frequency as the dongle warms up.
SMA F Antenna Port
On standard RTL-SDR’s the antenna port is either a MCX or PAL connector. MCX connectors are relatively uncommon and are susceptible to connector strain when using an adapter. PAL connectors are common with some TV connections, but no decent radio or antenna will use PAL due to its high insertion losses above ~100 MHz.
We’ve made these RTL-SDR dongles with SMA female antenna connectors. SMA is a very common connector in the radio field and provides a sturdy and secure connection. In addition SMA antenna adapters are much easier to find and insertion losses are lower.
We know some people prefer the F-type connector used in the previously sold ThumbNet dongles, but from our previous polling we believe the majority (~80%) of users prefer SMA. We may bring out F-type RTL-SDR’s again in the future if there is demand.
Note: Remember to not get confused between RP-SMA and SMA! RP-SMA or “reverse polarity SMA” is used for WiFi equipment only. In the normal radio world, most devices use standard SMA. RP-SMA is reversed, it has the male pin on the female connector, and the female hole on the male end. To be clear: This device is a radio device so it uses normal SMA connectors.
As discussed when we brought out our previous generation, the R820T2 tuner has slightly better sensitivity than the R820T and also works better at frequencies around 1.5 GHz. It also works better with the experimental HF drivers.
Improved Component Tolerances
We have these units manufactured with tighter tolerances on all passive components.
In our $24.95 USD package we provide two telescopic antennas. The smaller one goes from 6 cm to 20 cm, and the larger one goes from 20 cm to 1.5 m. The antenna base is also larger with a 4.5 cm diameter, when compared to the smaller bases shipped with most models. This provides more stable operation when using the larger antenna.
With antennas, usually the larger the antenna is the lower the frequency it can receive. These two antennas allow you to tune to almost the entire range of the RTL-SDR. Of course the antenna should be placed outdoors and up as high as possible to get the best performance. Placing the magnetic mount on a metal surface can also help complete the antenna as a quarter wave ground plane.
When fully collapsed the small antenna works decently at 1090 MHz for ADS-B frequencies.
Experimental 4.5V Bias Tee
A bias tee allows you to power external RF devices such as Low Noise Amplifiers (LNA’s) and active antennas through the coax cable. Since LNA’s should be placed right after the antenna, it can be sometimes hard to get power to them if a bias tee isn’t used.
We have included a simple (experimental) bias tee option in our latest units, inspired by mods madeby other experimenters. The bias tee is disconnected by default, but it can be activated by soldering two pads together on the PCB. Connecting the pads connects the antenna output to the USB 5V rail. The resistance in the fuse and inductor can reduce the output voltage to about 4.5V.
The USB power rail is protected from over current and shorts through a PTC resettable fuse with a hold current of 80 mA and trip current of 200 mA. This means that the fuse will become a short circuit if greater than 200 mA tries to flow through it, which may happen during a short or with faulty equipment. Between 80 mA and 200 mA is an unknown state, where the fuse may or may not trip, depending on the temperature. In practice we’ve tested it with a hold current of 120 mA in a ~16 degree ambient environment (and much hotter inside the dongle casing) and had no issues with premature tripping.
We used a 4.7 uH 250 MHz SRF inductor as the bias tee choke. At the highest frequency tunable by the RTL-SDR (~1700 MHz) this should only give a (simulated) ~1-2 dB loss through the inductor. For better performance at frequencies above 1 GHz you could experiment with a smaller value inductor and possibly with removing the static protection diode, though in our tests we saw very little difference with the diode removed.
We have tested the bias tee with an LNA4ALL and HABAMP both in bias tee mode. Both worked fine running for a number of hours. The HABAMP really improved ADS-B reception a lot and we highly recommend it. We also tested the unit with two LNA’s connected together, both powered by the bias tee and this also worked fine. An LNA like the LNA4ALL draws about 60 mA of current, so running two at once is pushing the hold current of 80mA on the fuse, but we had no trouble with about 120 mA of current, though we need to note that people in hot climates may have different results as the trip current reduces with higher temperatures. We also tested an active GPS antenna (active antennas contain built in LNA’s) which also worked.
With the bias tee and LNA’s we were able to improve weak signal reception and also receive several signals not usually receivable by the RTL-SDR alone such as L-band satellites like Inmarsat, GPS and Iridium with an appropriate antenna.
Experimental break out pads for direct sampling
The direct sampling mod is a hardware modification that allows you to tune to HF frequencies with an RTL-SDR. The best way to apply this mod is to directly solder your antenna or matching transformer to pins 4 & 5 of the RTL2832U chip. However, these pins are very small and so the mod requires extreme soldering ability.
These units have break out pads for these pins which make soldering to them much easier.
