[Date Prev][Date Next][Thread Prev][Thread Next] - [Date Index][Thread Index][Author Index]

Re: DSP based High Speed Data

My definition of high speed is too low. I was thinking of minimizing power
and antenna requirements and transmitting a 3 KB JPEG-encoded image in 10
seconds at 300 baud or 2400 bits/sec. This would require 39 kHz out of the
400 kHz available on P3D. 3 KB is the size of a JPEG file with an
SSTV-quality (120x160 pixel) color picture.

For very high data rates, a multi-tone system takes a lot of spectrum. FSK
requires about 1.7 times the bit rate in bandwidth so 200 kbps would occupy
350 kHz which is larger than the P3D analog or digital passbands.

A multitone system spreads the signal over a wider bandwidth to lower the
SNR required for a given error rate. With 128 tones, 8-bits are transmitted
in each symbol after error correction, so the baud rate would need to be
25,000 and the tone spacing would have to be 25 kHz at a minimum. The system
requires 3.2 MHz or 9 times the FSK bandwidth. It would also require 10 dB
less power at the receiver. However, a very fast DSP chip requires about 100
us to do the FFT calculation so it can't keep up with the baud rate.

As the number of tones increases, it gets worse. At 16-bits per symbol, the
baud rate halves and the number of tones goes up by 256 so the bandwidth
occupies increases by 128 to 409.6 MHz. The FFT calculation also gets much

If we keep the baud rate below 2000 symbols/second a cheap DSP chip can do
the FFT calculations. The data rate is limited to 16,000 bits/second and the
bandwidth required is only 128 kHz.

Higher speeds should be done with QPSK modulation. This requires 1 Hz of
bandwidth per bit/sec. before adding error corection overhead, requires
minimal processing on a DSP and is more efficient than FSK. FSK is actually
the worst case for a DSP chip -- it has to do about 20 arctangents per baud.
This requires 1260 instructions on an Analog Devices DSP per bit.

>Date: Tue, 12 Sep 2000 18:00:05 -0400
>From: bronson@eece.maine.edu
>Subject: [amsat-bb] DSP based High Speed Data
>Being curious about this High Speed Data stuff, I thought I'd try to put
>some numbers to the idea
>presneted by John Stephensen, KD6OZH.
>I'm new to DSP, but here goes anyway;
>Given the following:
>No info in the amplitude.
>20KHz audio upper limit.
>20Msps 10 bit A/D converter.
>DSP chip that can do the calculations fast enough.
>   a) ignore the DSP processing overhead
>A minimum of 2 data points on the highest Tone (20 kHz).
>   a) and the system can resolve the difference between tones reliably.
 >  How do I get some serious bit rates (200Kbps or better) out of this
>An 8bit (128 tone) system has a speed limit of <160Kbps.  If the signal
>BW is <=1KHz (from 20KHz down) you will have to resolve tones that
>differ by 7.8Hz.  You will have to wait at least 52.6us (19KHz, lowest
>frequency) for the entire waveform to arrive before you can decompose
>the signal.  This gives you a bit rate of about 152Kbps.  Decreasing the
>signal BW doesn't help all that much since you can't break 160Kbps with
>an 8bit tone structure anyway.
>So what about a 16bit word?
>Now there are 32768 tones.  For the same signal BW (1KHz) you will have
>to resolve tones to 31mHz.  Given the above constraints your A/D
>converter can give you 2mHz, double it (Nyquist times 2) and you are
>still only at 4mHz.  The same latency period of 52.6us holds since 19KHz
>is still the lowest frequency.  Now the data rate is 304Kbs.
>If we relax the signal BW to 5KHz the difference between tones jumps to
>153mHz, the latency period goes up to 66.67us (15KHz) and the bit rate
>drops to 240Kbps.
>Have I overlooked anything?
>Absolutely... Noise (in the RF channel and Audio channel),  Throughput
>of the DSP system,  Additional tuning tones, Separation of each
>composite (all 32768 tones) audio signal in the data stream, My
>ignorance and lack of experience in the DSP field.
>How bout someone else taking a crack at it?

Via the amsat-bb mailing list at AMSAT.ORG courtesy of AMSAT-NA.
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org