The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> ideas for design of single pulse detection
https://designers-guide.org/forum/YaBB.pl?num=1201254035

Message started by vivkr on Jan 25^th, 2008, 1:40am

Title: ideas for design of single pulse detection
Post by vivkr on Jan 25^th, 2008, 1:40am

Hi All,

I have the following design problem and would appreciate inputs in solving it:

There is a "radar kind" of application (but not radar) where a transceiver device sends out a pulse and then listens to the echo, which may or may not come.
Usually, a pulse train is transmitted.

If there is an echo from the sent pulse train, the receiver on the device must detect it and already detect the very first pulse reliably. So the problem is one
of single pulse detection. A matched filter provides an ideal solution. However, it is still quite hard to reliably detect the sent pulse in the presence of
background interference (akin to offset) which changes over time (though not very fast), and in presence of other similar transceivers.

What would be a robust and reliable (but as always not overly expensive method) to detect the single pulse reliably in presence of a limited number of
other transceivers? Is it possible to shape the pulse train or use some special modulation to achieve the desired goal. The frequency of operation is relatively
low (pulses of length 1 us say).

Ideas are welcome.

Regards
Vivek

Title: Re: ideas for design of single pulse detection
Post by didac on Jan 25^th, 2008, 10:10am

Hi Vivek,
I've worked a little bit with UWB-Impulse Radio systems(using Time-Hopping for multiple access) so maybe I can help a little bit(although maybe you are not in UWB field).
The problem you are describing it's more from architectural level than circuit level, detection of pulses it's "the easy part"(although timing synchronization in the matched filter can be quite difficult if you want something more than the detection of a pulse, i.e. maximum energy it's obtained only when the local generated template pulse an the receiver pulse are perfectly synchronized) the problem is the ability to discriminate between desired pulses and "interferences", I think that several approaches are valid:
1)Use an scheme like CSMA(but in this case without carrier):the transceiver don't transmit always, they listen the channel if free they sent,the others listen that somebody it's transmitting and then wait a predefined amount of time(duration of pulse + margin for the echo), it's similar to CSMA-CD in Ethernet wired systems but have the problem of the hidden terminal like in WLAN systems. The approach of WLAN it's CSMA-CA but needs a device that acts as "acces point" and a sequence of RTS/CTS(Request To Sent/Clear to Sent) but it avoids hidden terminal problem.
2)Using a topology Master/Slave where one device controls the others and authorizes for transmission, then the device has a temporal window for sending the train of pulses and listening, after that the master allows the following terminal.
3)Use of CDMA to code the train of pulses, each transceiver will have it's own pseudo-orthogonal sequence(I think Gold sequences are the ones used in 3G systems), this approach complicates the receiver since it's necessary to perform correlation of the whole received sequence and needs a perfect alignment of the the template with the received train end but in theory it's feasible to implement using Pulse-Position Modulation(PPM) where the code it's applied as a variable delay to the position of the pulse inside the chain(0 no delay,1 delay).

As you can see 1) & 2) fall inside what it would be TDMA techniques,3) it's CDMA one.

And I don't know, I will try to scratch my head a little bit more,
Hope it helps,

Title: Re: ideas for design of single pulse detection
Post by vivkr on Jan 28^th, 2008, 3:06am

Hi didac,

Indeed, I am not working in UWB, in fact not even in RF as you can guess from the pulse frequency levels I mentioned (low MHz).

Regarding your suggestions, I could not use (3) because using CDMA would mean that I would need to decode a train of pulses
first. My aim is to be able to detect a single pulse event already at the first pulse. For redundancy, 2-3 pulses are transmitted,
but ideally, the detection is accomplished on receiving the very first pulse. Synchronization is not a very big problem as some
amount of calibration is possible to ascertain the nominal return time. The application is kind of an object sensor.

Normally, the signal is reflected back by a reflector at fixed distance. In case there is an object in the path, this path is modified
and that gives the signal.

If another transmitter is lurking nearby, then the transmission from it is monitored during empty time slots and the internal
oscillator synchronized to transmit in a way as to avoid the interferer. This would be similar to your TDMA schemes.

The problem is that detection is still quite difficult and a very large dynamic range is needed upfront merely to accomodate
the very large amount of background noise and interference.

