Group Delay measurement with oscilloscope vs VNA

I am interesting in measuring small changes in group delay through an RF receiver module. The module is simply just a string of amps, attenuators, limiters, filters etc (no frequency conversion). Frequencies of interest are around 100MHz.

I initially used a vector network analyser to measure the group delay but this gives a very noisy result, even after using low IF bandwidth, increasing aperture points etc and making sure not to use too many sweep points (as this increases noise since GD=-dP/dF).

The VNA measurement gives a good idea of the absolute group delay and group delay change over wider frequency spans, but I need to try and measure down to 0.05ns/MHz and the noise is too high using a VNA for this.

I have now tried this measurement in the time domain - taking a CW signal from a signal generator, splitting the signal, with one path going to CH1 of the scope, and the other path going through the DUT and then into CH2. I then simply measure the time delay between the waveforms.

The problem is I am getting very different results to the VNA. With the VNA I was seeing about 0.5ns change between 60-80MHz, but with the scope I am measuring more like 3.5ns, which is WAY out of spec.

Does anyone have any idea why I am seeing such different results? Also, am I really measuring group delay by simply doing a time delay measurement on the waveforms? Or do I need to measure phase change and divide by frequency change? (I have done this and the results are even more strange).

I have removed the DUT from the measurement to check if it is a inherent measurement issue with either the scope or splitter but it doesn't seem to be.

Thanks in advance,
Mark

How are you comparing the time on your CRO? often they display the channels on alternate sweeps, so the time measurements are nonsense. If the input and output are meant to be anti phase, then view them on the A+B display. This so the in phase bit cancels and the residual out of phase bit can be seen and its angle from the amplitude, and from its angle,its delay. If the input and output are in phase do as above but pull the "invert" button on one channel.
Frank

what exactly are you measuring? A sine wave, or some broadband modulation or digital signal? You would want to measure a stepped sine wave...i.e. CW tone, make the measurement at 60 MHz, step freq to 61 MHz, remeasure, and so on.

Also, is there some isolation between the splitter and the DUT? like an amplifier with low S12? The reflected group delay of the DUT might be pulling the apparent time of the C1 reference, causing erroneous group delay measurement.

I know that modern Tektronix scopes display the true time relationship of all channels, I have carefully confirmed that. What scope type are you using?

The relation between the phase shift and the time delay is given by:
Time delay (sec) = [Phase Shift (°)] / [360 x frequency (Hz)]

Group delay is proportional to the rate of phase shift at each frequency of interest.
Group delay (sec) = [1 / 360] x [ΔΦ / Δf ]

Perhaps the VNA requires an optional high stability LO with low phase noise. What model is yours?

have you tried averaging? say compute the group delay from 100 frequency sweeps, add them all point by point, and divide by 100.

chuckey said:

How are you comparing the time on your CRO? often they display the channels on alternate sweeps, so the time measurements are nonsense. If the input and output are meant to be anti phase, then view them on the A+B display. This so the in phase bit cancels and the residual out of phase bit can be seen and its angle from the amplitude, and from its angle,its delay. If the input and output are in phase do as above but pull the "invert" button on one channel.
Frank

Click to expand...

Hi Chuckey, I'm not sure I understand. If the scope is displaying the two channels on separate sweeps this would be an issue. In this case the input and output are not necessarily meant to be in or out of phase, I simply want to know the delay through the signal chain. Can you clarify what you mean by A+B display on the oscilloscope? I assume I will still need the input and output connected to separate channels?
Thanks, Mark

- - - Updated - - -

biff44 said:

what exactly are you measuring? A sine wave, or some broadband modulation or digital signal? You would want to measure a stepped sine wave...i.e. CW tone, make the measurement at 60 MHz, step freq to 61 MHz, remeasure, and so on.

Also, is there some isolation between the splitter and the DUT? like an amplifier with low S12? The reflected group delay of the DUT might be pulling the apparent time of the C1 reference, causing erroneous group delay measurement.

I know that modern Tektronix scopes display the true time relationship of all channels, I have carefully confirmed that. What scope type are you using?

Click to expand...

Hi Biff44,

I am measuring a stepped sine wave. I don't think there is any pulling as I have a 40dB attenuator between the splitter and the DUT so match should be pretty good.

I'm using a R&S RTO1044.

Regards,
Mark

vfone said:

The relation between the phase shift and the time delay is given by:
Time delay (sec) = [Phase Shift (°)] / [360 x frequency (Hz)]

Group delay is proportional to the rate of phase shift at each frequency of interest.
Group delay (sec) = [1 / 360] x [ΔΦ / Δf ]

Click to expand...

Hi Vfone,

Thanks, I understand this. However what I'm unsure about is when doing a "time-delay" measurement on a scope, is this the same as "group delay" (both with units of time), or do I need to display the "time-delay" on the scope rather as "phase-delay" (in degrees) and from this use the equations for group delay (as you have given) to calculate group delay?

Regards,
Mark

SunnySkyguy said:

Perhaps the VNA requires an optional high stability LO with low phase noise. What model is yours?

Click to expand...

I'm using a R&S ZNC3. Yes you may be correct about a lower phase noise LO in the VNA, but unfortunately I don't have this option available hence my attempt to measure GD with a scope!

Thanks,
Mark

- - - Updated - - -

biff44 said:

have you tried averaging? say compute the group delay from 100 frequency sweeps, add them all point by point, and divide by 100.

Click to expand...

Yes I've tried averaging on the VNA but I still get a considerable level of noise in the group delay (larger than I'm trying to measure) even with the DUT out of the loop (VNA port 1 straight into VNA port 2, after calibration).
Mark

if you are using a "stepped sine wave" instead of a cw tone, then you are certain to have a difference between your oscilloscope and vna measurements. You DO understand what a stepped sine wave spectrum looks like, right?

biff44 said:

if you are using a "stepped sine wave" instead of a cw tone, then you are certain to have a difference between your oscilloscope and vna measurements. You DO understand what a stepped sine wave spectrum looks like, right?

Click to expand...

Hi Biff44,

By stepped sine wave I assumed you meant stepping a CW sine wave in frequency steps and measuring the delay for each step; "...make the measurement at 60 MHz, step freq to 61 MHz, remeasure, and so on", in which case in terms of spectrum it would be a single tone at any point in time. Is this not what you meant?

As far as I understood (and I may be wrong) I was effectively performing the same measurement that the VNA is doing but instead with a scope, i.e. measure the phase shift at discrete frequency points and then calculating the group delay from this data. However I'm not sure if I need to calculate the GD from the phase shift measured by the scope, or is the time-delay (given in ns) ,displayed by the scope, in fact already the group delay.

The problem is the VNA is telling me my system has about 0.5ns GD change over 60-80MHz, but with the scope I am measuring about 3.5ns GD change over the same span.

Mark