Hi Guys, Since last year the online "Electronic Design" magazine has published a series of online signal processing articles discussing the concept of "Signal Processing in the Density Domain." Have any of you read those articles *AND* been able to make sense out of them? Here are the first three articles: http://electronicdesign.com/analog/signal-processing-density-domain-part-i http://electronicdesign.com/analog/signal-processing-density-domain-part-ii http://electronicdesign.com/analog/signal-processing-density-domain-part-iii See Ya', [-Rick-]
Has anyone seen these articles?
Started by ●June 9, 2014
Reply by ●June 9, 20142014-06-09
On Mon, 09 Jun 2014 17:50:59 -0700, Rick Lyons wrote:> Hi Guys, > Since last year the online "Electronic Design" > magazine has published a series of online signal processing articles > discussing the concept of "Signal Processing in the Density Domain." > > Have any of you read those articles *AND* been able to make sense out of > them? Here are the first three articles: > > http://electronicdesign.com/analog/signal-processing-density-domain-part-i> > http://electronicdesign.com/analog/signal-processing-density-domain-part-ii> > http://electronicdesign.com/analog/signal-processing-density-domain-part-iii> > See Ya', > [-Rick-]I can see that the author is hugely proud of what strikes me as being a fairly simple and exceedingly niche application. Perhaps you're desperately looking for the answer to "why does this matter?" and you're not finding it because it's not there? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●June 9, 20142014-06-09
On 10.06.2014 02:50, Rick Lyons wrote:> > Hi Guys, > Since last year the online "Electronic Design" > magazine has published a series of online signal > processing articles discussing the concept of "Signal > Processing in the Density Domain." > > Have any of you read those articles *AND* been able to > make sense out of them? Here are the first three articles: > > http://electronicdesign.com/analog/signal-processing-density-domain-part-i > > http://electronicdesign.com/analog/signal-processing-density-domain-part-ii > > http://electronicdesign.com/analog/signal-processing-density-domain-part-iii > > See Ya', > [-Rick-] >Yep. I just read them, and they somehow make sense. The idea is to represent a signal by modulating the duty cycle. Just like PWM but instead of simple PWM the modulation source can be anything, even some random bit-stream. Just the ratio of high to low over time is what makes up the value of the signal. He then shows some ways to do arithmetic such as addition, multiplication etc with these signals. I haven't followed his writing on how he implements filters though. I do not see much advantage of doing things his way. There is one thing that I found interesting though: Using this approach you can represent signals as precise as you want without being limited by register size like we are in the digital domain or by noise in the analog world. Instead your trade-off becomes "computation speed" or "signal representation time". E.g. to represent a higher precision signal on a 1 bit data-stream you need more time to represent it using a duty-cycle on/off ratio. I think the author needs some kind of "killer use-case" to show how his design approach is better to standard analog or digital designs. Until then I'll keep doing it the tried and trusted way. /Nils
Reply by ●June 9, 20142014-06-09
"Nils" wrote in message news:bvn6otF60a8U1@mid.uni-berlin.de... On 10.06.2014 02:50, Rick Lyons wrote:> > Hi Guys, > Since last year the online "Electronic Design" > magazine has published a series of online signal > processing articles discussing the concept of "Signal > Processing in the Density Domain." > > Have any of you read those articles *AND* been able to > make sense out of them? Here are the first three articles: > > http://electronicdesign.com/analog/signal-processing-density-domain-part-i > > http://electronicdesign.com/analog/signal-processing-density-domain-part-ii > > http://electronicdesign.com/analog/signal-processing-density-domain-part-iii > > See Ya', > [-Rick-] >Yep. I just read them, and they somehow make sense. The idea is to represent a signal by modulating the duty cycle. Just like PWM but instead of simple PWM the modulation source can be anything, even some random bit-stream. Just the ratio of high to low over time is what makes up the value of the signal. He then shows some ways to do arithmetic such as addition, multiplication etc with these signals. I haven't followed his writing on how he implements filters though. I do not see much advantage of doing things his way. There is one thing that I found interesting though: Using this approach you can represent signals as precise as you want without being limited by register size like we are in the digital domain or by noise in the analog world. Instead your trade-off becomes "computation speed" or "signal representation time". E.g. to represent a higher precision signal on a 1 bit data-stream you need more time to represent it using a duty-cycle on/off ratio. Sort of like multi-precision arithmetic, 32 bits with an 8 bit arithmetic unit just takes more time... I think the author needs some kind of "killer use-case" to show how his design approach is better to standard analog or digital designs. Until then I'll keep doing it the tried and trusted way. /Nils
