Sampling and Bandwidth error correction

[Overflow]

Hi

I've written a program to decode the NOAA APT weather satellite images,
and its working grand...

My problem is, i don't know why!!

specifically, theres 4160 'pixels' transmitted every .5 seconds,
assuming that any pixel can take on any value(0-255, it's an 8 bit
signal) i.e. that its not smooth, then is the max frequency of the
signal 4160 Hz??

problem lies in the fact that this is modulated on a 2400Hz carrier,
would modulation (any type) not be the same as sampling in that would
it not need a carrier frequency double the maximum frequency in the
signal??

One of these cases of thinking ya understand something and looking back
over it it doesn't make sense!!!

Is it possible to correct for bandwidth errors? the program works on
signals recorded with a reciever of the correct bandwidth (50kHz). I'd
like it to be able to work with signals recorded recievers with higher
bandwidth.

I'm guessing that filtering won't work, but the only other way to do it
that i can think of is to remodulate it at the higher bandwidth then
demodulate it at the correct bandwidth.(incidently the signal is
transmitted on ~137.6MHz, so i'd have to change the sample rate too)
essentially i'm looking for a less expensive (even if more involved)
algorithm

[Jerry Avins]

Overflow wrote:
> Hi
>
> I've written a program to decode the NOAA APT weather satellite images,
> and its working grand...
>
> My problem is, i don't know why!!
>
> specifically, theres 4160 'pixels' transmitted every .5 seconds,
> assuming that any pixel can take on any value(0-255, it's an 8 bit
> signal) i.e. that its not smooth, then is the max frequency of the
> signal 4160 Hz??

Yes. The pixels come at 8320/sec (and the bits at 66520/sec). If they
alternate between two values, the frequency would be half that

> problem lies in the fact that this is modulated on a 2400Hz carrier,
> would modulation (any type) not be the same as sampling in that would
> it not need a carrier frequency double the maximum frequency in the
> signal??

First, your conception is incorrect. The signal is not AM modulated on a
carrier. IIRC, it is similar in character to the signal on a 38400 bps
telephone modem.

Second, although the maxim possible visible frequency that can be sent
at 8320 pixels/sec is 4160 lines/sec, the actual maximum might be
restricted to less than that.

> One of these cases of thinking ya understand something and looking back
> over it it doesn't make sense!!!
>
> Is it possible to correct for bandwidth errors? the program works on
> signals recorded with a reciever of the correct bandwidth (50kHz). I'd
> like it to be able to work with signals recorded recievers with higher
> bandwidth.

What's the problem? A garage door can't be too wide to let the car in.
Excess bandwidth lets in extra noise, and selectivity is needed to
exclude adjacent signals. but what other ill effects do you anticipate?

> I'm guessing that filtering won't work, but the only other way to do it
> that i can think of is to remodulate it at the higher bandwidth then
> demodulate it at the correct bandwidth.(incidently the signal is
> transmitted on ~137.6MHz, so i'd have to change the sample rate too)
> essentially i'm looking for a less expensive (even if more involved)
> algorithm

Unless you want to build an entirely digital receiver, the carrier
frequency is immaterial. The digital data are encoded onto the baseband
FSK, QPSK, or whatever.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

[Mark]

Are you receiving the APT format signal?

http://www2.ncdc.noaa.gov/docs/klm/html/c4/sec4-2.htm


The APT signal looks like it is a 2400Hz subcarrier that is AM
modulated with the intensity of the image. The 2400 Hz subcarrier is
FM modulated onto the RF carrier around 137 MHz.

I think it is 4160 analog pixels in 1 second.



Mark

[Overflow]

Mark wrote:
> Are you receiving the APT format signal?
>
> http://www2.ncdc.noaa.gov/docs/klm/html/c4/sec4-2.htm
>
>
> The APT signal looks like it is a 2400Hz subcarrier that is AM
> modulated with the intensity of the image. The 2400 Hz subcarrier is
> FM modulated onto the RF carrier around 137 MHz.
>
> I think it is 4160 analog pixels in 1 second.
>
>
>
> Mark

Yes sorry it is 4160 pixels/second. but even still my point still
applies

[Overflow]

Jerry Avins wrote:
> Overflow wrote:
> > Hi
> >
> > I've written a program to decode the NOAA APT weather satellite
images,
> > and its working grand...
> >
> > My problem is, i don't know why!!
> >
> > specifically, theres 4160 'pixels' transmitted every .5 seconds,
> > assuming that any pixel can take on any value(0-255, it's an 8 bit
> > signal) i.e. that its not smooth, then is the max frequency of the
> > signal 4160 Hz??
>
> Yes. The pixels come at 8320/sec (and the bits at 66520/sec). If they

> alternate between two values, the frequency would be half that

Sorry, as Mark points out below it's 4160 pixels per second. and no its
not 66520 bits/sec, my discription of it as a '8bit' signal is wrong,
the idea i'm trying to get across is that as it's AM modulated and
theres 256 discrete amplitude levels.

