Book Signal Integrity for PCB Designs available on net
in [Design]
|
Prev: Toshiba laptop aggravation
Next: Quad LVDS to LVPECL?
From: Additya on 11 Mar 2010 12:05 Hello friends , I came across a book for Signal Integrity which is available on internet. Check out http://www.ezdia.com//Signal-Integrity-for-PCB-Designers/Content.do?id=1435 The complete book will be made available over a period of two weeks.
From: Joel Koltner on 11 Mar 2010 13:51 "Additya" <addy.ind(a)gmail.com> wrote in message news:86c70eaf-2b15-4a4d-b1a6-6a65a06de33f(a)m35g2000prh.googlegroups.com... > http://www.ezdia.com//Signal-Integrity-for-PCB-Designers/Content.do?id=1435 From the above: --- Tom - I was reading this rise time and knee frequency stuff. It says that the rise time is a significant thing and determines the highest frequency content. Does it mean that if I have a 1 KHz square wave that has a rise time of 100 ps, it should be treated as a high frequency signal with spectral frequency of or 3.5 GHz? Bob - Ideally, yes, this 1 KHz signal should be treated as a 3.5 GHz signal at least from some of the SI angle, e.g., EMI. But Tom tell me why will anyone design a circuit with 1 KHz frequency with such a fast rise edge? Does it make sense? A very fast rising edge makes sense only when you want to achieve high data rate. --- Seems like that needs a bit more qualification... more like, "while not too many applications require super-fast edges but have low repetition rates, sometimes this will occur..."
From: John Larkin on 11 Mar 2010 14:58 On Thu, 11 Mar 2010 10:51:27 -0800, "Joel Koltner" <zapwireDASHgroups(a)yahoo.com> wrote: >"Additya" <addy.ind(a)gmail.com> wrote in message >news:86c70eaf-2b15-4a4d-b1a6-6a65a06de33f(a)m35g2000prh.googlegroups.com... >> http://www.ezdia.com//Signal-Integrity-for-PCB-Designers/Content.do?id=1435 > >From the above: > >--- > >Tom - I was reading this rise time and knee frequency stuff. It says that the >rise time is a significant thing and determines the highest frequency content. >Does it mean that if I have a 1 KHz square wave that has a rise time of 100 >ps, it should be treated as a high frequency signal with spectral frequency of >or 3.5 GHz? > >Bob - Ideally, yes, this 1 KHz signal should be treated as a 3.5 GHz signal at >least from some of the SI angle, e.g., EMI. But Tom tell me why will anyone >design a circuit with 1 KHz frequency with such a fast rise edge? Does it make >sense? A very fast rising edge makes sense only when you want to achieve high >data rate. > >--- > >Seems like that needs a bit more qualification... more like, "while not too >many applications require super-fast edges but have low repetition rates, >sometimes this will occur..." 3.5 GHz is the 3.5 millionth harmonic of 1 KHz. So the power spectral density up there will be zilch. But a fast edge can ring at any frequency, and that could, for instance, mis-clock a flipflop. So just because a signal is 1 KHz doesn't mean it's "low frequency" and safe to use in a digital system. John
From: Fred Bartoli on 11 Mar 2010 15:43 John Larkin a �crit : > On Thu, 11 Mar 2010 10:51:27 -0800, "Joel Koltner" > <zapwireDASHgroups(a)yahoo.com> wrote: > >> "Additya" <addy.ind(a)gmail.com> wrote in message >> news:86c70eaf-2b15-4a4d-b1a6-6a65a06de33f(a)m35g2000prh.googlegroups.com... >>> http://www.ezdia.com//Signal-Integrity-for-PCB-Designers/Content.do?id=1435 >>From the above: >> --- >> >> Tom - I was reading this rise time and knee frequency stuff. It says that the >> rise time is a significant thing and determines the highest frequency content. >> Does it mean that if I have a 1 KHz square wave that has a rise time of 100 >> ps, it should be treated as a high frequency signal with spectral frequency of >> or 3.5 GHz? >> >> Bob - Ideally, yes, this 1 KHz signal should be treated as a 3.5 GHz signal at >> least from some of the SI angle, e.g., EMI. But Tom tell me why will anyone >> design a circuit with 1 KHz frequency with such a fast rise edge? Does it make >> sense? A very fast rising edge makes sense only when you want to achieve high >> data rate. >> >> --- >> >> Seems like that needs a bit more qualification... more like, "while not too >> many applications require super-fast edges but have low repetition rates, >> sometimes this will occur..." > > 3.5 GHz is the 3.5 millionth harmonic of 1 KHz. So the power spectral > density up there will be zilch. > Let's push it a bit further... 350MHz is still 350Kth harmonic of 1kHz and there's still almost nothing there. So we can safely ignore this part of the spectrum... .... and, once provided the signal fundamental frequency is low enough we can design a rather band limited channel while still achieving fast rise time. Cute! -- Thanks, Fred.
From: Joel Koltner on 11 Mar 2010 16:30
"John Larkin" <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote in message news:pfiip55lel2963k90umda3cubdhg2bli0o(a)4ax.com... > But a fast edge can ring at any frequency, and that could, for > instance, mis-clock a flipflop. So just because a signal is 1 KHz > doesn't mean it's "low frequency" and safe to use in a digital system. I think that's (sort of) what he's trying to say... it's just that there's plenty of application for fast pulse generators with lot repetition rates (e.g., TDR!), where you do need to treat the interconnects as "high speed." |