I have some data that I think is single side band (it comes from an ultrasound scanner). I wish to demodulate this so that the frequency spectrum is centred on th ecentre of the sideband. I know the original carrier frequency, but not the sideband width or centre. How can I demodulate to the centre of the sideband? One way i see used is to do an envelope detection (basically, square/ sqrt followed by low pass filter). But that modifies as well as shifts the spectrum, and I am interested in the spectral data unmodified. Chris ================== Chris Bore www.bores.com
Demodulation of single side band?
Started by ●February 2, 2009
Reply by ●February 2, 20092009-02-02
Chris Bore wrote:> I have some data that I think is single side band (it comes from an > ultrasound scanner). > > I wish to demodulate this so that the frequency spectrum is centred on > the centre of the sideband. I know the original carrier frequency, but > not the sideband width or centre. > > How can I demodulate to the centre of the sideband? > > One way i see used is to do an envelope detection (basically, square/ > sqrt followed by low pass filter). But that modifies as well as shifts > the spectrum, and I am interested in the spectral data unmodified.Bringing the center of a single-sideband signal to DC seems to me a strange operation. Half the demodulated signal will consist of negative frequencies (pardon the expression!) that overlap the other half. Won't an ordinary AM detector do that if the carrier is suppressed? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●February 2, 20092009-02-02
On Mon, 02 Feb 2009 14:42:48 -0500, Jerry Avins wrote:> Chris Bore wrote: >> I have some data that I think is single side band (it comes from an >> ultrasound scanner). >> >> I wish to demodulate this so that the frequency spectrum is centred on >> the centre of the sideband. I know the original carrier frequency, but >> not the sideband width or centre. >> >> How can I demodulate to the centre of the sideband? >> >> One way i see used is to do an envelope detection (basically, square/ >> sqrt followed by low pass filter). But that modifies as well as shifts >> the spectrum, and I am interested in the spectral data unmodified. > > Bringing the center of a single-sideband signal to DC seems to me a > strange operation. Half the demodulated signal will consist of negative > frequencies (pardon the expression!) that overlap the other half. Won't > an ordinary AM detector do that if the carrier is suppressed? > > JerryJerry, Jerry, Jerry! And you're and old-time amateur radio operator! Bringing the center of a SSB signal to DC is at the heart of the "Weaver method" of SSB reception and transmission -- on the receive side, the signal is quadrature downconverted to DC, the two channels are independently filtered with plain old low-pass filters of 1500Hz or so bandwidth, then they are quadrature up-converted by 1500Hz and added. Aside from the 1500Hz whistle, I understand that it works very well. -- http://www.wescottdesign.com
Reply by ●February 2, 20092009-02-02
On Mon, 02 Feb 2009 08:30:10 -0800, Chris Bore wrote:> I have some data that I think is single side band (it comes from an > ultrasound scanner). > > I wish to demodulate this so that the frequency spectrum is centred on > th ecentre of the sideband. I know the original carrier frequency, but > not the sideband width or centre. > > How can I demodulate to the centre of the sideband? > > One way i see used is to do an envelope detection (basically, square/ > sqrt followed by low pass filter). But that modifies as well as shifts > the spectrum, and I am interested in the spectral data unmodified. > > Chris > ================== > Chris Bore > www.bores.comYou'll have to take some measurements and do some guessing then. Google on "Weaver method" for how this was done in the '50s. If you can translate from vacuum tubes to DSP, you'll be good to go. -- http://www.wescottdesign.com
Reply by ●February 2, 20092009-02-02