> > problem lies in the fact that this is modulated on a 2400Hz
carrier,
> > would modulation (any type) not be the same as sampling in that
would
> > it not need a carrier frequency double the maximum frequency in the
> > signal??
>
> First, your conception is incorrect. The signal is not AM modulated
on a
> carrier. IIRC, it is similar in character to the signal on a 38400
bps
> telephone modem.

it is AM modulated what i pionted out above should make this clear

> Second, although the maxim possible visible frequency that can be
sent
> at 8320 pixels/sec is 4160 lines/sec, the actual maximum might be
> restricted to less than that.

that's true, given it's a natural signal(image) the highest freq
components would be at the water land edges, so i guess i have a moot
point here, but the way the signal is transmitted limits the possible
quality of the image?

>
> > One of these cases of thinking ya understand something and looking
back
> > over it it doesn't make sense!!!
> >
> > Is it possible to correct for bandwidth errors? the program works
on
> > signals recorded with a reciever of the correct bandwidth (50kHz).
I'd
> > like it to be able to work with signals recorded recievers with
higher
> > bandwidth.
>
> What's the problem? A garage door can't be too wide to let the car
in.
> Excess bandwidth lets in extra noise, and selectivity is needed to
> exclude adjacent signals. but what other ill effects do you
anticipate?

that the signal will be swamped in noise( actually given that the noise
is demodulated with the signal, that they'll be completely
inseperable)...don't know how to show what i think will happen... and
don't have any software to do a quick prototype...

> > I'm guessing that filtering won't work, but the only other way to
do it
> > that i can think of is to remodulate it at the higher bandwidth
then
> > demodulate it at the correct bandwidth.(incidently the signal is
> > transmitted on ~137.6MHz, so i'd have to change the sample rate
too)
> > essentially i'm looking for a less expensive (even if more
involved)
> > algorithm
>
> Unless you want to build an entirely digital receiver, the carrier
> frequency is immaterial. The digital data are encoded onto the
baseband
> FSK, QPSK, or whatever.

so how do i go about seperating the signal from the noise??

[Jerry Avins]

Overflow wrote:
> Jerry Avins wrote:
>
>>Overflow wrote:
>>
>>>Hi
>>>
>>>I've written a program to decode the NOAA APT weather satellite
>
> images,
>
>>>and its working grand...
>>>
>>>My problem is, i don't know why!!
>>>
>>>specifically, theres 4160 'pixels' transmitted every .5 seconds,
>>>assuming that any pixel can take on any value(0-255, it's an 8 bit
>>>signal) i.e. that its not smooth, then is the max frequency of the
>>>signal 4160 Hz??
>>
>>Yes. The pixels come at 8320/sec (and the bits at 66520/sec). If they
>
>
>>alternate between two values, the frequency would be half that
>
>
> Sorry, as Mark points out below it's 4160 pixels per second. and no its
> not 66520 bits/sec, my discription of it as a '8bit' signal is wrong,
> the idea i'm trying to get across is that as it's AM modulated and
> theres 256 discrete amplitude levels.

I used your figure of 4160 pixels per half second. I guess that was
inaccurate. There are two frequencies to consider, the spatial
frequencies in the image and the electrical frequencies.

>>>problem lies in the fact that this is modulated on a 2400Hz carrier,
>>>would modulation (any type) not be the same as sampling in that would
>>>it not need a carrier frequency double the maximum frequency in the
>>>signal??
>>
>>First, your conception is incorrect. The signal is not AM modulated on a
>>carrier. IIRC, it is similar in character to the signal on a 38400 bps
>>telephone modem.
>
>
> it is AM modulated what i pionted out above should make this clear

You're right. I had the wrong signal in mind.

>>Second, although the maxim possible visible frequency that can be sent
>>at 8320 pixels/sec is 4160 lines/sec, the actual maximum might be
>>restricted to less than that.
>
>
> that's true, given it's a natural signal(image) the highest freq
> components would be at the water land edges, so i guess i have a moot
> point here, but the way the signal is transmitted limits the possible
> quality of the image?

Transmission bandwidth always imposes a limit on image resolution.

>>>One of these cases of thinking ya understand something and looking> back
>>>over it it doesn't make sense!!!
>>>
>>>Is it possible to correct for bandwidth errors? the program works on
>>>signals recorded with a reciever of the correct bandwidth (50kHz). I'd
>>>like it to be able to work with signals recorded recievers with higher
>>>bandwidth.
>>
>>What's the problem? A garage door can't be too wide to let the car in.
>
>>Excess bandwidth lets in extra noise, and selectivity is needed to
>>exclude adjacent signals. but what other ill effects do you anticipate?
>
> that the signal will be swamped in noise( actually given that the noise
> is demodulated with the signal, that they'll be completely
> inseperable)...don't know how to show what i think will happen... and
> don't have any software to do a quick prototype...
>
>
>>>I'm guessing that filtering won't work, but the only other way to do it
>>>that i can think of is to remodulate it at the higher bandwidth then
>>>demodulate it at the correct bandwidth.(incidently the signal is
>>>transmitted on ~137.6MHz, so i'd have to change the sample rate too)

Hopeless. Once the noise mixes into the same band as the signal, the
game is over.

>>>essentially i'm looking for a less expensive (even if more involved)
>>>algorithm
>>
>>Unless you want to build an entirely digital receiver, the carrier
>>frequency is immaterial. The digital data are encoded onto the baseband
>>FSK, QPSK, or whatever.

Forget that: wrong protocol. But the noise is mixed in to stay anyhow.

> so how do i go about seperating the signal from the noise??

Keep it out in the first place with a more selective receiver.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