Reply by ●June 10, 20142014-06-10
On Mon, 09 Jun 2014 17:50:59 -0700, Rick Lyons <R.Lyons@_BOGUS_ieee.org> wrote:> >Hi Guys, > Since last year the online "Electronic Design" >magazine has published a series of online signal >processing articles discussing the concept of "Signal >Processing in the Density Domain." > >Have any of you read those articles *AND* been able to >make sense out of them? Here are the first three articles: > >http://electronicdesign.com/analog/signal-processing-density-domain-part-i > >http://electronicdesign.com/analog/signal-processing-density-domain-part-ii > >http://electronicdesign.com/analog/signal-processing-density-domain-part-iii > >See Ya', >[-Rick-]It did remind me of this: https://www.youtube.com/watch?v=s7A5xLk3hIs Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●June 10, 20142014-06-10
I don't follow the logic - first state that conventional digital filters "generally need multipliers up to 32 bit", then present a sinc filter which can be conventionally implemented without multipliers (i.e. CIC). Regardless of the articles, there is general interest in "non-conventional" signal processing aka PWM, sigma-delta, thermometer-coded because it could be done very efficiently on modern CMOS (small gates and compact structures with short interconnects). What is often forgotten is that the SNR increases very slowly with oversampling - I need to quadruple my rate to gain another bit. It just happens to work for audio which is "almost DC" but it gets difficult at higher rates. It's very noisy business, and managing the noise in general structures is IMHO difficult to impossible. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 10, 20142014-06-10
On Tuesday, June 10, 2014 4:49:23 PM UTC+12, mnentwig wrote:> I don't follow the logic - first state that conventional digital filters > > "generally need multipliers up to 32 bit", then present a sinc filter which > > can be conventionally implemented without multipliers (i.e. CIC). > > > > Regardless of the articles, there is general interest in "non-conventional" > > signal processing aka PWM, sigma-delta, thermometer-coded because it could > > be done very efficiently on modern CMOS (small gates and compact structures > > with short interconnects). > > > > What is often forgotten is that the SNR increases very slowly with > > oversampling - I need to quadruple my rate to gain another bit. It just > > happens to work for audio which is "almost DC" but it gets difficult at > > higher rates. > > It's very noisy business, and managing the noise in general structures is > > IMHO difficult to impossible. > > > > _____________________________ > > Posted through www.DSPRelated.comWell the power world has gone that way. What used to be dc servos are now PWM all the way and so are dc-dc convertors, invertors etc - all use PWM. Reminds me of this film https://www.youtube.com/watch?v=Doz5w2W-jAY
Reply by ●June 10, 20142014-06-10
On Tue, 10 Jun 2014 00:35:17 -0700 (PDT), gyansorova@gmail.com wrote: [Snipped by Lyons]> >Well the power world has gone that way. What used to be dc servos are now PWM all the way and so are dc-dc convertors, invertors etc - all use PWM. Reminds me of this film >https://www.youtube.com/watch?v=Doz5w2W-jAYHi gyansorova, Thanks for posting that video URL. I forgot just how cute that scene was. I'm gonna send that URL to my ex. [-Rick-]
Reply by ●June 10, 20142014-06-10
Reply by ●June 10, 20142014-06-10
Den tirsdag den 10. juni 2014 06.49.23 UTC+2 skrev mnentwig:> I don't follow the logic - first state that conventional digital filters > > "generally need multipliers up to 32 bit", then present a sinc filter which > > can be conventionally implemented without multipliers (i.e. CIC). > > > > Regardless of the articles, there is general interest in "non-conventional" > > signal processing aka PWM, sigma-delta, thermometer-coded because it could > > be done very efficiently on modern CMOS (small gates and compact structures > > with short interconnects). > > > > What is often forgotten is that the SNR increases very slowly with > > oversampling - I need to quadruple my rate to gain another bit. It just > > happens to work for audio which is "almost DC" but it gets difficult at > > higher rates. > > It's very noisy business, and managing the noise in general structures is > > IMHO difficult to impossible. >doubling the sample gives you an extra 3dB, doubling the sample rate with a second order delta sigma gives you ~9dB (in the band of interest) I did hear something at Uni some 15 years ago about using delta sigma to do filters with single bit multipliers one reference I could quickly find was this: http://www.eecg.toronto.edu/~johns/nobots/papers/pdf/lewis91.pdf -Lasse
