Products > Test Equipment

Agilent 33500 series AWG and Trueform technology

(1/2) > >>

jpb:
The Agilent 33500 series of arbitrary waveform/function generators come up on e-bay fairly often - and on Agilent prime and the Trueform technology looks interesting
see here for details:

http://cp.literature.agilent.com/litweb/pdf/M8190-91050.pdf

They get around the jitter problems of fixed clock arbs (like the Rigol DG4000 series) by upsampling and then filtering and resampling which allows a fixed clockrate.

They claim it has all the advantages of fixed clock DDS (simple and effective image reject filtering and low cost) and of variable clock arbs (low jitter and no missed points).

BUT I am slightly cynical and notice the following two things about the specs for the 33500 series:

1.) Arbs cannot be swept and can only be amplitude or phase modulated NOT frequency modulated.
2.) A settling time of <200 ns to 0.5% of final value is listed.

I have a feeling that these two are related but I struggle to understand what this means in practice.

Settling time is easy to understand in terms of say a step change, but what exactly does it mean in terms of a constantly changing waveform? I would guess it means that the upsampling and filtering and resampling introduces a 200nsec delay - you essentially have some sort of pipeline (in computer terms) which perhaps needs to be flushed if frequency is changed.

If this is true then I'm not sure how great this approach is when compared with say using a RigolDG4062 at step-by-step mode and just presampling the waveform before loading it into arb memory. (I'm ignoring the advantage of sequenced waveforms and the fact that it has a lot more memory than the Rigol - I'm looking to understand the fundamentals.)

It also seems to be a bit of a disadvantage compared to a true arb which can rapidly sweep the sampling rate to give frequency changes and modulation.

Some unbiased technical assessment from clever forum members would be much appreciated!

nctnico:
In a nutshell: A played back waveform must be filtered depending on the samplerate otherwise you would see steps in a sine wave. The output filter is usually taken into account when creating an arb. waveform so the samplerate must be fixed.

jpb:

--- Quote from: nctnico on May 11, 2013, 12:20:55 pm ---In a nutshell: A played back waveform must be filtered depending on the samplerate otherwise you would see steps in a sine wave. The output filter is usually taken into account when creating an arb. waveform so the samplerate must be fixed.

--- End quote ---
Thanks for replying, but I don't understand what point you are making. Yes Agilent (as do others) does pre-distort some waveforms to account for the characteristics of the filter and this is much easier with a fixed sample rate. Also an image reject filter to remove the effects of the steps in the DAC is needed which is true also for variable clock arbs such as the Tabor WW5061, but that doesn't really answer my question - probably I didn't make myself clear enough.

To rephrase my queries:

In the trueform technology the process is :

i.) the arbitrary waveform samples from memory are expanded on the timescale by introducing blanks between and then filtering to essentially produce a lot of
interpolated points in-between - the memory point separation is dependent on the selected sample rate which is arbitrary but less than or equal to the maximum rate of
250MS/sec

ii.) this new waveform (or rather set of sampled points but with a very small time step equivalent to perhaps several GS/s) is resampled at 250MS/sec

iii.) the output is now filtered (again) to remove images arising from the 250MS/s DAC.

So the process allows arbitrary data sampling rates up to 250MS/s but always operates the DAC at 250MS/s - a clever system.

My questions are -

1.) Given that you can have any sample rate, why can't you sweep through a range of sample rates? (You can manually do runs at different sample rates but you
can't apply a sweep I think - at least not according to the data sheet.)

2.) What is this settling time of 200nsecs? Data sheets for other arbs don't mention settling time so it seems to be an intrinsic part of the Trueform process.

I guess I ought to address my questions to Agilent but I was wondering if anyone on the forums had experience of the system and its pros and cons versus true arbs with variable clocks and standard DDS systems with fixed clocks.

jpb:
Just as an update.

I asked this question on the Agilent forum and Agilent has confirmed that you CAN do a sweep of an arb so the data sheets are currently incorrect.

I still don't have an answer as to what the 200nsec settling time means in practice in the context of a time varying waveform.

Electro Fan:

--- Quote from: jpb on May 11, 2013, 11:10:02 am ---The Agilent 33500 series of arbitrary waveform/function generators come up on e-bay fairly often - and on Agilent prime and the Trueform technology looks interesting
see here for details:

http://cp.literature.agilent.com/litweb/pdf/M8190-91050.pdf

They get around the jitter problems of fixed clock arbs (like the Rigol DG4000 series) by upsampling and then filtering and resampling which allows a fixed clockrate.

They claim it has all the advantages of fixed clock DDS (simple and effective image reject filtering and low cost) and of variable clock arbs (low jitter and no missed points).

BUT I am slightly cynical and notice the following two things about the specs for the 33500 series:

1.) Arbs cannot be swept and can only be amplitude or phase modulated NOT frequency modulated.
2.) A settling time of <200 ns to 0.5% of final value is listed.
I have a feeling that these two are related but I struggle to understand what this means in practice.

snip

Some unbiased technical assessment from clever forum members would be much appreciated!

--- End quote ---

0.5% of 10 million is 50,000.  Does .05% of the final value mean that a 10MHz signal setting could wind up being 9.950MHz to 10.050MHz?

Navigation

[0] Message Index

[#] Next page