Oh you're of course right :) My rule of thumb was "7 to 9 times the fundamental frequency in bandwidth for reliable edges".
Cheers, Marcus On Mon, 2020-02-03 at 14:05 +0000, Chesir, Aaron M. wrote: > Your statement " A signal with 0.1 microsecond rise time definitely > has definitely more than 6 MHz bandwidth" is not necessarily true. > > Simple proof: > > The first 3 components of a 1 Mcycle/sec square wave are: > sin(2*pi*1e6*t) + (1/3) sin(2*pi*3*1e6*t) + (1/5) sin(2*pi*5*1e6*t). > > If you add just the above 3 components, the highest component of the > resultant signal is obviously 5 MHz. > > Its rise time is 0.1 microseconds. > > Aaron > > -----Original Message----- > From: Discuss-gnuradio < > discuss-gnuradio-bounces+achesir=mitre....@gnu.org> On Behalf Of > Müller, Marcus (CEL) > Sent: Monday, February 3, 2020 6:18 AM > To: mike.nel...@rdss.com; discuss-gnuradio@gnu.org > Subject: [EXT] Re: Recommendation for high sample rate receiver? > > Hi Mike, > thanks for following up on this: > A signal with 0.1 microsecond rise time definitely has definitely > more than 6 MHz bandwidth :) so that's where my confusion stems from. > > It will have something upwards of 20 MHz, rule of thumb should be > around 100 MHz bandwidths, and that fits nicely with the bare > necessary > 40 MHz bandwidth for do anything at a 25ns timescale. > As said, if you can capture that amount of bandwidth, you can infer > the timing – you really do not need 500 MS/s, as far as I can tell > from your description of the signal. > Then again, there might be complicating factors – extreme dynamic > range, for example. > > Could you tell us more details about your signal? And also, "as > accurately as possible" is not a spec; could you say "I need a timing > accuracy of X ns, given an SNR of Y dB", so that we could help you > there? > > Best regards, > Marcus > > On Sat, 2020-02-01 at 02:58 -0500, Mike wrote: > > Thank you to all who commented. > > > > The target signal of interest uses pulse modulation where each > > pulse > > is 1 microsecond in duration, with a rise time of less than 0.1 > > microsecond and a decay time of less than 0.2 microseconds. The > > goal > > is to identify the start (arrival) of a transmission at each > > receiver > > site as accurately as possible (better than 25 ns). > > > > Interpolation adds no useful information regarding start time, of > > course. Lower sampling rates lose arrival time resolution. > > > > No affordable SDR supports 500 MS/sec; I'm looking at A/D > > evaluation > > boards with an RF front end. > > > > > > Thanks! > > > > > > > > On 1/29/2020 10:34 PM, Kyeong Su Shin wrote: > > > To whom it may concern: > > > > > > Forgot to mention: There is a Wikipedia article, listing SDR > > > receivers with various capabilities ( > > > https://en.wikipedia.org/wiki/List_of_software-defined_radios ). > > > There's also something called OneRadio ( > > > http://www.oneradiocorp.com/ ). I saw an actual build of > > > OneRadio, > > > and it was pretty impressive (but expensive, of course). > > > > > > Do not expect these receivers to be well-supported by GNU Radio, > > > however. However (I think it is not necessary, but), if you > > > still > > > want to get a fast receiver and do not want to roll out your own > > > receiver using oscilloscopes or FPGAs, then I guess these > > > are possible alternatives. > > > > > > Regards, > > > Kyeong Su Shin > > > 보낸 사람: Kyeong Su Shin <kss...@postech.ac.kr> 대신 Discuss-gnuradio > > > < > > > discuss-gnuradio-bounces+ksshin=postech.ac...@gnu.org> > > > 보낸 날짜: 2020년 1월 30일 목요일 오후 12:10 > > > 받는 사람: discuss-gnuradio@gnu.org <discuss-gnuradio@gnu.org>; > > > mike.nel...@rdss.com <mike.nel...@rdss.com> > > > 제목: Re: Recommendation for high sample rate receiver? > > > > > > To whom it may concern: > > > > > > It is already well-discussed, but I would like to add a few > > > points: > > > > > > -If you absolutely want to have a such receiver (it's pretty > > > meaningless, as discussed already, but if you still want to), > > > then > > > you can grab a digital oscilloscope or a similar hardware and > > > attach > > > a RF frontend to it. You will end up losing some (actually, most > > > of) > > > samples, but you cannot run non-trivial data processing chains > > > at > > > 500MS/s in real-time with a generic desktop CPU anyway. > > > > > > -Regarding on why this is pretty meaningless (not using the > > > Nyquist > > > criterion or maths, but using intuitions): consider two > > > consecutive > > > samples, sampled by your receiver. Since the sampling rate is > > > way > > > higher than the bandwidth of the signal, these values are going > > > to > > > be nearly identical. There could be a bit of differences in the > > > amplitude and the phase, but the differences will be pretty > > > small > > > and will be easily washed out by the noise. You cannot expect to > > > get reliable TDOA results from that. You will have to > > > use > > > more samples to get more reliable results.. or just use a slower > > > receiver, anything that satisfies the Nyquist criterion. > > > > > > -If you know the structure of the transmitted signal (like PRNs > > > in > > > GPS), and if you are dealing with CDMA-like signals, then maybe > > > you > > > want to review the GPS receiver design principles and apply that > > > to > > > your design. Not sure if that's the case, though.. > > > > > > -Please consider power difference of arrival or phase > > > interferometry > > > as alternative methods. > > > > > > Regards, > > > Kyeong Su Shin > > > 보낸 사람: Qasim Chaudhari <qasim.chaudh...@gmail.com> 대신 Discuss- > > > gnuradio <discuss-gnuradio-bounces+ksshin=postech.ac...@gnu.org> > > > 보낸 날짜: 2020년 1월 30일 목요일 오전 11:05 > > > 받는 사람: discuss-gnuradio@gnu.org <discuss-gnuradio@gnu.org>; > > > mike.nel...@rdss.com <mike.nel...@rdss.com> > > > 제목: Re: Recommendation for high sample rate receiver? > > > > > > Hi > > > A high sample rate for such ns times of arrival resolution is > > > impractical. Same holds for high SNR and longer times of > > > measurement. GPS and most other high resolution positioning > > > systems > > > stitch the information together from the signal time of arrival > > > with > > > the carrier phase of arrival. Since carrier frequencies are > > > incredibly high, their phase can provide such ns accuracy > > > because > > > the phase information is preserved across the downconversion > > > stages > > > with sufficient linearity. For this purpose, the algorithms also > > > need to determine the integer number of arriving wavelengths. > > > > > > Cheers, > > > Qasim
