A container for 23cm biquad antenna

There is a bit of local activity on 23cm and the RTLSDR dongle allows me to receive that HAM band. An improvised indoor dipole brought in few interesting signals, increasing my interest for 1296 MHz. I have already built a biquad antenna, which needs a (cheap!) housing before being installed outside.

My biquad size is about 20x30x6 cm and I have had troubles locating suitable plastic containers in the kitchen department of local supermarkets.

During a visit to IKEA lower floor I spotted the SAMLA series of plastic transparent containers. The 11 litres one is large enough to host my biquad (38x28x14 cm) and it costs 1+1.75 = 2.75€ (lid+box).

The antenna will go on the internal side of the cover, which is easier to work on and cheaper in case the experiment fails. In that case the SAMLA container will be repurposed in the house.

RTLUSB in VirtualBox virtual machine

Months ago I was advocating the creation and distribution of a Linux virtual machine image with pre-installed software for RTLSDR work (drivers and end user applications). The release of Zadig and SDR# for MS Windows has quickly obsoleted my idea. But how about running the whole SDR thing within a virtual machine? Would the "virtual USB" software layer be fast enough to pass through RTLSDR IQ samples?

A test within the virtual machine is quickly done with the rtl_test.exe utility distributed with RelWithDebInfo.zip file (google). My setup:

• host computer i5 processor Win7 64bit
• guest computer WinXP 32bit
• R820T dongle

I have tried running rtl_test.exe at several sampling rates and always got lost bytes, even at the lowest value of 250 kSPS! And the CPU usage was not impacted at all, it remained low/idle.

My conclusion: unless you accept data loss in your RTLSDR decodes, a VirtualBox virtual machine does not guarantee enough "USB bandwidth" for successful reception within the VM itself.

SDR on a netbook

Since sdr# runs fine on the 2004 laptop (Centrino CPU), I wanted to test the RTLSDR hardware on the eeePC 901 with a dual-core Atom N270 processor. That is not an hyperfast processor, but the netbook is small and it has a solid-state drive (SSD, as opposed to "spinning" hard disk or higher capacity).

First test involved using ADSBsharp to upload data to the common hub. This piece of software runs the RTLSDR dongle at 2048 Msamples/sec, and the machine was coping with the load, delivering as high as 120 decoded packets per second.

Then I tried latest sdr# development release and, despite the additional load for the display, the eeePC 901 behaved exactly as the older laptop. At 250 ksamples/s (total bandwidth of 250 kHz) there is CPU to spare.

Sure, the screen is tiny, but for a portable operation the netbook is small and solid. Note: I forced the "Super performance" mode, which apparently consists in a 5% overclock. With a reduced sample rate the CPU can be slowed down (to be tested).

New RTLSDR dongle has arrived

Simply amazing. The sensitivity difference between my first RTLSDR dongle based on E4000 and today's with R820T is about 3 times, in favour of the latter. Even with the stock DVB-T antenna I could pick up more signals than the E4000 and the external GP. Then I "sacrificed" the stock antenna to use the cable with MCX connector for building an adapter to the TV plug: with the GP antenna outside I could receive airplanes as far as 300km/160nm. The ADS-B software was showing an average of >50 signals per second, against less than 10 for E4000.

So, as many others have reported, the R820T stick is more sensitive above GHz than the E4000 counterpart. I also noticed a much stronger frequency drift in the first minutes of operation, which also result in a warmer RTLSDR dongle.

Two experiments will follow:
- retry 23 cm SWL'ing during the monthly activity event
- test the ADS-B reception in a location with a 270 degrees wide horizon

ADS-B ground plane antenna

This is not rocket science: dipole and ground plane are the most basic antennas you can build. And they can be good performers too! Here is a picture of the 1090 MHz GP I assembled for my ADS-B reception tests.

It is built around a panel-mount BNC socket. Each of the three solid copper wires is about 6.6 cm long and there is a 10 kohm 1/4W carbon resistor across "hot" and "cold" points to discharge static electricity right at the source.

