Oscilloscope high frequency amplitude compensation

[cyberfish]

Just a random thought -

Why don't oscilloscopes compensate for high frequency attenuation in software if they know the scope's frequency response?

For example, if a 50MHz (-3db) scope sees a 50MHz signal with amplitude x, why doesn't it just multiply the output by 2 in software if they know the input signal must have amplitude 2x?

[Lightages]

This would require math in real time and slow down update rates. It would also double the noise.

[ejeffrey]

Because a 50 MHz scope has its bandwidth deliberately limited so they can charge more for a 100 MHz scope

[tekfan]

Compensating in software isn't a good idea. Even the Rigol has a varactor diode to make a low pass filter. All done in hardware not software. I guess the -3dB response is just a natural response of the amplifier. Anyway it's not guaranteed that the amplitude will drop by 3dB because every IC and transistor has a different frequency response.

[alm]

The -3dB point is not calibrated, the only calibration step is verifying that the level remains above -3dB over the bandwidth. Performance (eg. noise, distortion, phase shift, aliassing) tends to get worse with frequency, this will be amplified.

[Mechatrommer]

for the sake of measurement! this is still one of my mission, and still in failure due to lack of "budget" and "knowledge". i've somehow transported my thought to you cyberfish! we have to be aware that by recompensating a "profiled/attenuated" scope will as well recompensate/amplify noise and signal inaccuracy as janne mentioned. but why not we make it as feature we can turn ON and OFF? by keeping in mind that signal above scope BW cannot be trusted 100%. by nature, we cannot trust what we see in the region near the scope BW, ie -3dB and more attenuation. this feature also can give us rough idea of what the original signal is rather than calculate it ourself if we need to. this is also a kind of re-boosting scope BW.

[jahonen]

Note that it is not enough to measure the amplitude of the frequency response, in order to preserve the pulse shape, one needs phase also. That can be somewhat difficult. Only knowledge of complex frequency response (magnitude + phase) enables the calculation of proper compensating filter. Of course, this can't be stretched very far, and all factors to be compensated should be stable, otherwise the compensation will make things worse. And like said, the noise will increase.

But, this can be done offline without any changes to the scope in case somebody wants to try it out.

Regards,
Janne

[alm]

So you just need a vector network analyzer which is much more expensive than the scope to improve its frequency response .

If you have a square wave or sweep of known spectrum, you may be able to fit both the attenuation and phase shift curves to to the acquired signal. Make a model of how the frequency response and phase shift is expected to behave (at least first order approximation). Then figure out the parameters of your model with a non-linear curve fit. Fitting in the frequency domain (full FFT result, not just the magnitudes) would probably be easier.

You probably will need a scope with significantly more bandwidth to establish the reference signal.

I believe scope vendors use DSP to improve the performance for the high-end models. It's probably not really worth the trouble in the low end.

[chscholz]

That's a good idea, in fact that's what Danaher/Tektronix does with their high-end scopes (72004 series scopes).
These scopes have a 16 GHz bandwidth and "DSP boosts" the bandwidth to 20 GHz. 7 or so years ago this used to be the only way to
build high-speed realtime oscilloscopes.
The problem with this method is (among others) noise enhancement of the high frequencies.
Today Agilent goes to 32 GHz and LeCroy to 45 GHz both without DSP boosting.

[alm]

Don't they still use DSP to improve frequency response flatness / transient response? I believe that Tektronix was actually the last to adopt it, and that Agilent/Lecroy were using it first.

[chscholz]

Yes, that's my understanding, too; my guess is that DSP for flattening frequency/phase response of scopes (and high speed active probes!) is here to stay.

[w2aew]

That's a good idea, in fact that's what Danaher/Tektronix does with their high-end scopes (72004 series scopes).
These scopes have a 16 GHz bandwidth and "DSP boosts" the bandwidth to 20 GHz. 7 or so years ago this used to be the only way to
build high-speed realtime oscilloscopes.
The problem with this method is (among others) noise enhancement of the high frequencies.
Today Agilent goes to 32 GHz and LeCroy to 45 GHz both without DSP boosting.

These ultra-wideband realtime scopes, even the new 32GHz Agilent, all use DSP to flatten the amplitude/phase response.  Agilent and LeCroy do not let you turn this off, Tektronix does.  And yes, the 20GHz Tek 72004 scope also uses the DSP to boost the upper end of the frequency response to 20GHz, and that does raise the noise in this region.

The 45GHz LeCroy not only uses DSP - but it also uses something called Digital Bandwidth Interleaving.  This a technique where the full 45GHz of signal bandwidth is actually reconstructed from several frequency bands.  The input is split up into several frequency bands.  The higher frequency bands are downconverted using a broadband block conversion to translate to a lower frequency.  Each of these bands are then digitized, and then a massive amount of DSP is employed to "re-assemble" the signal for analysis. 

[gregariz]

My guess would be that since these high GHz scopes are such specialized bits of gear, with corresponding price tags, that they would be individually calibrated to correct for the non-linearities of their input responses.

[Mechatrommer]

My guess would be that since these high GHz scopes are such specialized bits of gear, with corresponding price tags, that they would be individually calibrated to correct for the non-linearities of their input responses.

we can do this on cheapo line.. but only if we have proper equipment(s).

[cyberfish]

Ah I guess phase is the problem.

Amplitude should be easy to calibrate. Just need a constant amplitude PLL built into the scope.

[Mechatrommer]

Ah I guess phase is the problem.
Amplitude should be easy to calibrate. Just need a constant amplitude PLL built into the scope.

duh!