PULSE DURATION ESTIMATION FROM THE OUTPUT OF MATCHED FILTER

Hi to everybody,

As you can understand from the subject, i am trying to estimate durations
of the pulses from the output of the matched filter. Let me explain you
the problem. I have pulses that are seperated in time enough under 10 dB
SNR white gaussian noise. The nominal pulse durations are 0.45
microseconds. However they may come 0.15 microseconds less than 0.45
microseconds. These narrower pulses should be accepted too. However,
pulses which are 0.20(or more than this) microseconds shorter than the
nominal pulse duration should be rejected. My sampling rate is 80 MHz.
Therefore there are just 4 points (0.05us*80 MHz=4) difference between the
narrowest pulse that should be accepted and the the pulses that should be
rejected. Another thing is pulses might come with a dynamic range of
amplitude. I mean a narrower pulse with a larger amplitude can have the
same peak value at the matched filter output with an another pulse which
has a lower amplitude but a has longer duration.And the noise power is
adjusted according to the weakest pulse(i.e it is 10 dB according to the
weakest pulse).Finally, my pulses are of rectangular type, with a rise
time 0.07 microseconds and a fall time 0.12 microseconds. Durations are
defined between 0.5*amplitude and rise and fall time are defined between
0.1*amplitude and 0.9*amplitude. I have tried many things but, each time i
was unable to have large detection probability. I would be grateful if
anybody can help me about this problem. I hope 4 points is not very few to
accomplish this task.

Thank you very much,

okutan.

okutan wrote:

> Hi to everybody,
> 
> As you can understand from the subject, i am trying to estimate durations
> of the pulses from the output of the matched filter. Let me explain you
> the problem. I have pulses that are seperated in time enough under 10 dB
> SNR white gaussian noise. The nominal pulse durations are 0.45
> microseconds. However they may come 0.15 microseconds less than 0.45
> microseconds. These narrower pulses should be accepted too. However,
> pulses which are 0.20(or more than this) microseconds shorter than the
> nominal pulse duration should be rejected. My sampling rate is 80 MHz.
> Therefore there are just 4 points (0.05us*80 MHz=4) difference between the
> narrowest pulse that should be accepted and the the pulses that should be
> rejected. Another thing is pulses might come with a dynamic range of
> amplitude. I mean a narrower pulse with a larger amplitude can have the
> same peak value at the matched filter output with an another pulse which
> has a lower amplitude but a has longer duration.And the noise power is
> adjusted according to the weakest pulse(i.e it is 10 dB according to the
> weakest pulse).Finally, my pulses are of rectangular type, with a rise
> time 0.07 microseconds and a fall time 0.12 microseconds. Durations are
> defined between 0.5*amplitude and rise and fall time are defined between
> 0.1*amplitude and 0.9*amplitude. I have tried many things but, each time i
> was unable to have large detection probability. I would be grateful if
> anybody can help me about this problem. I hope 4 points is not very few to
> accomplish this task.

Since the question is so fuzzy, I suggest you using the fuzzy logic 
approach to distinguish the pulses.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

I hope there are people who can help me. Is this atoo complicated problem??
I would appreciate the replies which say that you don't have any idea for
the solution of this problem. Or I will think that nobody cares about this
problem??

Thanks a lot

Okutan.

>
>
>okutan wrote:
>
>> Hi to everybody,
>> 
>> As you can understand from the subject, i am trying to estimate
durations
>> of the pulses from the output of the matched filter. Let me explain
you
>> the problem. I have pulses that are seperated in time enough under 10
dB
>> SNR white gaussian noise. The nominal pulse durations are 0.45
>> microseconds. However they may come 0.15 microseconds less than 0.45
>> microseconds. These narrower pulses should be accepted too. However,
>> pulses which are 0.20(or more than this) microseconds shorter than the
>> nominal pulse duration should be rejected. My sampling rate is 80 MHz.
>> Therefore there are just 4 points (0.05us*80 MHz=4) difference between
the
>> narrowest pulse that should be accepted and the the pulses that should
be
>> rejected. Another thing is pulses might come with a dynamic range of
>> amplitude. I mean a narrower pulse with a larger amplitude can have
the
>> same peak value at the matched filter output with an another pulse
which
>> has a lower amplitude but a has longer duration.And the noise power is
>> adjusted according to the weakest pulse(i.e it is 10 dB according to
the
>> weakest pulse).Finally, my pulses are of rectangular type, with a rise
>> time 0.07 microseconds and a fall time 0.12 microseconds. Durations
are
>> defined between 0.5*amplitude and rise and fall time are defined
between
>> 0.1*amplitude and 0.9*amplitude. I have tried many things but, each
time i
>> was unable to have large detection probability. I would be grateful if
>> anybody can help me about this problem. I hope 4 points is not very few
to
>> accomplish this task.
>
>Since the question is so fuzzy, I suggest you using the fuzzy logic 
>approach to distinguish the pulses.
>
>Vladimir Vassilevsky
>
>DSP and Mixed Signal Design Consultant
>
>http://www.abvolt.com
>
>
>
>
>

