Every time I watch this guy, my knowledge is updated. Thanks Zach. It was very nice of you to formulate it, but it would be nice to see it on a sample design to reinforce the knowledge. May be a design from scratch that shows all steps through the determining the all parameters needed.
Zach, even though a digital trapezoidal waveform has infinite bandwidth, practical applications can consider that say after 20 harmonics, one can cut it off and claim "most" of the energy is in that portion. As such, there is still value in having a bandwidth equation for that situation. Of course it isnt the rule of thumb you posted for lossless short channels, but who has developed it?
The bandwidth determination requires solving a transcendental equation for your specific channel and input signal, so it's not something that has been universally developed. You can derive some cases for typical values of conductor and dielectric loss because the solution requires a numerical approach. Regardless of a rule of thumb or an equation, the point still stands: the channel modifies the signal bandwidth and there is some frequency where significant roll-off in the bandwidth begins (like a 3dB frequency). In fact if you look at the MSU PDF I posted, applying a knee frequency rule to a trapezoidal wave would tell you that the only signal bandwidth that matters is that which is the inverse of the symbol period, not the rise time, as that is where the 20dB/decade slope starts to intersect with the 0dB slope in the signal's power spectrum. If you were to apply that rule to a 56G bitstream, this 3dB frequency concept would incorrectly tell you that the channel can only support 19.6 GHz of bandwidth and rolls off everything at higher frequencies (it actually needs at least 28 GHz).
Every time I watch this guy, my knowledge is updated. Thanks Zach.
It was very nice of you to formulate it, but it would be nice to see it on a sample design to reinforce the knowledge.
May be a design from scratch that shows all steps through the determining the all parameters needed.
Damn Zach, you did an outstanding job addressing the misconceptions! Love it!
Thank you!
Just amazing 😮😮😮Jack
This formula can be used while selecting a DSO and its probes.
That is because those probes are effectively RC circuits. The output from that probe is then taken as the input to the oscilloscope.
Zach, even though a digital trapezoidal waveform has infinite bandwidth, practical applications can consider that say after 20 harmonics, one can cut it off and claim "most" of the energy is in that portion. As such, there is still value in having a bandwidth equation for that situation. Of course it isnt the rule of thumb you posted for lossless short channels, but who has developed it?
The bandwidth determination requires solving a transcendental equation for your specific channel and input signal, so it's not something that has been universally developed. You can derive some cases for typical values of conductor and dielectric loss because the solution requires a numerical approach. Regardless of a rule of thumb or an equation, the point still stands: the channel modifies the signal bandwidth and there is some frequency where significant roll-off in the bandwidth begins (like a 3dB frequency). In fact if you look at the MSU PDF I posted, applying a knee frequency rule to a trapezoidal wave would tell you that the only signal bandwidth that matters is that which is the inverse of the symbol period, not the rise time, as that is where the 20dB/decade slope starts to intersect with the 0dB slope in the signal's power spectrum. If you were to apply that rule to a 56G bitstream, this 3dB frequency concept would incorrectly tell you that the channel can only support 19.6 GHz of bandwidth and rolls off everything at higher frequencies (it actually needs at least 28 GHz).