Programmed frequency switching time

[precaud]

(I'm not sure this is the right forum for this question, maybe Metrology or Test Equip would be better?)

I'm considering setting up an automated "stepped sine wave" measurement. The simplified process would be:

Set new frequency --> allow to settle --> trigger digitizer/DSO --> collect data --> rinse and repeat

What I'm concerned about is the synthesizer/generator frequency settling time. After sending the new frequency (via GPIB or whatever), how long does it take to make the change and settle? For sake of discussion, lets' assume "phase-continuous switching", where the new freq's first phase point will be calculated from the old freq's last phase point.

I looked through some generator data sheets and most companies don't specify it. They specify squarewave settling time but that is something else.

The only generator I have here that specifies it is an HP 3325B. It specifies switching time as the time from receipt of command to when the output freq is within 1Hz of the programmed value, as:
<10mS for a 100kHz step,
<25mS for a 1MHz step
<70mS for a 10MHz step.

Are there others that spec better?

How does one measure this?

[JS]

Run your measurements with your sistem using a short piece of coax as your DUT. Take long samples in the range you expect this to be, analyze with a window till you find what you want...

JS

[precaud]

With that approach, there's no trigger from the receipt of command to start the clock.

[JS]

With that approach, there's no trigger from the receipt of command to start the clock.

scope's 2nd channel?

JS

[DaJMasta]

This is usually specified in RF generators, and there are definitely those that can do it faster.  I think the way to check would be just hook it up to the scope, then maybe single shot capture with a trigger pulse to step the frequency up and measure it there.  Add a little buffer and a little delay to process the command and you should have a good approach.

I suspect sending individual frequency commands would be more delay than just a settle time (also needs processing time), so I would setup the sweep (if this is available) then trigger it with trigger pulses and just offset your digitizer from taking its reading a fixed interval from the trigger pulse, the delay interval being whatever you found in looking at it with your scope plus some margin.

[Kleinstein]

For testing the manufacturer might look at the internal VCO control. It will show a step there and this step is easy to observe.

Modern generators based of DDS have essentially instant frequency steps - maybe some minor time for the filter to settle, but that is usually well below µs.

[precaud]

I suspect sending individual frequency commands would be more delay than just a settle time (also needs processing time),

Yes, and that's what I want to do. Capturing and examining the waveform period is the obvious part. The question is, how to trigger the scope to capture the entire event from the sending (or receipt) of the freq command. And mark it as T=0.

so I would setup the sweep (if this is available) then trigger it with trigger pulses and just offset your digitizer from taking its reading a fixed interval from the trigger pulse, the delay interval being whatever you found in looking at it with your scope plus some margin.

There are no external signals available that coincide with a freq change. Maybe I'll have to pick off one of the GPIB status lines?

For testing the manufacturer might look at the internal VCO control. It will show a step there and this step is easy to observe.

Given its resolution and stability, I don't think the 3325B is a VCO-based design. But I could be wrong.

Modern generators based of DDS have essentially instant frequency steps - maybe some minor time for the filter to settle, but that is usually well below µs.

That was my thinking as well. But I haven't seen any numbers to back it up. If marketing an instrument for ATE applications, one would think that would be a really important spec to give. The only instruments (which operate up to 20MHz) I've seen it spec'ed for are the 3325x family and near-relatives (3336x).