开云体育

A guy puts a Tek 2430A through the tests. (The real test starts at about the 4:00 mark.)

The 2430A runs right through 500 MHz, although the trigger stops around 300-plus.

Dave

On Wed, Jan 9, 2013 at 11:49 AM, Dave C <davec2468@...> wrote:

A guy puts a Tek 2430A through the tests. (The real test starts at about the 4:00 mark.)

The 2430A runs right through 500 MHz, although the trigger stops around 300-plus.

This is what you would expect. For a non-DSP front-end, in order to
avoid time domain aberrations, there has to be a smooth rolloff in the
frequency domain. Most Tek (and other) scopes have close to Gaussian
time response, so the frequency response is quite predictable.

--- In TekScopes@..., David DiGiacomo wrote:

On Wed, Jan 9, 2013 at 11:49 AM, Dave C wrote:

A guy puts a Tek 2430A through the tests. (The real test starts at about the 4:00 mark.)

The 2430A runs right through 500 MHz, although the trigger stops around 300-plus.

This is what you would expect. For a non-DSP front-end, in order to
avoid time domain aberrations, there has to be a smooth rolloff in the
frequency domain. Most Tek (and other) scopes have close to Gaussian
time response, so the frequency response is quite predictable.

True for scopes of this generation, but virtually all modern DSOs have moved to DSP, even the lost cost models. few modern high speed DSOs use Gaussian response. The rolloff attenuates the odd harmonics when looking at serial data eye diagrams, which is the most common application for the fastest scopes.
"Brick Wall" frequency response would be ideal, except such a filter has a lot of phase distortion which also distorts the eye. Better implementations use a maximally flat phase response filter or a higher order Bessel-Thompson filter to preserve the harmonics in the passband with minimal phase distortion.
Most sampling heads still have Gaussian response with a single pole roll off well through – 10 dB or more. They are very good for serial data as long as the -3 dB BW is well beyond the frequency content of the data (a rule of thumb for minimal distortion is a maximum of 1 dB attenuation at the fifth harmonic of the fundamental (5 X data rate/2 for NRZ, 5 X data rate for RZ))
Obviously, this guideline can not be met with the new standards that operate in the range of 28 Gb/s. However the eyes are nearly sinusoidal in these anyway, so there is little harmonic content. That said, the measured eye diagrams will not be truly representative of the real waveform.
- Steve

On Wed, 09 Jan 2013 21:16:01 -0000, "Steve" <ditter2@...> wrote:

--- In TekScopes@..., David DiGiacomo wrote:

On Wed, Jan 9, 2013 at 11:49 AM, Dave C wrote:

A guy puts a Tek 2430A through the tests. (The real test starts at about the 4:00 mark.)

The 2430A runs right through 500 MHz, although the trigger stops around 300-plus.

This is what you would expect. For a non-DSP front-end, in order to
avoid time domain aberrations, there has to be a smooth rolloff in the
frequency domain. Most Tek (and other) scopes have close to Gaussian
time response, so the frequency response is quite predictable.

True for scopes of this generation, but virtually all modern DSOs have moved to DSP, even the lost cost models. few modern high speed DSOs use Gaussian response. The rolloff attenuates the odd harmonics when looking at serial data eye diagrams, which is the most common application for the fastest scopes.
"Brick Wall" frequency response would be ideal, except such a filter has a lot of phase distortion which also distorts the eye. Better implementations use a maximally flat phase response filter or a higher order Bessel-Thompson filter to preserve the harmonics in the passband with minimal phase distortion.
Most sampling heads still have Gaussian response with a single pole roll off well through – 10 dB or more. They are very good for serial data as long as the -3 dB BW is well beyond the frequency content of the data (a rule of thumb for minimal distortion is a maximum of 1 dB attenuation at the fifth harmonic of the fundamental (5 X data rate/2 for NRZ, 5 X data rate for RZ))
Obviously, this guideline can not be met with the new standards that operate in the range of 28 Gb/s. However the eyes are nearly sinusoidal in these anyway, so there is little harmonic content. That said, the measured eye diagrams will not be truly representative of the real waveform.
- Steve

Here are my observations of a Tektronix MSO5204 (2 GHz at 10 GS/sec)
that I evaluated last year:

1. Within the limits of its record length, it was just as susceptible
to aliasing as any other DSO I have used including my 2232 and 2440
which is what I would expect for an instrument designed for clean time
domain response. Peak detect and FastAcq (histogram mode) prevented
aliasing.

2. Enabling the DSP filtering to limit the bandwidth introduced
significant time domain pulse distortion compared to the hardware
bandwidth limiting.

The Rigol oscilloscope I played with not long ago also had significant
time domain pulse distortion in all cases.