DiBcom Diversity Technology [2009-12-22 – 09:29:41]

One of the most innovative technologies during the past few years in the DVB area has been DiBcom’s "Diversity Technology", at least in my opinion. It aims at improving the sensitivity of COFDM modulated signals (e.g. DVB-T/DVB-H) and compensating the Doppler effect for more robust reception in borderlands and/or mobile applications. It does so by smartly combining the signals received by multiple (small bar) antennas. Figure 1 illustrates the ideas behind the technology.

Figure 1 - Diversity Technology Overview

Several COFDM demodulators (e.g. DVB-T) are daisy-chained to each other. The Maximum Ratio Combining algorithm (MRC) inside weights the signals received by all of them in "real-time" with respect to their Signal-to-Noise Ratio (SNR) and smartly "sums them up" to one single improved signal. According to DiBcom, a two-antenna MRC system can achieve a sensitivity improvement (increase in SNR) of up to 3 to 10 dB.

Theory and exercise often tell a different story. The reason for me to take a look at the practical site of the technology is because I am living in DVB-T borderland here in Austria. ORS, Austria’s leading broadcaster, recommends using a roof antenna properly aligned to the transmitter for clean reception, see figure 2 for their coverage map in my region.

Figure 2 - ORS Coverage Map

Marked red is the location of testing. Light blue color indicates that proper reception is only possible with a roof antenna, the area not colored is not covered by DVB-T at all. The darker blue it gets, the better the signal conditions in this very area. Judging from that, it should be quite obvious that conditions are not at their optimum with lots of dead spots.
The coverage map only tells the story for ORS multiplex A and B (DVB-T.at) from the transmitter Pfaender, which are aired with an ERP of 71 kW.
Also aired from the transmitter Pfaender is a SRG SSR multiplex with only 9 kW (see table 1).

	ORS MUX A	ORS MUX B	SRG SSR MUX
Carrier	498 MHz	474 MHz	578 MHz
Polarization	Horizontal	Horizontal	Vertical
Bandwidth	8 MHz	8 MHz	8 MHz
Modulation	16 QAM	16 QAM	16 QAM
Guard Interval	1/4	1/4	1/8
FEC Coderate	3/4	3/4	3/4
Netto Datarate	14.93 MBit/s	14.93 MBit/s	16.59 MBit/s
ERP	71 kW	71 kW	9 kW

Table 1 - ORS Transmitter Pfaender

The strong ORS muxes did not seem to be much of a challenge for Diversity, so I concentrated on properly receiving even the 9 kW SRG SSR mux where I am pretty sure no coverage would be attested at all in the above map region.

To evaluate the performance of Diversity, I used TransEdit to analyze the digital data (on TransportStream level) delivered by a TerraTec T5 USB DVB-T receiver with a Diversity enabled DiBcom DIB7700 chipset for about half an hour in both, single and dual antenna environments and compared the TS packet loss (also including broken packets which could not be repaired by the FEC circuit).
Also evaluated was the impact of the antenna alignment (position and horizontal/vertical mounting). The small bar antennas, those that came with the T5, were placed outside the window. A region that is not covered is enough to deal with. So I decided to eliminate the bad impact of a well isolated house.

Figure 3 - Single Antenna Setup 1

Figure 4 - Single Antenna Setup 2

Figure 5 - Diversity Setup

Single antenna setup performance ranges from bad (lots of garbled data, see figure 4) to very bad (no evaluable data at all, see figure 3), pretty much depending on the alignment of the antenna.
The Diversity setup (figure 5) was pretty close to perfect, not as much depending on the alignment of the antennas as the single antenna setup. The signal is clean most of the time, only every now and then, short error bursts are introduced, causing some packets to be lost and/or broken. Garbled video packets in the double-digit area leads to only little video information being lost (in the area of KBit), which in real-life translates to subtle video artefacts for only a split second, depending on the error concealment of the video format and decoder.

Unfortunately, the DiBcom driver (3.12a) does not offer the possibility to monitor the SNR, but only propagates a meaningless quality percent indicator via the IBDA_SignalStatistics interface. 0%-20% vs. ~80% sounds like a huge impact, which the performance analysis confirmed in an impressive manner, but it does not give the least idea of the real SNR.

Cool thing is, that no special software support is required. DiBcom driver is pure BDA, providing two DVB-T BDA tuner filters. If the DVB application only opens one tuner, Diversity is used automatically. If both are opened (if PiP or parallel recordings are allowed), the DiBcom chipset is configured as dual tuner, adding the possibility to lock two frequencies simultaneously. No Diversity is used in this case. So if you are living in DVB-T borderland, you better disable one tuner for DVB applications supporting multiple of them.

So finally someone came up with something practical concerning the DVB-T reception problem. Compared to active antennas, which amplify the power of a signal (often only resulting in more noise), Diversity improves its quality (SNR), effectively resulting in a less erroneous stream.