Can your arb generator do this?

[precaud]

It's actually a very simple thing but most can't do it. My Pragmatic/Tegam 2414A does it easily, and is mainly why I keep using it..

Generate a single burst of an arb waveform, with:
: variable clock rate (NOT mapping the waveform into a fixed-clock DAC)
: a separate sync pulse output coincident with the first point
: at defined intervals adjustable from roughly 2 secs to 40m Secs (that's milliSec).

The only other requirement would be at least a 12-bit DAC.

If yours can do this, what is it, and how do you set it up?

The reason I ask is, I have enquired directly to the manufacturers of a couple of current-model arb/function gens and they can not do it.

[exe]

The reason I ask is, I have enquired directly to the manufacturers of a couple of current-model arb/function gens and they can not do it.

It might worth listing manufacturers here, this may save time for somebody.

[joeqsmith]

It's actually a very simple thing but most can't do it. My Pragmatic 2414A does it easily, and is mainly why I keep using it..

Generate a single burst of an arb waveform, with:
: variable clock rate (NOT mapping the waveform into a fixed-clock DAC)
: a separate sync pulse output coincident with the first point
: at defined intervals adjustable from roughly 2 secs to 40m Secs (that's milliSec).

The only other requirement would be at least a 12-bit DAC.

If yours can do this, what is it, and how do you set it up?

The reason I ask is, I have enquired directly to the manufacturers of a couple of current-model arb/function gens and they can not do it.

Beyond just basic general requirement (memory depth, bandwidth, filters, clock rates, channels.....)  could you provide more details about these specific requirements you have.   

: variable clock rate (NOT mapping the waveform into a fixed-clock DAC)
Does this very while the waveform is being played back?  If so, I assume that function is programmable as well and may not be just a simple linear or log sweep.   Could you describe this?  What is the range of clock frequency you are after?

[precaud]

It might worth listing manufacturers here, this may save time for somebody.

One was GW Instek MFG-2000 series. Even with its separate built-in pulse gen, it can not trigger its own arb gen.

The other is pending confirmation from a tech to their CS.

[jpb]

Most arbs are fixed clock rate. I have an Agilent/Keysight one which is theoretically variable but it decimates and interpolates (in a clever way).

For true arbs you probably need something like a Tabor one where the clock rate does vary:-

https://www.taborelec.com/

but they are not cheap. I think that they used to (or still do) produce the more expensive Keysight arbs as OEMs.

[precaud]

Beyond just basic general requirement (memory depth, bandwidth, filters, clock rates, channels.....)  could you provide more details about these specific requirements you have.   

The basic requirements are easily met by any recent single-channel arb gen, except of course filter, because none except the the Keysights have programmable filters.

: variable clock rate (NOT mapping the waveform into a fixed-clock DAC)
Does this vary while the waveform is being played back?

In my application, no.

What is the range of clock frequency you are after?

Nothing outrageous - from say 30MHz down to maybe 5kHz, clock rate.

[joeqsmith]

So you run some number of sweeps of a wave form at some rate but between sweeps change the clock rate?   

Most arbs are fixed clock rate. I have an Agilent/Keysight one which is theoretically variable but it decimates and interpolates (in a clever way).

Strange.  I will at times I have had to sync the clock of the Arb to my circuits. 

I wonder for the Agilent if you are referring to the how they synthesize the clock from the DDS. 

[precaud]

So you run some number of sweeps of a wave form at some rate but between sweeps change the clock rate?

   

One ensemble of averages (typically 64) is taken at one clock rate, one set at another (lower) rate, and their FFT's are "spliced" together in the computer.

[DaJMasta]

It would take a fair amount of extra programming, I expect, but you could create your waveform, then write a script to expand or compress it to the right period and then load it into the arb memory for playback, so that your only requirement for your generator is a sample rate high enough to get you the fine gradations you need in adjusting frequency steps - and with a little bit of interpolation in your algorithm wouldn't make these requirements high.  That way at least you don't need a variable sample rate on the output.  If your variations are in a set pattern and aren't changing on the fly, then you could also just look for an arb with a lot of sample memory - then you can do that process making the variable frequencies into a single waveform and just play back the full test regimen.

Sort of hacky ways of doing it, but it would let you use a pretty wide variety of generators.

[joeqsmith]

Seems simple enough.  I'm not seeing the problem.  Are the Arb's clocks just not stable enough?  Not enough resolution?  Too much phase noise?  Could you use a stable RF generator to drive your Arbs sample clock if you need lower phase noise and better drift? 

Or are new low cost Arbs really so bad that you can't do much with them?   I haven't looked.

[precaud]

Seems simple enough.  I'm not seeing the problem.  Are the Arb's clocks just not stable enough?  Not enough resolution?  Too much phase noise?  Could you use a stable RF generator to drive your Arbs sample clock if you need lower phase noise and better drift? 