Tim Wescott wrote:> On Mon, 02 Feb 2009 14:42:48 -0500, Jerry Avins wrote: > >> Chris Bore wrote: >>> I have some data that I think is single side band (it comes from an >>> ultrasound scanner). >>> >>> I wish to demodulate this so that the frequency spectrum is centred on >>> the centre of the sideband. I know the original carrier frequency, but >>> not the sideband width or centre. >>> >>> How can I demodulate to the centre of the sideband? >>> >>> One way i see used is to do an envelope detection (basically, square/ >>> sqrt followed by low pass filter). But that modifies as well as shifts >>> the spectrum, and I am interested in the spectral data unmodified. >> Bringing the center of a single-sideband signal to DC seems to me a >> strange operation. Half the demodulated signal will consist of negative >> frequencies (pardon the expression!) that overlap the other half. Won't >> an ordinary AM detector do that if the carrier is suppressed? >> >> Jerry > > Jerry, Jerry, Jerry! > > And you're and old-time amateur radio operator! > > Bringing the center of a SSB signal to DC is at the heart of the "Weaver > method" of SSB reception and transmission -- on the receive side, the > signal is quadrature downconverted to DC, the two channels are > independently filtered with plain old low-pass filters of 1500Hz or so > bandwidth, then they are quadrature up-converted by 1500Hz and added. > > Aside from the 1500Hz whistle, I understand that it works very well.I guess it depends on what you mean by the center of the sideband. Suppose a voice channel 3 KHz wide, modulated SSSB on a 10 MHz carrier, upper sideband. That puts the energy between 10,000 and 10,003 KHz. It's a mess if I zero beat 10,001.5 KHz with my little BFO. Chris didn't mention quadrature conversion or anything like Weaver. BTW, unless the signal is much oversampled, envelope detection is best done as sqrt(I^2+q^2), so quadrature comes into it that way. Jerry -- Engineering is the art of making what you want from things you can get. ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by ●February 2, 20092009-02-02
Jerry Avins wrote:> Tim Wescott wrote: >> On Mon, 02 Feb 2009 14:42:48 -0500, Jerry Avins wrote: >> >>> Chris Bore wrote: >>>> I have some data that I think is single side band (it comes from an >>>> ultrasound scanner). >>>> >>>> I wish to demodulate this so that the frequency spectrum is centred on >>>> the centre of the sideband. I know the original carrier frequency, but >>>> not the sideband width or centre. >>>> >>>> How can I demodulate to the centre of the sideband? >>>> >>>> One way i see used is to do an envelope detection (basically, square/ >>>> sqrt followed by low pass filter). But that modifies as well as shifts >>>> the spectrum, and I am interested in the spectral data unmodified. >>> Bringing the center of a single-sideband signal to DC seems to me a >>> strange operation. Half the demodulated signal will consist of negative >>> frequencies (pardon the expression!) that overlap the other half. Won't >>> an ordinary AM detector do that if the carrier is suppressed? >>> >>> Jerry >> >> Jerry, Jerry, Jerry! >> >> And you're and old-time amateur radio operator! >> >> Bringing the center of a SSB signal to DC is at the heart of the >> "Weaver method" of SSB reception and transmission -- on the receive >> side, the signal is quadrature downconverted to DC, the two channels >> are independently filtered with plain old low-pass filters of 1500Hz >> or so bandwidth, then they are quadrature up-converted by 1500Hz and >> added. >> >> Aside from the 1500Hz whistle, I understand that it works very well. > > I guess it depends on what you mean by the center of the sideband. > Suppose a voice channel 3 KHz wide, modulated SSSB on a 10 MHz carrier, > upper sideband. That puts the energy between 10,000 and 10,003 KHz. It's > a mess if I zero beat 10,001.5 KHz with my little BFO.The Weaver method would have you beat it with cos(10,001.5MHz) and sin(10,001.5MHz).> Chris didn't mention quadrature conversion or anything like Weaver. BTW, > unless the signal is much oversampled, envelope detection is best done > as sqrt(I^2+q^2), so quadrature comes into it that way.I'm aware of that, and would like Chris to elaborate on just what he is thinking on this. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●February 3, 20092009-02-03
