Jim Lesh: As I
had mentioned earlier this morning in a comment to Dr. Townes' presentation:
Yes, there is a trade-off on how imperfect you make that aperture versus how
many spatial directions you become subject to background noise.
Right - and noise from all those other directions if there is noise. You still
lose something, that's what you do.
Jim Lesh: If
there is noise in there. If you have stellar images that are point sources and
not many other images in the vicinity, then what you do is you have a wider
field of view viewing that noise source; the star in this particular case. But
if it is a distributed source, then yes, you will automatically increase the
noise based on that ratio of the field of view. If it is a point source, or not
very distributed source, then there is a limit of how much additional noise you
pick up because of that wider field of view.
Let me just comment on this question of signal to noise: If you were to cut your
signal off, then there is no noise. The way I would define signal-to-noise is on
the basis of the strength of the signal compared with the noise that's there
when there is no signal. In that case this comes out as normal signal-to-noise
just the way you are describing it. It is a matter of language as to what you
call "signal-to-noise". I think it would be handy and a little more
logical to say "the noise is what is there without the signal" - when
there is no signal at all. In that case, the noise can be essentially zero.
Jim Lesh: My
caution is; beware of simply taking the ratio of those two parameters, however
you define the noise. The actual performance of a detection system depends on
the particular operating point as well as the ratio of those two values.
Well, I would say if there were no noise and there were no photons there at all,
and then one photon comes along which is your signal and you detect it, then you
have a very high signal- to-noise! If you would define the signal-to-noise the
way I would define it, I think there is no more worry about it. It come out
Jim Lesh: I
think that a lot of people that would agree with you. I think there are people
that would likewise disagree with you because of the quantum noise that you
typically get through fluctuation when the intensity is large. But my point is
that it is not a function that depends on the ratio. It's a function that
depends on the individual operating points for both whatever you define as noise
as well as what you define as signal.
I think what you are defining as "noise" I would define instead as
Kent Cullers: It
seems to me that the issue here is one of detection versus decoding. I tend to
side with Professor Townes here. I want to know my detection signal-to-noise
because of my special interests. If on the other hand, I am trying to
communicate what a signal-to-noise ratio is, it has to tell me how well in fact
I can get my message back. These are often different issues in the microwave
region, certainly the statistics of detection and decoding look very different.
Clive Goodall (Floor):
I just have a point I am trying to track in this debate here. Unfortunately, a
lack of breadth of technical knowledge on my part has caused me to lose track.
In your critique of the models underlying the traditional approach, are you
producing an argument for enhancement of detection possibilities? I'm not sure
if I got the direction of criticism right here. Is it a gloomier outcome you are
looking at based on what you are saying, or is it a more optimistic picture? It
sounds to me a more optimistic one.
Jim Lesh: In may
cases, I think it will turn out to be more optimistic situation. You did state
it correctly, it is ways of enhancing the detectability of the pulse that I am
trying to refer to. In order to communicate, we typically have to detect those
pulses. We then try to extract from that some information through the modulation
format. In order to do the best job of communicating and getting that
information, we also have to do the best job of detecting. It is from that
standpoint that I was bringing up these various models.
Clive Goodall (Floor):
I'd like to move along now. We'll go another couple of notches, then we'll allow
further discussion because some of these same issues may well come up again. OK,
would you like to make any comments?
Well, the only comment I'd like to make since I was speaking of such very short
pulses, is that I didn't intend to mean in my presentation that somehow there
was something magic about one nanosecond. What I meant to convey was that it is
in the arena of short pulses, that you gain this bits-per-pulse very readily,
and that you gain this potential discrimination against the star's background.
There is nothing magic about one nanosecond just as there is no magic frequency.
It is just a convenient parameter when one goes through the numbers.