Regards
Vivek

Title: Re: ideas for design of single pulse detection
Post by didac on Jan 30^th, 2008, 7:48am

Hi Vivek,
I was thinking of this issue and maybe two options are feasible to investigate(although I personally find them complicated):
1)Use of superregenerative receiver concept, the problem is that since it's like a tuned oscillator I think pulse must be shaped to be "sinus like". Also althought in theory it's the topology with the greatest sensibility it needs a good isolation(causes spurious transmission in reception) and the control loop should be well defined.
2)Use of autocorrelation for reception, it's similar to matched filter but the template pulse it's the received pulse(which avoid synchronization), I've seen two approches:
a)Use of a "squarer":the received pulse feds the two branches of the multiplier
b)Sent two pulses instead of one, with a delay of Td, in reception the path to the multiplier is divided in two:one directly and the other through a delay(which I think is the hardest part to implement), this way the first received pulse is used for the multiplication with the second pulse.
Just some ideas that I had, not sure if any of them will be really useful.
Hope it helps,

Title: Re: ideas for design of single pulse detection
Post by fran2k5 on Jan 31^st, 2008, 7:01am

Hi,

probably the following reference could help:

Thresholded samplers for UWB impulse radar
Hjortland, H.A.; Wisland, D.T.; Lande, T.S.; Limbodal, C.; Meisal, K.;
Circuits and Systems, 2007. ISCAS 2007. IEEE International Symposium on
27-30 May 2007 Page(s):1210 - 1213

it's an all-digital idea...

Hope it helps,

frank

Title: Re: ideas for design of single pulse detection
Post by vivkr on Feb 1^st, 2008, 5:44am

Hi everyone,

Thanks for your responses. I hope that I can find a good solution based on these inputs.

Regards
Vivek

Title: Re: ideas for design of single pulse detection
Post by loose-electron on Feb 1^st, 2008, 10:30am

There are all kinds of DSP ways of dearling with this, but do you want the overhead of ADC and a DSP digital engine?

If so, give some thought to PRML type architectures (Partial response, maximum likelyhood, in sequene - BW limiting, ADC followed by a Viterbi detector)

If you want a simpler analog approach, think if peak detection would work. (amplitude qualify the peak of the pulse, and do it based upon the zero crossing of the derivative of the pulse)

both of the above aproaches have been used in mass storage (disk and tape) with good success.

Integration detection, and quantized feedback methods are NOT suggested.

email me if you need further explanations on these: jerry@effectiveelectrons.com

Title: Re: ideas for design of single pulse detection
Post by didac on Feb 2^nd, 2008, 11:11am

Hi,
loose-electron could you post any reference on the use of Viterbi as a pulse detector?, it's because I've always seen it as sequence detector(decoder,error corrector as you would like to name it but deals with bit streams) based in Maximum likelihood, moreover I always senn it used with bit chains coded following a FSM(Finite State Machine) method(like convolutional codes or Trellis Coded Modulation) and I'm a bit curious about this.
Thanks,
PS: when you say integration detection is not suggested you include the matched filter?(that it's the detection method that maximizes the SNR in a AWGN channel if my memory is not failing,moreover in multipath channels with CSI(Channel State Information) the Rake receiver can be derived using n-branch multipliers and maximum ratio combining).

Title: Re: ideas for design of single pulse detection
Post by loose-electron on Feb 4^th, 2008, 12:15pm

The Viterbi detector is best known in certain applications - some of which you mention, but in a genral application, if you can define the sampling points, their numeric values a Viterbi detector trellis can be defined to detect the maximum likelihood for that pulse in a noisy background.

All a viterbi detector does (simplified version) is determine the ideal pulse vs. observed pulsed difference in energy and determines which decision (one or zero) is the highest probablity.

Integration detection looks good if you are a math major. Take the integral of the pulse, have the pulses alternate in polarity, and the integral looks like an NRZ data stream. In an applications environment it falls apart due to non-ideal pulses, amplitude variance and similar issues, making the DC operating point slowly vary. Not good.

If you need more on this topic let me know, I spent three years doing industry research on better detection methods for disk drives, and the problem is very similar in nature.

-- Jerry

Title: Re: ideas for design of single pulse detection
Post by didac on Feb 4^th, 2008, 12:33pm

Hi Jerry,
Thanks for the explanation, the mapping of the pulse values fits well on my background, I've been doing a little bit google search and will try to look at it in depth when I have a little bit time. I've did a little bit of ML and MAP(Bayesian theory) estimators so I will add this little piece to knowledge vault just in case I needed it the future.
Thank you very much once again,