On Feb 3, 5:30�am, Chris Bore <chris.b...@gmail.com> wrote:> I have some data that I think is single side band (it comes from an > ultrasound scanner). > > I wish to demodulate this so that the frequency spectrum is centred on > th ecentre of the sideband. I know the original carrier frequency, but > not the sideband width or centre. > > How can I demodulate to the centre of the sideband? > > One way i see used is to do an envelope detection (basically, square/ > sqrt followed by low pass filter). But that modifies as well as shifts > the spectrum, and I am interested in the spectral data unmodified. > > Chris > ================== > Chris Borewww.bores.comPossibly square the signal, then lock into twice the carrier frequency with a PLL. Then devide this freq by two (with a flip-flop) and multiply it back into the original signal.Garnish with a low-pass filter. Here's one I prepared earlier.. Hardy
Reply by ●February 3, 20092009-02-03
On Feb 2, 9:45�pm, Tim Wescott <t...@seemywebsite.com> wrote:> Jerry Avins wrote: > > Tim Wescott wrote: > >> On Mon, 02 Feb 2009 14:42:48 -0500, Jerry Avins wrote: > > >>> Chris Bore wrote: > >>>> I have some data that I think is single side band (it comes from an > >>>> ultrasound scanner). > > >>>> I wish to demodulate this so that the frequency spectrum is centred on > >>>> the centre of the sideband. I know the original carrier frequency, but > >>>> not the sideband width or centre. > > >>>> How can I demodulate to the centre of the sideband? > > >>>> One way i see used is to do an envelope detection (basically, square/ > >>>> sqrt followed by low pass filter). But that modifies as well as shifts > >>>> the spectrum, and I am interested in the spectral data unmodified. > >>> Bringing the center of a single-sideband signal to DC seems to me a > >>> strange operation. Half the demodulated signal will consist of negative > >>> frequencies (pardon the expression!) that overlap the other half. Won't > >>> an ordinary AM detector do that if the carrier is suppressed? > > >>> Jerry > > >> Jerry, Jerry, Jerry! > > >> And you're and old-time amateur radio operator! > > >> Bringing the center of a SSB signal to DC is at the heart of the > >> "Weaver method" of SSB reception and transmission -- on the receive > >> side, the signal is quadrature downconverted to DC, the two channels > >> are independently filtered with plain old low-pass filters of 1500Hz > >> or so bandwidth, then they are quadrature up-converted by 1500Hz and > >> added. > > >> Aside from the 1500Hz whistle, I understand that it works very well. > > > I guess it depends on what you mean by the center of the sideband. > > Suppose a voice channel 3 KHz wide, modulated SSSB on a 10 MHz carrier, > > upper sideband. That puts the energy between 10,000 and 10,003 KHz. It's > > a mess if I zero beat 10,001.5 KHz with my little BFO. > > The Weaver method would have you beat it with cos(10,001.5MHz) and > sin(10,001.5MHz). > > > Chris didn't mention quadrature conversion or anything like Weaver. BTW, > > unless the signal is much oversampled, envelope detection is best done > > as sqrt(I^2+q^2), so quadrature comes into it that way. > > I'm aware of that, and would like Chris to elaborate on just what he is > thinking on this.Chris may not be aware what Chris is thinking.. :-) My problem with this may arise partly from my difficulty in casting the problem into a domain with which I am more familiar. The ultrasound world uses terms whose meanings differ subtly from those with which I am familiar. :-) The signal is from an ultrasound scanner. There are at least two types of data. The first is quadrature, (I,Q) samples. The second is what they call 'the RF' by which they mean, samples of the returned signal - which is only real values. This second type is demodulated by using a Hilbert transform (to eliminate the negative frequencies and give us an analytic signal), and then demodulatingn with a complex oscillator at the (supposed) carrier frequency. This seems similar to the demodulation using the Weaver Method, except that I am not yet low pass filtering. My proble though, is to determine (or automatically lock to) the (unknown) center of the sideband. The carrier is typically at 10 MHz or so, and the sample rate is usually 4x this (for I,Q data) or 8x (for RF). These data are typically coded into the ultrasound scan data. (There are limitations on what is available, also on visibility into the precise working of the scanners due to commercial considerations). In the case of what I have called single sideband, I think (it is unclear..) that the pulse is modulated so that its spectrum is to one side of the carrier. So my