The clocks are stable enough, phase noise is not an issue. It just that, for fixed-clock-DACs, unless the clock rate needed (which is determined by the DSO) is an integer multiple of the arb's clock, then some interpolation will occur.

Or are new low cost Arbs really so bad that you can't do much with them?   I haven't looked.

I don't know if they're bad or good. I just enumerate the requirements and see if they'll do it. The point of this little exercise is to see if a more modern arb, with more bits and faster clock, can replace the Pragmatic. So far, nada. The Keysight 33522B is probably the most interesting and capable of 'em, but even it does not have a trigger rate generator and costs more than I'd like to spend...

[precaud]

Sort of hacky ways of doing it, but it would let you use a pretty wide variety of generators.

Yeah, I don't want to mess with interpolation. The waveforms are calculated for specific # of points and freq span combos.

[joeqsmith]

Seems simple enough.  I'm not seeing the problem.  Are the Arb's clocks just not stable enough?  Not enough resolution?  Too much phase noise?  Could you use a stable RF generator to drive your Arbs sample clock if you need lower phase noise and better drift? 

The clocks are stable enough, phase noise is not an issue. It just that, for fixed-clock-DACs, unless the clock rate needed (which is determined by the DSO) is an integer multiple of the arb's clock, then some interpolation will occur.

So you need higher resolution DDS synthesizer, or want one with a different technology all together? 


Looking at what appears to be the manual for yours, the clock appears fixed which is why I am confused about what you are asking.

"3.4.4 Sample Clock and Output Frequency
Output waveform frequency is a function of  both the clock frequency and the number of samples. (Output f = Clock f/# Samples.) Since the default clock frequency is 10 MHz and
the standard waveform default memory allocation is 1000 samples, the default output frequency is 10 MHz/1000 = 10 kHz.

Press the CLOCK/FREQ key. Either the clock frequency (SCLK) or the desired output frequency (FREQ) can be entered directly. Use the double-arrow key to move the selection to the top row of the display. Use the edit knob or keypad to set the desired value. (Any change in the number of waveform samples will also affect the output frequency, as discussed in Section 5.) "

[precaud]

So you need higher resolution DDS synthesizer, or want one with a different technology all together? 

OK, now you're forcing me to get really specific 

If the 2414A had output filters that tracked the clock rate (like the 33512B), I would be a happy camper and life would be good.

The newer arbs have faster clocks and typically 14 bits. So I figgered, if I could add two bits to the DAC, double (or more) the clock rate, and double my waveform sample length, that would reduce the errors caused by the stairstepping by at least a factor of two, probably more. It would be a reasonable thing to try, as long as the cost was not excessive and other functionality was maintained.

Looking at what appears to be the manual for yours, the clock appears fixed which is why I am confused about what you are asking.

"3.4.4 Sample Clock and Output Frequency
Output waveform frequency is a function of  both the clock frequency and the number of samples. (Output f = Clock f/# Samples.) Since the default clock frequency is 10 MHz and the standard waveform default memory allocation is 1000 samples, the default output frequency is 10 MHz/1000 = 10 kHz.

The clock is not fixed, it is variable from 0.1Hz to 20MHz. In that paragraph they're only explaining why the unit powers up with a "default" output frequency of 10kHz for standard (i.e. built-in) waveforms.

[precaud]

OK, so noone has answered the question in the positive. Since modern arbs all appear to be fixed-clock designs, let's relax the requirements, and allow for them. So now the question is, can your arb gen:

Generate a single burst of an arb waveform, with:
: at least 30MHz clock and 12 bits resolution
: a separate sync pulse output coincident with the first point (to trigger your DSO)
: at defined intervals adjustable from roughly 2 secs to 40m Secs (that's milliSec).

The latter feature is called "Trigger Generator Delay" in the 2414A. Other units may call it something else. It's basically the time between internally-generated triggers in Continuous or Free-Run mode.

The reason I first looked at the Instek MFG-2000's is that they all come with a separate pulse generator built-in. The pulse gen could generate the sync pulses at the desired rate AND trigger the arb. But their tech support says it can not do the latter.

[tautech]

I'm thinking a SDG2042X can do this but I'd need to run some checks.
I've asked the higher skilled users about your requirements in the SDG2042X thread so give a bit of time for those that have these units to respond.

Anyways, one of these:
https://www.siglentamerica.com/waveform-generators/sdg2000x-series-functionarbitrary-waveform-generators/

[precaud]

OK, thanks. BTW, it's not really an unusual use. If you're making stimulus-response measurements in an environment that is noisy and/or has a long settling time, such as with a transducer in a gas (like air) or liquid (like water), this is the exact sort of setup needed. Averages of single bursts with sufficient delay between them to allow for for settling before hitting it again. There are other uses too.

BTW2, I examined the SDG1032X manual last week and found no "trigger generator delay" setting.