Let us know if you have any questions about these units, or feature requests for future units. We’ve tried to make the most popular changes that don’t increase the cost too much, but we are always open to ideas for future improvements.
Now in order to reduce the number of spurious signals produced at multiples of 28.8 MHz by the external local oscillator, Milan has added a simple low pass filter to the oscillator output. He used a 3-pole Butterworth filter with a 30 MHz cut off point. This filter acts to the convert the Si535A’s square wave output into a sine wave, which should help reduce the number of spurious signals produced.
Milan tested his filtering by creating a wideband comparison sweep with rtl_power. With the low pass oscillator inserted, the spurious signals were clearly reduced a significant amount. Then by reducing the drive level from 8mA to 2mA, the spurious signals were reduced even more. By using a filter with more poles it’s possible that the spurious signals could be reduced even further.
With most low cost RTL-SDRs the provided 28.8 MHz local oscillator is of low quality and will often have a sizeable frequency offset and temperature drift. To fix this the oscillator can be replaced with a higher quality one.
Milan also writes how the Si535A has options for different drive currents and writes that by lowering the drive current from 8 mA to 2 mA the spurious signals found at multiples of 28.8 MHz in the RTL-SDR are reduced. He also writes the the Si535A also allows you to detune the frequency a bit which can help to tune around any spurious signals. This could be useful for example when using rtl_power as you could create a spurious free plot.
A TCXO is a temperature controlled oscillator. Its advantage over a standard oscillator like the one used in a normal RTL-SDR is that its frequency will not drift as the temperature of the dongle changes.
The modified TCXO dongle is based on a standard full sized R820T RTL-SDR PCB board. Inside we can see that the standard 28.8 MHz oscillator has been removed and in its place is a 28.8 MHz TCXO oscillator. The old oscillator has been removed and the new oscillator is carefully soldered in its place. The soldering job appears to be nice and tidy.
Next we tested the TCXO dongle against a control RTL-SDR which had a standard non-TCXO oscillator. First we measured the PPM offset at room temperature against a known ATIS signal. The TCXO dongle had an offset that was somewhere around 0.5 PPM. The control RTL-SDR had an offset of 60 PPM.
As the dongle heats up from use, the oscillator will experience thermal drift, causing the frequency offset to change. The TCXO should be immune to this problem due to it’s temperature compensation circuitry. To test the TCXOs temperature compensation capabilities we placed both dongles in a freezer for 30 minutes and then took them out and recorded their start and end PPM offsets after 30 minutes of operation. To simulate a warm environment the dongles were also placed under a warm tungsten light during operation.
The control RTL-SDR started with an PPM offset of 58 PPM and ended with an offset of 72 PPM, giving a total drift of 14 PPM. The TCXO RTL-SDR dongle started with a PPM offset of ~0.5 PPM and ended with an offset of ~0.5 PPM, giving a total drift of 0 PPM.
Below we have recorded animated GIFs of the drift observed in both dongles. The first GIF shows the control RTL-SDR. Note how the frequency offset oscillates at first and then slowly drifts away in one direction. The TCXO dongle exhibited no drift whatsoever.
The results show that the US TCXO has a very low overall PPM offset and is very stable over temperature changes. If you want one of these dongles they can be found for sale over on Ebay for $65 USD.
In the previous post some commenter’s raised the concern that this product was overpriced at $65 USD. However, we believe this price is reasonable. The reason the cost is much higher than a standard RTL-SDR is that it is very difficult to obtain TCXO oscillators with 28.8 MHz clocks. The company selling these needed to have the TCXOs custom made from the factory with a large minimum order quantity of 1000 pieces. Then after adding the labour, quality control and selling costs the profit margins become quite small.
Previously on our blog we posted about RTL-SDR dongles for sale in Japan that had been modified by replacing the low quality 28.8 MHz oscillator with a quality low ppm and high stability temperature controlled oscillator (TCXO).
Now there is a new USA based source for modified TCXO dongles over on Ebay. The new dongles are modified with a 1 ppm high stability TCXO and sell for $65 USD. The manufacturer of these modded dongles has been kind enough to send us a sample and we will have a review of the product up in a few days.
The oscillator on the RTL-SDR is prone to thermal drift, which means that as the dongle heats up from use, the frequency you are tuned to may change over time. A TCXO compensates for differences in temperature and thus keeps the frequency stable as the temperature changes.
Over on his blog, Nobu has uploaded a post showing his experiments with a forced air cooled RTL-SDR dongle (note the blog is in Japanese so use Google Translate). A tiny fan from RS components is attached to the case of the dongle and draws power from the dongles own power regulator. Nobu also replaced the crystal oscillator with a flat packaged oscillator for a better fit.