The approach taken for the Digital NBTV mode is completely different, in many senses:
- It uses separate programs for modem and codec in the transmitter and receiver
- TCP/IP communications is used between program modules
- The modem is high speed single-tone PSK, similar to NATO's STANAG 4285
- An equalizer system is included to compensate for ionospheric path variation
- Wavelet image compression and forward error correction are used
- It provides completely noise-free and error-free image reception, even on 80m
Modulation
Digital NBTV uses a modulation technique which is widely used by high speed HF radio modems.
A single 1500Hz phase modulated carrier is used to send both packet sync and payload.
Using BPSK modulation, a pseudo-random binary (PN) sequence starts each packet, and is used to identify an exact point in the
transmission from which the data can be synchronized. A cross-correlator is used
in the receiver to locate the one point in the whole message where the sequence
matches up with the local copy of the sequence. The cross-correlator
works with a known pattern to look for, and is a very powerful and sensitive tool.
These radio modems use the PN sequence technique to enable complex high speed data to be decoded accurately - the ranging information determined from the cross-correlator is used to correct the received data timing to reduce errors induced by the ionosphere, using a signal processing device called an 'equalizer'. The equalizer also corrects for Doppler errors which affect carrier phase, making the use of 8-PSK practical.
Digital NBTV uses a 31-bit PN sequence borrowed from STANAG 4285, with one chance in two billion of a perfect score being caused by noise. It uses 80 symbols (modulation time slots) to send this sequence about 2.5 times. Each packet is contained in a 336 symbol frame. 256 symbols are used for image data and FEC information. Since 4-PSK is used for the data, each packet could contain 512 bits of image data, or 3047 bps raw data rate. The data symbols are scrambled in an 8-PSK pattern to improve resistance to selective fades.
Each Digital NBTV packet takes 168ms to transmit
Like the STANAG 4285 system, single-tone PSK Digital NBTV can also operate at 2400 baud, using a sub-carrier frequency of 1800Hz. The corresponding bandwidth (just under 3kHz) is too much for most HF transceivers, but quite suitable for VHF, and gives a worthwhile speed improvement. However, to fit the signal into a normal amateur transceiver IF, it is usually operated at 2000 baud using a 1500Hz sub-carrier.
Modem
The modem section of the transmitter or receiver converts digital data into PSK audio for the transmitter, or
received audio into digital data, respectively. The receiver modem also has to manage sync and equalization.
Each transmitted packet commences with a BPSK pseudo-random (PN) sequence (same sequence as STANAG 4285), which is used to synchronize the receiver timing with the start of the packet, and also serves as a measuring point for the receiver equalizer software which measures and compensates for frequency offset and drift, and other code which compensates for timing errors. Detection of the PN sequence is achieved using a cross-correlator. This technique is extremely sensitive, so no matter how weak the signal is, packet synchronization is secure.
The packet data payload is transmitted as 4-PSK, to ensure a high data rate. There are nearly six packets per second, using a 2000 baud modem.
Codec
Because a digital system is inherently much less bandwidth-efficient than an analog one, in order to achieve even reasonable
frame rate, considerable effort must be made to reduce the amount of data transmitted to a minimum. Two coding and decoding (codec) strategies are used:
a pixel interpolation technique, and a wavelet compression technique.
The Digital NBTV system offers some flexibility of image size:
- 48 x 48 pixels zoomed
- 64 x 64 pixels
- 96 x 96 pixels zoomed
- 128 x 128 pixels
- 256 x 256 pixels
- 64 x 64 pixels
Interpolation
All transmitted pictures are square (1:1) in pixel ratio, but due to interpolation techniques used, are generated from
4:3 ratio images, and result in received images that are again conventional 4:3 landscape format. This gives a built-in compression to
3/4 of the original data. The images are always displayed the same size, but of course vary in resolution.
The image at the top of this page shows the receiver image view pane, with a 64 x 64
pixel image displayed. The image was received over a 500km path on 80m at night, and is completely noise-free.
Compression
Standard image compression algorithms such as JPEG, JP2 and MPEG are designed for
significantly higher image resolution than those used here, and do not work well on such small (low resolution) images.
A special series of 'Wavelet' compression algorithms was therefore developed, tuned for small images.
The wavelet transforms to work at their highest efficiency when operating with images that are 2n (a power of two) wide and high,
which is accomodated by the interpolation to 1:1 aspect ratio and the choice of image sizes.
Three alternative transforms are offered:
- Haar
- Daubechies D4
- Cohen-Daubechies-Feaveau 9/7 CDF97 (default)
- Daubechies D4
The transmitter codec illustrating transmission in 128 x 128 mode
In the above screen-shot of the transmitter codec program, you see at the left the image from the camera, already squashed horizontally by arithmetic compression into a 1:1 ratio. To the right of this is the actual image transmitted, and you can see the artifacts and blurriness generated by reducing the original large image to 128 x 128, and the application of the CDF97 compression. This will be exactly the same as the received image. To the right of this picture are the codec controls, which set the frame size, the compression transform used, and (further to the right) a vertical bar which shows how much of the available space in each packet is occupied by data (green) and Forward Error Correction (FEC) data (red). The green bar changes as you change picture size, and you can drag the red bar up and down to change the strength of error correction. All these controls can be used 'on the fly' - the change the appearance of the displayed picture immediately, and affect the transmitted pictures from the start of the next frame.
Error Correction
A digital system, especially one with compression, cannot recover an image at all if it is not received correctly. This is
a tall order on HF without some sort of error correction. The technique used here is the Reed-Solomon (RS) technique, which
is ideally suited to packetized data, and also has the advantage that the level of correction strength depends on how much
correction data is transmitted. The FEC decoder will simply throw away any image that cannot be completely corrected. Thus
reception using this NBTV system is always error free. You might not get many pictures, especially if conditions are poor,
but every picture received will be perfect. Two levels of Reed Solomon coding are used: an errors-only rate 2/3 RS on the data within each packet, and an errors and erasures variable rate RS on the data (which have been interleaved and thus spread over several packets to further reduce the effect of burst errors).
Data Transmission
Because this is a completely digital system, it is no trouble at all to also transmit text and other data. In this software
two simple options are offered. You can tag every packet with a callsign. There is no overhead involved, as there is free
space in each packet to allow this to be included. You can also send a text message up to 39 characters long. This is sent as
a special text packet, once every new image. This does slow down the image transmission rate slightly.
Transmission Speed
The actual time taken to send each frame depends on the complexity of the image, the compression used, and the amount
of FEC used. Image complexity alone can account for a wide variation. The table below shows typical values, provided you leave the compression transform and error correction level
at the default settings.
| Mode | TX Size | Image Size | Pixels/Frame | Frame Rate |
| 48 | 48x48 | 48x64 | 2304 | ~9 sec |
| 64 | 64x64 | 64x85 | 3072 | ~10 sec |
| 96 | 96x96 | 96x128 | 9216 | ~13 sec |
| 128 | 128x128 | 128x171 | 16384 | ~15 sec |
| 256 | 256x256 | 256x341 | 65536 | ~25 sec |
This performance seems at first glance to be not much better than SSTV image rates, but that's the price paid for using a digital (and noise free) system. Note also that the time taken to send each frame is not in direct proportion to the number of pixels in the image, as the wavelet compression is much more efficient on higher resolution images. The transmissions are always in colour, as the compression ratio for B&W is little better and thus the option is not justified.
The ITU 'Emission Designation' for all these modes is 2K40G1FNF.