[tautech]

OK, thanks. BTW, it's not really an unusual use. If you're making stimulus-response measurements in an environment that is noisy and/or has a long settling time, such as with a transducer in a gas (like air) or liquid (like water), this is the exact sort of setup needed. Averages of single bursts with sufficient delay between them to allow for for settling before hitting it again. There are other uses too.

BTW2, I examined the SDG1032X manual last week and found no "trigger generator delay" setting.

From your specs 1032X won't cut the mustard for you, instead focus on 2042X for its 16 bit and 8mpts.

The capabilities are varied, depending on waveform type like there is delay but only in Pulse that can't be swept. But you can sweep an arb waveform and most others.
Anyways, a dumb example of an AM modulated arb Stair with 10us pulses that took my fancy while stuffing around dead tired. 
I'll come back to this tomorrow.

[precaud]

BTW2, I examined the SDG1032X manual last week and found no "trigger generator delay" setting.

From your specs 1032X won't cut the mustard for you, instead focus on 2042X for its 16 bit and 8mpts.

I appreciate your help, but let's keep things in perspective. 16 bits is not necessary, and 8Mpts is practically irrelevant for my needs. The 1032 may not "cut the mustard" but its not because of those things...

The capabilities are varied, depending on waveform type like there is delay but only in Pulse that can't be swept. But you can sweep an arb waveform and most others.

The 2042X is two-channel, I don't need that. And I don't need to sweep the arb, either...

[jpb]

So you run some number of sweeps of a wave form at some rate but between sweeps change the clock rate?   

Most arbs are fixed clock rate. I have an Agilent/Keysight one which is theoretically variable but it decimates and interpolates (in a clever way).

Strange.  I will at times I have had to sync the clock of the Arb to my circuits. 

I wonder for the Agilent if you are referring to the how they synthesize the clock from the DDS.

I may be wrong on this - I'm referring to Keysight's True Form:
https://about.keysight.com/en/newsroom/backgrounders/33500B/

This is very different from a Tabor unit where the DDS clock itself is varied in rate (they have clock generators)
eg ww2571a (picked at random)

https://www.taborelec.com/ww2571a

250MS/s Performance
Higher performance test equipment and
systems are needed as products which
use increasing signal bandwidths are
developed. The sample rate generator can
be programmed from frequencies as low as
1.5Hz to 250MHz with superior waveform
quality and purity. For example, phase
noise is typically below 120dB/Hz at 10kHz
offset for a 10MHz sine wave

[Marco]

A fixed clock for the same investment/sophistication is always going to be cleaner than a PLL generated one, so it makes some sense to interpolate the signal data rather than PLL generate the clock. Especially with the higher clocked DACs.

How about TGA12100. Bit of a boat anchor, but Aim-TTi is still selling it.

[nfmax]

I had a play with my shiny new Agilent 33522B, and it is possible to do what you want, using a single channel (so a 33511B would also work).

Set up the ARB waveform with whatever sample rate you want, and select the preferred filter to give you smoothed signal output. Disable modulation & sweep, enable burst. Select burst on, N-cycle mode, #cycles to 1. On the trigger menu, set the source to 'timer', then on the 'trigger setup' menu set the time period to whatever you want. Back on the trigger menu, set Sync on, set the channel to whichever channel you are using. On the 'Sync setup' menu, choose normal or inverted as required. If you choose sync mode 'normal', you get a sync pulse that stays active for the duration of the ARB signal being bursted. If you choose the mode 'carrier', you can adjust the number of samples (at the ARB sample rate) for which the sync signal is active, down to a minimum of 4.

In both cases, the sync pulse starts at the start of the ARB waveform. You can adjust the burst phase to shift this, if your ARB signal starts & ends with a 'guard band' of constant DC level.

[tautech]

BTW2, I examined the SDG1032X manual last week and found no "trigger generator delay" setting.

From your specs 1032X won't cut the mustard for you, instead focus on 2042X for its 16 bit and 8mpts.

I appreciate your help, but let's keep things in perspective. 16 bits is not necessary, and 8Mpts is practically irrelevant for my needs. The 1032 may not "cut the mustard" but its not because of those things...

OK but to quote you:
 

errors caused by the stairstepping

Is the reason why I'd point you to 2024X models rather than 1032X and without the need to use any filters.

[precaud]

I had a play with my shiny new Agilent 33522B, and it is possible to do what you want, using a single channel (so a 33511B would also work).

Darn... that series of arbs looks better and better... maybe I'll keep an eye out for a used one...

Thanks for checking this.

[precaud]

OK but to quote you:
 

errors caused by the stairstepping

Is the reason why I'd point you to 2024X models rather than 1032X and without the need to use any filters.

My experience with wave generation is, the stairstepping is not significantly reduced just by having more bits; it's either by filtering at/near the clock rate, or a balanced combination of more samples and more bits. I.e. at the same sample rater, more bits doesn't buy you a whole lot unless the starting point is wholly inadequate, ind in this case, it is not...