[Namaste-dev] Re: Fw: Modulation Baseline

Phil Karn karn at ka9q.net
Mon Jun 16 14:28:13 PDT 2008

> 7. I missed the reasoning behind selection of IP and fixed channel
> assignment.  It seems the design is primarily for a full time assignment of
> a part of the satellite capacity for a "circuit switched" voice connection
> between two users?

Voice is the intended primary application, yes, but we don't want to
limit it to that. There may be need to send other data, such as still
photos or (very) slow scan TV, within the capacity of the satellite.

Besides, IP has many advantages even for voice. This kind of satellite
system is best suited for remote locations where no other communications
are available, but it won't span all the parties they may wish to talk
to, especially in an emergency or public service situation. IP is
extremely flexible, and it is specifically designed for interoperability
between users on dissimilar physical networks.

I can easily envision an emergency voice net (multi point round table)
that spans satellite users, users on terrestrial digital repeaters
adapted to VoIP, ad-hoc cooperative packet voice networks, and of course
VoIP nodes on the "real" Internet.

Protocol overhead can be made a non-issue with header compression,
already widely used in terrestrial VoIP.

> 8. Does RTP have any benefits in a primarily voice environment?

Yes, RTP (Real Time Protocol) was specifically designed for voice,
though it can carry other real time streams, such as video. I think it
also has some desirable properties for spacecraft telemetry -- another
realtime data stream.

> 9. Why was a choice made to put so much technology in orbit?  It seems a
> system with a linear transponder and all the complexity on earth would be
> less likely to experience failure, be less expensive to launch, and be more
> supportive of multiple transmission schema. I'll stop there.

This is a fundamental question. I agree that simple "bent pipe"
transponders have the significant advantages you give. But they also
have significant disadvantages:

a. Noise concatenation. The SNR of the demodulated down link is set by
both the down link and uplink SNRs. E.g., if they're equal, the down
link will be 3 dB worse than if the uplink had infinite SNR. Uplink SNRs
are usually higher than down link SNRs but they're still not infinite,
and achieving high uplink SNR can require a lot of uplink power that can
be expensive.

Depending on the transponder bandwidth and gain, repeated uplink noise
can represent a non-negligible fraction of the spacecraft down link
power. This is expensive wasted power.

b. Multiple access interference and spacecraft DC-RF efficiency. There
are many ways to share a simple bent pipe transponder but none of them
are ideal.

First, a bent pipe implies the same modulation, coding and data rates on
both uplink and down link. This is not necessarily optimum because the
uplink and down link can be quite different.

Nonlinear transponders provide maximum DC-RF efficiency, but that
implies some form of TDMA. It requires careful timing and relatively
high peak uplink power. The cost of the uplink HPA is a major cost
driver for the earth terminal, and this cost increases with peak power,
not average power.

Synchronous TDMA is ill suited for lots of small stations with irregular
traffic patterns, e.g., voice users. So you need some sort of dynamic
random access protocol, e.g., CSMA or CSMA/TDMA. Despite much research
that has produced some complex protocols, performance on long delay
paths is still poor.

FDMA requires a linear transponder to prevent intermodulation
distortion. AMSAT probably has more experience with efficient linear
transponder designs than anyone else, and it is largely disappointing.
They're very complex and require careful component matching and
adjustment (bad for space environments), yet a noise power ratio (NPR)
of only 15 dB is typical. FDMA also requires careful power control by
the uplink stations, and long AMSAT experience has shown this to be
nearly intractable. (Of course if we build the ground station we could
ensure power control.)

CDMA has much to recommend it for bent pipe transponders, but it still
requires linear transponders and tight uplink power control.

An "intelligent" transponder consisting of a bank of digital uplink
receivers feeding a computer and a digital down link transmitter has
many advantages over a bent pipe, starting with power efficiency.

a. The uplink and down link are very different problems and have
different optimum solutions. The modulation, coding and data rates can
be different. The uplink itself might support a variety of modes for
different applications and station classes.

The uplink needs efficient multiple access and a signal format that can
be rapidly acquired. Whatever FEC is used has to be decoded in the
spacecraft. Power efficiency is desirable but not critical. The cost of
the uplink HPA tends to dominate the ground station cost, so it's a
driving factor in the system design.

The down link isn't multiple access, so rapid acquisition isn't
required. It can be a single, high speed, priority-multiplexed stream
containing repeated user data, spacecraft telemetry and spacecraft data
(e.g., imagery). Every station can easily demodulate it all if desired.
Down link power efficiency is absolutely critical to system performance
and cost (see below).

b. The overriding design factor in any communications satellite is down
link power efficiency. This should be obvious since power generation is
extremely costly in space and the down link HPA is by far the biggest
spacecraft load. The figure of merit to be maximized is the aggregate
down link user data rate divided by the DC power consumption of the
spacecraft down link subsystem.

This means choosing a down link coding and modulation that minimizes
Eb/No requirements and maximizes the DC->RF conversion efficiency of the
down link HPA.

Minimizing Eb/No means strong FEC (turbo codes, low-density parity check
codes, or a concatenated RS/convolutional code), plus coherently
demodulated MSK or PSK (including BPSK, QPSK and offset QPSK).

Maximum HPA efficiency means nonlinear amplification and modulation that
can tolerate it: MSK or unfiltered PSK.

c. Unlike a bent pipe, there is no retransmitted noise. This improves
ground station performance and spacecraft power efficiency.

> 10. Has there been a "marketing assessment" to validate the concept and end
> user utility of this proposed satellite system?  In the 5-10 years it will
> take, at best, to launch, there will likely be many alternative
> communications techniques available to amateur radio operators.

There has been much discussion of this. The overriding goal as I see it
is to make satellites accessible to hams with severe antenna
restrictions. That means an antenna no larger than a DirecTV or Echostar
dish. Current ham satellites generally require large steerable arrays
that most hams simply cannot put up. This necessarily means abandoning
SSB, moving up to the gigahertz bands, going digital and making the
necessary ground station available at reasonable cost when the satellite
is launched.

Many of the unique features of ham radio have been overtaken by cellular
phones and the Internet, but two are left: education and
emergency/public service applications. Satellites are already best
suited for mobile or portable thin-route use from remote areas with no
other telecommunications, and this makes a system like ours attractive
to emergency services.

More information about the Namaste-dev mailing list