exciting signal is not a carrier, but a pulse whose central frequency is at some distance from the nominal carrier. This stimulus, as it reflects back from the system under investigation, is modulated further by the system. The ultrasound images usually (it seems..) do an envelope detection on this, which has the effect of demodulating so that the 'center' of the spectrum always sits at DC. I think this means that the demodulation effectively follows the spectrum's center (however this may be defined). This is fine for pictures, but I am interested in analysis of the spectrum, and its meaning. For me, I wish to shift the spectrum by a fixed amount, and I want that amount to be the center of the exciting sideband modulation. It is a problem for me because normally people are only concerned to get a 'nice' picture, not to measure frequency components. The pictures 'look' good under a variety of demodulation and (often non- linear) filtering schemes, but I need the original spectrum properly shifted by the central exciting frequency. This is not like speech, where we could hear a difference from an expectation. With images of this sort, a small difference in the frequency shift by demodulation simply makes a different spatial frequency 'look like' DC. For example a region whose texture is in fact high spatial frequency may 'look like' a smooth region - and since we don't really know what we are looking at, we can't easily tell if that is a reasonable representation of the underlying system's texture or not. (Ultrasound images of the sort I am dealing with often have very broad spatial frequency spectra across many areas, so making high frequency look smooth may still yield a 'nice' picture. I get confused even writing about this :-) Because I am to deal with a variety of scanners, I need a generic method that can be universally applied. :-) It must be a simple problem, solved many times. Isn't it...? Chris> > -- > > Tim Wescott > Wescott Design Serviceshttp://www.wescottdesign.com > > Do you need to implement control loops in software? > "Applied Control Theory for Embedded Systems" gives you just what it says. > See details athttp://www.wescottdesign.com/actfes/actfes.html- Hide quoted text - > > - Show quoted text -
Reply by ●February 3, 20092009-02-03
Chris Bore <chris.bore@gmail.com> wrote:>It is a problem for me because normally people are only concerned to >get a 'nice' picture, not to measure frequency components. The >pictures 'look' good under a variety of demodulation and (often non- >linear) filtering schemes, but I need the original spectrum properly >shifted by the central exciting frequency. This is not like speech, >where we could hear a difference from an expectation. With images of >this sort, a small difference in the frequency shift by demodulation >simply makes a different spatial frequency 'look like' DC. For example >a region whose texture is in fact high spatial frequency may 'look >like' a smooth region - and since we don't really know what we are >looking at, we can't easily tell if that is a reasonable >representation of the underlying system's texture or not.When I took a medical imaging course in college, the point that was repeatedly hammered home is that medical imagers are too often designed to give a "pretty" result rather than a meaningful result. Apparently a lot of the reconstruction algorithms out there border on being deceptive. (I realize this tangential comment does not help you with your immediate problem...) Steve
Reply by ●February 3, 20092009-02-03
On 3 Feb, 19:39, spop...@speedymail.org (Steve Pope) wrote:> When I took a medical imaging course in college, the point that > was repeatedly hammered home is that medical imagers are too > often designed to give a "pretty" result rather than a meaningful > result. �This is probably true. In the context of medical imaging, the 'normal' state of the subject unders study (the human body) is known. So the operators don't scan for a 'true' picture, but rather for one that reveals 'anomalies' or deviations from the 'normal' state. For diagnostic purposes it suffices to generate an image that is somehow different from a 'normal' reference image. There is usually no point in getting anything more than that form the data; the medics wouldn't know enough DSP to make use of such additional info anyway. Rune