This GP performs equally well with and without resistor, so leave it in place. This antenna is omnidirectional, and depending on its horizon it can bring up signals as far as 100 km away (E4000 tuner, airplane above 10'000 m). In this screenshot my receiver is below the ISSxyz airplane:

ADSB antenna

I am not a fan of aircraft monitoring, but their ADSB 1090 MHz transmissions represent a widely available and geographically distributed beacon for testing new antennas.

My antenna test-field is the usual balcony open to N-NE.

First I tried a biquad without reflector. This is somehow bidirectional in an "8" shape. The advantage of this antenna is the intrinsic short-circuit, which avoids the problem of static electricity build-up (and frying the RTLSDR dongle).

Then I read many people suggest a collinear antenna. Projects documented online do not agree on sizes of straight elements (13 vs 19 cm) and coils (2 vs 4 cm diameter), so I opted for a simpler ground-plane: one vertical element, and 2(4) radials. All 1/4th wavelength long = 300 / 1090 / 4 = 6.8 cm (a bit less, in practice). It is short, simple to tune and easy to handle. The 1090 MHz GP antenna can be made of 1mm dia. solid copper wire, even keeping the insulation. In order to discharge static I inserted a common 10kohm 1/4W carbon resistor across antenna terminals.

Performance? They receive signals from the same distance, the GP being slightly better (but I would need a side-by-side realtime comparison) probably due to the biquad needing some form of tuning, making it a bad match outside resonance window.

Conclusion. Unless you are absolutely sure in your interest in ADSB monitoring or you have access to proper instrumentation to measure antenna impedance above 1 GHz, go for the GP antenna. According to my observations, ADSB signals propagate in line-of-sight, so look for a good, open, position: I can draw my antenna electrical horizon by looking at what/where it can receive.
Use a low-loss coax (known quality SAT-TV coax is a good choice) and keep it short: this trick will easily save those couple of dBs so hard to achieve with a different antenna without tuning instrumentation.

IFR, Knob evolution

2012 Xmas season has taken me back to the IFRK code, so that transverter functions would be finally included, paving the way to a long-awaited public release. Curiously, after almost 3 months without looking at the source code, I was able to identify a simple way to support the transverter math.

The current IFRK state-of-the-art is:

• frequency computed down to 10 Hz resolution, like the FT8x7 display
• transverter LO 10 Hz precision
• maximum shown output frequency is above 21.4 GHz
• tuning knob supported!
• tuning knob has 11 steps and "fast" 10x mode
• direct dial; in transverter mode, only kHz digits can be inserted (1296.xyz.000)
• LO value <> memory location mapping
• FT817 proprietary functions for power level control and A/B/C repeat on IFRK display

In order to summarize current key mappings I have compiled a table with key functions (opens in new window).

Besides some cosmetics on the display and a thorough testing, I am planning to output in binary form the LO position and/or the band selected (2 bits each). These last features would control unused pins on the ATmega168 chip, so there will be no impact on existing IFR(K) circuits.

HNY 2013!

Beach40 - oscillator waveforms

The Beach40 is a simple QRP DSB transceiver for 40m, designed by Peter VK3YE who released 4 youtube videos about it. In the first video Peter draws the schematic diagram, which has then been re-drawn in electronic form and shared through the Minimalist_QRP_Transceivers Yahoo group.

It is a simple circuit, whose development can be monitored with the 100 MHz equipment of my shack/lab (oscilloscope, frequency counter). So I dediced to give it a try.

After I built the oscillator, buffer and diode-ring mixer I checked waveforms on two "test points" as identified on this modified schematic diagram:

On TP1 the waveform is almost a sinewave. The small knee close to the top occurs both with a ceramic resonator and a quartz. It also gets worse when the resonator is pulled upwards: what is causing it? How to fix it?

Probe was 10x, so 2V/div, AC coupling

Then I measured voltage on TP2, with diode ring mixer installed but without the output transformer; the effect of diodes is evident:

Probe was 10x, so 2V/div, AC coupling

And finally a glimpse at my ugly Manhattan style (probe on TP1):