On Jun 11, 8:28 am, "okutan" <maoku...@aselsan.com.tr> wrote:
> I hope there are people who can help me. Is this atoo complicated problem??
> I would appreciate the replies which say that you don't have any idea for
> the solution of this problem. Or I will think that nobody cares about this
> problem??
>
> Thanks a lot
>
> Okutan.
>
>
>
>
>
>
>
> >okutan wrote:
>
> >> Hi to everybody,
>
> >> As you can understand from the subject, i am trying to estimate
> durations
> >> of the pulses from the output of the matched filter. Let me explain
> you
> >> the problem. I have pulses that are seperated in time enough under 10
> dB
> >> SNR white gaussian noise. The nominal pulse durations are 0.45
> >> microseconds. However they may come 0.15 microseconds less than 0.45
> >> microseconds. These narrower pulses should be accepted too. However,
> >> pulses which are 0.20(or more than this) microseconds shorter than the
> >> nominal pulse duration should be rejected. My sampling rate is 80 MHz.
> >> Therefore there are just 4 points (0.05us*80 MHz=4) difference between
> the
> >> narrowest pulse that should be accepted and the the pulses that should
> be
> >> rejected. Another thing is pulses might come with a dynamic range of
> >> amplitude. I mean a narrower pulse with a larger amplitude can have
> the
> >> same peak value at the matched filter output with an another pulse
> which
> >> has a lower amplitude but a has longer duration.And the noise power is
> >> adjusted according to the weakest pulse(i.e it is 10 dB according to
> the
> >> weakest pulse).Finally, my pulses are of rectangular type, with a rise
> >> time 0.07 microseconds and a fall time 0.12 microseconds. Durations
> are
> >> defined between 0.5*amplitude and rise and fall time are defined
> between
> >> 0.1*amplitude and 0.9*amplitude. I have tried many things but, each
> time i
> >> was unable to have large detection probability. I would be grateful if
> >> anybody can help me about this problem. I hope 4 points is not very few
> to
> >> accomplish this task.
>
> >Since the question is so fuzzy, I suggest you using the fuzzy logic
> >approach to distinguish the pulses.
>
> >Vladimir Vassilevsky
>
> >DSP and Mixed Signal Design Consultant
>
> >http://www.abvolt.com- Hide quoted text -
>
> - Show quoted text -

is this for an ATC transponder or DME or other radar?

Mark

okutan wrote:

> I hope there are people who can help me.

Certainly. But if you want your problem solved for you, this is going to 
be the business.


> Is this atoo complicated problem??

No. I bet the problem is trivial.

> I would appreciate the replies which say that you don't have any idea for
> the solution of this problem.

I have no idea of what the problem is. You are talking some triffles 
about filters, frequencies and pulses instead of the clear statement of 
what is the signal, what is the noise, what should be done as the 
result, and what hardware is available.

> Or I will think that nobody cares about this
> problem??

The problem is yours. Why should anyone care?

> Thanks a lot

And how much exactly is "a lot"?


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

On 11 Jun, 14:28, "okutan" <maoku...@aselsan.com.tr> wrote:
> I hope there are people who can help me. Is this atoo complicated problem??
> I would appreciate the replies which say that you don't have any idea for
> the solution of this problem. Or I will think that nobody cares about this
> problem??

First, no one but you care about your problem.

What you need to do, then is to make your problem intersting
enough for others to start thinking about it. There are two
ways of doing that:

1) Do your homewrok first
2) Describe your problem in full

After reading your post twice I can'n figure out if the
tolerance is 4 samples or the pulse is 4 samples wide.

If the pulse is 4 samples, your problem is hard.
If the pulse is wider and the tolerance is 4 samples,
there are standard text you might want to consult.
I would suggest volume 1 of van Trees.

Rune